An Untargeted Exposure Study of Small Isolated Populations Using Atmospheric Gas Chromatography Coupled with High Resolution Mass Spectrometry

In this study, pooled plasma samples from individuals living in various small isolated coastal communities were analyzed using an exposomics approach to determine whether differences exist between the communities with regard to families and concentrations of contaminants. Samples were analyzed using atmospheric pressure chemical ionization gas chromatography (APGC) coupled to high resolution mass spectrometry (HRMS), operated in data independent acquisition (DIA) mode, where precursor and fragment information were collected in a single run.

Exposure studies involve complex data and subtle comparisons within the data sets. By utilizing the soft ionization of APGC and acquiring accurate mass data on both precursor and fragment ions in one method, a complete data set can be produced. This combined with the processing power of Progenesis QI Software allows complex statistical analysis to be performed quickly and easily. Progenesis QI also allows the searching of thousands of online databases and user generated libraries. This combination of hardware and software permits a simplified approach to exposomics workflows.

Benefits

• Generation of accurate mass measurements of low- and high-energy spectra allows targeted and untargeted data analysis in a single data set.
• “Soft” ionization using APGC results in preservation of the molecular ion. That combined with fragmentation after ionization produces comprehensive spectral details.
• Integrated MVA and elucidation tools aid in identification of markers of interest with automatic elemental composition, searching of online databases, and structural assignments.

Human exposure to environmental contaminants has been linked to various health problems. When analyzing known environmental contaminants of interest such as persistent organic pollutants (POPs) including dioxins, PCBs, and PAHs, targeted mass spectrometry methods are employed. Recently studies have been conducted using a metabolomic approach to determine differences of exposure between different populations. The term “exposomics” refers to studies that look at a wide array of contaminants in humans that may pose health risks. 

In this study, pooled plasma samples from individuals living in various small isolated coastal communities were analyzed using an exposomics approach to determine whether differences exist between the communities with regard to families and concentrations of contaminants. Samples were analyzed using atmospheric pressure chemical ionization gas chromatography (APGC) coupled to high resolution mass spectrometry (HRMS), operated in data independent acquisition (DIA) mode, where precursor and fragment information were collected in a single run.

One of the major challenges of this type of study is interpreting the massive amounts of data generated. In order to facilitate data interpretation, Waters Progenesis QI data analysis software was utilized. First, targeted analysis was performed against a defined contaminants database. Then, multi variant analysis (MVA) was carried out to determine any differences between the communities. Elucidation of unknown contaminants was also achieved using Progenesis QI Software, which involved searching online databases and matching structural information to the high energy data. Finally, confirmation of one of the findings was performed using a standard.

Sample preparation

2 mL of plasma was taken and spiked with ¹³C internal standard and mixed with ethanol and saturated ammonium sulphate  solution (for denaturation). The samples were then extracted with hexane. The extracts were evaporated and purified on a florisil column (1 g). POPs were eluted with 25% dichloromethane in hexane. Purified extracts can be concentrated up to 20 µL of hexane prior to GC-MS analysis. For this study, this protocol was suitable, since the goal was to study POPs and POP-like compounds and this protocol intends to extract and purify contaminants related to the chemical property of POP’s, such as  non-polar lipophilic molecules.

Before analysis of the samples, a standard GC mixture was run on the system. The mixture contained chlorinated pesticides that have similar physiochemical properties to the POPs of interest. As they were acquired with APGC and MS^E the spectra produced were specific and well conserved. The low energy spectra showed little fragmentation compared to traditional EI+ analysis due to the soft ionization produced by APGC. The high energy spectra yield fragmentation information that can be utilized to perform structural verification. Figure 1 shows the results for hexachlorobenzene analyzed in the standard and illustrates the intense precursor (low energy) with good fragmentation (high energy) from this soft ionization technique.

Once the standard was run on the system and it was verified that the sensitivity and mass accuracy were as expected, the samples were analyzed. The samples were injected in triplicate. Normally in a metabolomics experiment the samples would be randomized to prevent any build-up of compounds by injecting the same sample and to account for any drop in instrument sensitivity over time. In this case the sample volumes were 20 µL in hexane which is volatile. If the samples were randomized after the first injection puncturing the vial septum the samples could have been concentrated due to solvent evaporation and bias the experimental results. For this reason it was decided to run the samples in series. Once the sample data was collected within UNIFI Software, it was transferred to Progenesis QI for data interpretation.

Upon import into Progenesis QI, the possible adducts that may have been present in the data set were selected. In this case the M^+• and the (M+H)⁺ were selected due to the ionization mechanisms of APGC. The runs were then automatically aligned to account for any drift in retention time over long run periods such as in a metabolomics study. To ensure consistent peak picking  and matching across all data files, an aggregate data set was created from the aligned runs. This contained peak information from all of the sample files, enabling detection of a single map of compound ions. This map was then applied to each sample, yielding 100% matching of peaks with no missing values aiding the multivariate statistical analysis.

