Low-Cost Heating Modalities Allow the Detection of Biomarkers for Plant Infection Using Rapid Evaporative Ionization Mass Spectrometry (REIMS) That Are Pathogen Specific

To support a predicted global population of over 9 billion people by 2050, (1) agricultural production will need to increase markedly. One of the barriers to this is the burden of agricultural pathogens and pests, which reduce annual crop yields of some crops, such as rice, by over 40%. (2) The use of agricultural pesticides has allowed substantial increases in productivity, but comes with associated risks of the development of resistant pathogens and environmental damage due to runoff (3) and off-target effects. (4) Microbial pathogens (including bacteria, fungi, parasites, and viruses) are common pathogens in agricultural crops. Due to the broad range of plant pathogens, diagnostics can be a challenge, and visible signs of infections typically appear at a point where the infection is established. Earlier detection/diagnosis of plant pathogens could, therefore, allow for earlier and more effective interventions. This could decrease the overall use of agricultural pesticides as infection would be less established and affect a smaller proportion of crops, and also prevent the spread of pathogens within an agricultural setting. (5) Plant parasitic nematodes (PPNs) are globally distributed, soil-borne pests, which exert a significant economic burden on global agriculture. (6) Through costs associated with control and crop yield loss, it has been estimated that these parasites are responsible for economic losses greater than £100 billion per year. (7) Diagnosis of PPN crop infection is very challenging based on visual inspection of plant foliage. Therefore, it requires assessment of crop roots in the field and collection of PPN tissue for laboratory-based molecular diagnostics and/or analysis of soil samples for cysts. As a result, current PPN diagnostics lack throughput, convenience, and sensitivity. Bacterial plant pathogens pose additional challenges to agricultural productivity, largely linked to their faster growth rate and ease of dispersal compared to PPN. Symptoms of plant bacterial infections are typically more visually identifiable due to the impact on the above-soil plant structures, particularly leaves and fruiting bodies. Nevertheless, at this point, bacterial infections are typically well-established and will prevent early interventions to limit pathogen spread and associated impact on crop productivity. (8)

Current methods for diagnostics can be based on the observation of visual symptoms by trained personnel, but this can be costly and time-consuming and, for many bacterial pathogens, not species-specific. Molecular methods are increasingly being developed for the detection of plant pathogens, particularly the use of polymerase chain reaction (PCR) assays, loop-mediated isothermal amplification (LAMP) assays, and high-throughput sequencing. However, these are still costly and time-consuming, which prevents their broad adoption. (9) When they are used, due to their cost, it is not typically feasible for them to be used as a screening tool. Mass spectrometry has a much lower cost-per-sample than many molecular methods, albeit with a typically higher capital expenditure required for instrumentation. Nevertheless, this offers the potential for low-cost and potentially high-throughput analysis of plant material for screening and surveillance.

Mass spectrometry-based metabolomics has been used to elucidate host–pathogen interactions during plant infections and potentially identify metabolite-based traits that could be used to breed more resistant varieties of crop plants. (10) Fewer studies have used this approach as an alternative method of diagnosis. Success has been seen using both liquid chromatography (11) and gas chromatography, (12) or a combination of both separately, (13) coupled to mass spectrometry, but has typically focused on just one pathogen. These approaches also require time-consuming and resource-intensive extractions and analytical runs, which limit their potential for screening and/or rapid diagnostics. Techniques within the field of ambient ionization mass spectrometry are characterized by an ability to perform ionization on samples in a minimally modified form in their native environment, with many taking place in air. (14) Such approaches have the potential to make mass spectrometry analysis more accessible, less costly, and with higher-throughput than conventional methods. One such technique is rapid evaporative ionization mass spectrometry (REIMS) which uses heating to vaporize a sample and release gas-phase ions which are analyzed through a mass spectrometer. (15) The technique was initially developed as a tool for the intraoperative identification of cancerous tissues during surgery, (16) but saw broad application to clinical microbiology diagnostics, (17) food authenticity analysis, (18) and high-throughput screening, (19) among others, due to the utilization of various modalities for sample heating, including laser ablation. (20) These, however, have been typically costly and would likely pose a barrier to broad adoption. In this work, we explored a range of low-cost heating modalities (namely, a 450 nm laser and soldering iron), which cost approximately £150/$200 each, and optimized them for the analysis of plant leaf material. Each was then utilized in the analysis of plant leaf material from plants that had been infected with either a PPN (Meloidogyne incognita) or bacterial (Pseudomonas syringae) pathogen against uninfected controls, with the intention of identifying pathogen-specific biomarkers.

