Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Dec 1.
Published in final edited form as: J Occup Environ Med. 2019 Dec;61(Suppl 12):S25–S34. doi: 10.1097/JOM.0000000000001665

Metabolome-wide Association Study of Deployment to Balad, Iraq or Bagram, Afghanistan

Young-Mi Go 1, M Ryan Smith 1, Douglas I Walker 1,2, Karan Uppal 1, Patricia Rohrbeck 3, Pamela L Krahl 4, Philip K Hopke 5, Mark J Utell 5, Timothy M Mallon 4, Dean P Jones 1
PMCID: PMC6901093  NIHMSID: NIHMS1044759  PMID: 31800448

Abstract

Objective:

To use high-resolution metabolomics (HRM) to identify metabolic changes in military personnel associated with deployment to Balad, Iraq, or Bagram, Afghanistan.

Methods:

Pre- and post-deployment samples were obtained from the Department of Defense Serum Repository (DoDSR). HRM and bioinformatics were used to identify metabolic differences associated with deployment.

Results:

Differences at baseline (pre-deployment) between personnel deployed to Bagram compared to Balad or Controls included sex hormone and keratan sulfate metabolism. Deployment to Balad was associated with alterations to amino acid and lipid metabolism, consistent with inflammation and oxidative stress, and pathways linked to metabolic adaptation and repair. Difference associated with deployment to Bagram included lipid pathways linked to cell signaling and inflammation.

Conclusions:

Metabolic variations in pre- and post-deployment are consistent with deployment-associated responses to air pollution and other environmental stressors.

Keywords: Environmental Toxicants, Exposome, Inflammation, Metabolic response, MWAS

INTRODUCTION

Concern about occupational exposures to military personnel and adverse health outcomes resulted in 2006 Department of Defense (DoD) Instruction 6490.03 requiring Military Services to perform comprehensive deployment health risk assessments that includes baseline, routine, and incident-related health surveillance and documentation of deployment-related exposures (1, 2). To support this effort, we assembled a multidisciplinary research team to assess the utility of serum in the Department of Defense Serum Repository (DoDSR) for biomonitoring chemical exposures and identifying biomarkers associated with exposures. Results established usefulness of DoDSR samples to link metabolomics (3-7), cytokines (8), microRNAs (8, 9), and environmental chemicals (7) to deployment-associated environmental exposures and health risk (3, 5, 7). The results show utility of the current DoDSR and support expansion of the specimen profile, possibly adding blood, urine, and DNA, to enhance usefulness.

With utility of the DoDSR established, we redirected efforts toward answering critical questions related to comprehensive assessments of deployment health risks. In support of this goal, we built upon previous research that was designed to evaluate the risk of adverse health outcomes in Service Members deployed to regions of Iraq and Afghanistan (1, 2). In this design, the personnel roster of deployed troops was used to select matching pre- and post-deployment serum samples (Cases) for 200 Service Members from a large cohort who deployed to Iraq and Afghanistan and provided security around burn pits. A group of 200 demographically matched, non-deployed Service Members (Controls) without similar deployment exposures but with specimens in the DoDSR were selected as controls.

In recent years, we have advanced development of sensitive, high-throughput chemical profiling methods for precision medicine (5, 10-12). These high-resolution metabolomics (HRM) methods provide information relevant to a broad spectrum of environmental exposures, including both known environmental chemicals and a large number of currently unidentified chemicals. Previous research shows that occupational exposure to trichloroethylene is associated with plasma and urinary metabolites that were also correlated with disease-related biomarkers (13). Perhaps more critically for evaluation of health risks associated with deployment, HRM data has the potential to be used within the emerging framework of exposome research to test for associations of occupational exposures with diet and other factors (Fig 1). In the present study, we used HRM data from the above 800 DoDSR samples (200 each pre- and post-deployment samples for Cases and Controls) to perform a metabolome-wide association study (MWAS) of deployment to areas with burn pit exposures in Balad, Iraq or Bagram, Afghanistan.

FIGURE 1.

FIGURE 1.

Open in a new tab

Central model for exposome research. The human exposome consists of many exposures (left). Internal doses of many chemicals (middle left), along with metabolic responses (middle right), are measured by high-resolution metabolomics. The internal metabolic responses are correlated with outcomes (right).

METHODS

Source of Samples

Eight-hundred de-identified serum samples collected from Armed Forces personnel during active duty were obtained from the Department of Defense Serum Repository (DoDSR) as part of a larger study evaluating exposure and response biomarkers (2). The repository consists of approximately 50 million serum samples originally collected for mandatory armed forces personnel HIV testing (14). Samples were stored at −30°C and underwent a single thaw/freeze cycle to aliquot 500 μL into separate microfuge tubes to ship from the DoDSR to Emory University for analysis. Samples were shipped on dry ice and maintained at −80°C until analysis.

In the current study, serum samples were excluded from data analysis if they met any of the following criteria: deployment length was less than 150 days, samples for pre-deployment analysis were collected while currently deployed at another site, or samples were excluded with implausible collection dates. Using these criteria, data for 13 individuals were excluded from analysis for the Case group while 14 individuals were excluded from the Control group. Therefore, the overall numbers in the current study consisted of paired pre- and post-deployment data for 373 individuals, with 187 in the Case group and 186 in Control group.

