Introduction
Behcet’s disease (BD) is a chronic and relapsing vascular autoimmune/autoinflammatory disease of unknown cause, displaying involvement of multiple organs [
1,
2]. BD is highly prevalent in countries along the “Silk Road”; it can be induced by persistent and excessive immune reactions via autoantigen-activated dendritic cells and T or B cells and leads to endothelial cell damage and vasculitis [
3]. Increases in human IgM, IgG and IgA levels have been found in BD patients, and the production of immunoglobulin isotypes is associated with mucocutaneous, ocular and systemic involvement in naïve active BD patients [
4,
5].
Clinically, corticosteroids and immunosuppressants are employed in rheumatoid immune diseases to attenuate the inflammatory response and tissue damage and relieve clinical symptoms due to their anti-inflammatory and immunosuppressive effects on the immune system [
4,
6‐
9]. However, their long-term use may lead to immune disorders [
4,
7] and increased susceptibility to viral infection [
10]. For example, Djaballah-Ider et al. found that corticosteroid therapy significantly reduced serum immunoglobulin isotype markers [
4] and inflammatory mediators related to disease pathogenesis, including IL-18 and IFN-γ regardless of the clinical manifestations in BD [
7]. Direskeneli et al. found that thalidomide has both anti-inflammatory and regulatory effects in BD, decreasing the levels of the TNF-α receptor, CD8/CD11b + T cells and natural killer cells in early treatment while increasing CD4 + CD45RO+ memory T and γδ + T cells in later treatment [
9].
The systems in the human body (e.g., coagulation, inflammation, etc.) are known to execute their functions cooperatively instead of alone, which can be reflected by clinically measured variables [
11‐
15]. For example, in a study of 48 participants with cirrhosis and nonalcoholic fatty liver disease, Niu et al. revealed the association of five plasma proteins (DPP4, ANPEP, TGFBI, PIGR and APOE) with liver enzymes through a global correlation map of clinical and proteomic data, implying their associations with cirrhosis and nonalcoholic fatty liver disease [
12]. Nathan et al. investigated the personal, dense and dynamic data from 108 individuals during a 9-month period and generated a correlation network between clinical variables, proteomes and genome sequences, which revealed communities of related analytes associated with physiology and disease. For example, the negative correlation between levels of cystine in plasma and polygenic risk scores for inflammatory bowel disease revealed that genetic predisposition of diseases may be manifested by analyte changes and suggested that supplementation with cystine in a healthy population at high risk may stop the transition to disease by preventing inflammation and oxidative damage. Nathan proposed that measurement of personal data clouds over time can improve understanding of health and disease and are the essence of precision medicine.
Evidence has suggested that multiple pathological pathways are involved in BD with no single common denominator related with BD [
2]. However, no common or dominant pathological factor for BD has been identified until now. Moreover, the correlations of clinical variables in BD and their associations with BD diagnosis, progression and therapy are largely unknown. The hypothesis of this work is to explore the interactions between different clinical variables by correlation analysis to determine the associations between the functional linkages of different paired variables and potential diagnostic biomarkers of BD. To address this issue, we first measured the immunoglobulin proteome (IgG, IgG1–4, IgA, IgA1–2) using a plasma microarray and performed a comprehensive correlation analysis of the immunoglobulin proteome and 29 clinical variables. We defined the physiological, pathological and pharmaceutical relationships based on the correlations of all variables in the healthy controls (HCs) and BD patients without and with immunomodulatory therapy. Furthermore, we calculated the ratio changes between clinical variables to identify the specific indicators for the diagnosis of BD and differential diagnosis from other types of vasculitis.
Materials and methods
Demographic and clinical characteristics of subjects
All plasma samples were obtained from the Peking Union Medical College Hospital (Table
1), where BD patients were diagnosed according to the 1990 International Study Group (ISG) criteria [
16] and the International Criteria for Behcet’s Disease (ICBD) [
17], and patients with Takayasu arteritis (TA) and those with ANCA-associated vasculitis (AAV) were diagnosed respectively according to [
18,
19].Furthermore, all patients with BD were assigned to four groups according to medication use, which included BD patients without treatment (BD-N), treatment with corticosteroids (BD-C), treatment with immunosuppressants (BD-I) or treatment with both (BD-C&I). Patients using immunosuppressants are defined as those who are under drug treatments, including Azathioprine, Cyclosporine, Thalidomide, Cyclophosphamide, Leflunomide, Hydroxychloroquine and Tripterygium glycosides. Blood samples were anticoagulated with EDTA, centrifuged at 12,000 rpm for 10 min, and the upper plasma layer was collected and frozen at − 80 °C until use. This study was approved by the Medical Ethics Committee of Peking Union Medical College Hospital (JS-2049), informed consent was obtained from all subjects. All research on humans was performed in accordance to the Declaration of Helsinki.
