Abstract

   Ankylosing spondylitis (AS) is a type of spondyloarthropathies, the
   diagnosis of which is often delayed. The lack of early diagnosis tools
   often delays the institution of appropriate therapy. This study aimed
   to investigate the systemic metabolic shifts associated with AS and TNF
   inhibitors treatment. Additionally, we aimed to define reliable serum
   biomarkers for the diagnosis. We employed an untargeted technique,
   ultra-performance liquid chromatography-mass spectroscopy (LC-MS), to
   analyze the serum metabolome of 32 AS individuals before and after
   24-week TNF inhibitors treatment, as well as 40 health controls (HCs).
   Multivariate and univariate statistical analyses were used to profile
   the differential metabolites associated with AS and TNF inhibitors. A
   diagnostic panel was established with the least absolute shrinkage and
   selection operator (LASSO). The pathway analysis was also conducted. A
   total of 55 significantly differential metabolites were detected. We
   generated a diagnostic panel comprising five metabolites (L-glutamate,
   arachidonic acid, L-phenylalanine, PC (18:1(9Z)/18:1(9Z)),
   1-palmitoylglycerol), capable of distinguishing HCs from AS with a
   high AUC of 0.998, (95%CI: 0.992–1.000). TNF inhibitors treatment could
   restore the equilibrium of 21 metabolites. The most involved pathways
   in AS were amino acid biosynthesis, glycolysis, glutaminolysis, fatty
   acids biosynthesis and choline metabolism. This study characterized the
   serum metabolomics signatures of AS and TNF inhibitor therapy. We
   developed a five-metabolites-based panel serving as a diagnostic tool
   to separate patients from HCs. This serum metabolomics study yielded
   new knowledge about the AS pathogenesis and the systemic effects of TNF
   inhibitors.

   Keywords: ankylosing spondylitis, metabolomics (OMICS), TNF inhibitor,
   liquid chromatography-mass spectroscopy, biomarker

Introduction

   Ankylosing spondylitis (AS) belongs in the group of diseases called
   spondyloarthropathies, presenting with chronic back pain, which
   predominantly affects the spine and the sacroiliac joints. AS is more
   common in males, presenting with inflammatory back pain, the onset of
   which typically occurs in the third or fourth decade of life. At the
   advanced stage, the disease progression may result in spinal deformity,
   limitations of spinal mobility and inevitably impaired quality of life.
   The etiology of AS remains unclear. According to studies in twins,
   genetic factors are now thought to account for over 90% of the risk for
   AS ([37]1). Despite the numerous disease-associated variants identified
   in AS with genome-wide association studies, they cumulatively explain
   only a small proportion (<28%) of the heritability of these diseases
   ([38]2). Environmental exposures have been suspected to play a role in
   AS, such as mechanical stress, infections ([39]3), smoking ([40]4), and
   breast feeding ([41]5) in the early life.

   Due to the insidious onset or ignorance of the lower back pain, the
   diagnosis for AS is often delayed by 5–10 years ([42]6). The
   administration of biologics therapies, which has been proven highly
   effective in multiple clinical trials, could achieve a high rate of
   clinical remission. However, the lack of early diagnosis tools often
   delays the institution of appropriate therapy. Consequently, there is
   an unmet need for more effective and sensitive biomarkers of diagnosis.
   The introduction of tumor necrosis factor (TNF) inhibitors nearly two
   decades ago opened a new chapter of the treatment of AS, especially in
   patients with insufficient response to conventional treatment and
   inhibit radiographic progression ([43]7). Nonetheless, the underlying
   pathophysiology targeted by anti-TNFα therapy has not yet been
   elucidated.

   As one of the ‘omics’ technologies, metabolomics is a fast-developing
   research area in the post-genomic era. It has emerged to be a powerful
   comprehensive approach to characterize convoluted metabolic changes and
   evaluate the biochemical mechanisms involved in such changes in a
   systematic fashion ([44]8). Currently, metabolomics has been utilized
   in biomarker discovery in multiple rheumatic diseases, including
   rheumatoid arthritis (RA) and reactive arthritis ([45]9, [46]10).
   However, the application of metabolomics to AS is still in its infancy,
   although several studies have been reported recently ([47]11–[48]13).
   Overall, the sample size of most studies is small and the types of
   samples are diverse including plasma, fecal, and urine. The two most
   common detection methods in metabolomics, nuclear magnetic resonance
   and mass spectrometry, have not yet been employed, but the findings are
   contradictory among studies. Moreover, the follow-up data regarding
   metabolic alteration in patients treated with TNF inhibitors are
   scarce.

   Therefore, this study aimed to investigate the systemic metabolic
   shifts associated with AS and define reliable serum biomarkers for the
   diagnosis of AS. Additionally, we aimed to further investigate the
   influence of 24-week TNF inhibitor treatment on metabolic profiles in
   AS. To accomplish these objectives, we employed an untargeted technique
   with high sensitivity and specificity, namely ultra-performance liquid
   chromatography-mass spectroscopy (LC-MS), to analyze the serum
   metabolome of 32 AS individuals and 40 healthy controls (HCs).
   Furthermore, we attempted to construct a metabolites-based diagnostic
   panel to distinguish AS from the healthy controls. Besides, metabolome
   profiles were also compared before and after treatment with TNF
   inhibitors. We propose that the serum metabolite signatures can assist
   in diagnosis and provide insight into the underlying pathophysiology of
   AS and the systemic effects of TNF inhibitors.

