Introduction

Behcet’s disease (BD) is a chronic systemic disease characterized by recurrent attacks of oral aphthous ulcers, genital ulcers, skin lesions, and uveitis1,2. It has a global distribution, and more frequently occurs in the countries along the ancient “Silk Road”3. Behcet’s uveitis (BU), one of the common types of uveitis with the highest rate of blindness in China, manifests as relapsing and remitting panuveitis and retinal vasculitis. Until now, BU is of great public health concern due to its poor visual outcome or even blindness caused by the destructive and recurrent attacks on ocular structures and sensory retina, which warrants investigation of the underlying mechanism of this disease4. It has been shown that BD is closely linked to T helper (Th) responses5,6, particularly, the hyperactivity of Th1 and Th17 responses is reported to be the cornerstone of BD pathogenesis7,8. There is evidence that genetic factors and metabolic disorders have also been implicated in the development of this disease9,10. However, the exact pattern of global Gene-Protein-Metabolism alterations in BD is unclear.

The search for a depiction of panoramic change from large numbers of studies is expected to reveal the molecular alterations and the underlying mechanism in BD. However, there are several issues about exploring the molecular changes of this disease. Firstly, despite extensive research has been carried out on this disease, it is still difficult to systematically explore the consistently molecular changes in individual studies. Secondly, the generalizability of many existing publications is rather controversial, as experimental data are less convincing because of small sample sizes or the differences of study designs. In the face of these challenges, current research on this issue has rarely explored the reproducibility and dependability of published data across large-scale studies in BD. As the vote-counting approach has now been widely used to explore consistently dysregulated molecules in various diseases11,12, this approach has attracted increasing attentions and is expected to explore the altered molecular panorama of BD.

In this study, we performed the vote-counting approach to explore consistently dysregulated molecules in the circulation and adaptive immune-related samples from patients with BD, and aimed to depict the pattern of molecular alterations in this disease.

Materials and methods

Data source

A thorough assessment of literature was performed using the vote-counting approach, as previously described13. The BD-related search terms were developed from specialized vocabulary, and searched in PubMed. The key terms for search are listed in Supplementary Table 1. Firstly, studies published between January 1946 and April 2024 in any language were selected for preliminary screening according to abstracts, and experimental articles were chosen for subsequent analysis. Secondly, articles were screened based on the inclusion criteria that (i) studies compared the levels of gene, protein, or metabolite between patients with BD and controls; (ii) positive results; (iii) original research, and exclusion criteria that (i) intervention studies that investigated genetic, proteinic, or metabolic alterations associated with anti-BD treatment; (ii) animal model and cell studies; and (iii) gene, protein, or metabolite with unknown regulation direction. Thirdly, differential genes, proteins, and metabolites were extracted from included studies after reviewing full-text. Relevant information of eligible molecules was recorded, including molecular name, tissue type, up- or down-regulation, sample size, and criteria for BD. Finally, eligible molecules were standardized naming using UniProt and Human Metabolome Database. Genes and proteins were uniformly labeled as gene symbols since the corresponding relationship between genes and proteins. If a molecule was validated at both the transcript and protein levels in the same samples from the same study, it was counted only once. Duplicate molecules in one study but from different samples were regarded as individual entries. Two researchers received training in the pilot test prior to independent data collection and management. For the purpose of resolving any misunderstandings or disputes, regular meetings were organized.

Analysis approach

The vote-counting approach is based on the hypothesis that a differential molecule’s probability of up- or down-tendency obeys a binomial distribution13. It is an effective way to identify molecules that were independently validated by external studies, with a presumption that a reliable differential molecule is robust enough to exhibit significance across various studies14. Currently, this is the most practical way to avoid a dilemma regarding the lack of statistics, such as raw quantitative data. And it has now been widely used to integrate massive biological information from large-scale studies13,14. Therefore, this strategy was used in this study to explore the molecular alteration of patients with BD. Candidate molecules in the blood, peripheral blood mononuclear cells (PBMCs), CD4+T cells, and aqueous humor (AqH) from all available studies were selected for binomial analysis, whereas other samples were excluded due to the limited data amounts.

