P.-H.
Stefanuto†
*a,
K. A.
Perrault†
b,
R. M.
Lloyd
c,
B.
Stuart
b,
T.
Rai
d,
S. L.
Forbes
b and
J.-F.
Focant
a
aCART, Organic and Biological Chemistry Group, Chemistry Department, University of Liège, Allée de la chimie B6c, B-4000 Liège, Belgium. E-mail: phstefanuto@ulg.ac.be
bCentre for Forensic Science, University of Technology Sydney, PO Box 123 Broadway, NSW 2007, Australia
cDepartment of Chemistry, University of Leicester, University Road, Leicester LE1 7RH, United Kingdom
dSchool of Mathematical Sciences, University of Technology Sydney, PO Box 123 Broadway, NSW 2007, Australia
First published on 25th February 2015
This study demonstrates the first documented use of comprehensive two-dimensional gas chromatography – high-resolution time-of-flight mass spectrometry (GC×GC-HRTOFMS) for volatile organic compound analysis in the forensic sciences. High-resolution mass spectral data provided higher confidence in analyte identification. GC×GC-HRTOFMS will be valuable for future studies of decomposition odour and other complex volatile matrices.
The odour produced by decomposing remains is a complex mixture of various VOCs. Due to the complexity of the decomposition VOC profile, traditional one-dimensional gas chromatography (1D GC) often suffers from insufficient peak capacity to accurately separate the large number of analytes present, thus providing low mass spectral library matches and leading to the potential for compound misidentification. The decomposition VOC profile also exhibits a wide dynamic range whereby the presence of both high-level and low-level VOCs contribute to the complexity of the matrix. Background VOCs (of similar structure and behaviour to decomposition VOCs) are also present in the profile. For these reasons, decomposition VOC profiling benefits substantially from analysis by comprehensive two-dimensional gas chromatography (GC×GC). This comprehensive analysis allows two independent separation mechanisms to occur by attaching two GC columns together at a junction known as a modulator. Due to the narrow peaks that are produced by modulation between the columns, time-of-flight MS (TOFMS) is often coupled with GC×GC in order to provide the required acquisition rate (>50 Hz). Recently, GC×GC coupled to low-resolution (LR) TOFMS has been employed in this field leading to a more developed understanding of cadaveric decomposition odour.1–8
Recent developments in environmental monitoring have investigated the coupling of GC×GC with high-resolution time-of-flight mass spectrometry (GC×GC-HRTOFMS).9–13 Unit mass resolution using a low-resolution TOFMS may be insufficient for compound detection in complex matrices that exhibit interferences.14 GC×GC-HRTOFMS provides the potential to combine the high peak capacity of the GC×GC system with accurate mass spectral data. The mass measurement accuracy provides the ability to estimate the chemical formula of analytes, providing higher confidence in analyte identification.14 Differences in the mass measurement between analytes and interferences also affords improved deconvolution in complex matrices.14
In the non-targeted analysis of decomposition odour, criticism of compound identification is often met. There are various challenges encountered in decomposition odour profiling that make this concern challenging to address including: (1) the substantial costs associated with obtaining a large database of chemical reference standards; (2) the impossibility of obtaining chemical standards for every compound identified; (3) the lack of certified reference materials for the sample matrix; and (4) the limitation in replicating generated samples because of ethical, legal, and logistical challenges associated with obtaining human or animal remains. Thus, tools for increasing the confidence in mass spectral identifications are required in order to improve the accuracy of decomposition odour profiling.
In this study, the use of GC×GC-HRTOFMS was investigated for the analysis of decomposition VOCs in soil beneath carrion. Assessment of the instrumentation on a restricted data set was desired in order to determine the added value this instrumentation could have for the field of decomposition VOC analysis, specifically for future longitudinal studies involving large set-up and collection of field trial data. Determining whether absolute quantification would be possible from samples using the developed method was also of interest via the evaluation of the linear range of calibration curves for representative standards. As GC×GC-HRTOFMS is a novel and developing technique, quality control information of the mass accuracy data was required to ensure consistency and reliability of the results for future studies. The use of GC×GC-HRTOFMS for the VOC profiling of complex biological matrices presented herein may also be applied to other areas of forensic VOC monitoring as well as environmental, metabolomics and food science applications.
