Distinct late-stage osteoarthritis profiles identified through NF-κB, TNF-α, and TGF-β–driven synovial inflammation and pain

Abstract

This study delineates distinct late-stage osteoarthritis (OA) profiles, characterized by NF-κB, TNF-α, and TGF-β–mediated synovial inflammation and pain, in a cohort of 31 patients undergoing joint arthroplasty. Histopathological analysis demonstrated synoviocyte hyperplasia, increased stromal cellularity, inflammatory infiltrates, and enhanced vascularization, findings that correlated with higher synovitis scores and exacerbated movement-associated nociception. Increased expression of NF-κB and TNF-α in the synovial membrane confirmed their association with stronger inflammation and pain perception. Bayesian networks revealed relationships among NF-κB, TNF-α, and pain scores, suggesting that NFκB is a primary driver of pain in low-grade synovitis, while TNF-α becomes more influential in high-grade synovitis. Cluster analysis identified four distinct patient subgroups: (1) males with severe radiographic OA, elevated NF-κB in both the synovium and fluid, and the absence of synovitis; (2) predominantly females with metabolic comorbidities, a variable degree of synovitis, and elevated TNF-α in synovial fluid; (3) patients with low pain scores and no histopathologically confirmed synovial inflammation; and (4) patients with high-grade synovitis, pronounced tissue alterations, increased NF-κB, TNF-α, and TGF-β expression in the synovial membrane, but without metabolic comorbidities. The findings clarify the influence of synovitis in late-stage OA and suggest that the identified profiles may assist in earlier disease stratification and the development of tailored treatments in the future.

Introduction

There is increasing research attention on osteoarthritis (OA), the most prevalent form of arthritis¹. Over time, there has been a significant shift in the interpretation of OA, moving beyond the conventional wear-and-tear paradigm^2,3. OA is now recognized as a complex joint disease, which includes the contribution and crosstalk of degenerative changes and inflammatory processes to disease progression⁴. In this context, elucidating the role of synovial inflammation in the pathogenesis of OA is essential for optimizing disease management – alleviating major symptoms such as pain and delaying progression toward disability⁵.

Previous studies combined clinical, radiological, and laboratory data to stratify OA patients into distinct clinical phenotypes. This distinguishes the following phenotypes: minimal joint disease, bone and cartilage metabolism, chronic pain, mechanical overload, post-traumatic, inflammatory, aging-associated, and metabolic syndrome^4,6,7. Despite these attempts to offer promises of more personalized care, the current clinical phenotypes do not accurately represent the underlying pathophysiological processes of OA, as some cases remain unclassified or exhibit overlapping phenotypes^6,7. Furthermore, researchers found that a single clinical phenotype may result from various molecular endotypes or represent a combination of them⁷. To better systematize the subgroups of OA, explain the symptoms seen in clinical phenotypes, and develop individualized, multimodal treatments, experts emphasize the importance of understanding the disease at the molecular level^4,7,8.

All joint-forming elements, especially the cartilage compartment, contribute to OA pathogenesis. However, synovial inflammation, or synovitis, is an active component of OA that significantly influences symptom severity and disease progression and often precedes radiologically detectable changes in OA-affected joints^9–12. Although synovitis typically presents as low-grade, its severity can vary and may even mimic changes seen in rheumatoid arthritis. This suggest that the expression levels and types of synovitis-associated signaling molecules are not uniform¹³. Studies have shown that synovial inflammatory conditions are involved in generating different OA pain patterns, indicating that diverse molecular pathways coordinate the development of synovitis, leading to varied OA manifestations^10,12,14,15.

Synovitis arises from changes in both layers of the synovial membrane, with alterations in the lining synoviocytes, as well as modifications in vascularization, cellular composition of resident synoviocytes, and immune cell infiltration in the sublining stroma¹⁶. Immune cells, in particular, are a source of various signaling molecules that drive tissue remodeling and inflammatory responses. Among these are transcription factors, such as nuclear factor kappa B (NF-κB); cytokines, like tumor necrosis factor alpha (TNF-α); and growth factors, like transforming growth factor beta (TGF-β), which could serve as potential biomarkers for OA endotyping^2,13.