Table 1
Demographic and clinical characteristics of subjects
Discovery set | Age(y) | 35.40 ± 10.16 | 27.25 ± 13.18 | 30.50 ± 6.80 | 31.04 ± 9.800 | 49.17 ± 13.51 | / | / |
Sex(M/F) | 10/5 | 7/5 | 4/4 | 20/8 | 31/35 | / | / |
ICBD | 5.73 ± 2.63 | 5.75 ± 1.36 | 5.00 ± 1.53 | 6.643 ± 1.890 | / | / | / |
Validation set | N | 27 | 5 | 16 | / | 30 | 18 | 18 |
Age(y) | 37.81 ± 11.60 | 34.20 ± 12.40 | 36.94 ± 13.69 | / | 38.40 ± 11.40 | 51.89 ± 14.07 | 33.94 ± 10.46 |
Sex(M/F) | 8/19 | 3/2 | 5/11 | / | 14/16 | 7/11 | 7/11 |
ICBD | 5.19 ± 1.62 | 6.60 ± 0.89 | 5.50 ± 1.32 | / | / | / | / |
Quantification of the immunoglobulin proteome using plasma microarray
All plasma samples were retrieved from the − 80 °C freezer, thawed on ice and centrifuged at 12,000 rpm for 10 min at 4 °C. Five microliters of each plasma sample was diluted with 0.02% BSA (phosphate-buffered saline, PBS, pH = 7.4) according to the immunoglobulin subtype (IgG2 IgG3, IgG4 and IgA2:10×; IgA1: 100×; IgG, IgA, IgG1: 500×). Standard immunoglobulin proteins were obtained commercially, including IgG (ZSGB-BIO, Beijing, China), IgA (Bersee Technology Co. Ltd., Beijing, China), IgG1 (Sino Biological Inc., Beijing, China), IgG2 (Sino Biological Inc.), IgG3 (Sino Biological Inc.), IgG4 (Sino Biological Inc.), IgA1 (Fitzgerald Industries International, Massachusetts, USA) and IgA2 (Fitzgerald Industries International). BSA (1 mg/ml) and 1x PBS (pH 7.4) were used as blank controls. All the samples were printed on the modified slide surface (CapitalBio Technology Co., Ltd., Beijing, China) in duplicate by a Smart-Arrayer™ 136 microarrayer (CapitalBio Technology Co., Ltd., Beijing, China).
Prior to the assay, the plasma microarray was first blocked with 1% BSA at room temperature for 1 h. The detection of immunoglobulin proteins in plasma was performed by incubation for 30 min with the appropriate fluorescein-labeled detection antibodies, including Donkey anti-hIgG(Fc) Alexa Fluor 555 and Rabbit anti-hIgA(Fc) Alexa Fluor 647 (Jackson Immuno Research, Pennsylvania, USA), Mouse Anti-Human IgG1 Hinge-Alexa Fluor® 488, Mouse Anti-Human IgG2 Fc-Alexa Fluor® 488, Mouse Anti-Human IgG3 Hinge-Alexa Fluor® 647, Mouse Anti-Human IgG4 Fc-Alexa Fluor® 647, Mouse Anti-Human IgA1-Alexa Fluor® 647 and Mouse Anti-Human IgA2-Alexa Fluor® 488 (SouthernBiotech, Birmingham, USA). The unbound molecules were removed by washing the slide with 0.05% PBST three times and deionized water two times in the dark. Then, the resulting slide was air-dried and scanned by GenePix® 4300A (Molecular Devices, California, USA) at a wavelength of 488 nm (IgG1, IgG2 and IgA2), 532 nm (IgG) or 635 nm (IgA, IgA1, IgG3 and IgG4).
The quantification of immunoglobulin proteins in plasma was performed by using a standard curve fitted with a 4- or 5-parameter logistic model using the “nplr package” in R as previously described [
20].
