Metabolomic and metallomic profile differences between Veterans and Civilians with Pulmonary Sarcoidosis

Sarcoidosis is a disorder characterized by granulomatous inflammation of unclear etiology. In this study we evaluated whether veterans with sarcoidosis exhibited different plasma metabolomic and metallomic profiles compared with civilians with sarcoidosis. A case control study was performed on veteran and civilian patients with confirmed sarcoidosis. Proton nuclear magnetic resonance spectroscopy (1H NMR), hydrophilic interaction liquid chromatography mass spectrometry (HILIC-MS) and inductively coupled plasma mass spectrometry (ICP-MS) were applied to quantify metabolites and metal elements in plasma samples. Our results revealed that the veterans with sarcoidosis significantly differed from civilians, according to metabolic and metallomics profiles. Moreover, the results showed that veterans with sarcoidosis and veterans with COPD were similar to each other in metabolomics and metallomics profiles. This study suggests the important role of environmental risk factors in the development of different molecular phenotypic responses of sarcoidosis. In addition, this study suggests that sarcoidosis in veterans may be an occupational disease.


Sample Preparation
200 µl of samples were thawed on the ICE and were them filtered using 3 kDa NanoSep microcentrifuge filters at 12,000 x g for 1 hour at 4ºC. Filters were prewashed before adding samples using ddH2O to reduce preservative contamination. After collecting samples into clean 1.5 ml vials, filters were additionally rinsed using 100 µl of D2O. 80 µl NMR phosphate buffer was added to samples and sample were adjusted using D2O to final volume 400 µl. NMR phosphate buffer consisted of 0.5 M NaH2PO4 buffer solution at pH 7.0) containing 2.5 mM 2,2dimethylsilapentane-5-sulfonate (DSS, final concentration 0.5 mM) as an internal reference compound. To prevent bacterial contamination, 10 µl sodium azide (1 M NaN3) was added to each sample. pH was adjusted to 7.0 ± 0.04 at room temperature.

Data Acquisition
NMR spectroscopy was carried out in a blinded manner using an automated sample changer hold at 4ºC on a a 600 MHz Bruker Ultrashield Plus NMR spectrometer (Bruker BioSPin Ltd., Canada). NMR spectra were acquired by pre-saturation pulse sequence (noesypr 1d) with an optimized water suppression and a mixing time of 100 ms [1,2]. NMR spectra were obtained by 1024 scan, zero filled and Fourier transformed to 128k points. NMR spectra were processed by Topspin software program to line broadening, phasing, baseline correction and referencing peaks compared to the DSS peak at 0.0 ppm using the (Bruker BioSpin Ltd., Canada).

Metabolite Concentration Profiling
NMR spectra were randomly ordered and analyzed to avoid progressive bias. ChenomX NMR Suite 7.1 software (Chenomx Inc., Edmonton, Alberta, Canada) was used to identify and quantify metabolites of NMR spectra [1]. NMR spectra was manually phased in the processor module of the ChenomX software. Water region and baseline correction was deleted to ease profiling of the peaks. All peaks were automatically quantified compare to the DSS concentration, as an internal reference [3].
Fifty µl of plasma samples were centrifuged at 13200 rpm for 10 minutes at 4°C temperature. Supernatant was then transferred to a 15 ml tubes containing 10 ml of 2% nitric acid to obtain a 1:200 dilution. Diluted samples in acid nitric were centrifuged at 4000 rpm for 10 minutes to make the samples as clean as possible without debris. SeronormTM trace elements (SERO AS, Billingstad, Norway) serum level 2 (lot 1309416) was used during experiment of samples as the quality controls (QCs) to evaluate the reproducibility and lowering of sensitivity due to contamination of the device. The preparation of Seronorm was done based on the manufacturers' instruction by dissolving in water followed by dilution in 2% nitric acid.

Data analysis
Both NMR and HILIC-MS datasets were normalized by median fold change normalization data [9]. Log transformed and centering univariance scaling (UV) were applied to preprocessed datasets. Multivariate data analysis wes performed then analyzed using the SIMCA-P+ program (Version 13.0, Umetrics AB, Umeå, Sweden).

Verification of the OPLS-DA Model
CV-ANOVA, R2Y and Q2Y were considered to verify the quality of OPLS-DA, a supervised multivariate separation analysis. CV-ANOVA determine the reliability and assess the significance of discrimination model. R2Y and Q2Y explain the goodness of variation of group status and the goodness of prediction, respectively [10]. Both R2Y and Q2Y is obtained through a cross validation method which was based on the internal sevenfold cross validation by leaving out of one seventh (1/7th) to test R2Y on remaining portion and measuring the prediction (Q2Y) using 1/7th removed.
This process is repeated 7 times for each for each left out group of samples. R2Y and Q2Y score can vary between 0 and 1, where the scores are closer to 1 showing an excellent with a high level of predictability. Q2Y more than 0.5 is considered a good model and high predictability for human samples. Q2Y> 0.7 is a very good model.

Variables important on projection (VIP)
VIP was used to choose the best OPLS-DA model regarding the highest predictability. The VIP is used to select the metabolites with the highest effects in separation of the two groups. VIP works in a weighted fashion using a quantitative measure of discriminatory power of the metabolites that is ranked by a unitless number. Table S1, Unpaired t-test analysis of the 1H-NMR dataset showing significant (p < 0.05) metabolites between the veteran and civilian sarcoidosis cohorts. (Highlight are metabolites with FDR < 0.05)         Figure S13. PCA model (2D and 3D plots) of metallomic data obtained by ICP-MS shows the contribution of 33 elements measured in separating veteran subjects with sarcoidosis from COPD controls.

Supplementary (Figures and Tables)
Figure S14. Coefficient plot shows alterations of 33 elements in veteran and civilians with confirmed sarcoidosis.
A B Figure S15, OPLS-DA of ICP-MS-based metallomics data using all identified metal ions comparing veterans with COPD with A. civilian subjects with sarcoidosis (n=33 elements) and B. veterans with sarcoidosis (N=33 elements).
A B Figure S16. Partial least square regression (PLSR) analysis shows a strong relation between most important metabolites in relation to the sarcoidosis cohort type (veteran or civilian). A. using metabolites detected by 1H-NMR. B. Using metabolites detected by HILIC-MS. Figure S17. Partial least square regression (PLSR) analysis shows a relative strong relation between 33 elements detected by ICP-MS in relation to the sarcoidosis cohort type (veteran or civilian).
Figure S18. Coefficient plot shows relative correlation of measured metabolites between subjects with sarcoidosis radiologic stage 4 versus subjects with stages 1-3. Figure S14, Cytoscape-based pathway analysis using 1H-NMR data showing several metabolic pathways that differ between veteran sarcoidosis and civilian sarcoidosis cohorts. Figure S15, Cytoscape-based pathway analysis using HILIC-MS data showing several metabolic pathways that differ between veteran sarcoidosis and civilian sarcoidosis cohorts.