NF-κB is a transcription factor for numerous pro-inflammatory and catabolic mediators, including interleukin (IL)-1β and IL-6, which play key roles in OA pathogenesis. NF-κB is associated with radiographic OA severity and contributes to synovial hyperplasia and cartilage destruction^17–19. Additionally, NF-κB contributes to pain intensification through its interaction with TNF-α, thereby establishing OA-associated pain^20,21. TNF-α is one of the earliest pro-inflammatory cytokines in the inflammatory cascade and is released abundantly by synoviocytes, especially in the early and progressive stages of OA²². It contributes to synovitis, cartilage degradation, fibrosis and OA-associated pain^15,17,23. TGF-β displays a multifaceted nature in inflammation and tissue remodeling, thus is involved in both protective and pathological processes. At low levels, it promotes tissue maintenance and repair, but at high levels, it drives fibrotic changes in the synovial membrane and may contribute to the formation of osteophytes^24,25. Its involvement in pain mechanisms is also recognized, as TGF-β signaling can lead to changes in tissue stiffness and neuropathic responses^26,27. These signaling molecules pool in the synovial fluid, diffusing from different joint compartments, including synovial membrane, thus adversely affecting the extracellular matrix and cellular components of all adjacent joint tissues, which intensifies disease severity and clinical manifestations²⁸.

We aimed to comprehensively assess the complex expression, distribution, and interrelations of NF-κB, TNF-α, and TGF-β in the synovial membrane in relation to synovial inflammation. We sought to determine whether the immunohistochemical findings in the synovial tissue correspond to the levels of these biomarkers in the synovial fluid and to associate these findings with pain patterns, the major clinical manifestation of late-stage OA. Based on this information, we also explored the potential for identifying different late-stage OA profiles.

Materials and methods

Study population

The study included thirty-one adult patients with clinically and radiologically confirmed late-stage OA of the knee or hip, who underwent either total knee or total hip replacement surgery between January 2023 and December 2023 at the Hospital of Traumatology and Orthopedics. All patients fulfilled predefined inclusion and exclusion criteria, had complete clinical and laboratory data, and had both synovial membrane tissues and synovial fluid samples successfully obtained during surgery. Patients who did not meet the eligibility criteria, had incomplete clinical or laboratory data, or had no synovial tissues or synovial fluid due to intraoperative limitations were excluded from the study to ensure the parameters are consistent and comparable. Inclusion criteria were based on the American College of Rheumatology (ACR) guidelines for hip²⁹ and knee OA³⁰. Patients with clinically confirmed inflammatory arthritides or autoimmune rheumatic disorders were excluded from the cohort. Clinical data collected from the patients and their medical records included information on the duration and clinical characteristics of the disease, as well as with age, sex, body mass index (BMI), hyperlipidemia, and diabetes mellitus. A BMI of 25 kg/m² or higher was considered elevated. The presence or absence of hyperlipidemia and diabetes mellitus was determined by reviewing medical records to confirm the diagnoses or by evaluating laboratory analyses of the lipid profile and glycosylated hemoglobin. These conditions were categorized as either “yes” or “no”. We identified and grouped these characteristics as metabolic comorbidities, which include increased BMI, hyperlipidemia, and diabetes mellitus. Prior to surgery, OA-related pain in the affected joint, both at rest and during movement, was assessed using the pain subscales of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)³¹. Two questions from the rest pain subscale and three questions from the movement pain subscale were analyzed, with each item rated on a scale from 0 (no pain) to 4 (extreme pain). The composite score for rest pain ranged from 0 to 8 and was categorized as follows: 0 (no pain), 1–2 (mild pain), 3–4 (moderate pain), 5–6 (severe pain), and 7–8 (extreme pain). Similarly, the composite score for movement pain ranged from 0 to 12, with categories as follows: 0 (no pain), 1–3 (mild pain), 4–6 (moderate pain), 7–9 (severe pain), and 10–12 (extreme pain). Higher scores in both the rest and movement pain subscales indicated greater severity of OA-related pain. Additionally, subjective pain intensity in the affected joint was assessed using the patient-reported Visual Analogue Scale (VAS), ranging from 0 to 10 points, with categories defined as: 1–2 (mild pain), 3–6 (moderate pain), and 7–10 (severe pain). The radiographic severity of the OA-affected joint prior to surgery was determined according to the Kellgren–Lawrence (K/L) grade (0–none; I–doubtful; II–minimal; III–moderate; IV–severe), assessed by an experienced musculoskeletal radiologist blinded to the patient’s condition. Laboratory analyses to monitor systemic inflammation included measuring serum C-reactive protein (CRP) levels in the blood before surgery, with values less than 5 mg/L considered normal.

