Huimin Guoab,
Jiaqing Chenab,
Yin Huangab,
Wei Zhangc,
Fengguo Xu*ab and
Zunjian Zhang*ab
aKey Laboratory of Drug Quality Control and Pharmacovigilance, China Pharmaceutical University, Ministry of Education, Nanjing 210009, China. E-mail: zunjianzhangcpu@hotmail.com; fengguoxu@gmail.com; Tel: +86-25-83271454 Tel: +86-25-83271021
bState Key Laboratory of Natural Medicine, China Pharmaceutical University, Nanjing 210009, China
cState Key Laboratory of Quality Research in Chinese Medicines, Macau University of Science and Technology, Taipa, Macau, China
First published on 1st June 2016
Time-series metabolomics studies can provide snapshots of the metabolic status during the progress of disease and drug treatment. In order to reveal the dynamic development of a disease, a pseudo-kinetics approach based on calibration and integration of absolute fold changes (FCk) along the time points was proposed. A distance based discriminative trajectory was applied to assess global metabolic alterations during ileus progress upon principle component analysis (PCA). The area under the curve (AUC) and peak time (Tmax) of the absolute FCk–time curve were integrated as complementary thresholds to screen sensitive biomarkers compared with the typical ileus parameters of gastric emptying (GE) and gastrointestinal transit (GI). As a result, docosahexaenoic acid, glycocholic acid, glutamine and acetyl-carnitine were identified as sensitive biomarkers for early prediction of paralytic ileus, which were related to intestinal inflammation and mucosal injury. Furthermore, the discrimination abilities of these four biomarkers were validated using receiver-operating characteristic (ROC) analysis. Collectively, this study demonstrated that our proposed pseudo-kinetics approach is effective to screen more reliable and sensitive biomarkers along a pathologic progress.
The most commonly used method for time series is PCA based trajectory analysis to highlight the metabolic shifts.3–6 Although this plot can depict the metabolic variation trends that consisted of different groups during disease progress, it is difficult to screen biomarkers based on the established models. Batch modeling is another widely used method for dynamic study of biological processes.7 It is possible to establish a control chart to monitor the deviations between the normalities and models. But a drawback with batch modeling is that all study objects must have a similar metabolic and response rate.8 Zhang et al. proposed a weighted method based on the means and variations along the time points in order to identify discriminative metabolites reflecting dynamic development of disease.9 In spite of this, cumulative variation degrees of metabolites along disease progress were ignored, which is convictive to express organic disturbance for pathophysiology. Thus, it is critical to establish a kinetics processing method to deal with time series data where the integral effects of metabolites during disease progress must be considered.
Vincristine (VCR), an alkaloid derived from the Catharanthusroseus (L.) G. Don, has been widely used as a chemotherapeutic agent for the management of hematological malignancies and solid tumors.10,11 However, its clinical setting has been limited due to unavoidable neurotoxicity induced by the property of high binding affinity towards neuronal cytoskeleton protein and disruption of microtubule polymerizations.12 Paralytic ileus and constipation are the most serious side effects of VCR on the gastrointestinal tract.13 However, their mechanism remains obscure because of the complicated interactions of multiple factors.
In this study, a pseudo-kinetics approach combining with multivariate projection methods based on calibration and integration of absolute fold changes (FCK) along disease progress was proposed. Area under curve (AUC) and peak time (Tmax) of absolute FCk–time curve were calculated and compared between metabolites and ileus typical parameters with the aim of revealing more reliable and sensitive diagnosis biomarkers and metabolic dysregulations contributing to gastrointestinal toxicity induced by chemotherapy.
To verify the histological progress of paralytic ileus, five rats in each group of model and control were subjected to phenol red gelatin for investigation the influence of gastric emptying (GE) and gastrointestinal transit (GI) in time points of 0, 4, 8 and 11 day, respectively. Furthermore, intestinal gross and histological examinations were also recorded as visible monitors of ileus. The sera were sequentially collected after 8 h of overnight fasting. This collection was conducted in day 0, 4, 8 and 11. The sera were stored at −80 °C until analysis.