Experimental Section

Rapid Evaporative Ionization Mass Spectrometry Analysis

All heating modalities were fitted with a bespoke-fitted aspirator resin head made with an SL1S three-dimensional (3D) printer (Prusa Research, Czech Republic), which gave a clearance of approximately 2 mm between the sample and opening. Approximately 2 m of poly(tetrafluoroethylene) (PTFE) tubing with 1.5 mm internal diameter linked the aspirator head to the inlet of the Xevo G2-XS quadrupole time-of-flight (QToF) mass spectrometer (Waters, U.K.). Just prior to the inlet, the analyte-containing smoke was mixed with 2-propanol solvent (21) containing leucine enkephalin at a concentration of 0.1 ng/μL in a stainless-steel T-piece. The 2-propanol solvent was introduced at a flow rate of 200 μL/min using an Acquity I-Class Plus binary solvent manager (Waters). The combined mixture entered the REIMS interface (Waters, U.K.), where it was heated to approximately 700 °C via collision with a Kanthal ribbon surface to remove the 2-propanol solvent prior to entry into the ion guide of the mass spectrometer. Prior to use each day, the mass spectrometer underwent calibration and detector voltage setup using the manufacturer’s recommended procedure. Mass spectra were acquired over a 50 to 1200 m/z range at a scan rate of 2 scans per second in negative ion detection mode with an instrument resolution of approximately 20,000.

Data Processing and Statistical Analysis

Acquired mass spectra were imported into Abstract Model Builder software (Version 1.0.1966.0, Waters, Hungary), where individual acquisition windows were manually selected and then underwent lockmass correction to 554.2615 m/z, background subtraction, and peak binning to 0.1 Da width bins. Any features that significantly correlated (Spearman’s coefficient < −0.5 or >0.5, with Bonferroni corrected P value <0.05) with leucine enkephalin intensity and/or run order were removed using a custom R pipeline. For parameter optimization, data was analyzed in GraphPad Prism (version 10.0.1). For statistical analysis, data was uploaded to the online MetaboAnalyst 5.0 (22) workflow. Data was subjected to total ion count normalization, Log10 transformation, and Pareto scaling prior to univariate analysis through Mann–Whitney U test with FDR multiple hypothesis correction (with a significance threshold of p < 0.05) and multivariate principal component analysis (PCA) and sparse partial least-squares discriminant analysis (sPLS-DA), which were validated with leave-20%-out cross-validation based on replicates across all samples. MetaboAnalyst does not support a leave-one-leaf-out cross-validation, which would be more statistically robust. Biomarker analysis was completed within the MetaboAnalyst workflow using the same preprocessing parameters. Where required, tentative metabolite annotations were assigned from the Human Metabolome Database (23) (HMDB), with a mass accuracy threshold of <10 ppm alongside literature searches of potential matches.

Results and Discussion

Optimization of Heating Modalities

Three modalities were tested against tomato plant leaf material to optimize the heating parameters. A total of six replicates for each combination of parameters were completed. Summary figures for the best-performing parameter combinations, based on total ion count intensity (TIC) and number of features above set intensity thresholds (above 10³, 10⁴, 10⁵, and 10⁶, respectively), for each heating modality are given in Figure 1, with a full breakdown given in Supporting Figure S1. Across all three, relatively similar levels of TIC and feature threshold intensities were observed. For the CO₂ laser, Figure 1a–c, a higher frequency of laser pulsatile operation was associated with both a lower TIC and a lower number of features meeting a high intensity threshold, particularly at lower laser powers. This suggests that longer laser pulses are needed in order for the heating temperature to reach a critical point for the release of metabolites for ionization, which has been observed in other biomasses using a CO₂ laser for heating during REIMS analysis. (24) The analysis region from the CO₂ laser, Figure 1a, can be seen to be highly reproducible due to its integration with an automated XYZ gantry robot. For the CO₂ laser, 3W at 5 Hz was chosen as the optimal heating parameter. Fewer parameter combinations were available for testing for the 450 nm laser, with only the laser depth setting available to change in the control software, which is used as a proxy of power, with settings of 50, 75, and 100% available. This means an approximate range of 1.5–3.0W of laser power was tested. Based on all three parameters as shown in Figure 1d–f, the 100% depth setting resulted in a considerably higher TIC and number of features above set threshold intensities, substantially at 10⁵ and 10⁶ thresholds. A higher TIC variability is, however, observed at this setting, which may be linked to the observed variability in analytical repeats, as shown in Figure 1h. Similar levels of variability were observed for soldering iron analysis which, due to its hand-held operation, is largely expected. Although variability was higher for both 450 nm laser and soldering iron heating, it is still within a relatively small window for the 450 nm laser (between ≈ (1–2) × 10⁹ TIC), while for soldering iron heating it was a 4-fold (between ≈ (2–8) × 10⁹ TIC) range. For soldering iron heating, a temperature of 400 °C was determined as optimal through a combination of the highest TIC intensity and observed variability of measurements.