Patients were divided into three study groups based on the intensity of histopathologically confirmed synovitis, as described in detail in the Histopathology section.

Ethics, consent and permissions

The study protocol, including the use of patient data, joint tissue material, and synovial fluid, was approved by the Ethical Committee of Rīga Stradiņš University (Decision No. 4/36/2023) and conducted in accordance with the Declaration of Helsinki. The study was carried out in strict compliance with the fundamental principles of scientific research in medicine, and participation was entirely voluntary and free from external constraints. Informed, written, and signed consent was obtained from all patients prior to joint replacement surgery.

Sample collection

Synovial membrane specimens were fixed in 10% formalin, embedded in paraffin, and serially sectioned into 4–5 μm histological slices. These sections were mounted on adhesive microscope slides (Menzel GmbH, Braunschweig, Germany) for subsequent histopathological and immunohistochemical analyses.

Synovial fluid samples were collected via direct aspiration with a syringe and needle at the time of joint arthroplasty. The aspirated fluid was centrifuged at 4000 rpm for 15 min to remove cellular debris, aliquoted into 100 µL vials, and stored at − 20 °C until use in the planned assays.

Histopathology

Tissue sections stained with hematoxylin and eosin were used to evaluate synovitis based on the Krenn and Morawietz scoring system, which assesses hyperplasia of the lining layer, sublining cellularity, and inflammatory infiltrate density¹⁶. Each parameter was graded on a 0–3 scale, with the total synovitis score (0–9). Based on the synovitis grade, patients were divided into three groups as follows: 0–1 (first group: no synovitis), 2–4 (second group: low-grade synovitis), and 5–9 (third group: high-grade synovitis)¹⁶.

Vascularization was assessed by grading the presence of congested blood vessels in the sublining layer on a scale from 0 to 3, defined as follows: 0 (few vessels), 1 (small number of vessels), 2 (moderate number of loosely distributed vessels), and 3 (high number of densely distributed vessels). The presence of giant cells was evaluated in 10 randomly selected fields, and scored as 0 (none), 1 (few, up to 5), or 2 (numerous, > 5).

Immunohistochemistry

For immunoexpression analysis of NF-κB, TNF-α, and TGF-β in the synovial membrane, tissue sections were deparaffinized, rehydrated, and blocked for endogenous peroxidase activity using 30% hydrogen peroxide in methanol. Antigen retrieval was performed with sodium citrate buffer (pH 6.0) at 96 °C, and non-specific binding was blocked using 1% bovine serum albumin in phosphate-buffered saline.

Sections were incubated overnight with primary antibodies against NF-κB p65 subunit (Abcam, Cambridge, UK; dilution 1:250), TNF-α (Abcam, Cambridge, UK; dilution 1:100, clone P/T2), and TGF-β1 (Bio-Rad Laboratories, Hercules, CA, US; dilution 1:2500, clone TB21). Primary antibody amplification was carried out using the HiDef Detection HRP Polymer system (Cell Marque, Rocklin, CA, USA), followed by DAB substrate for antigen visualization (Cell Marque, Rocklin, CA, USA). Sections were counterstained with Mayer’s hematoxylin, dehydrated, mounted, and scanned for further image analysis using Glissando Slide Scanner (Objective Imaging Ltd., Cambridge, UK).

Negative controls were included by omitting the primary antibody. Cells exhibiting brown staining were considered immunopositive, and antigen expression was quantified in 10 randomly selected fields at × 400 magnification by two independent observers blinded to the clinical data.

