Article pubs.acs.org/jpr
Untargeted Mass Spectrometry-Based Metabolomic Profiling of Pleural Effusions: Fatty Acids as Novel Cancer Biomarkers for Malignant Pleural Effusions Ching-Wan Lam* and Chun-Yiu Law Department of Pathology, The University of Hong Kong, Hong Kong, China ABSTRACT: Untargeted mass spectrometry-based metabolomic profiling is a powerful analytical method used for broad-spectrum identification and quantification of metabolites in biofluids in human health and disease states. In this study, we exploit metabolomic profiling for cancer biomarker discovery for diagnosis of malignant pleural effusions. We envisage the result will be clinically useful since currently there are no cancer biomarkers that are accurate enough for the diagnosis of malignant pleural effusions. Metabolomes of 32 malignant pleural effusions from lung cancer patients and 18 benign effusions from patients with pulmonary tuberculosis were analyzed using reversed-phase liquid chromatography tandem mass spectrometry (LC−MS/MS) using AB SCIEX TripleTOF 5600. MS spectra were analyzed using XCMS, PeakView, and LipidView. Metabolome-Wide Association Study (MWAS) was performed by Receiver Operating Characteristic Curve Explorer and Tester (ROCCET). Insignificant markers were filtered out using a metabolome-wide significance level (MWSL) with p-value < 2 × 10−5 for t test. Only compounds in Human Metabolome Database (HMDB) will be used as cancer biomarkers. ROCCET analysis of ESI positive and negative MS spectra revealed free fatty acid (FFA) 18:1 (oleic acid) had the largest area-under-ROC of 0.96 (95% CI = 0.87−1.00) in malignant pleural effusions. Using a ratio of FFA 18:1-to-ceramide (d18:1/16:0), the area-under-ROC was further increased to 0.99 (95% CI = 0.91−1.00) with sensitivity 93.8% and specificity 100.0%. Using untargeted metabolomic profiling, the diagnostic cancer biomarker with the largest area-under-ROC can be determined objectively. This lipogenic phenotype could be explained by overexpression of fatty acid synthase (FASN) in cancer cells. The diagnostic performance of FFA 18:1-to-ceramide (d18:1/16:0) ratio supports its use for diagnosis of malignant pleural effusions. KEYWORDS: Metabolomic profiling, mass spectrometry, pleural effusions, fatty acids
■
INTRODUCTION Lung cancer is the most common cancer with a worldwide incidence of 1.2 million and a mortality of 1.1 million.1 About 40% of exudative pleural effusions are related to lung cancer2−4 and about 55% are related to pulmonary tuberculosis.5 Clinically, both conditions showed overlapping symptoms including chronic cough, shortness of breath, unexplained weight loss, fatigue, loss of appetite, low-grade fever, etc. The diagnosis is most often made by lung biopsy and bacterial culture. However, lung biopsy is an invasive procedure, while bacterial culture for tuberculosis usually takes a long time. However, there are no current diagnostic cancer biomarkers for accurate diagnosis of malignant pleural effusions. To date, various markers, such as cytokeratin fragment (CYFRA) 21-1, cancer antigen (CA) 15-3, and CA 19-9 have been proposed as cancer biomarkers for malignant effusions, but the majority of these cancer biomarkers have low diagnostic sensitivity of less than 60%.6,7 A better novel cancer biomarker with higher sensitivity for malignant pleural effusions will be useful clinically. To identify novel cancer biomarkers for malignant pleural effusions, our strategy is to apply state-of-the-art technology for metabolomic profiling, i.e., high-resolution tandem mass spectrometry (MS/MS), to profile the metabolomes of pleural effusions of patients with lung cancer. Metabolomic profiling is an effective tool © XXXX American Chemical Society
for disease biomarker discovery by observing changes in metabolite concentrations in various biofluids in disease state.8,9 Through untargeted metabolomic analyses, various metabolic pathways that are activated in cancer can be identified and the associated pathway metabolites can be used as diagnostic cancer biomarkers. In this study, we studied the metabolomes of malignant pleural effusions by high-resolution liquid chromatography (LC)−MS/ MS. We hypothesize that this novel approach will subclassify exudative pleural effusions caused by malignancy. The most diagnostic marker from each group was identified using wholemetabolome receiver operating characteristic curve (ROC) analysis, and the ratio of the two biomarkers was used to assess their diagnostic performances in subclassifying exudative pleural effusions into malignant and benign pleural effusions.