Chemicals and HPLC Columns

Acetonitrile (HPLC grade), formic acid (HPLC grade), and water (HPLC grade) were obtained from Sigma–Aldrich (St. Louis). A mixture of internal standard stable isotopic chemicals (15, 16) from Cambridge Isotope Laboratories, Inc. (Andover, Pennsylvania). Analytical columns consisted of C18 column (Higgins Analytical, Targa, 2.1 × 10 cm) with a short, end-capped C18 pre-column (Higgins Analytical, Targa guard).

High-Resolution Mass Spectrometry and Data Pre-processing

For analysis, a 65 μL serum aliquot was added to 130 μL of acetonitrile containing a mixture of stable isotope-labeled internal standards and prepared for metabolomics analysis using established methods (15-17). A quality control pooled reference sample (QStd3) was included at the beginning and end of each analytical batch of 20 samples for quality control and quality assurance (16). Samples were analyzed in triplicate by liquid chromatography with Fourier transform mass spectrometry (Dionex Ultimate 3000, Q-Exactive, Thermo Fisher; Waltham MA) with positive electrospray ionization (ESI) mode and resolution of 70,000 (4). Spectral m/z features were acquired in scan range 85-1,250 mass-to-charge ratio (m/z) over a ten minute window. LC-chromatography with C18 column was performed as follows: flow rates were set to 0.35 ml/min for the first 6 min and then changed to 0.5 ml/min for the remaining 4 min. The buffer system consisted of the first 2-min of 5% A (2% (v/v) formic acid in water), 60% water, and 35% acetonitrile, followed by a 4-min linear gradient to 5% A, 0% water, 95% acetonitrile. The final 4-min period was maintained at 5% A, 95% acetonitrile. Raw data files were extracted using apLCMSv6.3.3 (18) with xMSanalyzerv2.0.7 (19), followed by batch correction with ComBat (20). Uniquely detected ions consisted of m/z, retention time and ion abundance, referred to as m/z features.

Metabolic Feature Selection

Prior to data analysis, triplicate injections were averaged and only m/z features with at least 80% non-missing values in either of the groups and more than 50% non-missing values across all samples were retained. After filtering based on missing values, data were log2 transformed and quantile normalized (21). Selection of differentially expressed m/z features was performed based on one-way ANOVA, one-way repeated measures ANOVA, and two-way repeated measures ANOVA using the limma package in R (22). Benjamini-Hochberg false discovery method was used for multiple hypothesis testing correction at FDR<0.2 threshold (23). Visualization of the data, which was based on similarity in expression, was performed using unsupervised two-way hierarchal clustering analysis (HCA) utilizing the hclust() function in R to determine the clustering pattern of selected m/z features and samples. Principal component analysis (PCA) was performed using the pca() function implemented in R package pcaMethods. To evaluate systemic metabolic alterations due to deployment, a metabolome-wide association analysis was performed for discriminatory metabolites at P < 0.05 and characterized for pathway enrichment using Mummichog2.0 (24).

Targeted Analysis of Environmental Chemical Detection

Environmental chemicals were from an in-house library of selected environmental chemicals (see Smith et al, (25)) validated at Level 1 identification (26).

Metabolite Annotation

Metabolic features were annotated using xMSannotator (27); confidence scores for annotation by xMSannotator are derived from a multistage clustering algorithm. Identities of selected metabolites were confirmed by co-elution relative to authentic standards and ion dissociation mass spectrometry. Confidence Level 1 indicates identification by criteria of Schymanski et al. (26). Additional annotations were made (≥ 2) based on adduct form, and using database including KEGG, (Kyoto Encyclopedia of Genes and Genomes) (28), HMDB (Human Metabolome Database) (29), T3DB (Toxin and Toxin Target Database) (30) and Lipid Maps (31) at 10ppm tolerance.

Integrative Analysis using xMWAS

The metabolomics data and various environmental xenobiotics from the same set of samples were integrated using xMWAS based on the sparse partial least-squares (sPLS) regression method for data integration (12). sPLS is a regression modeling approach which performs variable selection and data integration simultaneously, and was originally designed for problems where the sample size (n) is significantly smaller than the total number of variables (p) and where the variables are highly correlated (32). xMWAS also performs community detection using the multilevel community detection algorithm (33) to identify groups of nodes that are heavily connected with other nodes in the same community, but have sparse connections with the rest of the network. The input for xMWAS included cotinine (cotinine measures in300 samples ) and the metabolome (1002 metabolic features in 300 samples) with thresholds for determining significant associations must have met the correlation threshold criteria (∣r∣ > 0.35) and P < 0.05 as determined by Student's t-test. A selected list of environmental chemicals (35 chemicals in 300 samples) and the metabolome (1002 metabolic features in 300 samples) with thresholds for determining significant associations must have met the correlation threshold criteria (∣r∣ > 0.50) and P < 0.05 as determined by Student's t-test. PAH’s (17 PAH in 300 samples) and the metabolome (1002 metabolic features in 300 samples). Thresholds for determining significant associations must have met the correlation threshold criteria (∣r∣ > 0.25) and P < 0.05 as determined by Student's t-test. Metabolic features and environmental chemicals had previously been selected by LIMMA analysis as significant using the criteria described above, and had been quantile normalized and log2-transformed.

