by Brian Musselman, DART Technical Advisor, Bruker Applied Markets
In many instances, scientific advancement and analytical utility rely on characterizing a sample in its native state. This is especially true when it comes to the identification and investigation of animal products and plant materials, which ideally call for the detection of chemical compounds at trace levels in bugs, leaves, feathers, bark, soil, or any number of other complex matrices. A defining characteristic of these complex matrices is that they are far from pristine, typically incorporating a diverse array of compounds other than the analyte of interest. Applying techniques such as GC-MS to such samples can be useful, but the associated sample preparation is a limiting factor with respect to throughput that can simultaneously erode the relevance of the resulting data, especially with respect to the identification of easily fragmented molecules.
Direct analysis in real-time mass spectrometry (DART-MS) can characterize solids, liquids, and gases with no sample preparation, making it inherently fast. This is a defining characteristic as our ability to analyze larger data sets advances. The throughput and capabilities of DART-MS make it feasible to measure a significant number of samples in relatively short timeframes and then apply advanced statistical techniques, chemometric processing tools and artificial intelligence to advance learning. In this article we summarize work in this field from Professor Rabi Musah's lab at the State University of New York at Albany, and look at some of the exciting results that this approach is delivering in areas such as chemical ecology and forensic science.
Introducing DART-MS
The critical components of a DART ion source are shown in figure 1. DART operates at ambient pressure with the sample suspended in the open air at ground potential. Inert gas - such as helium or nitrogen - flowing through the system is converted into plasma by a corona discharge around the high voltage needle. Subsequent lens filtration removes metastables leaving a stream of neutral, highly energetic atoms and molecules, which flow through the heater and out of the ion source via the grid voltage. Increasing temperature facilitates the desorption of molecules from the sample surface while the grid voltage prevents ion-electron recombination; it also serves as a source of electrons for negative-ion formation.
Figure 1: DART is a rapid, efficient, and broadly applicable ionization source for MS
Releasing this electronically and/or vibronically excited gas stream initiates a cascade of reactions that create reagent ions from water or solvent vapor close to the analytical target, affecting its chemical ionization at the inlet to the mass spectrometer; either positive or negative ions can be created. The resulting MS spectra provide a complete chemical fingerprint of the sample.
From a practical perspective, this technology delivers an array of benefits including:
- Speed and convenience: With little to no requirement for sample preparation
- Soft ionization: Enabling the representative analysis of fragile, easily fragmented molecules, though collision induced dissociation (CID) is achievable (where required) through the application of higher voltages
- Highly sensitive: Measurements with nanogram level detection
- Representatively analyze compounds: The ability to analyze compounds that have been deposited or adsorbed onto the surface of a sample, or that are being desorbed into the atmosphere
Figure 2: DART-MS in practice - simple and rapid, even for complex matrices.
These features are particularly advantageous relative to chromatographic methods – GC-MS and LC-MS - when it comes to generating large sets of relevant data for the analysis of complex matrices, to access the information needed to advance scientific understanding. Figure 2 illustrates the ease-of-application of DART-MS; the following studies demonstrate the value of the resulting data.
Case study 1: Do plants contribute to environmentally problematic sulfur in the environment?
A study was carried out to investigate mechanisms by which plants might release sulfur compounds, notably carbonyl sulfide (OCS), thereby contributing to sulfur-based pollution in the environment. OCS is a long-lived compound with notable abundance in the stratosphere that has been securely linked with the formation of sulfuric acid. Mimosa pudica (M. pudica) was selected as a model for the study. As a touch sensitive plant - its leaves close in response to stimuli - it releases a foul-smelling odor when uprooted.
From experiments involving GC-MS headspace analysis following the crushing of plant samples, this odor has been attributed to the release of carbon disulphide (CS2). In these analyses, only CS2 was detected and no other sulfur compounds were observed. CS2 oxidizes to OCS and is a known precursor of sulfur dioxide in the environment, so these results have been used to ascribe a direct contribution from M. Pudica to sulfur compounds in the environment.
In this study, DART-MS was used to further elucidate the sulfur-releasing mechanism of the plant.
Shedding new light on sulfur release mechanisms with DART-MS
M. pudica seedlings were germinated in agar under sterile conditions to eliminate the possibility of introducing sulfur by, for example, bacterial or microbial colonization of the roots. Volatiles in the headspace above the growing seedlings were sampled and concentrated using a solid phase microextraction (SPME) polydimethylsiloxane (PDMS) fiber and then analyzed in both positive and negative ion mode DART-MS. Figure 3 shows the resulting spectra.
