Background

Psoriatic arthritis (PsA) is a musculoskeletal disease that can affect multiple domains, accompanied by other comorbidities such as metabolic syndrome. Some patients with PsA may exhibit axial involvement, referred to as axial PsA (axPsA), while those without axial joint involvement are known as peripheral PsA (pPsA). There is an ongoing controversy regarding whether axial PsA is a distinct disease from axial Spondyloarthritis (axSpA), but multiple studies have pointed out clinical distinctions between axPsA and axSpA [1]. These differences include a less common presence of Inflammatory back pain (IBP), lower rates of HLA-B27 positivity, characteristic axial imaging changes, and so on [2]. IBP, a critical symptom of inflammation in the axial joints for axSpA diagnosis, is found in only 26% of axPsA patients who meet imaging criteria [3]. Secondly, unlike the high positivity rate of HLA-B27 in ankylosing spondylitis [4], the rate of HLA-B27 positivity in patients with PsA is only 20–35% [5, 6], whereas in axPsA this number rises to 43% [7]. All these differences from axSpA may delay the identification of axial involvement in PsA. Conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), which are commonly used in the treatment of pPsA, are considered ineffective for treating axial disease. Guidelines recommend that in patients with predominantly axial disease which is active and has insufficient response to nonsteroidal anti-inflammatory drugs (NSAIDs), therapy with biological disease-modifying antirheumatic drugs (bDMARDs) should be considered [8]. Misdiagnosis of axPsA can result in the progression of axial involvement, leading to joint damage and a significant impact on prognosis [9]. Besides, some patients with PsA only develop axial disease in the late stage of the disease [10, 11], which also implies the need to pay attention to screening during the follow-up period. However, the frequent use of imaging exams undoubtedly increases the radiation and economic burden on patients.

The utilization of reliable biomarkers aids in the early diagnosis of diseases and understanding of the pathogenesis of diseases [12, 13]. Therefore, the discovery of a reliable biomarker in patients with PsA for predicting or identifying involvement of the axial disease would greatly benefit clinicians in prescribing imaging examinations and initiating early use of biologic agents. This, in turn, would improve the prognosis, physical function, and overall quality of life for these patients. Multiple types of samples, such as urine, feces, and saliva can be used as potential samples for research; however, serum and plasma are commonly used due to their ease of collection and stable composition [14]. Techniques for analyzing the serum proteome include mass spectrometry (MS), Multiplex bead- or aptamer-based assays (Slow off-rate modified aptamer scan), and Proximity extension assay (Olink) [15]. Although the latter two techniques offer higher sensitivity, as targeted proteomic techniques, they only monitor the presence or absence of target proteins, whereas MS techniques allow for the hypothesis-free approach with shotgun untargeted MS workflows [16]. Therefore, this study employed untargeted proteomic technology to explore serum biomarkers of axPsA.

Methods

Study design and collection of clinical samples

The research study was divided into two phases: the discovery phase and the verification phase. The workflow of the study is illustrated in Fig. 1. The study included patients who were diagnosed with PsA based on the Classification Criteria for Psoriatic Arthritis (CASPAR) [17] and also had accessible results of sacroiliac joint computerized tomography (CT) and spinal X-ray, for the formation of the discovery and verification cohorts. The definition of axPsA was based on a previous study [11], which included the presence of New York criteria sacroiliitis (unilateral grade ≥ 3, or bilateral grade ≥ 2 sacroiliitis), and/or≥ 1 marginal/paramarginal syndesmophytes of the cervical or lumbar spine. The grading of sacroiliac arthritis on CT examination was done according to the study by Ye et al. [2b).