Materials and Methods

Study Participants

   A total of 32 AS patients were enrolled in this study from the
   Rheumatology department in the Third Affiliated Hospital of Sun Yat-sen
   University between June 2016 and June 2018. Additional 40 sex- and
   age-matched HCs from the physical examination center in our hospital
   were consecutively recruited. Inclusion criteria of AS patients were as
   follows:1) aged over 16 years old; 2) fulfilled the modified 1984 New
   York diagnosis criteria; 3) patients did not take any TNF inhibitors
   treatment before enrollments; 4) active disease (Bath Ankylosing
   Spondylitis Disease Activity Index (BASDAI) ≥4.0 or Ankylosing
   Spondylitis Disease Activity Score-CRP (ASDAS-CRP) ≥1.3); 5) the
   patients administered TNF inhibitors (Etanercept) over 24 weeks.
   Patients with other rheumatic diseases, other systemic diseases or
   tumors were excluded from this study. Patients who took conventional
   disease-modifying antirheumatic drugs (DMARDs) or medicine impacting
   serum metabolites (such as insulin and statin) were also excluded. All
   HCs had no history of chronic disease or rheumatic diseases.
   Demographic and clinical parameters including age, sex, symptom
   duration, BASDAI, Bath Ankylosing Spondylitis Functional Index (BASFI),
   ASDAS-CRP, and laboratory indicators, such as HLA-B27, C-reactive
   protein (CRP) and erythrocyte sedimentation rate (ESR) were recorded.
   The serum samples of AS patients were collected before and after the
   24-week TNF inhibitors treatment. All procedures involving human
   participants in the study were performed in accordance with the 1964
   Helsinki declaration. The protocol was approved by the Ethics Committee
   of the Third affiliated Hospital, Sun Yat-Sen University ([2013]2-93).
   All patients signed informed consent prior to study enrollment.

Sample Collection and Preparation

   The detailed protocols of samples collection, preparation, metabolomics
   profiling and data pre-processing are available in [49]Supplementary
   Data.

Data Processing and Statistical Analysis

   The data processing procedures comprised filtering, imputation of
   missing values (R package “DMwR”) and area normalization. Then the data
   matrix was imported into SIMCA-P software (version 14.1, Umetrics AB,
   Umea, Sweden) for multivariate statistical analysis including principal
   component analysis (PCA) and orthogonal partial least-squares
   discriminant analysis (OPLS-DA). Metabolites were further applied to
   the univariable Wilcoxon rank-sum test. The differential metabolites
   that satisfied the criterion of variable importance in the projection
   (VIP) values of >1.0 and false discovered rate (FDR) of <0.05 were
   considered as biomarker candidates. A diagnostic model was established
   with Least absolute shrinkage and selection operator (LASSO) regression
   (R package “glmnetcr”). The results were presented as mean ± standard
   deviation (SD) for continuous variables and as percentage for
   categorical variables. GraphPad Prism (version 6.02, San Diego, CA,
   USA) was used for statistical analysis of the data. P < 0.05 was
   considered statistically significant.

Results

Demographic Characteristics of the Study Population

   A total of 72 individuals (32 AS patients and 40 HCs) were enrolled in
   the serum metabolic profiling study. Flow diagram of the overview for
   the study design and analytical pipeline was depicted in [50]Figure 1.
   All patients and HCs were included in the discovery stage. The
   treatment stage consisted of 32 follow-up patients who received 24-week
   TNF inhibitors therapy. The general demographic and clinical
   characteristics of the study participants were presented in [51]Table
   1. Age and gender were matched between AS patients and HCs. Most of the
   participants in the two groups were male (92.5% vs. 90.6%, p>0.05) and
   the mean ages were 27.05 ± 5.64 years vs. 28.63 ± 7.53 years (p>0.05).
   The mean disease duration of patients was 94.28 mouths. Indicators of
   clinical assessments significantly improved after treatment, while the
   acute phase reactants (ESR and CRP) significantly decreased (p<0.001
   respectively). Besides, a significant reduction of BASDAI, BASFI, and
   ASDAS-CRP was observed (p<0.001 respectively). These results indicated
   the therapeutic benefit of TNF inhibitor treatment.

Figure 1.

   [52]Figure 1
   [53]Open in a new tab

   The workflow for the study design and analytical pipeline.

Table 1.

   The demographic characteristics of the study population.
   Characteristics HC (n = 40) AS (pre-treatment) (n = 32) AS
   (post-treatment) (n = 32) p
   Male, n (%) 37 (92.5) 29 (90.6) – 1.000*
   Age, year, mean ± SD 27.05 ± 5.64 28.63 ± 7.53 – 0.314*
   HLA-B27 positive, n (%) – 30 (93.8) – –
   Duration, month, mean ± SD – 94.28 ± 48.79 – –
   BASDAI, mean ± SD – 6.85 ± 2.03 3.28 ± 1.70 <0.001
   BASFI, mean ± SD – 4.46 ± 2.22 2.37 ± 2.01 <0.001
   ASDAS-CRP, mean ± SD – 4.36 ± 0.79 1.07 ± 0.73 <0.001
   ESR, mm/H, median (IQR) – 29.5 (12.5–46.5) 5.0 (3.0–10.8) <0.001
   CRP, mg/L, median (IQR) – 21.1 (12.4–44.0) 2.6 (1.1–9.5) <0.001
   [54]Open in a new tab

   *P-values were calculated with HCs as references. Other p-values were