Coupling Urinary Trihalomethanes and Metabolomic Profiles of Type II

Subscriber access provided by UNIV OF NEWCASTLE

Article

Coupling urinary trihalomethanes and metabolomic profiles of type II diabetes: a case-control study Xanthi D. Andrianou, Pantelis Charisiadis, and Konstantinos Christos Makris J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.6b01061 • Publication Date (Web): 16 Jun 2017 Downloaded from http://pubs.acs.org on June 21, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

Coupling urinary trihalomethanes and metabolomic profiles of type II diabetes: a casecontrol study Xanthi D. Andrianou+, Pantelis Charisiadis+ and Konstantinos C. Makris* Water and Health Laboratory, Cyprus International Institute for Environmental and Public Health, Cyprus University of Technology, Limassol 3041, Cyprus +

Equal contribution

*Corresponding author: Konstantinos C. Makris Email: [email protected] Tel: +35725002398

ACS Paragon Plus 1 Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT Abiding by the exposome paradigm, incorporation of external and internal exposure metrics using metabolomics tools is warranted to refine the etiology of type II diabetes (T2D). A small (n=51) age- and sex-matched case-control study was conducted in Cyprus coupling urinary trihalomethanes (THM) with T2D. The objectives were to: i) perform a comparative assessment of different deconvolution parameters in compound identification, and ii) evaluate the association between differentially expressed metabolites and either urinary THM or T2D status. Untargeted urinary metabolomics was performed with a GC/MS triple quadrupole mass spectrometry system. Results of three deconvolution searches each yielding >130 metabolites were used in subsequent analyses. The number of differentially expressed compounds by T2D status or the urinary THM levels (above or below median) differed among the three searches. The identity of these compounds was also confirmed using known standards (level 1 identification). In multivariate logistic regression, 3-aminoisobutyric acid was an important predictor of lower odds of T2D, after adjusting for known risk factors. The widespread incorporation of metabolomics in population studies accounting for environmental exposures will eventually pave the way for the exposome characterization, improving also our understanding of the disease process. Keywords: metabolomics, type 2 diabetes, trihalomethanes, exposures, exposome


Page 2 of 30

Page 3 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction The comparative prevalence of type II diabetes (T2D) is expected to exponentially increase in the next couple of decades with a concomitant rise in health care costs1. This will particularly affect the Eastern Mediterranean region that has one of the highest rates of the disease, globally 1. In addition to the classical risk factors of T2D, such as obesity, lack of physical activity, smoking, and high alcohol consumption 1, emerging risk factors, such as exposures to environmental chemicals, including pesticides, are gaining attention by the scientific community 2. Tackling T2D at the population level requires a holistic view of the disease, which can be complemented with the application of the exposome principles. The exposome aims to capture and measure all environmental/behavioral factors and the concomitant physiological responses occurring during one’s lifetime 3,4. The variability of phenotypes studied within the context of the exposome requires novel tools that are agnostic and allow the generation of hypotheses about disease risk factors that have been so far ignored or unidentified. At the same time, the application of the exposome framework requires the use of high throughput and high resolution omics methodologies 5. Metabolomics tools facilitate the comprehensive study of an organism’s physiological state as influenced by external stimuli and captured by the metabolic profiles. Advances in metabolomics methodologies allow metabolite profile panels to be investigated, utilizing both hypothesis-driven and hypothesis-generating approaches, especially in prospective study designs such as prospective cohorts which can include information on external exposures. Within this framework, top-down metabolomics could play a prominent role in the quest of the chronic disease etiology, the design of effective intervention measures, as well as in personalized medicine 5,6.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 30

Several population studies from established cohorts have already compared the metabolic profiles of T2D patients and healthy individuals 7. These studies have employed both targeted and untargeted approaches, using biofluids, such as serum/plasma, urine or both 8–10. Besides the incorporation of demographics, anthropometrics or food frequency data, the inclusion of exposure measurements, either measured in the environment or via biomarkers, into metabolomics-based profiling in T2D studies has been less frequently practiced. Two of the first studies of this kind combined urinary arsenic or plasma organochlorine pesticides measurements with the metabolic fingerprints of participants

while Patel et al (2010)

presented the approach of the environmental-wide association studies to assess links between environmental exposures and T2D11–13. The biomatrices most frequently used in population-based metabolomics studies are blood or urine. Human urine contains a wealth of “metabolic information” and it has been extensively studied in high-throughput metabolomics analyses. For the assessment of the complete human urine metabolome, bioanalytical platforms (i.e. NMR, GC-MS, DFI/LC-MS/MS, ICPMS, UV- and FD-based HPLC) and information from the literature have been combined 14. In total, 2651 metabolites were identified, 2206 of which were found in the literature and 445 metabolites were identified from the experimental analysis (378 of which were quantified) 14. Using the chemical classification system developed for the human metabolome database (HMDB), human urinary metabolites fall into 230 different chemical classes out of the 356 chemical classes annotated for the entire human metabolome, demonstrating the enormous chemical diversity of the urine matrix, when compared with that of serum or plasma 14,15. The integration of metabolomics data into exposome analyses requires high quality tools and algorithms that provide robust data assessment. Given that the process of chromatogram deconvolution

and

metabolite

identification

can

depend

on

computational

methods/algorithms, the parameters used during sample processing could influence the