The HRTOFMS was operated in electron ionization (EI) mode with an ionizing voltage of 70 eV. An acquisition rate of 50 Hz was used with a mass range of 35–400 m/z and an acquisition delay of 1 minute. The plate voltage was 2150 V with a sampling interval of 0.25 μs. Data were acquired using MassCenter version 2.6.2b (JEOL Ltd.). Instrument tuning was performed using perfluorokerosene (PFK) (Tokyo Chemical Industry Co. Ltd., Tokyo, Japan) and the mass resolution was 7637 at m/z 293. Data was analysed in GC Image 2.5 HR (Zoex Corporation) using the GC Project and Image Investigator features. A cumulative image was created using all samples being analysed. A template was built using this image and removal of column bleed and other artefacts that were not specific to the analysis was performed. The feature template was then applied to the centroided sample files (.7rw). The configuration used to generate the cumulative image involved baseline correction, a 0.8 second baseline shift, and a minimum blob volume of 40000. Compound identification was made by comparison to the 2011 NIST library with a manually-applied threshold match of 700.5,17 Linear retention indices from the 1D column were used to further verify analyte identifications along with mass measurement error from the raw profile data (.7rw). Compounds of interest were identified by comparing the Fisher ratio (FR) for the normalised volume of each compound to a critical F value (Fcrit). Where FR > Fcrit, the variance between the two classes was significant and the compound was considered to be decomposition-specific. This data analysis approach has been previously demonstrated using complex multivariate GC×GC data.18 Principal component analysis (PCA) was performed in The Unscrambler X version 10.3 (CAMO Software, Oslo, Norway) for visualisation based on scores (samples) and loadings (analytes).
The GC×GC-HRTOFMS instrumentation exhibited dissatisfactory modulation of early-eluting analytes. This likely occurred due to their high volatility and the lack of cryogenic modulation. As a result, the oven method from previous work was adjusted by removing the 5 minute hold at 35 °C at the beginning of the sample run. Cryogen-free cold jet cooling has the potential to be problematic in profiling the entire decomposition odour profile. Cryogenic modulation would provide the best possible results due to the lower temperature reached in the cold jets (i.e. between −196 and −210 °C), allowing improved modulation of lighter compounds below C7. However, the use of a cryogen-free cooling system significantly reduces operational costs. Many low boiling point compounds detected in previous work still remained detectable using the method in this study (e.g. benzene, dimethyl disulphide, etc.).
Previous reports have demonstrated that HRTOFMS has a sufficient acquisition rate at 25 Hz to be coupled with GC×GC at no detriment to quantification.19 The maximum acquisition rate of the GC×GC-HRTOFMS was 50 Hz, which further expands its utility as a GC×GC detector. A slice of the 1D and 2D traces are represented in Fig. 1. Although the HRTOFMS detector exhibits a lower acquisition rate (i.e. maximum 50 Hz) than that of the LRTOFMS detector (i.e. typically 100 Hz or higher), its maximum acquisition rate was capable of providing a sufficient number of scans across the width of the narrow peaks generated by the modulator. For a typical peak eluting at the detector with a 200 millisecond width, 10 scans would be performed across the peak providing a high quality GC×GC chromatogram for performing analyte quantification and mass spectral deconvolution. Although past research demonstrated that HRTOFMS has a sufficient scan rate to be a good candidate for GC×GC detection,19 this confirms the assertion for the wide dynamic range experienced with decomposition odour samples.
Chromatogram alignment was necessary in order to ensure that the compounds reported in one sample were aligned with those from other samples. This alignment was conducted based on retention time (tR) and mass spectral matches between chromatograms. This was carried out based on the high-resolution mass spectral data. Therefore, alignment of high-resolution data provides a more robust alignment process, thus further increasing confidence in analyte identification across samples within a data set. In decomposition odour analysis, studies are often performed longitudinally and can last months or even years. Slight changes in response and conditions over time can result in difficulty comparing results over experimental days. High-resolution data alignment facilitates a robust comparison of analytes across an entire study, thereby providing more reliable results.