Targeted analysis

A MetaScope database containing precursor and fragment ion information for 98 expected compounds was searched for all samples present in the data set. The search parameters, shown in Figure 2, used a 5 ppm mass error, a 1-minute retention time window, and a 2 mDa fragment mass error to determine identifications. This yielded 24 positive results in the pooled plasma samples. Figure 3 shows the identifications and the sample in which the compound was found in the highest abundance. One of the things to note is that Community 1 seemed to have a high abundance of POP detections. This community may be of particular interest to investigate for  the presence of other untargeted compounds.

p,p’-DDT, p,p’-DDE, and o,p’-DDE show a slight up regulation in Population 5; however the compounds were found to have no significant variation in concentration between the communities according to the p-values. Dichlorodiphenldichloroethylene (DDE) is formed by the dehyrdrohalogenation of dichlorodiphenyltrichloroethane (DDT). Due to DDT's historically wide use as an insecticide in agriculture, it is commonly seen in animal tissue as DDT is fat-soluble and bioaccumulative. It is also regularly found in fish that constitute a major part of the diet in these small communities.² DDT and DDE are endocrine disruptors and considered possible human carcinogens. DDE and DDT provide relevant POPs exposure markers in several populations, and therefore they are important to identify. This was possible using the targeted approach highlighted here. 

The list of identified compounds was then subjected to a filter to show only the compounds that had a max fold change higher  than 2, which highlighted the compounds that had significant differences between communities. This yielded a list of 11 compounds shown in Table 2. The up regulation of PCB 118 in Population 1 is shown in Figure 3. As Community 1 had the highest abundance of these target compounds, it was decided that further untargeted analysis should be performed on this community.

Untargeted analysis

To investigate the data further all filters were removed. Progenesis QI software automatically generates a principle component analysis (PCA) plot that clearly depicts the separation of the communities (Figure 5). In order to perform further statistical tests the data was automatically exported to EZinfo. Community 1 was compared to all of the other communities using an orthogonal partial least squared discriminate analysis (OPLS-DA) model. This allowed an S-plot to be generated where significant compounds of interest could be identified at the extremes of the S-Plot. Figure 5 shows the S-plot generated from the OPLS-DA model. 17 significant markers were selected and imported directly into Progenesis QI.

The compounds of interest were then subjected to a database search. During this search the precursor accurate masses were searched against selected ChemSpider databases within  a 5 ppm mass error. The structures of the possible compounds resulting from the ChemSpider search were then subjected to in silico fragmentation and compared to the experimental fragment peaks within the high energy spectra for the compound that was within a 10 ppm mass error. These results were then ranked using an accurate mass matching score and fragmentation score. This process was automatic and took less than 1 minute to complete. The search parameters are shown in Figure 6.

A number of interesting results were obtained from the ChemSpider database search, the first being tocopherol, which yielded a good fragmentation match where 37% of the fragments could be accounted for from the high energy spectrum. Tocopherols (TCPs) are a class of organic chemical compounds, many of which have Vitamin E activity. TCPs are found at high levels in vegetables and berries.³ The isolated communities with a mostly vegetarian diet would explain the higher concentrations of TCPs in this population. The results from the database search are shown in Figure 7.

Another strong identification from the database search was 1,3-Benzothiazole. Benzothiazoles (BTHs) are a class of compounds that are produced in high volumes. They are used as corrosion inhibitors and found in rubber materials, herbicides, azo dyes, and food flavoring.⁴ This finding is remarkable as these isolated communities would not directly be using materials containing BTH; hence it could be concluded that the exposure was due to environmental contamination. Another hypothesis is that Population 1 is the only population connected to modern food supply via a direct airport in the south. The presence of BTH may be a bio indicator of processed food consumption, as BTH is widely used as a freshness preservative in packaging. Both of these theories could be investigated further to verify the source of the BTH exposure. Figure 8 shows the possible identification results for 1,3-Benzothiazole.

As BTHs are a group of compounds, correlation analysis was performed using Progenesis QI. A dendrogram was automatically generated in the software showing the related compounds. A few compounds selected from the dendrogram were related to 1,3-Benzothiazole. The software visualization of these relationships is shown in Figure 9. These compounds were tagged  and searched against the ChemSpider databases. This resulted in another possible identification of a thiazole compound, 4-phenyl-2-propyl-1,3-thiazole, (Figure 10).

Confimation of results

A standard of 1,3-Benzothiazole (BTH) was obtained in order to confirm the fragmentation pattern and the identification of this compound in the samples. A different GC method was employed for this analysis which was carried  out at a later date than the initial analysis.  The spectra from the standard matched that of the proposed identification of BTH from the ChemSpider search and the spectra from the sample. This allows the initial database to be updated to include BTH as a target exposure compound for further investigations of the population studies.

Exposure studies involve complex data and subtle comparisons within the data sets. By utilizing the soft ionization of APGC and acquiring accurate mass data on both precursor and fragment ions in one method, a complete data set can be produced. This combined with the processing power of Progenesis QI Software allows complex statistical analysis to be performed quickly and easily. Progenesis Q1 also allows the searching of thousands of online databases and user generated libraries. This combination of hardware and software permits a simplified approach to exposomics workflows. 

1. Atmospheric Pressure GC (APGC). Waters White Paper, Part no. 720004771EN. August, 2013.
2. G E Burdick, E J Harris, H J Dean, T M Walker, J Skea, D Colby. The Accumulation of DDT in Lake Trout and the Effect on Reproduction. Transactions of the American Fisheries Society. 93:2, 1964.
3. Piironen et al. Tocopherols and Tocotrienols in Finnish Foods: Vegetables, Fruits, and Berries. J Agric Food Chem. 3086, 34 I: 742–746.
4. Asimakopoulos et al. Benzotriazoles and benzothiazoles in human urine from several countries: A perspective on occurrence, biotransformation and human exposure. Env Intl. Volume 59, September, 2013.