In this study, the vote-counting procedure was carried out to compute vote-counting statistic (VCS) for each molecule. Each entry for a molecule showing a trend toward up- or down-regulation received a count score of 1 or − 1, respectively. Cumulative scores were evaluated by calculating the VCS of each molecule. VCS > 0 indicated that candidate molecule was prone to upregulation across studies, whereas < 0 indicated that it was prone to downregulation. The binomial probability test was conducted in R software v3.6.3 (https://www.r-project.org/) with the function binom.test to compare assumed counts of up- and down-regulations with actual counts. A P value < 0.05 was regarded as significantly different. Since molecules with total counts below four had a minimum P value of 0.125, they were not introduced into the vote-counting procedure.

Results

Data set

A total of 13,285 citations from PubMed were screened. After data filtering, 745 eligible studies that identified differential genes, proteins, or metabolites in patients with BD were cumulatively gathered, whereas 12,540 studies were excluded based on the screening criteria. The annotations of excluded studies are displayed in Supplementary Table 2. And the full information of included studies, including title, authors, citation, journal, publication year, and criteria for BD is provided in Supplementary Table 3. In total, 2,354 differential molecules with 3,574 molecule entries in blood, PBMCs, CD4+T cells, and AqH samples were extracted from the eligible studies. The full information of candidate molecules in the above samples, including molecule name, sample, sample size, and regulation direction is displayed in Supplementary Table 4. Eighty-two unique molecules in blood, 6 in PBMCs, 2 in CD4+T cells, and 4 in AqH samples were reported as dysregulated in at least four studies, and were manually curated as a candidate set (Fig. 1). And the full information of other samples excluded due to limited data is shown in Supplementary Table 5.

Fig. 1
figure 1

Distribution of candidate molecules in the blood, PBMCs, CD4+T cells, AqH samples from patients with Behcet’s disease. The darker color represented the larger value of total dysregulation counts.

Full size image

Molecular alterations in blood

In blood, a total of 1,347 differential molecules with 2,322 molecule entries were reported. Eighty-two candidate molecules were voted and visualized in Fig. 2A. After calculation by the vote-counting procedure, 38 molecules were consistently upregulated, and the top 10 molecules were C-reactive protein (CRP; VCS = 131, p < 0.001), tumor necrosis factor-α (TNF-α; VCS = 35, p < 0.001), malondialdehyde (MDA; VCS = 28, p < 0.001), homocysteine (Hcy; VCS = 27, p < 0.001), C-X-C motif chemokine ligand 8 (CXCL8; VCS = 23, p < 0.001), interleukin 6 (IL-6; VCS = 23, p < 0.001), interferon-γ (IFN-γ; VCS = 26, p < 0.001), interleukin 17 (IL-17; VCS = 21, p < 0.001), α2-macroglobulin (α2-M; VCS = 15, p < 0.001), and fibrinogen (VCS = 15, p < 0.001). And 4 molecules were consistently downregulated, namely high density lipoprotein cholesterol (HDL-C; VCS = − 15, p < 0.001), apolipoprotein A (Apo-A; VCS = − 8, p = 0.004), hemoglobin (Hb; VCS = − 9, p = 0.011), and glutathione (GSH; VCS = − 5, p = 0.031) (Table 1).

Table 1 The vote-counting results in blood.
Full size table

In plasma, the alterations of 442 differential molecules with 648 molecule entries were manually curated, and 20 candidate molecules were visualized in Fig. 2B. After calculation by the vote-counting procedure, 11 molecules were consistently upregulated, and the top 5 molecules were CRP (VCS = 23, p < 0.001), MDA (VCS = 15, p < 0.001), Hcy (VCS = 14, p < 0.001), fibrinogen (VCS = 10, p < 0.001), and von Willebrand factor (vWF; VCS = 7, p = 0.008) (Table 2).

Table 2 The vote-counting results in plasma.
Full size table

In serum, 376 differential molecules with 858 molecule entries were reported. Forty-six candidate molecules were voted and visualized in Fig. 2C. After calculation by the vote-counting procedure, 18 molecules were consistently upregulated, and the top 5 were CRP (VCS = 70, p < 0.001), TNF-α (VCS = 30, p < 0.001), CXCL8 (VCS = 20, p < 0.001), IL-6 (VCS = 18, p < 0.001), and IL-17 (VCS = 17, p < 0.001). And 2 molecules were consistently downregulated, namely HDL-C (VCS = − 12, p < 0.001) and Apo-A (VCS = − 6, p = 0.016) (Table 3). In addition, the overlapping dysregulated molecules among blood, plasma, and serum are displayed in Fig. 2D.