Measurement of clinical variables
All plasma samples were removed from the − 80 °C freezer, thawed on ice and centrifuged at 12,000 rpm for 10 min at 4 °C. The basic and clinical information of patients was obtained from the Hospital Information System of Peking Union Medical College Hospital, including age, sex, disease history, clinical symptoms and clinical treatment information with corticosteroids and/or immunosuppressants (Table
1). The results of the laboratory tests at the time of sample collection were obtained from the Laboratory Information Management System, including clinical chemistry, clinical immunology, hematology, etc. The abbreviations and full names of all clinical variables are shown in Table S
1. Routine blood tests were completed by a Siemens ADVIA2120 or Sysmex XN9100 analyzer (Siemens, Munich, Germany; Sysmex America, Illinois, USA); ESR tests were completed by a Greiner MONITOR-S analyzer (Greiner Bio-one GmbH, Kremsmünster, Austria); CRP tests were completed by an Orion QuikRead go Instrument (Orion Corporation, Espoo, Finland); biochemical variables were completed by a Beckman AU5821 analyzer (Beckman Coulter, California, USA); and routine urinalysis was completed by Siemens Bayer Clinitek 500 analyzers (Siemens, Munich, Germany).
Statistical analysis
R version 3.5.2 and Prism 8.2.0 were used to perform all the statistical analyses. Descriptive statistics were presented as the mean ± standard deviation for continuous data or frequencies for categorical variables. Student’s t test or one-way ANOVA was applied to test the mean differences between two groups or multiple groups with normal distributions, respectively; otherwise, the Wilcoxon rank sum test or Kruskal-Wallis test was performed.
The linkage analysis of the immunoglobulin proteome and all variables was performed by calculating the Pearson’s or Spearman’s correlation coefficient between two variables according to their normality. Pearson’s correlation coefficient was performed when the data of both variables had a normal or log-normal distribution. Hierarchical clustering analysis of the correlation coefficient matrices was performed using Euclidean distance and the complete method in the pheatmap package in R, with which the positively and negatively correlated variables were clustered together in a heatmap. A P value of < 0.05 was considered to be statistically significant.
In addition, ratios between every two clinical variables were calculated. Ratios more than one with significant difference when comparing BD-N and HC were retained for further comparison between groups in the validation set. A P value of < 0.05 was considered to be statistically significant.
Discussion
BD is an inflammatory disease of unknown etiology that affects the epidermal, mucocutaneous, vascular, ophthalmologic, gastrointestinal, pulmonary, and central nervous systems. Corticosteroids and immunosuppressants are frequently employed clinically to treat BD patients by regulating inflammation and immune disorders. As an option for long-term treatment for autoimmune diseases [
33], immunosuppressants have an inhibitory effect on the immune response to weaken attacks on body’s own tissue by inhibiting the proliferation and function of T cells or B cells [
34]. In contrast, corticosteroids work quickly but have significant side effects. Corticosteroids can affect almost all kinds of immune cells and multiple points of the immune response. For example, they prevent lymphocyte recycling and the production of antibody-producing and cytotoxic effector cells, but they also have significant anti-inflammatory effects. They inhibit the adhesion of neutrophils to vascular endothelium in inflammatory sites and inhibit monocyte function, among other effects [
35]. However, the pathogenesis of BD and its therapeutic influence by immunomodulatory medication are largely unknown. To address this question, we comprehensively measured and analyzed the changes in clinical variables related to immunity, inflammation, coagulation and nutrition in HCs and BD patients without and with immunomodulatory medication.
We observed an overall increase in immunoglobulin proteome expression in BD patients without treatment (Fig.
2), which demonstrates the existence of an immune disorder during BD development [
4]. However, the expression of the immunoglobulin proteome, especially IgG1, IgG2 and IgG4, can be suppressed by corticosteroids and immunosuppressants. The results can be further confirmed by the correlation analysis, in which the correlations of pathological linkages (IgA-IgG3 and IgA1- IgG3) were increased in the BD group and decreased under immunomodulatory therapy (Figs.
2c and
3e). The results demonstrate that corticosteroids and immunosuppressants exert their effects by inhibiting immune and inflammatory responses [
4,
7].
The same results were observed in inflammation, in which WBC and NEUT were significantly increased in the BD groups, which was consistent with the functions of WBC and NEUT in mediating vessel damage through enhanced migration in the circulatory system. While corticosteroids are used to inhibit inflammation and the immune response in certain clinical situations, they may also cause an increase in the WBC count and predominantly neutrophils (NEUT) mainly by the demargination of the neutrophils from the endovascular lining [
36,
37]. In addition, the use of corticosteroids may promote the maturation of neutrophils in the bone marrow and mobilization into the blood circulation by expression of key receptors such as Annexin A1 [
38,
39], as also observed for the pathological link of BASO-LY% and LY%-WBC. Although immunosuppressants such as azathioprine are reported to causes dose-related bone marrow suppression and leukopenia [
40], we did not observe significant difference in blood cells in this study. All these results suggest that clinical evaluations of inflammation should consider medication use as well as the clinical symptoms and signs of the patients.