ELISA assay

Levels of NF-κB1 (Biorbyt Ltd., Cambridge, UK), TNF-α (Thermo Fisher Scientific, Waltham, MA, USA), and TGF-β1 (Biorbyt Ltd., Cambridge, UK) in synovial fluid were quantitatively measured using commercial enzyme-linked immunosorbent assay (ELISA) kits, following the manufacturer’s instructions. The sensitivity and detection ranges for each ELISA kit were as follows: NF-κB, sensitivity < 0.063 ng/mL, detection range 0.156 ng/mL–10 ng/mL; TNF-α, sensitivity < 2.3 pg/mL, detection range 7.8 pg/mL–500 pg/mL; TGF-β, sensitivity < 1 pg/mL, detection range 15.6 pg/mL–1.000 pg/mL). Due to the low volume of synovial fluid in some samples, each molecule was replicated in a single assay.

Statistical analysis

Statistical analyses were performed using JMP Pro 17 (SAS, Cary, NC, USA), JASP 0.16.1 (JASP Team, 2022), and GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). Sample size calculations were based on Cohen’s methodology for detecting large effect sizes. Descriptive statistics were calculated for patient demographics and baseline characteristics. Categorical outcomes were analyzed using Fisher’s exact test (F) when chi-square test assumptions were violated (expected cell counts < 5). This non-parametric test was chosen to ensure accurate p-values given our small sample size of 31 patients distributed across three groups. The D’Agostino–Pearson and Shapiro–Wilk tests were used to assess the normality of the numerical data. One-way ANOVA (A) or the nonparametric Kruskal-Wallis (KW) test (when the data did not meet normal distribution criteria) with a two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli as post-hoc tests were applied. To explore relationships between variables, Spearman’s rank correlation was used, with correlation strength classified as weak (0.2 to 0.4), moderate (0.4 to 0.7), and strong (0.7 to 0.9). Statistical significance was set at p < 0.05.

Results

Clinical profile of the patient cohort

A total of 31 patients (6 males [19.4%] and 25 females [80.6%]; mean age 67.7 years, range 48–83 years, standard deviation ± 8.31 years) with clinically, radiologically, and histologically confirmed late-stage OA were enrolled in the study. Radiologically, the patients in the cohort presented with moderate to severe late-stage OA; among them, 42% and 58% were diagnosed with K/L grade III and IV, respectively. Of the cohort, 19 patients were diagnosed with knee OA, and 12 with hip OA. The study groups were defined based on histopathological assessment as follows: no synovitis, 10 patients (32.3%); low-grade synovitis, 12 patients (38.7%); and high-grade synovitis, 9 patients (29.0%). Patient characteristics and clinical profiles across the study groups are summarized in Table 1. No significant differences in BMI, obesity, hyperlipidemia, CRP levels, or the presence of concomitant diabetes were observed between the groups. The duration of OA-related symptoms prior to joint replacement surgery was comparable between the groups, although a trend towards a shorter duration was observed in the high-grade synovitis group.

Table 1.

Comparison of patients’ characteristics, clinical and laboratory data across no-synovitis, low-grade synovitis and high-grade synovitis groups. Abbreviations: K/L–Kellgren-Lawrence grade, BMI–body mass index, CRP–C-reactive protein, VAS–visual analogue scale, WOMAC–Western Ontario and McMaster Universities Osteoarthritis Index, p_A–p-values from one-way ANOVA (A) test, p_F–p-values from Fisher’s exact test, p_KW–p-values from Kruskal-Wallis test.

