A Map Of Metabolic Phenotypes in Patients With Myalgic Encephalomyelitis/ Chronic Fatigue Syndrome Ⅱ

The lipidomes of ME/CFS 

A comprehensive analysis of complex lipids was performed in serum from the same study cohort. Of the included 892 lipid species, 153 presented significantly different serum concentrations in the ME/CFS  group relative to the HC group, of which 34 had lower levels and 119 had higher levels (Figure 6A; Supplemental Data Set 2, sheets 2 and 4). The lipid class sum concentrations indicated an elevation of diacylglycerols  (DAG), TAG, and dihydroceramide (DCER) and reduction of lysophosphatidylcholines (LPC) in the  ME/CFS group compared with the HC group (Supplemental Data Set 2, sheet 3). However, distinct differences in lipid profile were evident when subgrouping according to the proposed ME/CFS metabotypes.  Overall, there was extensive coherence between lipid-related effects observed on the global metabolomics  (HD4) and lipidomics (complex lipidomics platform [CLP]) platforms. The following is a summary of the uniform and metabotype-specific changes in serum lipids in patients with ME/CFS relative to HC subjects. Uniform lipidome effects in the ME/CFS group. Only 16 of the 892 lipid species presented uniform effects in the ME-M1 and ME-M2 subsets relative to the HC group (Supplemental Data Set 2, sheet 5). Of these,  13 compounds had lower serum concentrations in ME/CFS patients, and all of these were phospholipid derivatives of phosphatidylcholine (PC) and phosphatidylethanolamine (PE), including LPC and lysophosphatidylethanolamine (LPE). This included several PE plasmalogen compounds, which represent a chemically and biologically unique phospholipid subclass (49). Ten of the 13 uniformly lowered phospholipid species contained at least one 18:2 acyl moiety, which represents linoleic acid, an essential fatty acid and arachidonic acid (20:4) precursor. Overall, the level of linoleic acid tended to be low in the phospholipid subclasses  (PC, PE, phosphatidylinositol [PI], LPC, and LPE; Figure 6B). Only 3 single lipid species were uniformly elevated — the sphingolipids, CER(18:1), SM(18:0), and SM(18:1) — yet additional sphingolipids showed similar trends. Regarding effects on total serum fatty acid content (i.e., esterified and nonesterified), the only significant uniform change in the ME-M1 and ME-M2 subsets was a reduced level of 14:0 (Figure 6B). Metabotype ME-M1 lipidome. This subset had a significantly elevated serum concentration of 23 of 892  and lowered level of 227 of 892 lipid metabolites compared with the HC group (Figure 6; Supplemental  Data Set 2, sheet 6). Most of the elevated metabolites were NEFAs (18 of the 25 measured free fatty acids;  Figure 6A), contributing to the higher total NEFA concentration. The remaining 5 elevated compounds in  the ME-M1 subset were all sphingolipids containing 18:0 or 18:1 (CER[18:0], CER[18:1], DCER[18:0],  SM[18:0], SM[18:1]). The 227 lowered metabolites predominantly included species of glycerolipids (138  TAG, 5 DAG, 2 monoacylglycerols [MAG]) and phospholipids (79, including 16 lysophospholipids),  in addition to SM(14:1), SM(22;1), and CE(14:1). Although the sum concentration of TAG was not significantly lowered relative to the HC group, the lowered levels of multiple single TAG are coherent with the 

Figure 6. Overview of lipid classes. The serum lipidome data set was analyzed to overall compare the ME/CFS group (ME/All) and the 3 metabotype subsets  (M1–M3) relative to the HC group. (A) The volcano plots give a general overview of all lipid molecules in a given class, with the –log10 of the P-value on the y axis and log2 fold change on the x axis. Each dot represents a metabolite and is colored according to the direction of change and significance level relative to HC as indicated (P < 0.05, 2-tailed Welch’s test; q < 0.05, adjusted P value). The number in each quadrant provides the respective counts of q-significant metabolites, and additional p-significant metabolites are shown in parentheses. (B) The relative amount of specific fatty acids (FA, first column) in the different lipid classes (column title) in the total ME/CFS group (bottom label, A) and according to metabotype subsets (M1–M3). The color of the heatmap cells display the  log2 fold change relative to HC, as indicated. *P < 0.05, 2-tailed Welch’s test; **q < 0.05, adjusted P-value. Total, the total sum of fatty acids across all lipid classes;  CE, cholesterol ester; CER, ceramide; DAG, diacylglycerol; DCER, dihydroceramide; FFA, free fatty acid (synonymous to nonesterified fatty acid, NEFA); HCER,  hexosylceramide; LCER, lactosylceramide; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; LPL, lysophospholipid; MAG, monoacylglycerol;  PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PL, phospholipids, SM, sphingomyelin; SP, sphingolipids; TAG, triacylglycerol.

