Contributed equally
To identify the major serum biomarkers predicting the response to methotrexate (MTX) treatment in patients with early rheumatoid arthritis (RA), we evaluated the relationships between the individual response to MTX and various associated factors utilizing the 1H nuclear magnetic resonance (1H NMR)-based metabolomic method. Thirty-eight early RA patients were enrolled in this cohort study, and they received MTX (10 mg/week) orally as monotherapy for 24 weeks. According to the American College of Rheumatology criteria for improvement, clinical evaluation following MTX treatment was carried out at baseline and at the end of 24 weeks. Furthermore, collected serum samples were analyzed using 600 M 1H NMR for spectral binning. The obtained data were processed by both the unsupervised principal component analysis (PCA) and the supervised partial least squares discriminant analysis (PLS-DA). Lastly, multivariate analyses were performed to recognize the spectral pattern of endogenous metabolites related to MTX treatment. Differential clustering of 1H NMR spectra identified by PCA was found between the effective (n=25) and non-effective (n=13) group of RA patients receiving MTX treatment. Multivariate statistical analysis showed a difference in metabolic profiles between the two groups using PLS-DA (R2=0.802, Q2=0.643). In targeted profiling, 11 endogenous metabolites of the effective group showed a significant difference when compared with those of the non-effective group (p<0.05). Serum metabolites correlated with MTX treatment in patients with early RA were identified, which may be the major predictive factors for evaluating the response to MTX treatment in patients with early RA. Furthermore, our results highlight the usefulness of 1H NMR-based metabolomics as a feasible and efficient prognostic tool for predicting therapeutic efficacy to MTX treatment.
Rheumatoid arthritis (RA) is a systemic chronic inflammatory joint disease, which is characterized by persistent synovitis, systemic inflammation and autoantibodies (
Recent advances in metabolomics offer us a chance to identify changes in the concentration of small endogenous metabolites in biological samples caused by drug therapy. Metabolomics refers to the total metabolites present in biofluids (e.g., blood and urine) and chemometric techniques combining 1H nuclear magnetic resonance (1H NMR) has been applied for metabolomic profiling and characterization of individually tailored therapeutic approaches (
In this study, we categorized the subsets of early RA patients receiving MTX treatment for 24 weeks into two groups according to the clinic assessment criterion, i.e., effective group and non-effective group. Furthermore, 11 endogenous metabolites of the effective group were significant different than the non-effective group based on 1H NMR metabolomic analysis and multivariate statistical analysis (p<0.05). As a potential strategy and a more convenient technique, the 1H NMR-based metabolomic approach deserves further evaluation for assessing the efficacy of MTX treatment in patients with early RA.
This cohort study was conducted from January 2009 to September 2010 at a single investigational site, the Department of Integrated Chinese Traditional and Western Medicine, Tongji Hospital (Wuhan, China). Patient demographic and clinical details were collected on standardized data collection forms. All patients underwent physical examination, electrocardiography and routine blood testing. The duration of RA, age at which MTX administration was initiated, gender, body weight, laboratory data; clinical data including anti-cyclic citrullinated peptide (anti-CCP), serum C-reactive protein (CRP) concentration, rheumatoid factor (RF), erythrocyte sedimentation rate (ESR), serum creatinine clearance (SCR) and alanine transaminase (ALT) were obtained from the medical records.
Patients 18 years of age and older with early (<0.5 year) active RA, as defined by the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) (
Ethical approval was obtained from the Ethics Committee of Tongji Medical College, China. Written informed consent was obtained from each participant in accordance with the principles outlined in the Declaration of Helsinki.
Treatment efficacy was evaluated in early RA patients receiving MTX monotherapy for 24 weeks, according to the European League Against Rheumatism (EULAR) response criteria (
Venous blood samples were collected in the morning preprandial, after overnight fasting, using vacuum tubes (BD, 8.5 ml, USA) and left at room temperature to coagulate for 45 min. After this, samples were centrifuged at 1000 x g for 10 min, and serum was collected and immediately stored at −80°C until being used for metabolomic analysis (
Preparation of serum samples for metabolomic analysis by 1H NMR was performed manually according to Beckonert
The NMR spectra were acquired using a Bruker Avance III 600 spectrometer (Bruker BioSpin, Rheinstetten, Germany), operating at a 1H frequency of 600.13 MHz, and equipped with a 5-mm 1H TXI probe. Samples were afforded 5 min within the spectrometer to equilibrate before acquisition. Standard one-dimensional (1D) 1H NMR spectra were acquired using a single 90º pulse experiment with water presaturation using a recycle delay of 3 sec. Each data set was averaged over 64 transients using 32 K time domain points. The data were Fourier transformed, and spectra were referenced to the TSP signal at 0 ppm.