Fig. 2
figure 2

Molecular alterations in blood, plasma and serum from patients with BD. Dysregulated molecules were identified by the vote-counting procedures with a significance threshold of one-tailed p < 0.05. (A) Volcano plots for consistently upregulated (red) and downregulated (blue) molecules in blood. (B) Volcano plots for consistently upregulated (red) and downregulated (blue) molecules in plasma. (C) Volcano plots for consistently upregulated (red) and downregulated (blue) molecules in serum. (D) Venn plot of shared dysregulated molecules among blood, plasma, and serum.

Full size image
Table 3 The vote-counting results in serum.
Full size table

Molecular alterations in PBMCs

A total of 1,036 differential molecules with 1,099 molecule entries were reported in PBMCs. Six candidate molecules were voted, and the result of molecular alterations is shown in Fig. 3A. After calculation by the vote-counting procedure, 3 molecules were consistently upregulated, namely TNF-α (VCS = 7, p = 0.008), IFN-γ (VCS = 5, p = 0.031), and toll like receptor 4 (TLR4; VCS = 5, p = 0.031).

Fig. 3
figure 3

Molecular alterations in PBMCs, CD4+T cells, and AqH from patients with BD. (A) Volcano plots for consistently upregulated molecules (red) in PBMCs. (B) Volcano plots for consistently upregulated molecules (red) in CD4+T cells. (C) Volcano plots for consistently upregulated molecules (red) in AqH.

Full size image

Molecular alterations in CD4+T cells

In CD4+T cells, there were 61 differential molecules with 82 molecule entries reported. Two candidate molecules were voted, and the result of molecular alterations is displayed in Fig. 3B. After calculation by the vote-counting procedure, 2 molecules were consistently upregulated, namely IL-17 (VCS = 9, p = 0.002) and IFN-γ (VCS = 7, p = 0.008).

Molecular alterations in aqueous humor

A total of 43 differential molecules with 71 molecule entries were reported in AqH samples. Four candidate molecules were voted, and are shown in Fig. 3C. After calculation by the vote-counting procedure, 3 molecules were consistently upregulated, namely CXCL8 (VCS = 7, p = 0.008), IFN-γ (VCS = 5, p = 0.031), and TNF-α (VCS = 5, p = 0.031).

Integrated molecular alterations in blood, PBMCs, CD4+T cells, and aqueous humor

Integrated analysis was performed to identify target molecules associated with the pathogenesis of BD. By overlapping molecules that were consistently dysregulated in blood, PBMCs, CD4+T cells, and AqH samples, IFN-γ was identified as the target molecule in all samples. The numbers of consistently dysregulated molecules among blood, PBMCs, CD4+T cells, and AqH samples are shown in Fig. 4.

Fig. 4
figure 4

Venn plot for shared dysregulated molecules among blood, PBMCs, CD4+T cells, and AqH samples in patients with BD.

Full size image

Discussion

Based on evidence from large-scale studies in BD, we demonstrated that 42 molecules in blood, 3 molecules in PBMCs, 2 molecules in CD4+T cells, and 3 molecules in AqH were consistently dysregulated by performing the vote-counting procedure. Our study is the first systematic analysis to explore consistently dysregulated molecules in BD. These findings exhibit the alteration of molecular pattern in BD patients, which contributes to reveal the underlying molecular mechanisms and provide potential therapeutic targets for this disease.

Contrary to our expectations, the main results found in this study revealed the insufficiency of reproducibility in biological research. Although more than 700 studies were investigated, only about 40 molecules were identified to be consistently dysregulated in blood, whereas 6 of them were common between plasma and serum investigations. Moreover, only several molecules were consistently dysregulated in PBMCs, CD4+T cells, and AqH. This situation may be due to various parameters, such as population, region, sample preparation, experimental operation, data collection and analysis, which can affect the identification of differential molecules. Indeed, as the decisive factor influencing the reproducibility of measurements, any alteration in parameter settings during the procedures may substantially effect on the results of both quantitative and qualitative analyses. Previous studies have also reported similar results of the inadequate reproducibility of data across independent studies13,14, which is consistent with this study. Despite these concerns, the results of this study demonstrate that certain molecules are consistently dysregulated across independent studies, which helps to understand the physiopathology of BD.