Patients’ characteristics	No synovitis, n = 10	Low-grade synovitis, n = 12	High-grade synovitis, n = 9	p-value
Age, years, mean (SD)	62.10 (10.13)	69.67 (6.25)	71.22 (5.51)	pA=0.027
Sex Male, n (%) Female, n (%)	2 (20.0) 8 (80.0)	3 (25.0) 9 (75.0)	1 (11.1) 8 (88.9)	pF=0.855
Affected joint Knee, n (%) Hip, n (%)	7 (70.0) 3 (30.0)	8 (66.7) 4 (33.3)	4 (44.5) 5 (55.5)	pF=0.583
Radiographic OA severity K/L 0, n (%) K/L I, n (%) K/L II, n (%) K/L III, n (%) K/L IV, n (%)	0 0 0 6 (60.0) 4 (40.0)	0 0 0 5 (41.7) 7 (58.3)	0 0 0 2 (22.2) 7 (22.6)	pKW=0.261
BMI, kg/m2, median (IQR)	34.95 (30.10-36.55)	33.00 (28.38–36.52)	26.20 (22.50–36.30)	pKW=0.081
Obesity (BMI ≥ 30.0), n (%)	7 (70.0)	8 (66.7)	3 (33.3)	pF=0.205
Hyperlipidemia, n (%)	5 (50.0)	9 (75.0)	2 (22.2)	pF=0.064
CRP level prior to surgery, mg/dL, median (IQR)	3.30 (2.30–5.42)	2.05 (1.00-4.03)	1.70 (1.10–10.30)	pKW=0.452
Diabetes mellitus, n (%)	5 (50.0)	6 (50.0)	2 (22.2)	pF=0.430
Duration of the disease symptoms, median (IQR)	60.00 (30.00-117.00)	66.00 (33.00-120.00)	48.00 (36.00–60.00)	pKW=0.576
VAS pain score, median (IQR)	5 (4.25–6.75)	6 (4.75-8.00)	5 (5.00–7.00)	pKW=0.847
WOMAC movement pain, median (IQR)	6 (5.25-7.00)	8 (6.00–9.00)	8 (8.00–9.00)	pKW=0.125
WOMAC rest pain, median (IQR)	4 (4.00–4.00)	3.50 (2.75-5.00)	4 (3.00–4.00)	pKW=0.747

Patient-reported VAS pain scores indicated moderate to severe pain across all groups. Although higher WOMAC movement pain scores were observed in the low- and high-grade synovitis groups, suggesting more severe pain compared to the moderate pain reported in the no-synovitis group, this difference was subtle. Similarly, WOMAC rest pain scores were comparable across all groups, consistently indicating moderate pain.

Synovial tissue histopathology across study groups

Patients with morphologically detectable low- and high-grade synovitis demonstrated synoviocyte hyperplasia in the lining layer compared to the no-synovitis group (p = 0.0074 and p = 0.0063, respectively). The cellularity of the sublining layer was not elevated in the no-synovitis group but increased with the progression of synovitis in the low- and high-grade synovitis groups (p = 0.0104 and p = 0.0021, respectively). The density of inflammatory infiltrates ranged from none in the no-synovitis group to few, mostly perivascular lymphocytes in the low-grade synovitis group, and to numerous inflammatory cells, forming follicle-like aggregates in the high-grade synovitis group (p = 0.0356 and p < 0.0004, respectively). Vascularization grade tended to increase with synovitis grade (Fig. 1a–c), with few vessels in the no-synovitis group, small to moderate numbers in the low-grade synovitis group, and a moderate increase in the high-grade synovitis group. A significant difference in vascularization grade was found between the no-synovitis group and the high-grade synovitis group, while no significant difference was observed between the no-synovitis and low-grade synovitis groups (p = 0.0201 and p = 0.2164, respectively) (Fig. 1d). No giant cells were observed in the no-synovitis group, whereas numerous giant cells were present in the high-grade synovitis group (p = 0.0001).

Fig. 1.

Representative histological images illustrating the differential features of synovial tissue obtained from OA patients exhibiting no synovitis, low-grade synovitis, and high-grade synovitis. (a) Synovial membrane from the no-synovitis group exhibits a non-thickened lining layer with no evidence of increased cellularity. The sublining layer appears loosely organized with sparse vascularization. (b) In the low-grade synovitis group, the synovial lining is moderately thickened, accompanied by increased sublining cellularity. The presence of inflammatory cell infiltrates and dilated, congested capillaries is evident within the sublining stroma. (c) Synovial tissue from the high-grade synovitis group reveals a markedly thickened lining layer composed of up to seven cell layers. The sublining region is densely cellular, with prominent vascular congestion and numerous capillaries surrounded by aggregates of inflammatory and immune cells, predominantly lymphocytes and plasma cells, arranged in follicle-like structures. (d) Violin plots illustrate the extent of synovial vascularization across the three groups. The accompanying table provides detailed descriptive statistics for each group.

