Differentiai Diagnosis by Metaboiic Profiie: A Dream or Reaiity?* Samuel K. Appavu, MD, FCCM Department of Surgery University of Illinois College of Medicine at Rockford Rockford, IL . ualitative and quantitative measurement and interpreItation of metabolic profile in health and disease are the current area of research and are a recent progression ñ'om t ^ studies of genomics and proteomics. Advanced analytical technologies including high-performance liquid chromatography gas chromatography (GC), mass spectrometry (MS), and nuclear magnetic resonance (NMR) spectroscopy either alone or in combination are being widely used to analyze samples from natural sources, including tissues and biofiuids. Chemometrics, a system of statistical analytical applications, is used to evaluate and understand the complex data derived from these experiments. A quick look at current research on metabolomics will show studies involving every human organ system, including prediction of embryo implantation based on spent culture medium fi'om in vitro fertilization. There are also studies on discovery of new chemicals and analyses of the structure of such new chemicals. Human metabolomic studies aim to develop metabolic profiles of diseases wdth the hope that such multiparameter profiles may help to more accurately diagnose specific disease processes than a single biomarker. Metabolomic research is being applied to study of both chronic and acute illnesses. Some examples wiU be highlighted here. In oncologic research, application of the above technologies has yielded promising results. In 1993, Schiebler et al (1), using 'H NMR spectroscopy of prostate tissue, demonstrated the metabolic profiles of normal prostate, benign prostatic hyperplasia, and adenocarcinoma. Their work demonstrated that depressed levels of citrate in adenocarcinoma may be the discriminating factor. Sitter et al (2) studied breast cancer tissue from 85 patients along with adjacent uninvolved tissue in 18 of these patients using high-resolution magic-angle-spinning NMR spectroscopy. The resulting spectra were examined by three different approaches: 1) Relative intensities of glycerophosphocholine (GPC), phosphocholine (PC), and choline were compared for cancerous and uninvolved specimens. 2) Eight metabolites, choline, creatine, ß-glucose, GPC, glycine, myo-inositol, PC, and taurine, were quantified from the recorded spectra and compared with tumor histological type and size, patient's lymph node status, and tissue composition of the 'See also p. 1140. Key Words: inflammatory bowel disease; metaboiomic research; nuclear magnetic resonance spectroscopy; septic shock; urinary tract infection The author has disclosed that he does not have any potential conflicfs of interest. Copyright ® 2014 by the Society of Criticai Care Medicine and Lippincott Williams & Wiikins DOI: 10.1097/CCM.0000000000000254
1300
www.ccmjournal.org
sample. 3) The spectra were also compared with tumor histological type and size, lymph node status, and tissue composition of samples using principal component analysis. The researchers showed that tumor samples could be distinguished from noninvolved samples based on relative intensities of signals from GPC, PC, and choline with 82% sensitivity and 100% specificity. Using a combination of NMR and two-dimensional GCMS (GCxGC-MS), Asiago et al (3) analyzed the metabolic profiles of 56 breast cancer patients during follow-up after surgical therapy. Several blood samples were drawn during this follow-up. The cancer recurred in 20 patients, whereas the remaining 36 patients were cancerfi'ee.NMR and GCxGC-MS data of blood samples from both these groups were compared. Eleven metabolite markers (seven from NMR and four from GCxGC-MS) were analyzed. These metabolite markers identified samples from patients with disease recurrence with a sensitivity of 86% and a specificity of 84% (area under the receiver operating characteristic curve = 0.88). Eifiy-five percent of the patients could be correctly predicted to have recurrence on an average of 13 months before the actual recurrence was clinically diagnosed, representing a large improvement over the breast cancer-monitoring assay CA 27.29. With respect to nononcologic chronic diseases, Williams et al (4) in a well-designed study examined urine samples of 86 patients with Crohn's disease (CD), 60 patients with ulcerative colitis (UC), and 60 healthy controls using high-resolution 'H NMR spectroscopy. They evaluated urinary metabolites that are infiuenced by gut microfiora. They found hippurate levels to be the lowest in CD and differed significantly from the other two groups. Eormate levels