Fig. 2
figure 2

Unbiased L-MS/MS-based protein analysis. a PCA score plot of the serum samples of the discovery cohort. b PLS-DA score plot of the serum samples of the discovery cohort. c This Venn diagram shows the number of DEPs found in the pairwise comparison among the three groups. d The volcano plot shows the DEPs between axPsA and pPsA. PEDF was the top-upregulated DEP according to the p-value. axPsA axial psoriatic arthritis, pPsA peripheral psoriatic arthritis, HC healthy control, DEPs differentially expressed proteins, PEDF pigment epithelium-derived factor

Full size image

We identified a total of 130 DEPs when comparing healthy controls and all patients with PsA. When comparing axPsA and HC, pPsA and HC, and axPsA and pPsA, we identified 101, 120, and 45 DEPs, respectively. The overlap between these DEP group sets was analyzed with the Venn diagram, which is illustrated in Fig. 2c. Among these 45 DEPs between axPsA and pPsA, 11 proteins were found to be up-regulated, while 34 proteins were down-regulated in axPsA. A volcano plot was used to visualize the changes in protein expression between axPsA and pPsA, as shown in Fig. 2d. Among these proteins, PEDF was found to be the top significantly upregulated protein in axPsA, based on the adjusted P value.

Using the bioinformatics method, we conducted an analysis of the DEPs between axPsA and pPsA. The results of the GO analysis indicated that the DEPs were associated with biological processes related to innate immunity like complement, coagulation, and the regulation of proteolytic activity. Furthermore, these DEPs were involved in the regulation of enzyme activity, which was identified as the most important molecular function. KEGG analysis showed that these DEPs were mainly involved in the complement pathway and hemostasis (Additional file 1: Figure S1).

Identification of candidate biomarkers

To better identify clinically available biomarkers, we employed various approaches to select candidate biomarkers. Boruta analyses confirmed 28 biomarkers that are important for identifying patients. Random Forest (RF) analyses were conducted to objectively evaluate the importance of serum proteins, and the top ten proteins are displayed in Fig.3b, c. Additionally, LASSO regression selected 7 DEPs. Among these candidate biomarkers, PEDF appeared in all lists and ranked first in RF. Therefore, we selected PEDF as a potential marker for further verification. Based on quantitative analysis using mass spectrometry data, we observed a significant upregulation of PEDF in axPsA compared to pPsA. Receiver operating characteristic (ROC) analyses of PEDF yielded an area under the curve (AUC) value of 0.925.

Fig. 3
figure 3

Identification of candidate biomarkers based on MS data from the discovery cohort. a Feature selection based on the Buruta algorithm; Feature ordering based on mean decrease accuracy (b) and mean decrease gini (c) in random forest model. PEDF was the top rank DEP; d quantitative analysis of serum PEDF levels in two groups using mass spectrometry data (****p < 0.001). The intensity of PEDF was normalized. e Receiver operating characteristic curve analysis of candidate biomarkers for axPsA vs. pPsA based on quantitative analysis of mass spectrometry data. PEDF pigment epithelium-derived factor, DEPs differentially expressed proteins, axPsA axial psoriatic arthritis, pPsA peripheral psoriatic arthritis

Full size image

ELISA verification of MS-identified biomarkers

To further investigate the expression of PEDF in the serum of patients with axPsA and pPsA, we detected the level of PEDF in 37 patients with axPsA, and 51 patients with pPsA by ELISA in the verification phase. As shown in Fig. 4, PEDF expression was significantly higher in axPsA compared with pPsA (37.9 ± 10.1 vs. 30.5 ± 8.9 μg/mL, p < 0.001), the AUC score was 0.72 (95%CI 0.61–0.83). PEDF, BMI, and the clinical variables with significant differences in univariate analysis were included in the multivariate analysis. The results showed that PEDF remained significantly elevated in axPsA patients (P = 0.017, Additional file 1: Table S1). There were no significant differences observed in the level of PEDF among axPsA patients with different imaging types (Additional file 1: Figure S2). We performed a correlation analysis between the serum levels of PEDF and major disease manifestations in patients with PsA. The results showed that serum PEDF was positively correlated with BMI (r = 0.4, P < 0.001) and CRP (r = 0.42, P < 0.001). Additionally, a noticeable trend towards a positive correlation between PEDF and Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) was observed (r = 0.36, P = 0.064). There were no significant correlations found between serum PEDF levels and swollen joint count (SJC), TJC, and psoriasis area and severity index (PASI) (Fig. 5).