relatively low total serum TAG concentration was observed in laboratory measurements (Figure 3D). Furthermore, there was a significant or trending decrease in the sum concentrations of phospholipids (PC, PI, trend  for PE) and lysophospholipids (LPC, LPE) (Supplemental Data Set 2, sheet 3). There was no significant effect on total levels of sphingolipid subclasses, but there was a trend for lower total DCER level. In the  NEFA fraction, there was a particular relative enrichment (mole percent) of the 16:1, 18:1, and 18:2 fatty  acids, whereas the TAG fraction had significantly lower content of C12 and C14 fatty acids (Supplemental Data Set 2, sheet 10).

The findings are compatible with increased mobilization and oxidation of fatty acids, as suggested by increased concentrations of free fatty acids, glycerol, and ketone bodies. The observed changes in the ME-M1 subset may mimic metabolic adaptations known to occur in response to energy strain triggered by starvation or exercise (50, 51). Theoretically, such a response could be caused by aberrant regulation of glucose catabolism, a chronic high ATP demand, or uncoupling of mitochondrial respiration leading to excessive oxidative flux and inefficient ATP production. Metabotype ME-M2 lipidome. This subset had higher concentrations of 538 of 892 lipid compounds, and lower concentrations of 27 of 892, compared with the HC group (Figure 6; Supplemental Data Set 2, sheet  7). The vast majority of the elevated compounds were different species of either TAG (454 species) or DAG  (43 species; Figure 6); accordingly, there were significantly higher total serum concentrations of these 2 lipid classes. There were also higher levels of some single species — and the total class concentration — of cholesterol esters. Multiple single phospholipids (25 species of PC, PE, and PI) and sphingolipids (8 species of SM,  CER, DCER, and HCER) were elevated, but the total concentrations of these compound classes were mainly unaffected, apart from the PI class presenting significant increase compared with the HC group (Supplemental  Data Set 2, sheet 3). 

Similar to the ME-M1 subset, there was no significant effect on total levels of sphingolipid subclasses. Several single NEFAs (10 species) had lower serum concentrations in the ME-M2 subset compared with the HC group, contributing to the lower total NEFA level. Regarding phospholipids presenting lower levels, the effects were essentially coherent with the changes already described as uniform in the ME/ CFS group. The elevated serum TAG concentration and lowered NEFA concentration in the ME-M2 subset compared with the HC group were confirmed in supplementary laboratory analyses (Table 2 and Figure 3D).  In agreement with the higher TAG and DAG levels, there was a general increment in fatty acids esterified in these glycerolipids, across the fatty acid spectrum (Figure 6B). In summary, the metabolic phenotype of this subset was particularly characterized by a high serum concentration of TAG (women) and low NEFA (both  sexes). Increased trafficking of TAG and DAG in blood is often associated with metabolic imbalance, excessive peripheral lipid accumulation, and induction of cellular and mitochondrial stress responses. Such effects have been described as consequences of various chronic diseases, also involving contexts of inflammation (34). Metabotype ME-M3 lipidome. This small subset had higher concentrations of 22 of 892 lipids, and lower concentrations of 5 of 892 lipids, compared with the HC group (Figure 6A; Supplemental Data Set  2, sheet 8). Elevation was primarily seen for certain cholesterol esters (7 species) and phospholipids (12  species). Notably, the serum level of 6 PE plasmalogens was higher in this subset, which contrasted the effects on this phospholipid subclass in the 2 other ME/CFS subsets. The few lipids presenting lower serum  concentration in the ME-M3 subset compared with the HC group included 2 NEFA species (FFA[14:0],  FFA[20:4]) and 3 PC derivatives (LPC[20:2], PC[16:0/22:4], PC[18:1/20:2]). This subset displayed higher total SM level, and trends of higher total cholesterol ester and PE levels, compared with the HC group.  Relatively few effects were found regarding fatty acid composition of the different lipid classes (Figure 6B).  It appeared that several of the affected single traits were unique to this subset compared with the others,  but the possible implications are difficult to evaluate due to the low number of subjects. Viewing the overall tendencies, the ME-M3 subset appeared to be more similar to the ME-M2 subset than the ME-M1 subset,  as indicated by a trending elevation of TAG. Hormone signatures of metabolic stress We investigated the circulating levels of selected hormones controlling energy homeostasis in contexts of physiological strain, inflammation, and pathogenesis (Figure 7). Compared with the HC group, the ME/ CFS group had slightly elevated mean insulin and leptin serum levels, as well as lower high molecular weight  (HMW) adiponectin (Figure 7, A–C). However, this was primarily driven by the ME-M2 subset, as no significant or trending effect was seen for the others. Accordant with increased insulin, the ME-M2 subset also had an elevated mean concentration of C-peptide (Figure 7D). These specific effects on hormone levels support different regulatory contexts in the ME-M1 and ME-M2 subsets, and the findings are coherent with the observed metabolic phenotypes indicating reduced glucose utilization (ME-M1) and excessive lipid accumulation (MEM2). Furthermore, there were increased serum concentrations of FABP4 and FGF21 in the ME/CFS group compared with the HC group (Figure 7, E, and F). These factors are regarded as signals of energy strain and have been linked to metabolic disease as well as exercise (52, 53). Only the ME-M2 subset presented a significant elevation of FGF21. There was a small group of patients with a particularly high level of FGF21 in the  ME-M2 subset, and we suspect this may indicate an excessive burden of hepatic metabolic stress. The elevation in serum FABP4 concentration was similar in the ME-M1 and ME-M2 subsets. The observed endocrine signatures are consistent with deregulated metabolism and elevated energy strain in patients with ME/CFS,  and with links to regulatory networks that may explain context-dependent responses.