Each NMR spectrum was reduced to 0.04 ppm wide segments between 0.00 and 10.0 ppm using Chenomx NMR Suite Professional software version 4.6 (Chenomx Inc., Edmonton, Canada), giving a total of 230 integrated regions per NMR spectrum. The spectrum regions of water (δ=2.6–3.0 ppm) and urea (δ=4.3–4.7 ppm) were removed from the analysis for all groups in order to prevent variation in each sample. Each NMR variable was normalized to the total area in order to allow a spectrum-to-spectrum comparison. Metabolites were assigned based on comparisons with the chemical shifts of standard compounds in the Chenomx software. Custom library entries were created for unidentified resonances in order to carry them through the analysis for relative concentration comparison. Typically, the concentrations of serum metabolites were reported as ratios with creatinine.
The spectral data were converted to Microsoft Excel format and imported into SIMCA-P software (version 11.5, Umetrics, Umeå, Sweden) for multivariate analysis in which all spectral data were mean-centered with no scaling. Initially, principal component analysis (PCA) was applied to the clustering of all identified outliers. Partial least squares-discrimination analysis (PLS-DA) was then advanced for pattern recognition analysis from the converted spectral data (
Clinic data are expressed as the means ± standard deviation (SD), and the data were analyzed for statistical significance by ANOVA test using SPSS version 17.0. Post-hoc tests were applied to perform all pairwise comparisons between group means by least significant difference (LSD) t-test. A value of p<0.05 was considered to indicate a statistically significant difference.
Demographics and characteristics of the patients at baseline and at the end of 24 weeks of MTX treatment are provided in
A 600 MHz 1D 1H NMR spectrum of a typical human serum sample is shown in
The mean-centered data from 1D 1H NMR spectra were input into the SIMCA-P software and processed by the unsupervised method of the principal component analysis (PCA), and the resulting score loading plots for both the effective and the non-effective group are shown (
The spectral data of the serum samples were collected from the effective (n=25) and non-effective group (n=13), respectively, and the metabolic profiles of the serum samples are depicted in
The quality of the principal component models was evaluated with the parameters R2 and Q2, and the goodness of fit was quantified by R2 while the predictive ability was indicated by Q2. Generally, R2Y and Q2Y explained variation in the data, where 0 was no variation explained, and 1, for which 100% variation was accounted for. The results indicated that this PLS-DA model had a high R2Y value of 0.802 and Q2Y value of 0.643, which suggested that although each patient in the two groups was treated with the same concentration of MTX, and it was possible to metabolically discriminate with different physical constitutions (
The permutation testing was performed to validate PLS-DA models. It provided the statistical significance of the estimated predicted power of the models by comparing R2Y and Q2Y values of the original model with those of the re-ordered model, which was created newly whenever y data was permutated at random. It is known that models with the R2Y intercept <0.4 and a Q2Y intercept <0.05 indicate valid models. As shown in
VIP values of several metabolites assigned by using SIMCA-P are listed in
Metabolomics is a novel scientific discipline focused on the association between disease and metabolic profile in tissue and biofluids, as determined by techniques including 1H NMR spectroscopy, mass spectroscopy and liquid chromatography mass spectroscopy. Moreover, metabolomic analysis can also be utilized to identify biomarkers of drug treatment (
Biological fluids, such as blood and urine, contain a large number of metabolites that may provide valuable information on the metabolism of an organism, and thus concerning its health status. In the present study, 11 metabolites were found to vary with the differential response of the early RA patients receiving MTX treatment, which suggest they could be used to predict the therapeutic effects of MTX. It appeared that nucleic acid metabolism may be highly interrelated to the therapeutic effects, with uric acid and uracil being two of the most prominent metabolites noted. In addition, the changes in levels of TMAO and hypoxanthine suggest an alteration in purine and pyrimidine ratios, and in turn may affect cell metabolism, energy conservation and biosynthetic pathways, even signal transduction and translation (
The potential metabolic pathways highlight the complexity of the metabolic response to MTX treatment. Although conclusively defining roles for each metabolite were not feasible based on the data from this study, 11 metabolites presented here may be considered useful biomarkers for their discriminatory power to distinguish the effectiveness of MTX treatment in patients with early RA. We therefore hypothesized that several metabolic pathways are related to the therapeutic effects of MTX, such as nucleic acid metabolism, homocysteine metabolism, one-carbon metabolism, and metabolite analysis may be used to predict the effects of MTX treatment in RA patients.