GI = A/B |
GE = 1 − (amount of phenol red recorded after 20 min/amount of phenol red recorded after 0 min) |
Orthogonal projection to latent structures discriminant analysis (OPLS-DA) was carried out after Pareto scaling for equal metabolite weighting in the SIMCA-P 13.0 software package (Umetrics. Umea, Sweden). The variables with a variable importance in the projection (VIP) > 1 were subjected to nonparametric Wilcoxon–Mann–Whitney test using SPSS software in order to find the changed metabolites with statistical significance (p < 0.05) between the models and controls in each time points. Then, absolute fold change analysis (FC) was conducted to screen features with absolute FC > 1.2.
Identification of the metabolites based on GC-MS data was firstly performed by searching the NIST database installed in the Shimadzu QP2010 Ultra GC-MS system. Metabolites with a similarity of more than 80% were finally verified by available reference compounds. Metabolites in LC-MS was identified by four steps: (1) quasi-molecular ion was judged by molecule ions, adduct ions and (de)-protonate molecule ions; (2) Formula Predictor (Shimadzu, Japan) was used for metabolites prediction; (3) MS/MS experiments were conducted to get the structure information; (4) database search.17 Discriminative ions were identified based on accurate molecular weights, molecular formula and MS/MS fragmentation proposed by the online Human Metabolome database (http://www.hmdb.ca/), Lipidmaps database (http://www.lipidmaps.org/) and Massbank (http://www.massbank.jp/). Finally, commercial standards were adopted to support the metabolites identification.
Heat-map was constructed using MultiExperiment View V4.6.1 (http://www.tm4.org). Statistical correlations were determined by Pearson's correlation analysis. For pseudo-kinetics analysis, time dependent first principle component based on PCA analysis and absolute fold change curves for individual metabolites were established. Tmax was obtained from the actual measured values. AUC was calculated by the GraphPad Prism 5. Metabolite relationships were derived from KEGG (http://www.kegg.jp/), SMPDB (http://www.smpdb.ca/) and existing literatures.
Levels of GI and GE are typical parameters for ileus diagnose. Compared to controls of day 4 and 8, GI and GE values in models were down-regulated, respectively (Fig. S-2A and B,† p < 0.05, calculated by Student's t-test) and their decrease peaked at day 8, while recovered at 11 day.
Body weight profiles of controls and models during the modeling are shown in Fig. S-2D.† A significant decrease was observed in the models, while an increase in controls after day 4. This result was consistent with the pathological study and GI/GE measurements. It indicates that the artificial modeling of paralytic ileus induced by VCR was successful on day 8 and that pathological condition was recovered after injection of 11 day.
Tmax, curve peak time, is a widely used kinetic parameter to demonstrate the rate of variation. In Fig. 2B, Tmax of DIPC and DPC-1 is 8. It indicates the maximum variation derived from ileus modeling along PC-1 and integrated PCs presented at day 8, which is corresponding to the gross pathology. While Tmax of DPC-2 is 4, it represents metabolic shifts derived from ileus in the time points of day 4 also play important roles in clarifying the development of ileus. It is suspect that more susceptible metabolites to modeling presented peak variations at day 4 prior to the ileus of day 8. Thus, Tmax of DIPC DPC-1 and DPC-2 provides an accurate way to present the time point of maximum metabolic variations during the disease or toxicity globally.
To describe the dynamic variation trends, the average fold changes (FCKs) of the 39 metabolites and 2 ileus parameters were calculated (Table 1; FCk = If [Mi > , Mi/
, −
/Mi]; Mi represents models;
is the average of controls; i represents individuals in model group) and time dependent absolute FCK curves of 39 metabolites and 2 ileus parameters were fitted. Area under the curve (AUC), a popular kinetic parameter, is integrated to quantify the degree of variation. In this study, AUCs of FCK (GE) represented the integration of discrimination derived from modeling characterized by GE values. Hence, they were efficient thresholds to screen metabolites with higher AUC than that of GI and GE values as sensitive biomarkers. Tmax can represent peak time of metabolic variation quantitatively. Tmaxs of GI and GE were 8. This indicated the intestinal injury peaked in day 8. Thus, metabolites with lower Tmax were praecox biomarkers for predicting the disease.