Figure 7. Changes in serum signaling factors are linked to energy metabolism. 

Selected hormones potentially associated with energy strain and deregulated metabolism were measured in serum from patients with ME/CFS and HC subjects using immune-based methods. (A–F) The serum concentration was  measured in 83 patients with ME/CFS and 30 HC subjects for insulin (pg/mL) (A), leptin (ng/mL) (B), HMW adiponectin (μg/mL) (C), C-peptide (pM) (D),  FGF-21 (pg/mL) (E), and FABP4 (ng/mL) (F). The group median ± IQR is indicated. The shown P values (upper right) are for a 4-group comparison of HC, M1,  M2, and M3 using 1-way Kruskal-Wallis ANOVA. *P <0.05, Mann-Whitney U test; **q < 0.05, adjusted P-value

Discussion

This exploratory metabolomics study revealed a map of common and variable metabolic phenotypes of ME/ CFS. The observed metabolic changes mainly fit into the paradigm of direct and indirect effects of energy strain. The physiological relevance was supported by associations with endocrine and clinical characteristics.  Through the following discussion, we suggest that energy strain may result from exertion-sensitive tissue hypoxia and leads to the systemic patterns of metabolic adaptation and compensation. The common metabolic changes in the ME/CFS group were dominated by a relatively small number of pathways with a credible impact on energy homeostasis. Elevated circulating glycerol suggests that lipolysis is induced, and this is normally observed during fasting and exercise (35, 48). In addition, several of the findings agree with the altered utilization of amino acids in patients with ME/CFS, including BCAAs, tryptophan,  and others. The increase in breakdown products of purine nucleotides such as adenosine and xanthosine also appeared as a possible signature of energy strain, which normally reflects increased ATP demands and altered amino acid metabolism in muscle (36). Systemic metabolic stress was supported by elevated FABP4  and FGF21, which may signal compensatory programs, as well as tissue-specific responses (54, 55). Corticosteroids did not seem to be involved, since they depended on age and BMI instead of physical function scores  (Supplemental Data 3) and we did not find changes in morning cortisol and ACTH in ME/CFS compared with HC. There was a particular loss of some phospholipids containing linoleic acid (18:2). This fatty acid is  an essential precursor for arachidonic acid, a central messenger molecule linked to inflammation and vasodilatation. The lowering of arachidonic acid derivatives such as 12-HETE and 12-HHTrE further indicates that auto- and paracrine processes may be affected. As further discussed below, some of these effects may be associated with mechanisms likely to be involved in symptoms of ME/CFS.