Anal. Chem. 2025, 97, 27, 14230–14238: Figure 1. Optimization of three heating modalities for the plant leaves. Heating optimization was completed using different combinations of operating conditions for each heating modality tested. For the CO2 laser, a selection of the highest performing conditions is given, with all available in Supporting Figure S1. For each of the CO2 laser, 450 nm laser, and soldering iron respectively, panels (a–c) give a representative image of a leaf postanalysis with each modality with 2 mm scale bar; (d–f) show the number of features identified above a set threshold of intensity; and (g–i) show the total ion count intensity. The color legends are shared between panels for each heating modality.

Comparison of Optimized Parameters

The REIMS metabolite fingerprinting data from the three optimal parameters for each heating modality were compared, Figure 2. This was to determine whether there were differences in the detected profiles among the three. The CO₂ and 450 nm lasers showed similarities in both TIC intensity and frequencies of features above certain thresholds. The soldering iron had both a higher TIC intensity and features above all thresholds, except 10³, as shown in Figure 2a,b. The variability in signal intensity was, however, higher than that of both lasers, which may reflect the manual use of the soldering iron compared to the automated analysis workflows for both lasers. The weight of the ablated material for each modality is shown in Supporting Figure S2 and corresponds largely to the surface area of analysis. Similar levels of reproducibility are shown in ablated material, however, which may suggest that the higher variability in signal intensity for the soldering iron approach comes from impacts on ionization or ion fragmentation during heating. PCA modeling of the metabolite fingerprinting data, Figure 2c, shows clear separation between all three heating modalities, with the soldering iron having the largest degree of variation. When the top 1000 features, based on mean intensity across all replicates, found in the metabolite fingerprints of all three modalities are compared, Figure 2d, it can be seen that the majority of features (53.9%) are shared between the three. The remainder are either solely found in each modality’s top 1000 features based on intensity, with the 450 nm laser accounting for the highest number of these at 22.3%, or shared between two, with the greatest overlap between the CO₂ laser and soldering iron at 18.8% of features. Based on these results, all three heating modalities appear to be effective at heating and generating an analyte-containing smoke from plant material. Although all three produce different metabolite fingerprints, the majority of features are shared. Based on visual inspection of raw and processed mass spectra in Supporting Figures S3 and S4, respectively, the main differences are not the presence/absence of features but rather their relative intensities. We do not, therefore, believe that different ionization mechanisms are at work, but rather that each modality differs in the intensity of ions generated. This is further supported by the spread of PC loadings across the mass-to-charge range shown in Supporting Figure S5 from Figure 2c. The automated laser modalities have lower variation compared to the hand-held soldering iron but come with associated laser hazards that require additional considerations for containment. For subsequent analysis of leaves from infected plants, it was decided to continue only with the 450 nm laser and soldering iron heating modalities. This is because they are the lowest cost of the three, and smaller and more moveable than the CO₂ laser, which would potentially make them easier to adapt for in situ analysis.

Anal. Chem. 2025, 97, 27, 14230–14238: Figure 2. Comparison of REIMS spectral fingerprints between optimized parameters for all three modalities. Panel (a) shows the number of features above set intensity thresholds; panel (b) compares total ion count intensities; panel (c) gives a 2D scores plot from principal component analysis of the three modalities; and panel (d) shows the number of features solely found within the top 1000 features by intensity for each modality, or shared between modalities.

Conclusions

Ambient ionization techniques hold considerable promise in expanding the scope of applications in which mass spectrometry can be of benefit. REIMS offers a particularly flexible technique in that the method of heating can be easily adapted to suit the type of sample under analysis. Here, we have shown that two very low-cost methods of heating can be used for the analysis of plant material. The use of a soldering iron did not, however, perform well in identifying markers of infection, and this may be due to the hand-held nature of the tool and the inherent variability associated with it. Lasers have been shown to be capable of conducting REIMS analysis of a broad range of samples but have typically been very high cost. We used a low-cost 450 nm laser that is typically used for the etching of wood and other materials to successfully analyze plant material and identify pathogen-specific diagnostic biomarkers. Importantly, our models relied on individual univariate AUC models and, thus, would not require a mass spectrometer capable of untargeted analysis. This offers the potential that REIMS could be deployed with a low-cost 450 nm laser and a single or triple quadrupole instrument that is typically less expensive and more robust than profiling instruments. Although this would be beneficial for a diagnostic platform, our work not only shows the potential of low-cost lasers for broader REIMS analysis but also the potential of REIMS for the analysis of plant material in situ. This could offer a number of potential benefits associated with speed, resources, and reducing sources of potential bias associated with storage and processing when compared to conventional mass spectrometry techniques paired with preionization chromatographic separation.