Model | Metabolites | Average AUC | Tmax (day) | Average FCk | |||
---|---|---|---|---|---|---|---|
0 d | 4 d | 8 d | 11 d | ||||
LC-MS | Docosahexaenoic acid (DHA) | 24.89* | 4 | 1.21 | −3.04 | −2.25 | −1.62 |
LC-MS | Glycocholic acid (GCA) | 46.05* | 4 | 1.07 | 5.94 | 4.78 | 2.28 |
LC-MS | L-Acetylcarnitine | 23.81* | 4 | 1.11 | 2.75 | 2.34 | −1.60 |
GC-MS | L-Glutamine | 23.67* | 4 | 1.02 | 3.27 | 1.89 | −1.29 |
GC-MS | L-Serine | 13.93 | 4 | 1.21 | 1.45 | 1.01 | −1.45 |
LC-MS | LysoPC(18:2) | 11.82 | 4 | 1.02 | 1.15 | 1.02 | −1.07 |
LC-MS | LysoPC(O-16:0/0:0) | 14.25 | 4 | 1.03 | 1.41 | 1.31 | −1.31 |
LC-MS | PG(20:1(11Z)/0:0) | 13.49 | 4 | 1.04 | 1.41 | 1.16 | −1.14 |
GC-MS | Pyroglutamic acid | 13.22 | 4 | 1.02 | 1.34 | 1.14 | −1.22 |
GC-MS | Uric acid | 12.95 | 4 | 1.01 | −1.26 | −1.19 | −1.15 |
GC-MS | Citric acid | 20.09* | 8 | 1.04 | −2.00 | −2.34 | −1.21 |
GC-MS | Linoleic acid | 15.49* | 8 | 1.05 | −1.47 | −1.66 | −1.13 |
GC-MS | L-Tryptophan | 13.56 | 8 | 1.05 | −1.21 | −1.36 | −1.24 |
LC-MS | LysoPC(17:0) | 14.51 | 8 | 1.02 | −1.26 | −1.60 | −1.22 |
LC-MS | LysoPE(18:0) | 13.64 | 8 | 1.03 | 1.15 | 1.44 | 1.29 |
LC-MS | MG(18:1(9Z)/0:0/0:0) | 13.92 | 8 | 1.04 | −1.31 | −1.34 | −1.27 |
GC-MS | Oleic acid | 15.67* | 8 | 1.05 | −1.51 | −1.71 | −1.03 |
GC-MS | Palmitelaidic acid | 24.56* | 8 | 1.14 | −1.91 | −3.33 | −1.99 |
GC-MS | Palmitic acid | 14.85 | 8 | 1.11 | −1.24 | −1.65 | −1.26 |
LC-MS | PE(P-16:0/0:0) | 14.25 | 8 | 1.03 | 1.25 | 1.53 | 1.22 |
LC-MS | Propenoylcarnitine | 27.30* | 8 | 1.03 | 1.73 | 3.87 | 3.18 |
LC-MS | SAM | 18.83* | 8 | 1.15 | −1.62 | −2.06 | −1.89 |
LC-MS | 3-Hydroxy-octadecenoylcarnitine | 40.46* | 11 | 1.04 | 3.53 | 4.09 | 6.63 |
GC-MS | 4-Hydroxy-L-proline | 14.58 | 11 | 1.01 | −1.30 | −1.43 | −1.57 |
GC-MS | Butyric acid | 17.13* | 11 | 1.02 | 1.17 | 1.92 | 2.46 |
GC-MS | D-Glutamic acid | 12.12 | 11 | 1.17 | 1.04 | 1.06 | −1.27 |
LC-MS | HETE | 14.98* | 11 | 1.24 | 1.36 | 1.34 | 1.46 |
GC-MS | L-Aspartic acid | 15.05* | 11 | 1.08 | −1.15 | −1.58 | −1.84 |
GC-MS | L-Lysine | 12.64 | 11 | 1.01 | 1.19 | 1.12 | −1.29 |
GC-MS | L-Threonine | 12.33 | 11 | 1.06 | 1.03 | −1.14 | −1.40 |
GC-MS | L-Tyrosine | 13.23 | 11 | 1.09 | −1.17 | −1.25 | −1.33 |
LC-MS | LysoPC(14:0) | 11.64 | 11 | 1.01 | −1.05 | −1.05 | 1.16 |
LC-MS | LysoPC(17:1) | 13.97 | 11 | 1.09 | 1.29 | 1.20 | −1.62 |
LC-MS | LysoPC(19:0) | 17.44* | 11 | 1.02 | −1.51 | −1.65 | −2.39 |
LC-MS | LysoPC(22:5) | 15.61* | 11 | 1.19 | 1.03 | −1.63 | −2.27 |
LC-MS | LysoPC(O-15:0/0:0) | 13.02 | 11 | 1.05 | −1.17 | −1.20 | 1.36 |
LC-MS | LysoPE(16:0) | 11.94 | 11 | 1.02 | −1.03 | 1.14 | 1.19 |
LC-MS | LysoPE(16:1) | 11.92 | 11 | 1.04 | 1.00 | 1.14 | 1.23 |
LC-MS | Taurocholic acid (TCA) | 23.40* | 11 | 1.03 | 1.69 | 2.09 | −4.84 |
GI | 14.86 | 8 | 1.02 | −1.31 | −1.68 | 1.13 | |
GE | 21.65 | 8 | 1.00 | −2.04 | −2.85 | 1.01 |