The effects that were heterogeneous in the ME/CFS group expressed specific phenotypes of deregulated energy metabolism in subsets of patients. The 2 most distinct ME/CFS metabotypes, ME-M1 and ME-M2,  aligned with well-known phenotypes of chronic diseases with immuno-metabolic projections (34, 37, 38). It should be kept in mind that none of the patients of our study had a clinical prediabetic or diabetic diagnosis. The  ME-M1 subset presented elevated serum levels of free fatty acids (i.e., NEFA) and ketone bodies, despite normal glucose and insulin. This may resemble a context with physiological correlations to glucose starvation and exercise (33, 35). In the ME-M2 subset, a different metabolic profile was expressed by the elevated serum TAG and insulin mean levels, yet blood glucose was not affected. This may reflect low-grade signs of lipid-induced insulin resistance associated with ectopic peripheral lipid accumulation and inflammatory responses (56). Both the autoregulation of blood flow have been found in persons with unexplained exertion intolerance, possibly via mechanisms of endothelial and microcirculatory dysfunction effects on vascular endothelium (66, 76, 77, 79,  80). In summary, we find that the metabolic changes in patients with ME/CFS are compatible with disrupted energetics enforced by tissue hypoxia on exertion. Further investigations should be performed to pursue this theory and to identify possible support strategies for improved clinical care. Possible limitations of the study included metabolite stability and the limited accuracy of global untargeted metabolomics (81). The limitations of univariate feature selection for the purpose of clustering and identifying patient subtypes in high-dimensional data sets are well-known issues in the statistical community (82, 83)  (Supplemental Data 1). Our comprehensive evaluation of possible confounders such as sex, BMI, age, diet, or medication did not indicate that these were main drivers of the ME/CFS phenotypes, but they may contribute to individual variation (Supplemental Data 1). A strength of the study was that statistical analyses were strongly and independently supported by multiple layers of biochemical findings and rationale. In summary, the study provides a map of metabolic alterations occurring in patients with ME/CFS.  We find that the observed changes are compatible with elevated energy strain, for instance, caused by tissue hypoxia on exertion. The potential roles of specific pathways will have to be validated and explored in further targeted studies.

Methods 

Patients with ME/CFS and HC. In total, 83 patients with ME/CFS and 35 HC were included in this study.  All patients fulfilled the Canadian consensus criteria for ME/CFS (1). The blood samples were collected before intervention (baseline) in 2 separate clinical trials led by Haukeland University Hospital, the “RituxME” trial (ClinicalTrials.gov NCT02229942, 2014–2017; ref. 45) and the “CycloME” trial (ClinicalTrials.gov  NCT02444091, 2015–2020; ref. 12). The HC samples were collected (from 2015 to 2017) from subjects with no chronic disease or chronic medication — primarily from staff at the Department of Oncology, Haukeland  University Hospital, and the Department of Biomedicine, University of Bergen. The HC group was selected to approximately match the age and sex distribution of the ME/CFS group. Biometric characteristics for HC  subjects and patients with ME/CFS are summarized in Table 1, which also contains the data for subgroups of patients with ME/CFS annotated to the 3 different metabotypes developed in this study. Blood samples were collected by venous puncture, processed according to a standardized biobank procedure as described in the trial protocols, and stored at –80°C (12, 45). All HC subjects, and 71 patients with ME/CFS,  were nonfasting on sample collection. The remaining 12 ME/CFS samples were collected after overnight fasting and were included to facilitate the evaluation of the possible impact of the postprandial state. Clinical blood and serum analyses were performed according to standard laboratory routines at the hospital. Global metabolomics and lipidomics. For the metabolite analyses, we included serum samples from all 83  patients and 35 HC subjects. The analyses were performed by Metabolon Inc. using their standard methods. The data were acquired applying 2 mass spectrometry-based analytical platforms, one global metabolite platform (HD4) providing measurement and identification of 882 compounds covering a broad spectrum of molecules and one CLP that assessed 1005 different molecular species of 14 lipid classes. The samples were analyzed in daily blocks, with several levels of sample and data quality control, such as inter-day variation correction before data normalization. Metabolic hormone analyses. For metabolic hormone analyses, we included the same samples, except 5 HC,  where there was an insufficient amount of serum. The serum samples were diluted, and assays were performed according to the manufacturer’s guidelines. Kits based on the Luminex multiplex bead immunoassay technology  (Luminex 100 instrument, Luminex Corp.) were used to detect FABP4, insulin, HMW adiponectin, and leptin  (catalog LXSAMH, R&D Systems). ELISA was used to measure FGF21 (catalog DF2100, R&D Systems) and  C-peptide (catalog DICP00, R&D Systems). Measurements were performed using the Spark microplate reader  (Tecan Trading AG). Data analysis was done in Excel, GraphPad Prism, and R. Statistics. All data analyses were performed within the R studio environment (84), using the R programming language (85). We first removed xenobiotics and metabolites with more than 25% missing values from the data set, and the pattern of missing values was assessed by χ2 and Fisher exact test for the group- and sex-wise comparisons. Missing values were imputed using the half-minimum method (86). The imputed data were quantile normalized, log2 transformed and autoscaled. Fold changes, P values (2-tailed Welch’s test) and adjusted P values  (Benjamini-Hochberg adjustment) were calculated using base functions in R. The autoscaled data were filtered based on unadjusted significant metabolites (P < 0.05) to reduce dimensionality in data. Row- and column-wise k-means were performed on significant features to cluster metabolites and samples, and we selected partitions  of clusters guided by the within-cluster sum of square (wss), gap statistic, and the silhouette method. This was visualized in a heatmap using the ComplexHeatmaps package (87), and the resulting clusters were extracted and  HC organized into a separate HC cluster. These clusters were used as a visualization overlay on an independent  PCA, and a loading plot was used to visualize loadings from PC1 and PC2 using ggplot2 (88).