Compound name (previously referenced in) | Chemical formula | CAS # | Library forward match | Library reverse match | 1 t R (min) | 2 t R (s) | Mean signal-to-noise ratio | Fisher ratio (FR) | Linear retention index (LRI) | Exact m/z | Measured m/z | Mass error (ppm) | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | Heptadecane, 2,6,10,14-tetramethyl- | C21H44 | 18344-37-1 | 790 | 741 | 30.72 | 0.544 | 399.71 | 39.89 | 1403 | 198.2348 | 198.2315 | 16.2 |
2 | Benzonitrile3,5,6,8,15,16,20–22 | C7H5N | 100-47-0 | 915 | 878 | 20.35 | 1.980 | 299.14 | 31.80 | 1059 | 103.0422 | 103.0393 | 28.0 |
3 | Decane, 6-ethyl-2-methyl- | C13H28 | 62108-21-8 | 806 | 715 | 32.02 | 0.551 | 54.16 | 17.10 | 1450 | 140.1565 | 140.1553 | 8.45 |
4 | Thieno[2,3-c]pyridine | C7H5NS | 272-12-8 | 796 | 788 | 27.83 | 2.434 | 42.63 | 10.16 | 1298 | 135.0142 | 135.0105 | 27.7 |
5 | Naphthalene4,20,22–26 | C10H8 | 91-20-3 | 892 | 823 | 26.35 | 1.548 | 58.53 | 9.68 | 1248 | 128.0626 | 128.0587 | 30.8 |
6 | Benzenemethanol, α,α-dimethyl-22–24 | C9H12O | 617-94-7 | 745 | 743 | 23.43 | 2.043 | 60.02 | 8.81 | 1152 | 136.0888 | 136.0851 | 27.1 |
7 | Benzene, 2-methoxy-4-methyl-1-(1-methylethyl)- | C11H16O | 1076-56-8 | 742 | 648 | 27.20 | 0.926 | 26.45 | 8.23 | 1275 | 164.1201 | 164.1168 | 20.3 |
8 | Pentadecane6,15,16,20,26,27 | C15H32 | 629-62-9 | 838 | 806 | 33.35 | 0.560 | 713.76 | 8.23 | 1500 | 212.2504 | 212.2473 | 14.7 |
9 | Ethanol, 2-phenoxy-6 | C8H10O2 | 122-99-6 | 787 | 780 | 28.08 | 3.895 | 106.44 | 8.16 | 1307 | 138.0681 | 138.0643 | 27.5 |
10 | Nonane1,4,6,8,24–27 | C9H20 | 111-84-2 | 911 | 803 | 14.82 | 0.454 | 638.33 | 6.82 | 900 | 128.1565 | 128.1528 | 28.7 |
11 | Dimethyl trisulfide3,6,8,15,16,20,21,23 | C2H6S3 | 3658-80-8 | 782 | 742 | 18.77 | 1.129 | 51.61 | 6.74 | 1012 | 125.9632 | 125.9603 | 23.1 |
A low mass error (31 ppm or lower) was achieved for all compounds based on the m/z indicated in Table 1. The low mass error contributed to the overall confidence in analyte identification, which highlights one of the main benefits of using GC×GC-HRTOFMS. The main advantage of using this technique for VOC profiling was the ability to simultaneously monitor the library identification, accurate mass data, and linear retention index (LRI) for each compound. For example, in Table 1 the peak assignment for compound 8 (pentadecane) was originally identified with a high quality library match (>900) as 1-iodo-2-methylundecane, which has been previously reported as a decomposition VOC.4,8 However, based on the LRI of the analyte (i.e. 1500) and the lack of a molecular ion (m/z 296) present for 1-iodo-2-methylundecane in the analyte spectrum, it was possible to identify that this compound was, in fact, pentadecane. Pentadecane exhibited a lower match quality to the NIST library than 1-iodo-2-methylundecane and was one of the secondary peak hits. However, the low mass error obtained for pentadecane (14.7 ppm) and its LRI made it possible to confirm the correct identification (confirmed by standard injection). LRIs and mass measurement error further contributed to confirming other lower NIST library identifications with matches between 700 and 800. It is common in VOC profiling of complex matrices to obtain match factors between 700 and 800, and these compounds may not be reported in the interest of stringency. This may also occur when the analyte's reference mass spectrum within the NIST library database has been obtained by quadrupole mass spectrometry, since the fragmentation patterns can differ slightly to fragmentation by TOFMS. The combination of available information demonstrates the high confidence in analyte identification and the added value of obtaining high-resolution mass spectral data. The use of HRTOFMS in GC×GC analysis of complex VOC mixtures will provide confirmatory information regarding analyte identifications that can be useful for building reference databases for users to increase consistency in analyte reporting.