RESULTS

Metabolic Responses Associated with Pre and Post-deployment

The Balad (n = 150) and Bagram subgroups (n = 37) were considered together as Cases in sample selection, but these samples were not collected concurrently. Therefore, we initially performed a one-way ANOVA for pre-deployment samples to test whether metabolic differences existed at baseline for the Controls (n = 186) and the Balad and Bagram subgroups. Results showed 1034 mass spectral features differed, and we performed hierarchical cluster analysis (HCA; Fig 2A) and principal component analysis (PCA; Fig 2B) to examine patterns of mass spectral intensity changes (Annotated metabolic features provided in Supplementary Table 1 using xMSannotator (27)). Results from both HCA and PCA showed that samples obtained from service members deployed to Balad shared characteristics with Controls while those deployed to Bagram largely clustered together and were separate from Balad and Control groups.

FIGURE 2.

FIGURE 2.

Open in a new tab

Metabolic responses associated with pre- and post-deployment. The 373 serum samples of pre-deployment [Balad (n = 150), Bagram (n = 37), Control (n = 186)] were analyzed for HRM and differences of metabolic features in pre-deployment were presented by unsupervised HCA-heatmap after selection of significant features determined by LIMMA using FDR < 0.2 (A) and PCA [Green (control), Red (Bagram), Yellow (Balad), B]. C, Mummichog pathway enrichment analysis on 1034 m/z features showed that alterations in metabolic pathways of androgen and estrogen, keratin, N-glycan, and vitamin E were observed in the 3 groups prior to deployment. (Filled bars indicate significance and the cutoff (p<0.05) is indicated by the dotted line). D, The 374 serum samples of pre- and post-deployment [Balad (n = 150), Bagram (n = 37)] were analyzed for HRM and 1757 m/z features associated with deployment were presented by unsupervised HCA-heatmap.

Because there were different selection criteria for service members deployed to Bagram than for those deployed to Balad and no reason to suspect differences in sample collection or storage for Bagram, Balad and Control sera, we used mummichog pathway enrichment analysis to determine if the differentiating m/z features were enriched for specific metabolic pathways. Results (Fig 2C) showed that the top pathway was for steroid sex hormones and the next two pathways were glycan pathways which maintain extracellular matrix components and allow tissues to respond dynamically to biomechanical changes to protect from extrinsic forces (34). Annotations for the accurate mass matches to metabolites within these pathways can be found in Supplementary Table 2. The differences between Balad and Bagram Service Members in the pre-deployment samples led us to use two-way repeated measures ANOVA to test for pre- and post-deployment differences of these subgroups. Results showed that the pre- and post-deployment samples for Bagram were more similar to each other than either was similar to pre- or post-deployment samples for Balad (Fig 2D). Consequently, we proceeded to examine differences between pre- and post-deployment samples independently for Balad and Bagram.

Balad Subgroup

Following data filtering and paired analysis with limma, 1035 mass spectral features differed at FDR<0.2 in the pre- and post-deployment samples for Service Members deployed to Balad (Fig 3). These included 351 metabolic features with m/z below 300 that increased and 95 features between 500 and 600 that decreased (Fig 3A). Most mass spectral features that differed had relatively short retention times on the C18 column, indicating that the respective chemicals were relatively polar in character (Fig 3B). In addition, features with longer retention times (non-polar chemicals) also differed, with 156 features increased and 55 decreased with deployment (Fig 3B). HCA of the discriminatory features showed that there was some clustering of pre- and post-deployment samples yielding 4 cluster modules (Fig 3C), and PCA showed that there was partial separation of the groups (Fig 3D). The data show that despite the heterogeneity of metabolism in the pre-deployment samples, metabolic changes associated with deployment are identified (annotated features differed in groups are provided in Supplementary Table 3; Tab 1) Pathway enrichment analysis showed a large number of metabolic pathways changed in association with deployment to Balad (Fig 4A). The top pathway (aminosugars) and three other pathways (sialic acid, keratan sulfate, N-glycan) represented pathways to maintain extracellular matrix and respond biomechanical stress. The pathway linked to detoxification (pentose and glucuronate) was also found to be changed. Adaptive metabolic pathways were altered which included the amino acid pathways arginine and proline, alanine and aspartate, aspartate and asparagine, as well as nitrogen metabolism. One pathway linked to oxidative stress (pentose phosphate) was also changed. In addition to changes in intermediary metabolism, two lipid pathways (glycosphingolipids and glycerophospholipids) were also changed in association with deployment to Balad. For a full list of annotated features, please refer to Supplementary Table 4; Tab 1.

FIGURE 3.

FIGURE 3.