A univariate analysis was performed on the extracted clusters from the multivariate analysis. Metabolites were organized into pathways according to annotations provided by Metabolon Inc. The fold changes, P values, and adjusted P values from metabolites in each respective pathway were visualized using ggplot2. Selected biometric data were correlated to metabolites that displayed uniform changes between clusters using the Spearman method, and P <  0.05 was set as a cutoff. This was visualized using the ggplot package. Data from the CLP platform were preprocessed by removing variables with more than 10% of missing values in the data set and were imputed by using the minimum observed value for each metabolite. The samples were organized into the respective clusters obtained from earlier analysis of HD4, and fold changes, P values (Welch’s test), and adjusted P values (FDR) were calculated using base functions within R. Associated metadata concerning lipid classes were obtained from Metabolon Inc. and the R package Lipidr (89). 

The resulting data were visualized using the ggplot2. Descriptive statistics were performed using 2-tailed Welch’s test for normally distributed data, Mann-Whitney U tests for variables with skewed distributions, and χ2 and  Fisher’s exact tests for categorical variables (GraphPad Prism). A comparison of 3 or more groups was performed using 1-way ANOVA test. Study approval. The clinical trials from which the biobank samples were obtained, including samples from healthy controls, were approved by the Regional Ethical Committee (Tromsø, Norway; no. 2010/1318-4, no.  2014/365, and no. 2014/1672). All patients provided written informed consent.

FIND THE BEST BOTANICAL HOLISTIC CARE FOR RECOVERING DISEASE-RELATED FATIGUE

Our carefully hand-crafted Neuro Regen formula includes the best-researched herbs. Each shows amazing potential for nerve pain management, plus general nerve and neuron support overall.

Includes:

Cistanche- 

Cistanche is the most commonly used herb in traditional Chinese medicine prescriptions for relieving physical fatigue. In fact, Cistanche is a kind of food material. On the theoretical basis of traditional medicine, modern medicine has also conducted a lot of research on the anti-fatigue effect of Cistanche. Experiments have proved that the anti-fatigue effect of Cistanche is due to the improvement of the body organs of Cistanche. Cistanche can improve the functions of the kidney and liver, thus achieving an anti-fatigue effect. The following is an overview of how Cistanche can relieve fatigue.

4 Characteristics of Cistanche Glycosides

With the development of society, people's life is getting faster and faster. From the moment the sun rises, as a part of the "living machine", everyone has to rush to their posts. Whether you are a technician standing in the workshop, a stock trader sitting in front of a computer, or even a student playing in the playground...you have to face countless pressures. People's jobs are different, but they all have one common trouble-fatigue. Fatigue is like a cockroach hiding in the corner of the kitchen. It has no substantial harm, but it has always plagued humans.

If you have anything don't understand, don't hesitate to tell me: wallence.suen@wecistanche.com