One of the ultimate goals of the current method was to enable comparisons between RA patients with a favorable response to MTX and those who respond poorly in clinical settings. According to the different situations, we could rapidly adjust the therapy regimen in the clinic. But one problem faced in clinical investigations is the inherently greater variability in a human population compared with that seen in our study. Thus, the study factors such as age, gender, diet, individual difference and other environmental influences, can be more extensively controlled by investigators than when the enrolled human study populations are investigated. Although the exact metabolites identified as potential ‘biomarkers’ may vary among different individuals, more studies could use the methodology applied here to investigate the efficacy of MTX treatment in RA patients. Metabolomic analysis is undoubtedly applicable to the search for biological predictors of response to drug treatment in RA, but future studies should employ larger patient cohorts, more discriminatory analyses, and a less equivocal clinical phenotype.
In conclusion, the monitoring of entire sets of metabolic features is critical to evaluate the efficacy of drug treatments. A metabolic ‘bioprofile’, consisting of predictive serum metabolite features from 1H NMR spectral data of early RA patients receiving MTX treatment are presented in this study. Significantly, our study demonstrated that various serum biomarkers can be used to evaluate MTX treatment in patients with early RA, and the metabolomic approach including 1H NMR spectroscopy coupled with multivariate statistical analysis may be useful to predict the outcome of drug response and treatment.
1H nuclear magnetic resonance
rheumatoid arthritis
methotrexate
disease activity score
principal component analysis
partial least squares discriminant analysis
disease-modifying anti-rheumatic drugs
non-steroid anti-inflammatory drugs
The authors specially thank Dr Ming Jiang of the School of Pharmacy, Tongji Medical College for his technical assistance.
Representative 600-MHz 1H NMR spectrum of serum samples from the ‘effective’ group (red, top) and ‘non-effective’ group (blue, bottom). Typical serum samples are shown and peak assignments of major metabolites on 1H NMR spectra. The spectrum regions of water (δ=2.6–3.0 ppm) and urea (δ=4.3–4.7 ppm) were removed from the analysis for all groups in order to prevent variation in each sample. 1, alanine; 2, acetate; 3, citrate; 4, cysteine; 5, taurine; 6, methionine; 7, hypoxanthine; 8, uracil.
The score plot of PCA components in the effective and non-effective groups. Each point represented a single serum spectrum, with the position determined by the contribution of the tracked features. Black squares indicate the PCA plot of the ‘effective’ group in early RA patients receiving MTX treatment and grey triangles indicate the ‘non-effective’ group.
Loading plot from the PLS-DA analysis. Unknown resonances are indicated by the numerical ppm value or unmarked if not significant between the two groups. Each point represents the number of a metabolite included in the PLS-DA analysis (total, 20 metabolic data points are shown), and the axes are associated with the score plot of
Quality of the principal component models was evaluated with the parameters R2 and Q2. R2 is the explained variance, Q2 is the predictive ability of the model validation, and correlation is the degree of overlap between the permuted and original models. The validation model exhibited good predictive ability, as indicated by the high Q2 value. Grey triangles, the goodness of fit was quantified by R2; black squares, the predictive ability was indicated by Q2 (R2Y=0.802, Q2Y=0.643).
Absolute concentrations of 11 discriminate metabolites from the PLS-LDA plot are shown. Based on creatinine, the relative concentrations of 11 endogenous metabolites were determined, where the effective group showed a significant difference when compared to those of the non-effective group. Dark and light grey squares represent the concentrations of metabolites between the ‘effective’ and ‘non-effective’ group, respectively. A significant difference is indicated by *p<0.05; and **p<0.01.