Open in a new tab

Metabolic responses associated with deployment to Balad, Afghanistan. The 300 serum samples (n = 150 each pre and post-deployment) collected from pre- and post-deployment to Balad were analyzed by HRM. A, Type I Manhattan plot of m/z features plotted against the −Log10 P value indicates that 1030 features are altered after deployment to Balad [red (658 increased after deployment) and blue (377 decreased after deployment) at FDR < 0.2; gray (2180) were not affected by deployment], B, Type II Manhattan plot using retention time (RT, s) plotted against −Log10 P value. C, Unsupervised HCA-heatmap indicates that intensity of 1035 m/z features drive the separation between the pre and post-deployment to Balad. D, PCA plot of selected significant features using the above selection criteria showing separation of the pre-deployment (red) and post-deployment group (green), through the 1st (19% variation) and 2nd (6% variation) principal components.

FIGURE 4.

FIGURE 4.

Open in a new tab

Pathway enrichment analysis associated with deployment. Mummichog pathway analysis on metabolic features differentiating Pre and Post deployment to Balad (A), Bagram (B), and Control (C). (Filled bars indicate significance and the cutoff (p<0.05) is indicated by the dotted line). For complete annotation of metabolites in each respective pathway, please see Supplemental Table 4.

Bagram Subgroup

In the Bagram subgroup (n = 37), 75 features differed with paired LIMMA at FDR< 0.2. The patterns of metabolite changes differed from those seen in the Balad group in that metabolites with m/z below 300 mostly decreased while those in the 500 to 600 range increased (Fig 5A). Note that in Fig 5A and 5B, both the FDR< 0.2 and P < 0.05 are shown to facilitate visualization of possible differences given that the Bagram subgroup had only 37 individuals. The early elution indicated that these metabolites are relatively polar, although there were also many metabolites eluting later which are likely to be non-polar in character (Fig 5B). HCA (Fig 5C) and PCA (Fig 5D) showed that the differentiating metabolites mostly separated the pre- and post-deployment samples. The metabolic features which separated the pre- versus post deployment were also annotated, and can be found in Supplementary Table 3; Tab 2. In contrast to the pathways differing for Service Members deployed to Balad, the majority of pathways associated with deployment to Bagram included lipid-related pathways, such as arachidonic acid, prostaglandin formation, fatty acid biosynthesis, and linoleate metabolism. Other enriched pathways include retinal and folate metabolism as well as the amino acids arginine and proline metabolism. (Fig 4B). For a complete list of annotated features driving the metabolic pathways, please refer to Supplementary Table 4; Tab 2.

FIGURE 5.

FIGURE 5.

Open in a new tab

Metabolic responses associated with deployment to Bagram. The 74 serum samples (n = 37 each pre and post-deployment) collected from pre- and post-deployment to Bagram were analyzed for HRM. A, Type I Manhattan plot of m/z features plotted against the −Log10 P value indicates that 75 features are altered after deployment to Bagram [red (20 increased after deployment) and blue (55 decreased after deployment) at FDR < 0.2; gray (3255) were not affected by deployment], B, Type II Manhattan plot using retention time (RT, s) plotted against −Log10 P value. C, Unsupervised HCA-heatmap indicates that intensity of 75 m/z features drive the separation between the pre and post-deployment to Bagram. D, PCA plot of selected significant features using the above selection criteria showing separation of the pre-deployment (red) and post-deployment group (green), through the 1st (38% variation) and 2nd (9% variation) principal components.

Controls

The controls were stationed within the U.S. and were not deployed during the time between the first and second serum collection time-point. Paired analysis with LIMMA showed that 56 m/z features differed at FDR< 0.2 (n=186). As illustrated for the Bagram comparisons above, both FDR< 0.2 and P < 0.05 are shown to facilitate comparisons. Most importantly, few metabolic differences were present; these occurred with distributions of m/z (Fig 6A) and retention times (Fig 6B) similar to those observed for deployment to Balad or Bagram, indicating some time-dependent differences were present. HCA showed little clustering according to the differing metabolites (Fig 6C), but PCA showed fairly clear separation (Fig 6D). For the annotated list of metabolites driving the change before and after collection, please refer to please refer to Supplementary Table 3; Tab 3. Accordingly, we performed pathway enrichment analysis and found only one associated pathway (Fig 4C). This pathway, glycerophospholipid metabolism, is associated with changes in fatty acid and lipid metabolism. For a list of annotated features, please refer to Supplementary Table 4; Tab 3.

FIGURE 6.

FIGURE 6.

Open in a new tab

Metabolic responses associated with deployment in Control group. The 372 serum samples (n = 186 each pre and post-deployment) collected from pre- and post-deployment to Bagram were analyzed for HRM. A, Type I Manhattan plot of m/z features plotted against the −Log10 P value indicates that 56 features are altered after deployment to Bagram [red (25 increased after deployment) and blue (31 decreased after deployment) at FDR < 0.2; gray (3037) were not affected by deployment], B, Type II Manhattan plot using retention time (RT, s) plotted against −Log10 P value. C, Unsupervised HCA-heatmap indicates that intensity of 56 m/z features drive the separation between the pre and post-deployment to Bagram. D, PCA plot selected significant features using the above selection criteria showing separation of the pre-deployment (red) and post-deployment group (green), through the 1st (23% variation) and 2nd (7% variation) principal components.