Compared with AUC and Tmax of GI and GE values, four sensitive biomarkers with higher AUC (p < 0.05, calculated by Student's t-test) and lower Tmax (p < 0.05, calculated by non-parameter test) were screened including docosahexaenoic acid, glycocholic acid, glutamine and acetyl-carnitine (Fig. 4A–D). In addition, the asterisk labeled metabolites listed in Table 1, such as palmitelaidic acid and propenoyl carnitine, represented higher AUCs than that of GI and GE values (p < 0.05), which contributed to the ileus mechanism.
Furthermore, receiver-operating characteristic (ROC) analysis was further conducted with their AUC calculations (Table S-5†) to investigate the discriminative abilities of the four biomarkers. With all samples, AUC of ROC curves were more than 0.8 and the sensitivities were higher than 0.7. They could be suggested that four biomarkers have abilities to diagnose ileus. Collectively, docosahexaenoic acid, glycocholic acid, glutamine and acetyl-carnitine screened by pseudo-kinetics approach were sensitive and reliable biomarkers during the process of ileus.
The elevation of acetyl-carnitine was a prominent feature before ileus peak. Carnitine plays a vital role in cellular energy and transportation of fatty acids. It can be increased by oxidative stress and heavy weight loss.25 In addition, the amount of docosahexaenoic acid (DHA) was decreased in models prior to ileus. DHA, an important ω-3 fatty acid, can inhibit arachidonic acid metabolism to inflammatory eicosanoids and also can give rise to mediators that are less inflammatory. Down-regulations of DHA might induce inflammation.26 Therefore, alterations of carnitines and fatty acids before ileus suggest that oxidative stress and inflammation are vital factors for development of ileus. Notably, VCR can promote the release of inflammatory cytokines by modulating the cellular Ca2+ levels, free radical generation, TNF-α expression and myeloperoxidase activation, which further emphasize the inflammatory inducement in the process of ileus.27,28
Inflammation can promote the inflammatory destruction of the intestinal brush border. As a result, intestinal mucosa barrier protection is compromised and absorption function is disturbed.29 The accumulation of glutamine before ileus may attribute to the malabsorption for intestinal mucosa injury. Glutamine is the main respiratory substrate for enterocyte and plays a central role in colonocyte nutrition and integrity.30,31 In addition, glycocholic acid (GCA) and taurocholic acid (TCA) elevated significantly in the previous of ileus. They are products of cholesterol and involved in lipid intestinal absorption, energy homeostasis and inflammations. An accumulation of bile acids exacerbates experimental colitis.32 The increased conversion from cholesterol to GCA and TCA could be due to the damage of inflammation and resulted in an abnormal accumulation of toxic bile acids and thereby facilitated the pathological development of ileus. Therefore, inflammation may be a key factor for the formation of paralytic ileus.