Inflammation may cause damage to the vessel wall and initiate the coagulation pathway and thrombosis [
22,
41]. Platelets could be hyperactivated under inflammation, after which granules are released to further promote coagulation and inflammation [
42]. In this study, upregulation of MPV in BD patients and its downregulation by corticosteroids were observed (Fig.
3). MPV reflects alterations in the morphology of platelets. Elevated MPV means larger platelets with more dense granules that are therefore more thrombogenic than smaller ones, and it is a marker of platelet function and is involved in thrombosis and vascular damage in BD [
43,
44]. In contrast, the change in PLT was not obvious, as changes in MPV and PCT can be observed before detectable changes in platelets [
22]. There is also growing evidence that platelets are not only involved in fatal vascular events but also function in disease progression by interacting with neutrophils. Schrottmaier et al. proposed that direct interaction of platelets with neutrophils leads to neutrophil activation, recruitment and formation of neutrophil extracellular traps, further promoting the progression of vascular pathologies [
42]. Pamuk et al. found significantly higher levels of platelet-neutrophil complexes in BD patients with major vascular involvement than in those without vascular involvement and healthy controls [
45]. Consistently, interaction between WBCs or neutrophils and platelet was observed in this research in BD patients based on their positive correlation (Fig.
4). Platelets also showed the highest affinity for other innate and adaptive immune cells, including monocytes (as discussed in the next section) and lymphocytes, by soluble mediators [
42]. In this study, the pharmacological linkages of LY%-PLT and CRP-PLT further suggest that integrative analysis of granulocytes, platelets and related variables is likely to provide a comprehensive understanding of disease activity, thrombotic potential and potential tissue damage.
Based on linkage analysis, we further constructed a novel method according to the ratio changes between two clinical variables and demonstrated that four ratios – TP/MCV, PCT/MONO, TP/MCH, and TP/MCHC – have higher value in BD than in HC, TA and AAV, suggesting these four ratios as potential diagnostic indicators for BD. PCT, which is produced from PLT and MPV, reflects the total platelet mass [
46]. In our research, PCT and MPV were significantly increased in BD. Platelets play an important role in the pathogenesis of thromboembolic diseases. Platelets are more reactive in BD patients than in normal controls, which may contribute to the tendency for thrombosis. Moreover, increased MPV in an inflammatory state contributes to thrombosis, which may be an independent risk factor for vascular involvement in BD [
47]. Evidence has shown that monocytes in BD patients are activated and produce proinflammatory cytokines, causing increased adhesion of neutrophils to endothelial cells and chronic inflammation [
48]. Interactions between platelets and monocytes is also reported to relate to major vascular involvement in BD [
45], and platelets may induce monocyte differentiation into a more inflammatory phenotype [
49]. The higher value of PCT/MONO, consisting of platelets and monocytes, confirms the potentially close interaction between platelets and monocytes [
42]. This has been highlighted as an important pathophysiological link between inflammation, thrombosis and endothelial activation [
50], such as the concordance of platelets and monocytes in immune-thrombosis. Moreover, platelets are reported to interact with monocytes to propagate their differentiation into macrophages, and when activated, platelets stimulate monocytes to leave the blood vessel and enter tissues, causing a higher level of PCT/MONO [
51]. We propose that a higher level of PCT/MONO, representing aggregates and interaction between platelet and monocyte, is a potentially attractive and easily accessible marker in BD [
42].
MCV, MCH, and MCHC are useful biomarkers in the evaluation of anemia. MCV indicates the mean size of red blood cells, while MCH and MCHC indicate the mean amount and the mean concentration of hemoglobin in each red blood cell, respectively. It has been reported that chronic anemia is common in BD patients, especially with intestinal involvement [
52,
53], with contributors like bone marrow failure [
54] or serum prohepcidin and hepcidin, whose levels are also closely associated with disease activity [
55]. It is likely that the increasing trend of total protein and/or decreases in MCV, MCH or MCHC lead to high levels for the three ratios. However, our study demonstrated that corticosteroids and immunosuppressants do not function by decreasing these higher ratios, illustrating the stability of these indicators. However, other factors involved and the specific mechanisms of these interactions remain to be elucidated in future studies.
There are several limitations in our research. First, the coregulatory mechanisms of clinical variables through physiological, pathological and pharmacological linkages are not well understood and should be carefully interpreted according to the clinical symptoms of BD patients. Second, the numbers of samples and patients’ information employed in this study were limited. In the future, we will include more information to match the backgrounds of control patients and verify the utility of these functional linkages in diagnosis and prognosis in larger cohorts.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.