Figure 3: In contrast to GC-MS data, DART-MS spectra of volatiles in the headspace above M. Pudica seedlings detect the presence of multiple sulfur compounds with no evidence of either CS2 or OCS.
DART-MS headspace analysis reveals a range of stable compounds along with harder to detect intermediates (see figure 3). Some of these compounds are securely associated with odor and are known precursors to CS2 and OCS, neither of which were detected. A primary conclusion of the study was that the CS2 detected in GC-MS is a likely result of the transformation of these compounds in the injection port of the GC, which is typically held at substantially elevated temperature.
Harnessing the informational insight provided by DART-MS to advance chemical ecology
The ability of DART-MS to detect and monitor in real-time the wide range of sulfur compounds released by M. Pudica makes it possible to gain new insight into the sulfur chemistry associated with the plant and to develop a better understanding of its behavior.
Further tests were carried out to determine the conditions under which M. pudica roots released the odor. Stimulation of the roots by non-animate objects such as a metal spatula or cotton swab produced no effect while direct contact with a finger or dragging the seedling across soil (plant held by tweezers, roots in contact with the soil) produced a substantial odor in less than 5 minutes. DART-MS analysis indicated the same spectra of sulfur compounds pre- and post-stimulation. However, it was found that odor release could be securely associated with elevated levels of specific compounds (see figure 4). Though these compounds are always released by the plant, certain types of touch trigger a boost in their levels, which makes them detectable by odor.
Figure 4: DART-MS data reveal that touching M. pudica stimulates an increase in several sulfur compounds, raising them above the human odor threshold.
This novel finding is directly attributable to the application of DART-MS and could not have been made with GC-MS because of its inability to detect the cocktail of sulfur compounds associated with the behavior of the plant. Applying DART-MS robustly differentiated odor-causing stimuli and, crucially, revealed the sulfur compounds that the plant actually releases, providing new and valuable information, thus elucidating the role of plant-released sulfur compounds within the environment. It begs the question as to whether the CS2 release routinely attributed to other plants is correct or simply an artefact of GC-MS analysis and opens up a new avenue for chemical ecology studies.
Case study 2: Is it possible to rapidly identify and classify new psychoactive substances (NPS) using DART-MS?
Crime labs routinely face the challenge of identifying and classifying new psychoactive substances (NPS), since once a drug is scheduled, crime manufacturers carry out structural modifications to evade detection. In this second study, analyses were carried out to determine the potential of DART-MS within this context.
Figure 5: MS spectra produced via electron ionization (EI - top), DART ionization at 20 V and DART ionization at 90 V illustrate the greater informational output achievable with CID.
Figure 5, shows MS spectra for 3,4-methylenedioxy-α-pyrrolidinopropiophenone (MDPPP), a stimulant. Analysis following electronic ionization (EI), a hard ionization technique, is relatively uninformative. DART-MS analysis under soft ionization conditions (20 V) generates pristine spectra but is also limited with respect to structural determination. DART-MS analysis at 90 V, on the other hand, produces spectra with distinguishing peaks at multiple m/z values. Increasing the voltage during DART-MS measurements accelerates molecules within the electric field, giving rise to CID, thereby revealing a chemical signature for the drug to support precise structural determination. This is an important gain, however, for forensic scientists linking the NSP with a known class of drugs is also a primary goal.
Figure 6: Creating neutral loss spectra (bottom) provides a method for securely classifying drugs.
Neutral loss spectra can be generated from DART-MS data via a process of mathematical manipulation, more specifically the subtraction of m/z values for each fragment from a reference value, in this case the protonated precursor of the molecule. Figure 6 illustrates this process and its ability to securely identify drugs in the same class. In this example, DART-MS data provide structural insight for both MDPPP and 3,4-methylenedioxy-α-pyrrolidinobutiophenone (MDPBP), while neutral loss spectra identify them as close relations, both belonging to a specific class of cathinones.
Applying DART-MS in drug forensics
This ability to both structurally characterize and classify drugs is extremely valuable. Figure 7 shows two examples of large datasets for a range of drugs, generated by DART-MS and analyzed using advanced statistical methods: principal component analysis (PCA) and hierarchical clustering.
Figure 7: Applying advanced statistical techniques - PCA analysis (top) and hierarchical clustering (bottom) - to large datasets of neutral loss spectra produces a tool for drug classification.