were higher and 4-cresol sulfate levels were lower in CD compared with UC. They determined that specific urinary metabolites related to gut microflora significantly differed between patients with CD and UC. Stephens et al (5), also studying urinary metabolic profiles of patients with infiammatory bowel disease (IBD) and healthy individuals using NMR methodology, found discrimination between IBD patients and the healthy. They also showed that CD and UC could be differentiated when the total study population data are considered, but if patients who underwent surgical resection were removed from the analysis, this discrimination could no longer be detected. This difference could be from many reasons including sample size and study methodology. NMR technology has been widely used to study microbiology and infectious diseases. Gupta et al (6) studied 682 urine specimens fi-om patients suspected to have urinary tract infections (UTI), 50 healthy controls, and also, commercially available standard strains of Gram-negative bacilli and Gram-positive cocci known to cause UTI. The urine specimens were also studied by standard microbiological culture methods. They found acetate, lactate, succinate, and formate were able to differentiate between UTI and healthy controls with 99.5% accuracy. They were also able to differentiate between Gram-negative and Gram-positive infections in 96% of May 2014'Voiume 42 • Number 5
Editorials
the specimens. In another study, Cupta et al (7) were also able to distinguish pseudomonas UTI by 'H NMR spectroscopy on the basis that Pseudomonas aeruginosa converted nicotinic acid to 6-hydroxy-nicotinic acid. In a tliird study, they (8) were able to identity Klebsiella pneumoniae UTI based on its ability to metabolize glycerol to 1.3-propanediol, acetate, ethanol, and succinate. Lin et al (9) studied experimental sepsis using cecal ligation and puncture (CLP) in a rat model. Eorty septic rats receiving CLP were divided into the surviving group and nonsurviving group on day 6, whereas 20 sham-operated rats served as the control group. Serum samples were collected from septic and sham-operated rats at 12 hours after surgery and analyzed using 'H NMR spectroscopy, which indicated that NMR-based metabolic profiling could reveal pathologic characteristics in the serum of sham-operated, surviving, and nonsurviving septic rats. In addition, six metabolites, including lactate, alanine, acetate, acetoacetate, hydroxybutyrate, and formate, which are mainly involved in energy metabohsm, changed markedly in septic rats, especially in the nonsurvivors. A predictive model for prognostic evaluation of sepsis using these data demonstrated a prediction accuracy of about 87%. Stringer et al ( 10) studied sera of patients with sepsis-induced acute lung injury and healthy controls using 'H NMR spectroscopy. Their data revealed differences between patients with ALI and healthy subjects in the level of total glutathione, adenosine, phosphatidylserine, and sphingomyelin. Furthermore, myo-inositol levels were associated with acute physiology scores (APS) and ventilator-free days. There was also an association between total glutathione and APS. They concluded that NMR quantitative analysis yielded a physiologically relevant metabolite dataset that distinguished sepsis-induced ALI from health. In this issue of Critical Care Medicine, Mickiewicz et al (11) present their work. They studied blood samples that had been collected as part of Critical Care Epidemiological and Biological Tissue Resource, an ICU tissue bank. Tbese samples had been collected within 24 hours of admission to the ICU. They obtained these samples from two groups of ICU patients: the first group consisted of patients who met the criteria for systemic inflammatory response syndrome (SIRS) and septic shock (septic shock, intra-abdominal septic processes, pneumonia, and urosepsis). The second group of patients who served as controls met less than two SIRS criteria (postcardiac surgery, angina, spinal stenosis, splenic rupture, and burns). Data were collected on patients' clinical progress and outcome. The blood samples were analyzed by 'H NMR spectroscopy. Chemometric analytical methods were used to interpret the data. They show several important findings: based on blood sample metabolic profiles, they were able to differentiate between septic and nonseptic patients. They identified 60 metabolites of which 31 became important in this differentiation. Among septic shock blood samples, the concentration of 14 metabolites increased while 17 others decreased. They also compared data from septic shock nonsurvivors to those of age- and gender-matched septic shock survivors. They found excellent separation between survivors and nonsurvivors based on the profile of 20 metabolites and excellent validation Critical Care Medicine