Fig. 4
figure 4

The level of PEDF in the patients in the verification cohort. a The comparison of serum PEDF between axPsA and pPsA with ELISA. (***p < 0.001). b Receiver operating characteristic curve analysis of candidate biomarkers for axPsA vs. pPsA. PEDF pigment epithelium-derived factor, axPsA axial psoriatic arthritis, pPsA peripheral psoriatic arthritis

Full size image
Fig. 5
figure 5

Correlation analysis between the serum level of PEDF and major disease manifestations. The correlation between BMI (a), CRP (b), ESR (c), SJC (d), TJC (e), PASI (f), BASFI (g), BASDAI (h), and serum PEDF. The data of BASDAI and BASFI was only available in patients with axPsA. BMI body mass index, CRP C reactive protein ESR erythrocyte sedimentation rate, SJC swollen joints count, TJC tender joints count, PASI Psoriasis Area and Severity Index, BASFI Bath AS Functional Index, BASDAI Bath Ankylosing Spondylitis Disease Activity Index

Full size image

Discussion

In recent years, there has been a growing interest in studying axPsA. Several studies have identified distinct characteristics that differentiate axPsA from axSpA, which may delay diagnosis and treatment [20, 21]. While reliable biomarkers can aid clinicians in prescribing more targeted imaging tests and identifying asymptomatic individuals with axial involvement, it is imperative to ensure that these patients receive timely diagnoses and appropriate treatments, such as interleukin-17A inhibitors. These inhibitors have been proven to delay radiographic progression and prevent loss of function [22]. Thus, this study utilized mass spectrometry technology to explore biomarkers capable of distinguishing between axPsA and pPsA and confirmed the dependability of serum PEDF as a potential biomarker.

As we all know, the efficacy of a biomarker is related to the definition of the disease. An important initiative in this field is the Axial Involvement in Psoriatic Arthritis cohort (AXIS) study, which aims to establish classification criteria for axPsA [23]. Considering the objectivity of imaging examinations and the poor sensitivity and specificity of existing IBP criteria in patients with axPsA [24], we refer to the imaging criteria used in previous studies. For the identification of axPsA in our study, we employed imaging criteria from previous research, including the New York criteria for sacroiliitis and/or syndesmophyte of the spine [11]. Given the high accessibility of sacroiliac joint CT in our cohort, a large number of patients had access to this imaging data. Additionally, several studies have shown that CT demonstrates superior diagnostic accuracy for axSpA [25]. Another study comparing magnetic resonance imaging and CT evaluations of axial lesions in the sacroiliac joint also indicated that CT has excellent specificity and good sensitivity [39]. A suite of machine learning techniques, such as logistic regression, random forests, and support vector machines can be used in the identification of a multivariate biomarker panel [40]. Thirdly, the patients included in this study were not treatment-naive, and the therapeutic medications may have influenced the outcomes. However, the unselected patients in this study align more closely with the real clinical environment and are more conducive to clinical applicability.

Conclusions

In conclusion, we utilized mass spectrometry to analyze the serum proteome in patients with axPsA and pPsA, and identified several DEPs between the two groups. AxPsA and pPsA have distinct serum protein profiles that can be used as biomarkers to discriminate between them. Among these proteins, PEDF showed promise as a potential biomarker, and its validity was confirmed using ELISA in a larger verification cohort. However, further validation is still needed in patients from an expanded or independent cohort before it becomes a truly reliable marker for clinical practice. Additionally, considering the clinical heterogeneity and potential comorbidities in patients with PsA, a biomarker panel with multiple proteins may be a more ideal diagnostic tool.