Page 5 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


outcome of the analysis in terms of the number and quality of identified compounds, especially in small-sized studies. Therefore, the systematic application of standard metabolomics reporting schemes such as those proposed by the Metabolomics Standards Initiative

16,17

in population studies should be a priority to allow the comparative evaluation

of different sample processing protocols within the context of the exposome framework, as more studies and data accumulate. Moreover, the combination of metabolomics analysis and external exposure metrics into the exposome paradigm will allow for the assessment of more personal exposure profiles, the “personal exposomes”. Prospective assessment, i.e. in cohort studies, of these “personal exposomes” will allow the identification of specific metabolite perturbations to be linked with variation in exposures and possible health outcomes. To this extent, a pilot (n=51) age- and sex-matched case-control T2D study was undertaken coupling urinary trihalomethanes (THM) levels with the corresponding metabolite profiles, using various spectra processing parameters; the study took place in Cyprus, exhibiting one of the highest EU-27 comparative prevalence rates of T2D (9.6%)

18

. THM is a class of

environmental chemicals including trichloromethane (TCM), bromodichloromethane (BDCM), dibromochloromethane (DBCM), and tribromomethane (TBM) for which recent reports indicated possible implications with metabolic alterations in animals and humans

19–

21

. THM are formed in chlorine-disinfected tap water and they are routinely detected during

showering and swimming in pools. Moreover, they may also form in indoor environments as a result of tap water use in a suite of domestic cleaning activities where disinfectants are used (mopping, toilet cleaning, dish/clothes washing etc.), contributing to the body burden of THM for adults and children 19,22. The objectives of this study were to: i) perform a comparative assessment of different deconvolution parameters in compound identification for the processing and analysis of metabolic fingerprints, and ii) evaluate the association between differentially expressed



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 30

metabolites and urinary THM levels and T2D, by comparing and contrasting metabolic profiles with the external THM exposure metrics.

Experimental Section Study design A pilot case-control study was set up with participants of a cross-sectional study on exposures to disinfection by-products (THM) that was conducted in Nicosia, Cyprus during summer 2012 22,23. Cases were self-reported, doctor-diagnosed diabetics (n=23) and controls were also self-reported non-diabetics (n=28) who were matched for age and sex, when possible. The maximum number of available controls was selected for each case. When a control that would match for both age and sex was not available, only age matching was performed. Demographics and other participant characteristics were collected with questionnaires. First morning urine samples were analyzed for THM using gas chromatography coupled triple quadrupole mass spectrometry according to our previously published methodology

24

. The

limits of detection and quantification (LOD and LOQ) for the four THM species were: 27 (80) ng/L, 11 (32) ng/L, 24 (71) ng/L and 13 (40) ng/L for the TCM, BDCM, DBCM and TBM, respectively. Urinary creatinine was determined by the picric acid-based spectrophotometric method (Jaffe method) and it was used to account for creatinine dillution25. The study has been approved by the National Bioethics Committee of Cyprus (EEBK/EP/2012/17) and all participants provided their informed consent. All experiments were performed in accordance with relevant guidelines and regulations.

Untargeted metabolomics analysis Chemicals Agilent Fiehn GC/MS Metabolomics standards kit containing d27-myristic acid and fatty acyl methyl ester (FAME) [FAME/d27-myristic acid mixture] was purchased from Agilent