Association of environmental chemicals with the metabolome

To test for associations of environmental chemicals with metabolic effects associated with deployment, we used xMWAS, a data-driven approach to examine complex -omics data (12). Environmental chemicals were obtained from targeted analyses of 271 environmental chemicals in the same Service Members deployed to Balad as the current study (see Smith et al in the current Supplement) (25). Twenty-six of 271 chemicals differed in association with deployment to Balad, and xMWAS results showed that three of these chemicals including trichlorfon metrifonate, dinoterfuran and ametryn clustered with metabolites that differed in association with deployment (Fig 7A). These metabolites were present in two chemical-metabolite communities of association, with the largest community centered on trichlorfon metrifonate. This hub was positively associated with the putatively annotated metabolites, dipeptide arginyl-glutatmine, the macrolide oleandolide, as well as purine metabolites guanosine pentaphosphate, and guanosyl-methylene-triphosphate, dihydroneopterin-triphosphate, a metabolite within the biopterin synthesis pathway. Many of these metabolites showed a modest decrease as a result of deployment, indicating that they could potentially play a role in metabolic adaptation to environmental chemical exposure (see Supplemental Fig 1). The second largest community contained metabolites correlated with two additional environmental chemicals (dinoterfuran and ametryn), which were associated with only a subset of the metabolites in the first community (Fig 7A), Metabolites associated with these two environmental chemicals included the pesticide sulfuramid, and the naphthopyran metabolite Janthitrem C. For a complete list of putatively annotated metabolites found to be significantly associated with the environmental chemicals, please see Supplemental Table 5; Tab 1.

FIGURE 7.

FIGURE 7.

Open in a new tab

Association of environmental chemicals with the metabolome. The 1037 m/z features associated with deployment to Balad (Fig 3) were examined for association with 271 environmental chemicals (See Smith et al. in the present Supplement) (A), PAH (B) and cotinine (3) (C) using xMWAS (12). A, Two major metabolic communities are identified; blue, environmental chemicals Ametryn and Dinotefuran association with 32 m/z features, which included the putative annotations the pesticide Sulfuramid, and the bacterial metabolite Janthitrem C; orange, metabolic community with anthelmintic agent metrifonate are shown, which was associated 119 m/z features which included the purine metabolites guanosine pentaphosphate and guanosyl-methylene triphosphate, the biopterin pathway metabolite dihydroneopterin 3-triphosphate, and the dipeptide arginyl-glutamine (r < 0.5 at p < 0.05). B, Two metabolic communities were associated with PAH’s. Blue, community includes two m/z features associated with benzo(ghi)perylene, one of which was annotated as the alkaloid neurine. Orange, includes twenty-four m/z features associated fluoranthene, and included the (protein breakdown product glutamyl-glutamine and (iso)leucyl-(iso)leucine), the lipids sphingosine and sphinganine, and the omega-3 fatty acid eicosapentaenoic acid (r < 0.25 at p < 0.05). C, cotinine is associated with 19 m/z features and include Sulfuramid, Janthitrem C, metrifonate, and the alkaloid isococculidine r < 0.35 at p < 0.05).

For comparison, we also tested for deployment-related associations of PAH and cotinine. Results for 17 measured PAH’s showed only two that varied with deployment had correlations with metabolites that changed with deployment. Of these, fluoranthrene had 23 correlated metabolites and benzo(ghi)perylene [B(ghi)P] correlated with two of these (Fig 7B). The correlated metabolites included the dipeptides glutamyl-glutamine and (Iso)leucyl-(Iso)leucine, the sphingolipids sphinganine and sphingosine, and the omega-3 fatty acid eicosapentaenoic acid. The alkaloid neurine correlated with B(ghi)P. For a complete list of annotated metabolites, please see Supplemental Table 5; Tab 2. Cotinine was also correlated with pesticide sulfuramid the naphthopyran metabolite Janthitrem C as well as the alkaloid isococculidine from the features driving the separation from Balad (Fig 7C); however, our group has also previously reported other metabolic associations with cotinine (4) and only weak associations with PAH and environmental chemicals (data not shown). The full list of features found to be associated with PAH’s can be found within Supplemental Table 5; Tab 3.

DISCUSSION

This study provides an assessment of metabolic differences associated with deployment to Balad, Iraq, and Bagram, Afghanistan by untargeted HRM using DoDSR samples collected before and after deployment. The HRM methods employed include metabolites in most human metabolic pathways and therefore provide a global assessment of metabolic responses (5, 35). The results show that the pre-deployment metabolic profiles of Service Members deployed to Bagram differed from Controls and those deployed to Balad, specifically in testosterone and estrogen metabolism and glycosaminoglycan pathways. The samples were selected from the DoDSR to sex-match Cases and Control, and did not make any provision to match Service Members deployed to Balad with those deployed to Bagram. Thus, the difference in sex hormone metabolism may be a consequence of differences in proportions of males and females in these groups. The Service Members deployed to Bagram were mostly security forces and may have had a different level of conditioning or physical fitness, which could account for differences in keratan sulfate and N-glycan pathways linked to maintenance of extracellular matrix and dynamic responses to physical activity or conditioning. These issues need to be addressed in future studies using an increased number of subjects to determine specific vulnerabilities based on subgroups, such as ethnicity, branch of service, and sex.