In Table 1, molecular ions were used for mass error measurement where possible. The aromatic VOCs (e.g. benzene, toluene, and styrene) exhibited strong molecular ion peaks based on the stability of their structures. However, for structures like alcohols, the molecular ion peak was only found in trace levels or was absent from the spectra. The formula calculator of the data processing software was used to determine the exact m/z of a selected fragment to facilitate computing of the mass error. However, in the future it may be useful to recollect a portion of the split flow of the desorption step onto a secondary tube that would be analysed at a lower ionization voltage. This would provide a higher chance of calculating the mass error based on the molecular ion of analytes with less stability that often displayed low or non-existent molecular ions in their spectra.
In decomposition odour analysis, there are often a large number of compounds identified that are present in the background VOC profile and fluctuate based on “noise”. PCA is often used in this field of research, yet generating PCA plots from a large multivariate data set can result in poor discrimination based on the principal components (PCs) even when differences exist between the groups. This is often managed by conducting PCA on the sum of compound classes or by reducing the number of compounds input into the analysis based on statistical thresholds. For this reason, only significant compounds from the pairwise FR analysis were used for PCA in this study. Visualisation of the remaining data structure after compound selection ensured that the selection of significant compounds of interest yielded differentiation between the control and experimental classes.
Fig. 2a demonstrates that it was possible to make a clear discrimination between the two classes along the first principal component (PC-1) axis. Fig. 2b demonstrates that the significant compounds chosen by the pairwise FR analysis were those that were indicative of the decomposition odour samples and not of the control samples, providing the distinction between the two compound classes in Fig. 2a. This is apparent because the compound loadings are located on the outer ring of the correlations loading plot (i.e. demonstrating their significance), and that they are on the right side of the plot where the experimental samples are located in the scores plot.
Fig. 2 Principal component analysis (PCA) plots of (a) scores and (b) correlation loadings based on normalised peak volume of significant compounds. Compound loading labels are associated with the numbering in Table 1. |
Although data filtration considerably reduced the number of compounds evaluated from those detected during the original analysis, Fig. 2b shows that the significant compounds were those that were specific to decomposition. In addition, the number of compounds reported as being significant was expected to be low since these samples were collected three months post-mortem when the remains were skeletonised and a reduced odour was present. Extending this analysis to earlier stages of decomposition would be of interest in the future to compare the high-resolution identifications with previous work that characterised decomposition VOCs in early-stage decomposition using low-resolution TOFMS. Using a FR analysis of this nature will facilitate discrimination of the more complex, multivariate data expected under these circumstances.
Fig. 3 Mass errors (in ppm) of selected compounds based on the measured m/z in comparison to the accurate m/z over a two week period. |
Due to these challenges, at this time it is suggested that the “added dimension” afforded by the HRTOFMS in this study would be most valuable to this area of research for generating reference databases of reported decomposition VOCs. This would provide a tool for researchers to refer to when conducting studies using GC or GC×GC with LRMS to provide a secondary (or tertiary) confirmation of analyte identification. Generating reference databases for decomposition odour analysis was first proposed in 2004 by Vass et al.22 and provided reference data for pioneering studies in decomposition odour analysis using GC-MS. Following on from this work, more information has now been made available about the decomposition VOC profile through implementation of GC×GC-TOFMS.1–8 It is now possible to develop decomposition VOC databases with even higher confidence using GC×GC-HRTOFMS. This provides a promising outlook for delivering an updated database of compounds based on the advances in analytical technology that have been employed in recent years.
Footnote |
† These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2015 |