It has been shown that Th1 and Th17 cytokines play a crucial role in the development of BD15. Regarding Th1 cytokines, IL-12, IL-2R and IFN-γ are endogenous cytokines secreted by Th1 cells, and are closely associated with inflammatory responses16. Studies have demonstrated that IL-12 and IL-2R are involved in the secretion of IFN-γ16,17. And IL-1β and IL-18 are synergizing with IL-12 to stimulate IFN-γ production and regulate Th1 responses18,19. In this study, our results revealed that IL-12, IL-2R, IL-1β and IL-18 were consistently upregulated in blood, and IFN-γ as the target molecule was consistently upregulated in blood, PBMCs, CD4+T cells and AqH samples, suggesting that hyperactivity of Th1 response plays an important role in the pathogenesis of BD. Preclinical and clinical evidences show that modulation of Th1 cytokines is a potential effective therapeutic approach in BD20, for example, cyclosporine A which inhibits the IFN-γ, IL-2R, and IL-12 expression can effectively relieve inflammation in patients with BD21. Here, our study supports the notion that pharmacological medications exert anti-inflammatory effects by suppressing the activation of Th1-pathway axis in this disease. Regarding Th17 cytokines, IL-6, IL-23, IL-17 and TNF-α are Th17-related proinflammatory cytokines22. It is reported that IL-6 and IL-23 are associated with the productions of IL-17 and TNF-α22,23. In this study, we found that IL-6 and IL-23 in blood, IL-17 in blood and CD4+T cells, and TNF-α in blood, PBMCs and AqH samples were consistently upregulated, suggesting that dysregulation of Th17 cytokines has been involved in the pathophysiology of BD. It is suggested that modulation of Th17 pathway is an effective therapeutic approach for this disease24, for example, Tocilizumab (IL-6 blockade), Infliximab, Etanercept and Adalimumab (TNF replacement therapies) have been clinically validated to be effective in the treatment of BD25,26,27,28. And the successful application of cytokine inhibitors further emphasizes the important role of Th1 and Th17 cytokines in the immunopathogenesis of this disease. In addition, other blocking agents, e.g., Canakinumab (anti-IL-1β monoclonal antibody), Tadekinig alfa (IL-18 binding protein), Secukinumab (IL-17 A inhibitor), and Ustekinumab (IL-12/IL-23 antagonist) may also be applicable in the treatment of BD, which warrants further studies to confirm.

Interestingly, when analyzing the integrated molecular changes of different biological tissues, we are surprised to find that different tissues exhibited almost the identical biological significance, that is, Th1/Th17 activation is identified to be the most prominent signature of BD. In this study, we have observed that Th1 related cytokines (IFN-γ) and Th17 related cytokines (IL-17 or TNF-α) are consistently upregulated in blood, PBMCs, CD4+T cells, and AqH. Blood as a peripheral tissue is essential of internal environment, and its molecular changes reflect the systemic state of the body. PBMCs contain a variety of immune cells, and CD4+ T cells are an important subset of them, both of them are closely related to immune response. AqH, as a special intraocular fluid, maintains a unique association with ocular manifestations, its molecular changes accurately and specifically reflect the local immune response and pathological process of the eye. In agreement with previous findings29,30, we observed that only Th1 and Th17 cytokines were consistently upregulated in both systemic, cellular immune-related and local intraocular tissues, supporting the classical Th1/Th17 activation hypothesis of BD. It is worth mentioning that IL-10 as the cytokine of Treg was inconsistently dysregulated in this study, future studies with more candidate entries are needed to investigate. Overall, our results suggest that the hyperactivity of Th1 and Th17 responses is the most prominent feature of molecular alterations in BD, and pharmacotherapy that inhibits activation of the Th1/Th17-pathway axis could be a potential therapeutic target for this disease.

