¹H-NMR-based metabolic profiling of healthy individuals and high-resolution CT-classified phenotypes of COPD with treatment of tiotropium bromide

Abstract

BackgroundHeterogeneity of COPD results in different therapeutic effects for different patients receiving the same treatment. COPD patients need to be individually treated according to their own characteristics. The purpose of this study was to explore the differences in different CT phenotypic COPD by molecular metabolites through the use of metabolomics.MethodsAccording to the characteristics of CT imaging, 42 COPD patients were grouped into phenotype E (n=20) or phenotype M (n=24). Each COPD patient received tiotropium bromide powder for inhalation for a therapeutic period of 3 months. All subjects were assigned into phenotype E in pre-therapy (EB, n=20), phenotype E in post-therapy (EA, n=20), phenotype M in pre-therapy (MB, n=22), phenotype M in post-therapy (MA, n=22), or normal control (N, n=24). The method of metabolomics based on 1H nuclear magnetic resonance (1H-NMR) was used to compare the changes in serum metabolites between COPD patients and normal controls and between different phenotypes of COPD patients in pre- and post-therapy.ResultsPatients with COPD phenotype E responded better to tiotropium bromide than patients with COPD phenotype M in terms of pulmonary function and COPD assessment test scores. There were differences in metabolites in COPD patients vs normal control people. Differences were also observed between different COPD phenotypic patients receiving the treatment in comparison with those who did not receive treatment. The changes of metabolites involved lactate, phenylalanine, fructose, glycine, asparagine, citric acid, pyruvic acid, proline, acetone, ornithine, lipid, pyridoxine, maltose, betaine, lipoprotein, and so on. These identified metabolites covered the metabolic pathways of amino acids, carbohydrates, lipids, genetic materials, and vitamin.ConclusionThe efficacy of tiotropium bromide on COPD phenotype E is better than that of phenotype M. Metabolites detected by 1H-NMR metabolomics have potentialities of differentiation of COPD and healthy people, discrimination of different COPD phenotypes, and giving insight into the individualized treatment of COPD.

An Updated Overview of Metabolomic Profile Changes in Chronic Obstructive Pulmonary Disease

“…COPD was divided into phenotype E and phenotype M by high-resolution computed tomography scanning. Different phenotypes displayed different metabolomic features toward drug treatment, such as the difference of amino acids and carbohydrate metabolism [18]. Moreover, COPD patients with phenotype M had a more significant response to bronchodilator and corticosteroid treatment [2], while phenotype E was more sensitive to anticholinergic treatment [18].…”

“…Different phenotypes displayed different metabolomic features toward drug treatment, such as the difference of amino acids and carbohydrate metabolism [18]. Moreover, COPD patients with phenotype M had a more significant response to bronchodilator and corticosteroid treatment [2], while phenotype E was more sensitive to anticholinergic treatment [18]. Therefore, metabolomics might have the potential to distinguish different COPD phenotypes and give insights into the therapeutic evaluation of COPD.…”

Pharmacometabolomics by NMR in Oncology: A Systematic Review

“…Even though very large bucket widths are not recommended as they decrease the resolution of the NMR spectra, an extreme reduction in the bucket width could significantly contribute to data overfitting as a result of the imbalance between the number of samples and the variables included in the analysis [ 66 ]. In general, most NMR-based metabolomics studies used a bucket width between 0.01 and 0.04 ppm, depending on the spectra complexity and the signal overlapping, for binning [ 29 , 43 , 67 , 68 , 69 , 70 ]. After binning, different normalization approaches were followed in most of the PMx studies.…”

“…Although normalization details were not specified in all the studies, the normalization strategy was mainly dependent on the sample type. Overall, serum and plasma NMR data were preferentially normalized to total area [ 40 , 41 , 46 , 47 , 49 , 51 , 52 , 60 , 67 , 68 , 69 , 71 , 72 , 73 , 74 , 75 , 76 ], although probabilistic quotient normalization [ 26 , 48 , 77 ] and other normalization procedures, such as glucose [ 78 ] or internal standard normalization [ 39 ], were applied in some studies. For urine normalization, total area [ 42 , 67 ] and creatinine signal were the methods most frequently used [ 44 , 79 , 80 ].…”

“…Principal Component Analysis (PCA) [ 96 ] was the method of choice in most of the studies, although hierarchical clustering [ 97 ] was also used in other studies [ 26 ]. Then, supervised analysis methods, such as partial least square discriminant analysis (PLS-DA) [ 25 , 41 , 42 , 46 , 48 , 49 , 60 , 67 , 68 , 71 , 73 , 74 , 75 , 76 , 78 , 79 , 98 , 99 ] or orthogonal PLS-DA (OPLS-DA) [ 24 , 26 , 29 , 40 , 43 , 50 , 68 , 70 , 72 , 74 , 75 , 76 , 77 , 98 , 100 , 101 , 102 ] were pursued in the majority of the studies to evaluate the discriminatory potential of the metabolic profile between the groups of study. Herein, alternative supervised approaches were also followed, including multilevel Partial Least Square (mPLS) [ 27 ], Random Forest (RF) [ 45 , 47 , 53 ], K-nearest neighbors (kNN) [ …”