■
MATERIALS AND METHODS
Samples
In this study, 50 pleural effusions from 32 patients with malignant pleural effusions and 18 patients with pulmonary tuberculosis Received: April 17, 2014
A
dx.doi.org/10.1021/pr5003774 | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
Omics-Based Biomarker Identification
were recruited for LC−MS/MS analysis. The collections of the samples were before any treatments given to the patients including anticancer or anti-TB treatments. The samples were collected in sterile plain bottles, transferred immediately to the laboratory, and centrifuged at 3000 rpm at 4 °C for 10 min. One hundred microliters of the clear supernatant was mixed with 400 μL of a solution of acetone/ethanol/methanol (1:1:1, v/v). After centrifugation, the supernatants were subjected to LC−MS/MS analysis.
The LC−MS/MS data were processed using PeakView (AB SCIEX) for the examination of the extracted ion chromatograms (XIC) and LipidView (AB SCIEX) for the identification of lipids. The raw LC−MS/MS data acquired were converted into mzXML files by MSconvert (ProteoWizard). The converted file was further processed using the XCMS (http://www. bioconductor.org/packages/2.8/bioc/html/xcms.html) running under R (http://www.r-project.org/), a platform for XCMS package. The processed data were normalized based on the total intensity. A ROC analysis was performed using the online ROCCET (ROC Curve Explorer and Tester, http://www. roccet.ca).10 Low-quality variables, such as those containing empty/zero/missing values >50% and those with 5% near constant variables based on the relative standard deviation were removed. The classical ROC curve analysis was used. The positive identification of the compounds from the LC−MS/MS data was based on chemical formula generated from the accurate mass with an accuracy 0.05). For the malignant pleural effusions, the cytology assessment showed 23 cases of adenocarcinomas, 4 cases of non-small cell carcinomas, 2 cases of small cell carcinomas, 1 case of anaplastic carcinoma, and 1 case with atypical cells. One case was diagnosed clinically, the B
dx.doi.org/10.1021/pr5003774 | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
against m/z (Figure 1A) and area-under-ROC curves (Figure 1B). Insignificant markers were filtered out using a metabolome-wide significance level (MWSL) with p-value
Untargeted Mass Spectrometry-Based Metabolomic Profiling of Pleural Effusions: Fatty Acids as Novel Cancer Biomarkers for Malignant Pleural Effusions Ching-Wan Lam* and Chun-Yiu Law Department of Pathology, The University of Hong Kong, Hong Kong, China ABSTRACT: Untargeted mass spectrometry-based metabolomic profiling is a powerful analytical method used for broad-spectrum identification and quantification of metabolites in biofluids in human health and disease states. In this study, we exploit metabolomic profiling for cancer biomarker discovery for diagnosis of malignant pleural effusions. We envisage the result will be clinically useful since currently there are no cancer biomarkers that are accurate enough for the diagnosis of malignant pleural effusions. Metabolomes of 32 malignant pleural effusions from lung cancer patients and 18 benign effusions from patients with pulmonary tuberculosis were analyzed using reversed-phase liquid chromatography tandem mass spectrometry (LC−MS/MS) using AB SCIEX TripleTOF 5600. MS spectra were analyzed using XCMS, PeakView, and LipidView. Metabolome-Wide Association Study (MWAS) was performed by Receiver Operating Characteristic Curve Explorer and Tester (ROCCET). Insignificant markers were filtered out using a metabolome-wide significance level (MWSL) with p-value < 2 × 10−5 for t test. Only compounds in Human Metabolome Database (HMDB) will be used as cancer biomarkers. ROCCET analysis of ESI positive and negative MS spectra revealed free fatty acid (FFA) 18:1 (oleic acid) had the largest area-under-ROC of 0.96 (95% CI = 0.87−1.00) in malignant pleural effusions. Using a ratio of FFA 18:1-to-ceramide (d18:1/16:0), the area-under-ROC was further increased to 0.99 (95% CI = 0.91−1.00) with sensitivity 93.8% and specificity 100.0%. Using untargeted metabolomic profiling, the diagnostic cancer biomarker with the largest area-under-ROC can be determined objectively. This lipogenic phenotype could be explained by overexpression of fatty acid synthase (FASN) in cancer cells. The diagnostic performance of FFA 18:1-to-ceramide (d18:1/16:0) ratio supports its use for diagnosis of malignant pleural effusions. KEYWORDS: Metabolomic profiling, mass spectrometry, pleural effusions, fatty acids
■