Similarly, alterations of carnitines (propenoyl-carnitine), free fatty acids (linoleic, oleic, palmitelaidic and palmitic acids) and lipids (LysoPCs, LysoPEs and MGs) presented in the peak of ileus. Decreased lipid metabolism can increase the activations of pro-inflammatory transcription factor NF-κB and decrease the anti-inflammatory transcription factor PPAR-γ.26 In addition, amounts of L-tryptophan and L-proline were decreased in the peak of ileus. Tryptophan catabolism is a main pathway in acute intestinal inflammation. Specially, indoleamine-2,3-dioxygenase (IDO), the first enzyme in tryptophan catabolism, is up-regulated by pro-inflammatory stimuli, such as the cytokines IFN-γ and TNF-α.33 Moreover, tryptophan is known to protect the gastric mucosa from NSAID induced intestinal damage.34 Proline is an important amino acid in collagen. The decrease of proline is indicative of decreased collagen levels in the gastrointestinal tissue.35 Accordingly, inflammation and inflammatory mucosa injury exacerbates the intestinal lesion and paralytic ileus.
Furthermore, short chain fatty acids (SCFA), like butyrate, increased in models significantly after ileus. The SCFA are normally produced by the gut bacteria via the fermentation of complex carbohydrates such as fiber and starches. Butyrate provides energy to the intestinal cell wall and promotes epithelial cell growth.36 The increase of the microbiota-related metabolite suggests that gut bacterial ecology plays an important role during the recovery of ileus. In addition, tricarboxylic acid cycle (TAC) intermediates, such as malate, succinate, citrate and 2-oxoglutarate, were deceased at the time points of day 8 and 11. This observation suggests hyperactive energy metabolism may contribute to recovery of gastrointestinal motility.
Collectively, it is reasonable to speculate that VCR-treatment induced intestinal inflammation and mucosa injury are praecox and substantial molecular mechanism of paralytic ileus. Subsequently, abnormal energy metabolism and gut microbiota may reverse the ileus development. These findings will contribute to find the characteristic metabolic changes associated with the paralytic ileus. However, further study is needed to clarify the mechanism of change in these biomarkers, limitation of indications and extrapolation for application in humans.
Footnote |
† Electronic supplementary information (ESI) available: Gastrointestinal transit (GI) and gastric emptying (GE) measurement. Table S-1. The stability and repeatability of the proposed method. Table S-2. Parameters of OPLS-DA models based on the different time points in GC-MS and LC-MS. Table S-3. Metabolites identified by GC-MS. Table S-4. Metabolites identified by LC-MS. Table S-5. Sensitivity and specificity of ROC curves based on four biomarkers. Fig. S-1. Scheme and histology results of the animal experiment in the present study. Fig. S-2. Gastric emptying, gastrointestinal transit, urea nitrogen/creatinine ratio and body weight profiles during the model periods. Fig. S-3. Pearson's correlations of metabolites and ileus parameters. Fig. S-4. Altered metabolic pathways during ileus. See DOI: 10.1039/c6ra12641c |
This journal is © The Royal Society of Chemistry 2016 |