metrics. When compared with admission Acute Physiology and Chronic Health Evaluation and Sequential Organ Failure Assessment scores, metabolomics model had more accurate diagnostic and prognostic power. This is the first adult human study to demonstrate the advantages of metabolic data in distinguishing septic shock versus nonseptic patients and in predicting survival. This is a well-performed study. The design, methods, and analyses have been performed excellently. The data are consistent with studies of similar type. There are, however, several weaknesses. This is a study of convenient samples, analyzed retrospectively. The sample size is small, especially the comparison of septic shock survivors and nonsurvivors, which had just four patients each. More important issue is its clinical applicability in septic sbock. When a patient is admitted to the ICU with features of SIRS or septic shock, it is not difficult to make the diagnosis of the condition. The difficulty is to find the source or cause of the septic shock in a timely manner. Therapeutic measures are always undertaken even before a source can be definitively identified. Metabolomics research is still evolving. Data thus far show that there is a difference in the metabolic profiles between health and disease, which is hardly surprising. It is known that global metabolomic analysis wiU yield an enormous number of metabolites and not all of them will be important. They may also change at different stages of health without there being a disease process causing such a change. From critical care perspective, the applicability of metabolomics will depend not on its ability to differentiate between disease and health but on whether or not it can provide timely assistance in differentiating between the multitude of diseases with an accuracy and reliability tbat is clinically acceptable.
REFERENCES 1. Sohiebler ML, Miyamoto KK, White M, et al: In vitro high resolution 1 H-spectroscopy of the human prostate: Benign prostatic hyperplasia, normal peripheral zone and adenocarcinoma. Magn Reson Med 1993; 29:285-291 2. Sitter B, Lundgren S, Bathen TF, et al: Comparison of HR MAS MR spectroscopic profiles of breast cancer tissue with clinical parameters. NMR Biomed 2006; 1 9:30-40 3. Asiago VM, Alvarado LZ, Shanaiah N, et al: Early detection of recurrent breast cancer using metabolite profiling. Cancer Res 2010; 70:8309-8318 4. Williams HR, Cox IJ, Walker DG, et al: Characterization of inflammatory bowel disease with urinary metabolic profiling. Am J Gastroenterol 2009; 104:1435-1444 5. Stephens NS, Siffledeen J, Su X, et al: Urinary NMR metabolomic profiles discriminate inflammatory bowel disease from healthy. J Crohns Colitis 20^3; 7:e42-e48 6. Gupta A, Dwivedi M, Mahdi AA, et al: Broad identification of bacterial type in urinary tract infection using (1)h NMR spectroscopy. J Proteome Res 201 2; 11:1844-1854 7. Gupta A, Dwivedi M, Nagana Gowda GA, et al: (1)H NMR spectroscopy in the diagnosis of Pseudomonas aeruginosa-'mduceá urinary tract infection. NMR Biomed 2005; 18:293-299 8. Gupta A, Dwivedi M, Gowda GA, et al: 1 H NMR spectroscopy in the diagnosis of Klebsiella pneumoniae-'induced urinary tract infection. NMR Biomed 2006; 19:1055-1061 9. Lin ZY, Xu PB, Yan SK, et al: A metabonomic approach to early prognostic evaluation of experimental sepsis by {1)H NMR and pattern recognition. NMR Biomed 2009; 22:601 -608 www.ccrTnjournal.org
1301
Copyright of Critical Care Medicine is the property of Lippincott Williams & Wilkins and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
1300
www.ccmjournal.org
sample. 3) The spectra were also compared with tumor histological type and size, lymph node status, and tissue composition of samples using principal component analysis. The researchers showed that tumor samples could be distinguished from noninvolved samples based on relative intensities of signals from GPC, PC, and choline with 82% sensitivity and 100% specificity. Using a combination of NMR and two-dimensional GCMS (GCxGC-MS), Asiago et al (3) analyzed the metabolic profiles of 56 breast cancer patients during follow-up after surgical therapy. Several blood samples were drawn during this follow-up. The cancer recurred in 20 patients, whereas the remaining 36 patients were cancerfi'ee.NMR and GCxGC-MS data of blood samples from both these groups were compared. Eleven metabolite markers (seven from NMR and four from GCxGC-MS) were analyzed. These metabolite