The pre-deployment samples for Service Members deployed to Balad were similar to Controls (Fig 2A, B). Consequently, metabolic differences between pre- and post-deployment samples for Balad may reflect more generalizable differences associated with deployment than those for the Bagram subgroup. Changes in pathways associated with glycosaminoglycans, including keratan sulfate and N-glycan pathways in the Balad subgroup, are consistent with increased physical demands to maintain tissue structures and dynamic responses. Metabolic differences in the Balad group also included amino acid and nitrogen metabolism; these pathways were previously found to change in association with physical conditioning (36) and could reflect adaptive responses to increased physical activity during deployment. Additionally, a combined transcriptome-metabolome wide association study of environmental chemical toxicity showed adaptive responses in these pathways (37).

Changes in other pathways may reflect responses to adverse exposures associated with deployment. For instance, changes in niacin pathways and pentose phosphate are commonly seen with oxidative stress (38, 39). Additionally changes in detoxification pathways and mitochondrial metabolism could also reflect adverse events. The arginine amino acid pathway, which is commonly associated with inflammation, was identified as enriched (40, 41); however, one must note that the overall pathway differences with deployment to Balad do not include lipid pathways previously associated with air pollution (42). With the present study size and selected population, one cannot assess whether this is a limitation of the study design or a critical deviation in expected metabolic responses.

The pre-deployment samples for the Bagram subgroup differed from Controls and the Balad subgroup, justifying separate analysis of the Bagram subgroup. With the assumption that the Service Members deployed to Bagram were highly fit prior to deployment, the observed differences in lipid pathways associated with deployment could reflect stress signaling associated with airborne toxicants. Prior studies show that polycyclic aromatic hydrocarbons perturb lipid signaling pathways (7), and environmental air sampling near burn pit and incinerator operations at Bagram confirmed diverse sources of related air pollutants (43). Although the present results only provide circumstantial evidence for a relationship, the results establish feasibility to use DoDSR samples in conjunction with other monitoring methods to help understand responses to deployment-associated exposures.

Personnel deployed to Iraq and Afghanistan were exposed to air pollution, including sand, dust and fossil fuel combustion products and smoke generated from open pit burning (1, 2). The burn pits were used to dispose of solid waste materials, such as plastics, metals, rubber, paints, solvents, munitions, and wood, with associated emissions of particulate matter, volatile organic compounds, polycyclic aromatic hydrocarbons, and heavy metals (1, 2). Although not examined in the present study, about half of the mass spectral features measured by HRM do not have accurate mass matches in metabolomics databases and could include previously undescribed chemicals generated in these burn pits. Thus, more detailed study of the un-identified high-resolution mass spectrometry signals associated with deployment could provide a basis for new strategies to address this challenging problem.

Epidemiological studies have explored the possible contribution of deployment-related environmental exposures to post-deployment chronic illness among Service Members and Veterans. These studies included comparisons of deployed and non-deployed personnel and different locations but did not provide consistent evidence for associations with increased respiratory symptoms and specific lung conditions (44-46). Routine high-throughput exposure surveillance using an HRM approach could support more extensive epidemiologic research to better test for such associations of exposure and health outcomes. The use of universal exposure surveillance would build upon existing occupational medicine and exposure sciences to connect external exposures, internal body burden of environmental agents and related biological responses and health outcomes (2). In the present research and associated studies, we show that existing DoDSR samples provide a high-quality cross-sectional reference for deployment-associated exposures and biologic responses. Extension to large numbers of samples could provide a reference database to link specific exposures to health outcomes. Additionally, our previous studies show that DoDSR samples further allow evaluation of confounding risk factors, such as tobacco use (4) and diet and nutrition (6). Integration of HRM data with breathing zone and personal monitoring could further enhance evaluation of adverse exposures and responses in deployed Service Members.

In summary, the present metabolome-wide association study of deployment to Balad, Iraq, shows differences characteristic of tissue repair and metabolic adaptation to stress, but without altered tryptophan and lipid signaling pathways commonly associated with air pollution and airway inflammation. Service Members deployed to Bagram differed from others in pre-deployment characteristics, perhaps reflecting physical conditioning. Metabolic characteristics associated with deployment to Bagram included lipid pathways associated with inflammatory signaling and mitochondrial energy metabolism.

CONCLUSIONS

HRM provides detailed measures of biologic responses which can be integrated with exposure data to evaluate deployment-associated risks. Research is needed to more specifically link metabolic responses with personal exposure monitoring to evaluate impacts of deployment at an individual level.