Demographics and clinical characteristics of the RA patients at baseline and after MTX treatment (10 mg/week) for 24 weeks.
Group | Healthy controls (n=20) | RA patients (n=38) | Early RA patients receiving MTX treatment
| |
---|---|---|---|---|
Non-effective (n=13) | Effective (n=25) | |||
Female/male | 15/5 | 26/5 | 10/1 | 16/4 |
Age (years) | 55.3±3.6 | 56.4±2.8 | 54.7±3.4 | 54.6±2.5 |
Disease duration (years) | - | 4.5±1.9 | 4.4±1.2 | 4.2±0.8 |
DAS28 | - | 5.9±1.4 | 4.6±1.2 | 2.3±0.8 |
Anti-CCP (RU/ml) | 6.3±1.9 | 38.3±5.1 | 36.3±7.8 | 15.3±3.4 |
CRP (mg/l) | 1.3±0.6 | 30.2±2.7 | 28.2±3.6 | 3.5±1.2 |
RF (IU/ml) | 40.7±5.2 | 378.8±80.6 | 336.8±97.6 | 166.7±62.1 |
ESR (mm/h) | 8.3±4.2 | 57.3±9.6 | 52.8±5.3 | 18.2±2.1 |
SCR (mg/dl) | 53.6±9.5 | 58.6±7.4 | 68.2±9.8 | 60.7±8.2 |
ALT (U/l) | 15.8±5.1 | 16.2±5.4 | 21.3±8.2 | 23.9±6.3 |
MTX, methotrexate; DAS, disease activity score; anti-CCP, anti-cyclic citrullinated peptide; CRP, c-reactive protein; RF, rheumatoid factor; ESR, erythrocyte sedimentation rate; SCR, serum creatinine clearance; ALT, alanine transaminase.
p<0.05 compared with healthy controls;
p<0.05 compared with non-effective group.
List of the metabolites identified by the Chenomx NMR Suite in the serum samples.
Metabolites | Chemical shift (ppm) and multiplicity |
---|---|
α-oxoglutarate | δ2.44 (t), 3.07 (t) |
Glycine | δ3.55 (s) |
Citrate | δ2.55 (d), 2.65 (d) |
Aspartate | δ3.56 (s), 6.74 (m), 7.08 (m) |
Acetate | δ1.92 (s) |
Alanine | δ1.32 (d), 3.72 (q) |
Cholesterol | δ3.60 (s) |
Creatinine | δ3.00 (s), 3.96 (s) |
Cysteine | δ2.72 (s) |
Histidine | δ3.96 (d), 6.49 (t), 7.73–7.76 (m) |
Hypoxanthine | δ3.14 (d), 4.05 (t), 5.31–5.37 (m), 6.77 (d) |
Lactate | δ1.32 (s), 4.23 (s) |
Glutamine | δ2.04 (d) |
Methionine | δ1.36 (s), 2.42 (t), 3.40–3.42 (q), 3.56 (d), 3.98 (s) |
Serine | δ2.40 (s) |
Taurine | δ3.36 (t), 3.68 (t) |
Tryptophan | δ4.43 (s), 7.47 (t) |
Trimethylamine-N-oxide | δ3.18 (s) |
Uracil | δ5.81 (d), 7.53 (d) |
Uric acid | δ3.13 (t), 5.46 (m) |
s, singlet; d, doublet; t, triplet; q, quarter; m, multiplet.
VIP values of the major contributing metabolites for the separation in the score plots derived from PLS-DA.
Metabolites | VIP value |
---|---|
α-oxoglutarate | 1.328 |
Glycine | 1.942 |
Citrate | 0.704 |
Aspartate | 1.213 |
Acetate | 2.348 |
Alanine | 0.841 |
Cholesterol | 0.536 |
Cysteine | 0.964 |
Histidine | 2.815 |
Hypoxanthine | 1.224 |
Lactate | 0.872 |
Glutamine | 1.116 |
Methionine | 2.478 |
Serine | 0.836 |
Taurine | 3.624 |
Tryptophan | 1.462 |
Trimethylamine-N-oxide | 2.367 |
Uracil | 2.437 |
Uric acid | 3.837 |
VIP, variable importance in the projection.