PCA analysis was applied to neutral loss spectra for multiple drugs that can be classified as ethcathinones, methylenedioxy compounds, and pyrrolidine containing substances. Feature masses were securely identified, and the analysis clearly and correctly grouped the drugs into the three classes. For the hierarchical clustering analysis, neutral loss spectra were similarly generated for 59 cathinones and amphetamines. The resulting dendrograms clearly identify distinct classifications, linking the drugs on the basis of structural similarity. Both techniques illustrate how DART-MS can be used to establish a secure foundation for the rapid classification of NPS.
In this example, the ability to apply DART-MS in CID mode is crucial to access the structural information required for secure identification and to generate neutral loss spectra. These fragmentation spectra are the key to successfully classifying unknown drug samples. A further point here is that the ease and speed of DART-MS is critical to the generation of the large datasets required to apply advanced statistical methods and by extension generate robustly predictive models.
Case study 3: Can DART-MS be used to develop a convenient feather identification protocol for wildlife forensics?
The Convention on International Trade in Endangered Species (CITES) protects a wide range of plants and animals but presents those working in wildlife forensics with the task of securely identifying samples that violate the regulations. In this third study, analyses were carried out to determine whether DART-MS could differentiate detached feathers from endangered macaw species, thereby presenting a method for their detection in items such as ceremonial garments. Trained ornithologists can identify whole macaws and potentially feathers but this is obviously an expert task. Alternative identification techniques include microscopy, which has the limitation of not being species specific, and DNA testing where there are relatively few fully sequenced genomes in this area. A more convenient solution suitable for non-expert users would therefore be advantageous.
Figure 8: DART-MS spectra for feathers from two different macaws, both regulated under CITES Appendix 1, provide clear differentiation.
Figure 8 shows examples of DART-MS spectra for two different types of macaws – scarlet and military – both of which are regulated under CITES Appendix 1 (species threatened with extinction). These spectra clearly differentiate the feathers, confirming the potential value of the technique. In total, 651 feathers were analyzed from 26 individual birds, representing 5 different species; 10 replicates were generated for each feather*.
*Note - all feathers were collected from birds held in captivity with no harm caused to any wildlife.
Developing an identification tool for endangered macaws
Figure 9 shows results from two mathematical analyses of the resulting data: PCA-discriminant analysis (DA) and partial least squares (PLS)-DA. The DART-MS spectra exhibit intra-species similarities and inter-species differences and consequently relate clusters clearly in different areas of both plots. In canonical correlation analysis, the PCA scores were found to explain 81% of the variation observed in the results; latent variables are similarly successful in detecting the difference in the PLS-DA analysis.
Figure 9: PCA-DA and PLS-DA analyses of DART-MS spectra for feathers from five types of macaw show clear and relevant clustering, illustrating the potential to provide a secure predictive tool for feather identification.
An additional statistical analysis technique - support vector machine (SVM) - was also applied to the data and a “fused classifier” was then developed from all the results. Table 1 shows data from an external validation of all four models.
Table 1: External validation of the developed models illustrates that while all four offer good performance, the fused classifier is particularly effective.
Feathers from blue and yellow (BY), red and green (RG) and scarlet (SC) macaws were included in the original data set. In general, all the models are extremely good at correctly assigning these feathers when presented with new samples; an exception is the PCA-DA model, which exhibits poor performance for SC feathers. Feathers from the great green (GG) macaw were not included in the training set and none of the models should therefore identify these. The PLS-DA and fused classifier both perform well here; the other two, less so. Overall, the fused classifier performs best, providing extremely reliable and accurate identification for species included in the training set while at the same time delivering minimum false positives for the GG.
This study illustrates the ability of DART-MS to provide chemical fingerprints that securely differentiate macaw species. Using these data in combination with multiple statistical analysis techniques to make a fused classifier proved highly effective in providing a tool for rapid and robust species identification; feathers from confiscated ceremonial headdresses were securely identified as belonging to endangered species. Models such as this are easily transferrable to point-of-need and provide an effective solution for reliable detection, even by relatively unskilled DART-MS users. Furthermore, far from being a one-off, these results add to the growing number of example applications in which DART-MS has been successfully used to identify plant and animal specimens.
In conclusion
The three applications explored here are varied but all topical. Together they illustrate the broad applicability and utility of DART-MS, and they share some common features. In each case, the ability of the technique to produce sensitive, reproducible, highly differentiating data is critical. But equally important is ease-of-use and speed. The robust application of advanced statistical analysis techniques and chemometric processing tools relies on having large datasets for training. Generating such data sets with DART-MS is straightforward and practical, lessening the workload imposed by longer, more involved techniques. As these examples demonstrate, the unique capabilities of DART-MS play an important role in advancing scientific understanding and delivering practical tools for application at the point-of-need.