Supplementary Material

Supplemental Fig 1
Supplemental Table 1
Supplemental Table 2
Supplemental Table 3
Supplemental Table 4
Supplemental Table 5

REFERENCES

  • 1.Lushniak B, Mallon CT, Gaydos JC, Smith DJ. Utility of the Department of Defense Serum Repository in Assessing Deployment Exposure. J Occup Environ Med. 2016;58:S1–2. [DOI] [PubMed] [Google Scholar]
  • 2.Mallon CT, Rohrbeck MP, Haines MK, et al. Introduction to Department of Defense Research on Burn Pits, Biomarkers, and Health Outcomes Related to Deployment in Iraq and Afghanistan. J Occup Environ Med. 2016;58:S3–S11. [DOI] [PubMed] [Google Scholar]
  • 3.Liu KH, Walker DI, Uppal K, et al. High-Resolution Metabolomics Assessment of Military Personnel: Evaluating Analytical Strategies for Chemical Detection. J Occup Environ Med 2016;58:S53–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Jones DP, Walker DI, Uppal K, Rohrbeck P, Mallon CT, Go YM. Metabolic Pathways and Networks Associated With Tobacco Use in Military Personnel. J Occup Environ Med 2016;58:S111–116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Walker DI, Mallon CT, Hopke PK, et al. Deployment-Associated Exposure Surveillance With High-Resolution Metabolomics. J Occup Environ Med 2016;58:S12–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Accardi CJ, Walker DI, Uppal K, et al. High-Resolution Metabolomics for Nutrition and Health Assessment of Armed Forces Personnel. J Occup Environ Med 2016;58:S80–88. [DOI] [PubMed] [Google Scholar]
  • 7.Walker DI, Pennell KD, Uppal K, et al. Pilot Metabolome-Wide Association Study of Benzo(a)pyrene in Serum From Military Personnel. J Occup Environ Med 2016;58:S44–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Woeller CF, Thatcher TH, Van Twisk D, et al. Detection of Serum microRNAs From Department of Defense Serum Repository: Correlation With Cotinine, Cytokine, and Polycyclic Aromatic Hydrocarbon Levels. J Occup Environ Med 2016;58:S62–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Woeller CF, Thatcher TH, Van Twisk D, et al. MicroRNAs as Novel Biomarkers of Deployment Status and Exposure to Polychlorinated Dibenzo-p-Dioxins/Dibenzofurans. J Occup Environ Med 2016;58:S89–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Uppal K, Walker DI, Liu K, Li S, Go YM, Jones DP. Computational Metabolomics: A Framework for the Million Metabolome. Chem Res Toxicol. 2016;29:1956–1975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Uppal K, Walker DI, Jones DP. xMSannotator: An R Package for Network-Based Annotation of High-Resolution Metabolomics Data. Anal Chem. 2017;89:1063–1067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Uppal K, Ma C, Go YM, Jones DP, Wren J. xMWAS: a data-driven integration and differential network analysis tool. Bioinformatics. 2018;34:701–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Walker DI, Uppal K, Zhang L, et al. High-resolution metabolomics of occupational exposure to trichloroethylene. Int J Epidemiol. 2016;45:1517–1527. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Perdue CL, Cost AA, Rubertone MV, Lindler LE, Ludwig SL. Description and utilization of the United States department of defense serum repository: a review of published studies, 1985-2012. PLoS One. 2015;10:e0114857. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Soltow QA, Strobel FH, Mansfield KG, Wachtman L, Park Y, Jones DP. High-performance metabolic profiling with dual chromatography-Fourier-transform mass spectrometry (DC-FTMS) for study of the exposome. Metabolomics. 2013;9:S132–S143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Go YM, Kim CW, Walker DI, et al. Disturbed flow induces systemic changes in metabolites in mouse plasma: a metabolomics study using ApoE(−)/(−) mice with partial carotid ligation. Am J Physiol Regul Integr Comp Physiol. 2015;308:R62–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Park YH, Lee K, Soltow QA, et al. High-performance metabolic profiling of plasma from seven mammalian species for simultaneous environmental chemical surveillance and bioeffect monitoring. Toxicology. 2012;295:47–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Yu T, Park Y, Li S, Jones DP. Hybrid feature detection and information accumulation using high-resolution LC-MS metabolomics data. J Proteome Res. 2013;12:1419–1427. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Uppal K, Soltow QA, Strobel FH, et al. xMSanalyzer: automated pipeline for improved feature detection and downstream analysis of large-scale, non-targeted metabolomics data. BMC Bioinformatics. 2013;14:15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007;8:118–127. [DOI] [PubMed] [Google Scholar]
  • 21.Patel RM, Roback JD, Uppal K, Yu T, Jones DP, Josephson CD. Metabolomics profile comparisons of irradiated and nonirradiated stored donor red blood cells. Transfusion. 2015;55:544–552. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Smyth GK, Michaud J, Scott HS. Use of within-array replicate spots for assessing differential expression in microarray experiments. Bioinformatics. 2005;21:2067–2075. [DOI] [PubMed] [Google Scholar]
  • 23.Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of teh Royal Statistical Society Seriese B. 1995;57:289–300. [Google Scholar]