Page 7 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Technologies Santa Clara USA. Urease type III from Canavalia ensiformis (Jack Bean) from Sigma Aldrich USA, pyridine anhydrous 99.8% and methanol from Sigma Aldrich Germany, methoxyamine hydrochloride 98% from Sigma Aldrich Switzerland, toluene analytical reagent from BDH Chemicals Ltd England, sodium sulfate ACS reagent ≥99% anhydrous powder from Sigma Aldrich India, and MSTFA (N-methyl-trimethylsilyltrifluoroacetamide) with 1% (v/v) TCMS (trimethylchlorosilane) for derivatization from Fluka Switzerland. The standard compounds used for the confirmation of the identity of the differentially expressed compounds were the following: benzoic acid ≥ 99.5% ACS reagent, from Sigma-Aldrich UK; catechol ≥ 99.0% powder, from Sigma-Aldrich Japan; uracil ≥ 99.0%, from Sigma-Aldrich China; Threonine ≥ 98.0% reagent grade (HPLC), from Sigma-Aldrich USA; D,L-3Aminoisobutyric acid ≥ 98.0%, from Sigma-Aldrich India; creatinine hydrochloride ≥ 97.0%, from Sigma-Aldrich India; 3-hydroxyphenylacetic acid ≥ 99.0%, from Sigma-Aldrich Germany; D-(+)-arabitol ≥ 98.5%, from Dr. Ehrenstorfer GmbH; ribitol ≥ 99.0%, from Sigma-Aldrich UK; L-(-)-fucose ≥ 99.0%, from Sigma-Aldrich Slovakia; 5-hydroxyindole-3acetic acid ≥98% (HPLC) crystalline, from Sigma-Aldrich Switzerland.

GC-MS analysis Urine samples were subject to untargeted (no specific compounds were selected to be measured and quantified) metabolomics analysis using a gas chromatography coupled triple quadrupole mass spectrometry system

26

. Analysis was performed in an Agilent 7890A GC

coupled with an Agilent 7000B triple quadrupole mass spectrometer. An Agilent 7693A series automatic liquid sampler was used with a 10.0 µL ± 1% glass syringe (Agilent, Santa Clara, USA). One µL of sample was injected in splitless mode in a PTV injector at 250°C. Separation was performed with a DB-5MS+DG (30 m × 0.25 mm × 0.25 µm) 5%phenyl/95%-methylpolysiloxane capillary column from Agilent J&W USA, with helium as carrier gas at a flow rate of 1.6 mL min-1. The oven temperature program was set to 60 °C for 1 min, then to 300 °C with 10.5 °C min-1 for 12 min, following a post run at 60 °C for 1 min. The total chromatographic run time was 36 min. Electron impact ionization was performed at



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

electron energy of 70 eV in the full scan mode from 50 to 600 m/z with a solvent delay of 6 min. The quadrupole mass detector, the ion source temperature and the transfer line were set to 150 °C, 250 °C and 250 °C, respectively. The system was controlled by the software Mass Hunter Workstation (Agilent, rev. B.05.00).

Sample preparation Urine samples, previously stored in –80 ⁰C, were gently thawed, and they were subject to urea depletion, extraction and derivatization prior to obtaining full mass spectra. Briefly, the retention time lock (RTL) compound (5 µL of myristic acid 0.75 mg/mL) was added in 200 µL of urine sample, following the addition of 100 units of urease and incubation at 37 °C for 1h. 1.2 mL of methanol was added and the sample was placed in 4 °C for 20 min, to enzyme precipitation and urine metabolite extraction. After centrifugation (10 min at 10,000 G) 1 mL of the supernatant was evaporated to dryness at 30 °C under a gentle stream of nitrogen. Then 100 µL of toluene was added following gas nitrogen dryness process again. To the dried metabolic extract 50 µL of 20 mg/mL methoxylamine in pyridine was added and incubation at 60 °C for 2h followed. 100 µL of derivatization reagent, MSTFA with 1% TCMS, was added and the sample incubated at 60 °C for 1h. After the sample was cooled, 100 µL were transferred to a GC vials with insert and the sample was ready for analysis. Quality control (QC) samples were prepared from a pooled urine volume from all the study samples following the addition of FAME markers (20 µL of FAME/myristic acid d27 mixture). Details on the sampling processing and protocol along with the chromatography-related metadata are available in the Supplementary Analysis Protocol and in Supplementary Table S1 and Table S2. The workflow of the untargeted metabolomics analysis is provided in the Supplementary material. A QC mixture of 13 FAME markers was used throughout the process. In total, 18 QC samples were prepared and six were used in each batch (three before the samples, one in


Page 8 of 30

Page 9 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


the middle and two in the end). All QC samples were prepared with pooled urine from all subjects. The QC samples were used to monitor the analytical process. The QC chromatograms were overlaid during the analysis and inspected for major RT shifts. Clustering of nine of QC samples (three selected per batch) in unsupervised principal components analysis (PCA) was used to assess the variability (technical vs. biological)26,27.