INTRODUCTION Lung cancer is the most common cancer with a worldwide incidence of 1.2 million and a mortality of 1.1 million.1 About 40% of exudative pleural effusions are related to lung cancer2−4 and about 55% are related to pulmonary tuberculosis.5 Clinically, both conditions showed overlapping symptoms including chronic cough, shortness of breath, unexplained weight loss, fatigue, loss of appetite, low-grade fever, etc. The diagnosis is most often made by lung biopsy and bacterial culture. However, lung biopsy is an invasive procedure, while bacterial culture for tuberculosis usually takes a long time. However, there are no current diagnostic cancer biomarkers for accurate diagnosis of malignant pleural effusions. To date, various markers, such as cytokeratin fragment (CYFRA) 21-1, cancer antigen (CA) 15-3, and CA 19-9 have been proposed as cancer biomarkers for malignant effusions, but the majority of these cancer biomarkers have low diagnostic sensitivity of less than 60%.6,7 A better novel cancer biomarker with higher sensitivity for malignant pleural effusions will be useful clinically. To identify novel cancer biomarkers for malignant pleural effusions, our strategy is to apply state-of-the-art technology for metabolomic profiling, i.e., high-resolution tandem mass spectrometry (MS/MS), to profile the metabolomes of pleural effusions of patients with lung cancer. Metabolomic profiling is an effective tool © XXXX American Chemical Society
for disease biomarker discovery by observing changes in metabolite concentrations in various biofluids in disease state.8,9 Through untargeted metabolomic analyses, various metabolic pathways that are activated in cancer can be identified and the associated pathway metabolites can be used as diagnostic cancer biomarkers. In this study, we studied the metabolomes of malignant pleural effusions by high-resolution liquid chromatography (LC)−MS/ MS. We hypothesize that this novel approach will subclassify exudative pleural effusions caused by malignancy. The most diagnostic marker from each group was identified using wholemetabolome receiver operating characteristic curve (ROC) analysis, and the ratio of the two biomarkers was used to assess their diagnostic performances in subclassifying exudative pleural effusions into malignant and benign pleural effusions.
■
MATERIALS AND METHODS
Samples
In this study, 50 pleural effusions from 32 patients with malignant pleural effusions and 18 patients with pulmonary tuberculosis Received: April 17, 2014
A
dx.doi.org/10.1021/pr5003774 | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
Omics-Based Biomarker Identification
were recruited for LC−MS/MS analysis. The collections of the samples were before any treatments given to the patients including anticancer or anti-TB treatments. The samples were collected in sterile plain bottles, transferred immediately to the laboratory, and centrifuged at 3000 rpm at 4 °C for 10 min. One hundred microliters of the clear supernatant was mixed with 400 μL of a solution of acetone/ethanol/methanol (1:1:1, v/v). After centrifugation, the supernatants were subjected to LC−MS/MS analysis.
The LC−MS/MS data were processed using PeakView (AB SCIEX) for the examination of the extracted ion chromatograms (XIC) and LipidView (AB SCIEX) for the identification of lipids. The raw LC−MS/MS data acquired were converted into mzXML files by MSconvert (ProteoWizard). The converted file was further processed using the XCMS (http://www. bioconductor.org/packages/2.8/bioc/html/xcms.html) running under R (http://www.r-project.org/), a platform for XCMS package. The processed data were normalized based on the total intensity. A ROC analysis was performed using the online ROCCET (ROC Curve Explorer and Tester, http://www. roccet.ca).10 Low-quality variables, such as those containing empty/zero/missing values >50% and those with 5% near constant variables based on the relative standard deviation were removed. The classical ROC curve analysis was used. The positive identification of the compounds from the LC−MS/MS data was based on chemical formula generated from the accurate mass with an accuracy 0.05). For the malignant pleural effusions, the cytology assessment showed 23 cases of adenocarcinomas, 4 cases of non-small cell carcinomas, 2 cases of small cell carcinomas, 1 case of anaplastic carcinoma, and 1 case with atypical cells. One case was diagnosed clinically, the B
dx.doi.org/10.1021/pr5003774 | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
against m/z (Figure 1A) and area-under-ROC curves (Figure 1B). Insignificant markers were filtered out using a metabolome-wide significance level (MWSL) with p-value