  • 24.Li S, Park Y, Duraisingham S, et al. Predicting network activity from high throughput metabolomics. PLoS Comput Biol. 2013;9:e1003123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Smith MRU K; Walker DI; Utell MJ; Hopke PK; Mallon TM; Krahl PL; Rohrbeck P; Go YM; and Jones DP Environmental Chemicals Altered in Association with Deployment for High Risk Areas. Journal of Occupational and Environmental Medicine. 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Schymanski EL, Jeon J, Gulde R, et al. Identifying small molecules via high resolution mass spectrometry: communicating confidence. Environ Sci Technol. 2014;48:2097–2098. [DOI] [PubMed] [Google Scholar]
  • 27.Uppal K, Walker DI, Jones DP. xMSannotator: An R Package for Network-Based Annotation of High-Resolution Metabolomics Data. Anal Chem. 2017;89:1063–1067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Wishart DS, Tzur D, Knox C, et al. HMDB: the Human Metabolome Database. Nucleic Acids Res. 2007;35:D521–526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Wishart D, Arndt D, Pon A, et al. T3DB: the toxic exposome database. Nucleic Acids Res. 2015;43:D928–934. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Fahy E, Sud M, Cotter D, Subramaniam S. LIPID MAPS online tools for lipid research. Nucleic Acids Res. 2007;35:W606–612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Chun H, Keles S. Sparse partial least squares regression for simultaneous dimension reduction and variable selection. J R Stat Soc Series B Stat Methodol. 2010;72:3–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Blondel VD, Guillaume J- L, Lambiotte R, Lefebvre E. Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment. 2008;2008:P10008. [Google Scholar]
  • 34.Caterson B, Melrose J. Keratan sulfate, a complex glycosaminoglycan with unique functional capability. Glycobiology. 2018;28:182–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Jones DP, Park Y, Ziegler TR. Nutritional metabolomics: progress in addressing complexity in diet and health. Annu Rev Nutr. 2012;32:183–202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Kim JH, Banton SA, Awad M, et al. Training-Related Metabolic Adaptations in American-Style Football Participants. Annals of Sports Medicine and Research. 2015;2:1048. [Google Scholar]
  • 37.Roede JR, Uppal K, Park Y, Tran V, Jones DP. Transcriptome-metabolome wide association study (TMWAS) of maneb and paraquat neurotoxicity reveals network level interactions in toxicologic mechanism. Toxicol Rep 2014;1:435–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Kuehne A, Emmert H, Soehle J, et al. Acute Activation of Oxidative Pentose Phosphate Pathway as First-Line Response to Oxidative Stress in Human Skin Cells. Mol Cell. 2015;59:359–371. [DOI] [PubMed] [Google Scholar]
  • 39.Benavente CA, Jacobson EL. Niacin restriction upregulates NADPH oxidase and reactive oxygen species (ROS) in human keratinocytes. Free Radic Biol Med. 2008;44:527–537. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Chandler JD, Hu X, Ko EJ, et al. Metabolic pathways of lung inflammation revealed by high-resolution metabolomics (HRM) of H1N1 influenza virus infection in mice. Am J Physiol Regul Integr Comp Physiol. 2016;311:R906–R916. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Zhang R, Kubo M, Murakami I, et al. l-Arginine administration attenuates airway inflammation by altering l-arginine metabolism in an NC/Nga mouse model of asthma. J Clin Biochem Nutr. 2015;56:201–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Liang D, Moutinho JL, Golan R, et al. Use of high-resolution metabolomics for the identification of metabolic signals associated with traffic-related air pollution. Environ Int 2018;120:145–154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Blasch KW, Kolivosky JE, Heller JM. Environmental Air Sampling Near Burn Pit and Incinerator Operations at Bagram Airfield, Afghanistan. J Occup Environ Med 2016;58:S38–43. [DOI] [PubMed] [Google Scholar]
  • 44.Abraham JH, Eick-Cost A, Clark LL, et al. A retrospective cohort study of military deployment and postdeployment medical encounters for respiratory conditions. Mil Med. 2014;179:540–546. [DOI] [PubMed] [Google Scholar]
  • 45.Powell TM, Smith TC, Jacobson IG, et al. Prospective assessment of chronic multisymptom illness reporting possibly associated with open-air burn pit smoke exposure in Iraq. J Occup Environ Med. 2012;54:682–688. [DOI] [PubMed] [Google Scholar]
  • 46.Smith B, Wong CA, Boyko EJ, et al. The effects of exposure to documented open-air burn pits on respiratory health among deployers of the Millennium Cohort Study. J Occup Environ Med. 2012;54:708–716. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Fig 1
NIHMS1044759-supplement-Supplemental_Fig_1.pptx (664.1KB, pptx)
Supplemental Table 1
NIHMS1044759-supplement-Supplemental_Table_1.xlsx (164.6KB, xlsx)
Supplemental Table 2
NIHMS1044759-supplement-Supplemental_Table_2.xlsx (11.9KB, xlsx)
Supplemental Table 3
NIHMS1044759-supplement-Supplemental_Table_3.xlsx (125.4KB, xlsx)
Supplemental Table 4
NIHMS1044759-supplement-Supplemental_Table_4.xlsx (19.9KB, xlsx)
Supplemental Table 5
NIHMS1044759-supplement-Supplemental_Table_5.xlsx (31.2KB, xlsx)

RESOURCES