Chromatogram deconvolution and compound identification Raw chromatograms were imported in AMDIS (version 2.71) for pre-processing in batch mode. Chromatogram deconvolution was performed with eight different combinations of search settings (Table 1). According to the Metabolomics Standards Initiative guidelines, metabolite identification at level 2 (putatively annotated compounds identified using spectral similarities with libraries) was initially performed, using the Fiehn library (G1676AA Agilent Fiehn GC/MS Metabolomics RTL Library) 17. The Fiehn library may also contain non unique spectra of the same compound at different stages of derivatization with indicative numbers next to each compound name. For the purpose of this analysis, each identified feature was considered unique even if it belonged to the same initial compound (i.e., L-threonine 1 and Lthreonine 2 were considered as two different compounds)

28

. Standard compounds in urine

were used to confirm the identity (at level 1) of the compounds found to be differentially expressed in the group comparisons.

Table 1. Raw chromatogram processing settings in AMDIS.

60 12

Search 2 60 12

Search 3 60 12

Search 4 60 12

Search 5 60 12

Search 6 60 12

One

One

One

One

One

Two

Search 1 Match factor Component width Adjacent peak subtraction


60 12

Search 8 60 10

One

One

Search 7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Resolution Sensitivity Shape requirements

Page 10 of 30

Medium Medium

Medium Medium

Medium Medium

High High

Low Low

Medium Medium

Medium Very High

Medium Medium

Medium

High

Low

High

Low

Low

Low

Low

Data analysis Initial analysis of results using the eight different deconvolution searches was performed for the QC samples. The total number of FAME markers identified in all QC samples and the difference between the expected and observed retention time (RT) were used to select the searches that were subsequently used in the analysis of the participants (unknown) samples. Given that there were no major differences between the expected and observed RT for the FAME markers identified, three searches were selected, based on the total number of identified FAME compounds. A peak table was generated from the estimated peak areas as extracted from the AMDIS report using the Metab package in R (version 3.3.2)

29,30

. The

peak intensity tables were then imported in MetaboAnalyst 31. Following the MetaboAnalyst workflow, missing values were replaced with half the minimum positive values of the original data and compounds present in < 50% of the samples were removed. Data was normalized using myristic acid d27 (RTL compound) as reference, and they were also logtransformed and scaled (Pareto scaling). Then, study participants were divided in two groups based on the presence of the disease and the levels of total THM (TTHM, the sum of TCM, BDCM, DBCM, and TCM). Differentially expressed metabolites based on the disease status (diabetics, and non-diabetics) and the TTHM levels (above and below the median of the raw ng/L concentrations) were those abiding with the criteria: (i) significantly different peak areas (p-value 2. Compounds differentially expressed between diabetics and non-diabetics were used in logistic regression analysis as predictors of the odds of T2D. The models were adjusted for age, BMI, sex, and the biomarkers of exposure (log-transformed – to the natural logarithm – TTHM) while


Page 11 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


creatinine was used also as a predictor to adjust for urine dilution. Association between TTHM and the differentially expressed metabolites were assessed also with linear regression models adjusted for age, gender and creatinine. P-values 2, as mentioned before. We opted for a higher cutoff for the selection of the differentially expressed compounds, because of the study is exploratory and the aim was to retain more compounds which can be evaluated more thoroughly based on the available literature and in future studies where targeted measurement can also be performed. Statistical analysis was conducted in R (version 3.3.2) 30 with RStudio (version 1.0.44)

32

and the packages: ggplot2, tableone, plyr, knitr, stargazer,

and sjPlot 33–39.

Results Study population and quality control of the metabolomics analysis A total number of 23 subjects were self-reported T2D patients among the participants of this study (n=51). Twenty-eight non-T2D controls were matched for age and sex when possible. Seven cases also reported being hypertensive. For one case, the BMI information was not available and the sample for the metabolomics analysis was not available for another case. There were no significant (p>0.05) differences in age, sex, marital status, education, smoking status and alcohol consumption between cases and controls (Table 2). Total THM levels were higher in cases than controls although this difference did not reach statistical significance (Table 2). Creatinine concentrations showed the opposite trend (higher in controls than the cases, and statistically significant, p 2. Details on the exact fold change estimates and the p-values are provided in the Supporting information (Supplementary Table S5 and Table S6).


Page 15 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


When the participants were grouped based on the median TTHM levels, three compounds namely, fucose (PubChem ID: 17106), 5-hydroxyindoleacetic acid (PubChem ID: 1826), and benzoic acid (PubChem ID: 243) were found to be differentially expressed (down-regulated). However, the crude association between TTHM and specific endogenous metabolites did not hold after adjusting for possible confounders (Supplementary Table S7). Fucose was also differentially expressed between T2D cases and controls, using search 1 results (Supplementary Table S5) and the compound was used as the dependent variable in linear regression adjusting for TTHM, disease status and possible confounders (i.e. age, sex, BMI). TTHM levels were not a significant predictor of fucose, when used along with creatinine (to account for urine dilution) while creatinine was significant (p

Coupling Urinary Trihalomethanes and Metabolomic Profiles of Type II

Recommend Documents