Moreover, by integrating the results of molecular alterations from different biological tissues, we have another observation that CXCL8 or TLR4 were consistently upregulated in blood, PBMCs and AqH. CXCL8 as a major mediator of inflammatory responses, exerts its effect through guiding neutrophils to the site of infection31,32. TLR4 promotes the production of CXCL8, which can induce neutrophil recruitment and inflammatory cell accumulation, as well as the release of proinflammatory cytokines such as C-X-C motif chemokine ligand 10 (CXCL10) and C-C motif chemokine ligand 2 (CCL2)33,34,35. And complement 3 (C3) acts as a chemoattractant that induces neutrophil chemotaxis36. Growing studies have shown that CXCL8, TLR4, CXCL10, CCL2, and C3 play an important role in the chemotaxis of neutrophils37. In this study, our results revealed that C3, CXCL10 and CCL2 in blood, CXCL8 in blood and AqH, and TLR4 in PBMCs were consistently upregulated, suggesting that enhanced neutrophil chemotaxis has been involved in the pathogenesis of BD. It is worth noting that, apart from the aforementioned Th1/Th17 related cytokines, only CXCL8 or TLR4 (the mediators of the neutrophil chemotaxis) showed a consistent upregulation in systemic, cellular immune-related and local intraocular tissues, suggesting that enhanced neutrophil chemotaxis may be another signature of this disease. There is evidence that modulating the chemotactic function of neutrophils contributes to attenuate inflammation38, which is consistent with our findings. For example, colchicine can inhibit neutrophil chemotaxis and downregulate CXCL8 and TLR4 expression, which is now widely used in the treatment of BD39,40. TLR4 inhibitors, e.g., resatorvid (TAK-242), may also be applicable for the treatment of BD, which warrants further study. Overall, modulation of the neutrophil chemotaxis axis may be a potentially therapeutic approach for this disease, which deserves further study to confirm.

Vascular endothelial injury and activation of haemostatic system have been reported to be associated with the occurrence of BD vasculitis. There is evidence that vascular endothelial growth factor (VEGF), endothelin 1 (ET-1), and ischemia modified albumin (IMA) are involved in the feature of vascular endothelial injury41. VEGF and ET-1 have been shown to be essential for pathological angiogenesis42,43,44,45, and IMA is closely associated with vasculitis46,47. In this study, we found that VEGF, ET-1, and IMA were consistently upregulated in blood, suggesting that vascular endothelial injury is related to the pathological process of BD. Consistent with our findings, anti-VEGF monoclonal antibody, e.g., Bevacizumab, has been shown to be effective in alleviating vascular inflammation in patients with BD48. ET-1 antagonists may also be applicable in the treatment of BD, which warrants further studies to confirm. In addition, fibrinogen, plasminogen activator inhibitor (PAI), factor VIII (FVIII), vWF, and alpha-1-antitrypsin (SERPINA1) are participated in the activation of haemostatic system49. Fibrinogen is considered as the primary basis for thrombosis, and PAI is an inhibitor of fibrinolysis to accelerate thrombosis49. FVIII and vWF are regarded as procoagulant factors and essential for coagulant activity50,51. And SERPINA1 irreversibly inhibits plasminogen activator and reduces coagulation time52. Our results showed that FVIII, vWF and SERPINA1 in blood, and fibrinogen and PAI in plasma were consistently upregulated, suggesting that hypercoagulability is closely associated with the development of BD. In summary, vascular endothelial injury and activation of haemostatic system may play an important role in BD vasculitis.

It has been shown that oxidative stress induced by MDA, Hcy and GSH is involved in the pathological process of BD53,54,55. Growing studies have demonstrated that oxidative stress is characterized by elevated lipid peroxidation and reactive oxygen species production in association with reduced antioxidant activity56,57. MDA is a final product of lipid oxidation that reflects the degree of free radical attacks, and its overexpression exacerbates cell membrane damage58. Hcy, one of the sources of reactive oxygen species production, is related to intensification of oxidative stress59,60. And GSH is regarded as an endogenous antioxidant against free radical damage, and its reduction implies reduced antioxidant capacity61. In this study, our results showed that MDA and Hcy in blood were consistently upregulated, whereas GSH in blood was consistently downregulated, suggesting that oxidative stress has been implicated in the progression of BD. Considering the antioxidant defense mechanism of this disease is susceptible to impairment, the administration of antioxidants may be a viable alternative treatment option in BD, which warrants confirmation by further studies.

This study has some limitations. Firstly, due to the limitation of the vote-counting approach, we cannot identify dysregulated molecules outside of the candidate set. Secondly, despite integrating raw data or merging the average concentrations of candidate molecules could acquire more comprehensive statistical results, it is still difficult to accomplish. Thirdly, we only compared the molecular alteration in blood, PBMCs, CD4+T cells, and AqH samples, and the altered pattern of other samples is needed to be analyzed in future study.

Collectively, this study revealed for the first time the pattern of molecular alterations in BD, which contributes to a systematic understanding of panoramic changes in this disease. Our results suggest that the hyperactivity of Th1 and Th17 responses is the most prominent feature of molecular alterations in BD. In addition, enhanced neutrophil chemotaxis, vascular endothelial injury, activation of haemostatic system, and oxidative stress may play an important role in the pathogenesis of BD, which are expected to be confirmed in further study.