markers identified samples from patients with disease recurrence with a sensitivity of 86% and a specificity of 84% (area under the receiver operating characteristic curve = 0.88). Eifiy-five percent of the patients could be correctly predicted to have recurrence on an average of 13 months before the actual recurrence was clinically diagnosed, representing a large improvement over the breast cancer-monitoring assay CA 27.29. With respect to nononcologic chronic diseases, Williams et al (4) in a well-designed study examined urine samples of 86 patients with Crohn's disease (CD), 60 patients with ulcerative colitis (UC), and 60 healthy controls using high-resolution 'H NMR spectroscopy. They evaluated urinary metabolites that are infiuenced by gut microfiora. They found hippurate levels to be the lowest in CD and differed significantly from the other two groups. Eormate levels were higher and 4-cresol sulfate levels were lower in CD compared with UC. They determined that specific urinary metabolites related to gut microflora significantly differed between patients with CD and UC. Stephens et al (5), also studying urinary metabolic profiles of patients with infiammatory bowel disease (IBD) and healthy individuals using NMR methodology, found discrimination between IBD patients and the healthy. They also showed that CD and UC could be differentiated when the total study population data are considered, but if patients who underwent surgical resection were removed from the analysis, this discrimination could no longer be detected. This difference could be from many reasons including sample size and study methodology. NMR technology has been widely used to study microbiology and infectious diseases. Gupta et al (6) studied 682 urine specimens fi-om patients suspected to have urinary tract infections (UTI), 50 healthy controls, and also, commercially available standard strains of Gram-negative bacilli and Gram-positive cocci known to cause UTI. The urine specimens were also studied by standard microbiological culture methods. They found acetate, lactate, succinate, and formate were able to differentiate between UTI and healthy controls with 99.5% accuracy. They were also able to differentiate between Gram-negative and Gram-positive infections in 96% of May 2014'Voiume 42 • Number 5
Editorials
the specimens. In another study, Cupta et al (7) were also able to distinguish pseudomonas UTI by 'H NMR spectroscopy on the basis that Pseudomonas aeruginosa converted nicotinic acid to 6-hydroxy-nicotinic acid. In a tliird study, they (8) were able to identity Klebsiella pneumoniae UTI based on its ability to metabolize glycerol to 1.3-propanediol, acetate, ethanol, and succinate. Lin et al (9) studied experimental sepsis using cecal ligation and puncture (CLP) in a rat model. Eorty septic rats receiving CLP were divided into the surviving group and nonsurviving group on day 6, whereas 20 sham-operated rats served as the control group. Serum samples were collected from septic and sham-operated rats at 12 hours after surgery and analyzed using 'H NMR spectroscopy, which indicated that NMR-based metabolic profiling could reveal pathologic characteristics in the serum of sham-operated, surviving, and nonsurviving septic rats. In addition, six metabolites, including lactate, alanine, acetate, acetoacetate, hydroxybutyrate, and formate, which are mainly involved in energy metabohsm, changed markedly in septic rats, especially in the nonsurvivors. A predictive model for prognostic evaluation of sepsis using these data demonstrated a prediction accuracy of about 87%. Stringer et al ( 10) studied sera of patients with sepsis-induced acute lung injury and healthy controls using 'H NMR spectroscopy. Their data revealed differences between patients with ALI and healthy subjects in the level of total glutathione, adenosine, phosphatidylserine, and sphingomyelin. Furthermore, myo-inositol levels were associated with acute physiology scores (APS) and ventilator-free days. There was also an association between total glutathione and APS. They concluded that NMR quantitative analysis yielded a physiologically relevant metabolite dataset that distinguished sepsis-induced ALI from health. In this issue of Critical Care Medicine, Mickiewicz et al (11) present their work. They studied blood samples that had been collected as part of Critical Care Epidemiological and Biological Tissue Resource, an ICU tissue bank. Tbese samples had been collected within 24 hours of admission to the ICU. They obtained these samples from two groups of ICU patients: the first group consisted of patients who met the criteria for systemic inflammatory response syndrome (SIRS) and septic shock (septic shock, intra-abdominal septic processes, pneumonia, and urosepsis). The second group of patients who served as controls met less than two SIRS criteria (postcardiac surgery, angina, spinal stenosis, splenic rupture, and burns). Data were collected on patients' clinical progress and outcome. The blood samples were analyzed by 'H NMR spectroscopy. Chemometric analytical methods were used to interpret the data. They show several important findings: based on blood sample metabolic profiles, they were able to differentiate between septic and nonseptic patients. They identified 60 metabolites of which 31 became important in this differentiation. Among septic shock blood samples, the concentration of 14 metabolites increased while 17 others decreased. They also compared data from septic shock nonsurvivors to those of age- and gender-matched septic shock survivors. They found excellent separation between survivors and nonsurvivors based on the profile of 20 metabolites and excellent validation Critical Care Medicine
metrics. When compared with admission Acute Physiology and Chronic Health Evaluation and Sequential Organ Failure Assessment scores, metabolomics model had more accurate diagnostic and prognostic power. This is the first adult human study to demonstrate the advantages of metabolic data in distinguishing septic shock versus nonseptic patients and in predicting survival. This is a well-performed study. The design, methods, and analyses have been performed excellently. The data are consistent with studies of similar type. There are, however, several weaknesses. This is a study of convenient samples, analyzed retrospectively. The sample size is small, especially the comparison of septic shock survivors and nonsurvivors, which had just four patients each. More important issue is its clinical applicability in septic sbock. When a patient is admitted to the ICU with features of SIRS or septic shock, it is not difficult to make the diagnosis of the condition. The difficulty is to find the source or cause of the septic shock in a timely manner. Therapeutic measures are always undertaken even before a source can be definitively identified. Metabolomics research is still evolving. Data thus far show that there is a difference in the metabolic profiles between health and disease, which is hardly surprising. It is known that global metabolomic analysis wiU yield an enormous number of metabolites and not all of them will be important. They may also change at different stages of health without there being a disease process causing such a change. From critical care perspective, the applicability of metabolomics will depend not on its ability to differentiate between disease and health but on whether or not it can provide timely assistance in differentiating between the multitude of diseases with an accuracy and reliability tbat is clinically acceptable.
REFERENCES 1. Sohiebler ML, Miyamoto KK, White M, et al: In vitro high resolution 1 H-spectroscopy of the human prostate: Benign prostatic hyperplasia, normal peripheral zone and adenocarcinoma. Magn Reson Med 1993; 29:285-291 2. Sitter B, Lundgren S, Bathen TF, et al: Comparison of HR MAS MR spectroscopic profiles of breast cancer tissue with clinical parameters. NMR Biomed 2006; 1 9:30-40 3. Asiago VM, Alvarado LZ, Shanaiah N, et al: Early detection of recurrent breast cancer using metabolite profiling. Cancer Res 2010; 70:8309-8318 4. Williams HR, Cox IJ, Walker DG, et al: Characterization of inflammatory bowel disease with urinary metabolic profiling. Am J Gastroenterol 2009; 104:1435-1444 5. Stephens NS, Siffledeen J, Su X, et al: Urinary NMR metabolomic profiles discriminate inflammatory bowel disease from healthy. J Crohns Colitis 20^3; 7:e42-e48 6. Gupta A, Dwivedi M, Mahdi AA, et al: Broad identification of bacterial type in urinary tract infection using (1)h NMR spectroscopy. J Proteome Res 201 2; 11:1844-1854 7. Gupta A, Dwivedi M, Nagana Gowda GA, et al: (1)H NMR spectroscopy in the diagnosis of Pseudomonas aeruginosa-'mduceá urinary tract infection. NMR Biomed 2005; 18:293-299 8. Gupta A, Dwivedi M, Gowda GA, et al: 1 H NMR spectroscopy in the diagnosis of Klebsiella pneumoniae-'induced urinary tract infection. NMR Biomed 2006; 19:1055-1061 9. Lin ZY, Xu PB, Yan SK, et al: A metabonomic approach to early prognostic evaluation of experimental sepsis by {1)H NMR and pattern recognition. NMR Biomed 2009; 22:601 -608 www.ccrTnjournal.org
1301
Copyright of Critical Care Medicine is the property of Lippincott Williams & Wilkins and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