Relation of inflammatory infiltrates with pain assessment questionnaires

No association was observed between subjective VAS pain scores and the density of inflammatory infiltrates in the no-synovitis group, as some patients reported severe pain levels despite the absence of inflammatory infiltrates. Additionally, increasing inflammatory infiltrate density had no significant impact on VAS pain scores in either the low-grade or high-grade synovitis groups, with VAS pain scores remaining relatively stable regardless of infiltrate density (Fig. 2).

Fig. 2.

Relationships between the density of inflammatory infiltrates and VAS, WOMAC at rest, and WOMAC during movement scores within groups with no synovitis (1), low-grade synovitis (2), and high-grade synovitis (3). Blue lines between boxes represent correlations between the median values of these parameters. Abbreviations: VAS–visual analogue scale, WOMAC–Western Ontario and McMaster Universities Osteoarthritis Index.

No association was found between WOMAC rest pain scores and the density of inflammatory infiltrates in the no-synovitis and low-grade synovitis groups. However, in the high-grade synovitis group, there was a tendency toward higher WOMAC rest pain scores corresponding with an increased density of inflammatory infiltrates. Similarly, WOMAC movement pain scores tended to increase in relation to the density of inflammatory infiltrates across all groups (Fig. 2).

Immunohistochemical and correlation analysis of NF-κB, TNF-α, and TGF-β expression in the Synovial membrane across synovitis grades

Expression analysis of NF-κB, TNF-α, and TGF-β within the synovial membrane was performed to gain deeper insights into synovitis. Furthermore, correlation analyses were conducted to explore associations between biomarker expressions and to identify potential dominant molecular pathways across different synovitis grades.

Immunohistochemical assessment of NF-κB, TNF-α, and TGF-β expression in the synovial membrane

We assessed the immunolocalization of key biomarkers in both compartments of the synovial membrane: the lining and sublining layers (Fig. 3).

Fig. 3.

One-way analysis shows the median values of NF-κB, TNF-α, and TGF-β immunolabeled cells in the lining layer in the left column, while the right column shows the median values of these markers in the sublining layer within groups with no synovitis (1), low-grade synovitis (2), and high-grade synovitis (3). The diamond plot represents the group mean of each variable with a 95% confidence interval, while the box plot shows the median with the interquartile range. Red numbers indicate significant p-values (pKWChi–comparison of numeric ordinal data using the Kruskal-Wallis test with Chisquare approximation). The right side of each graph displays the normal quartile values regression line, evaluating a single factor. Abbreviations: NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta.

A significant overexpression of NF-κB was observed in the lining layer of the low-grade synovitis group compared to the no-synovitis group (p = 0.0177). In the sublining layer, NF-κB immunopositivity was significantly higher in both the low-grade and high-grade synovitis groups compared to the no-synovitis group (p = 0.0192 and p = 0.0305, respectively). NF-κB expression was significantly higher in the sublining layer compared to the lining layer in the high-grade synovitis group, while no significant difference was observed in the no-synovitis and low-grade synovitis groups (p = 0.0087, p = 0.2336, and p = 0.2336, respectively) (Fig. 4).

Fig. 4.

Violin plots illustrate the comparative expression of NF-κB, TNF-α, and TGF-β immunopositive cells within the synovial lining and sublining layers among patients with no synovitis, low-grade synovitis, and high-grade synovitis. The plots reflect the distribution and intensity of immunolabelling across the groups. Descriptive statistics corresponding to each parameter are provided in the tables below.

TNF-α immunolabeling in the lining layer ranged from weak to moderate, with a tendency toward higher expression in the high-grade synovitis group compared to the other two groups. In the sublining layer, TNF-α expression remained similar across all groups. A trend toward higher TNF-α expression in the sublining layer compared to the lining layer was observed in the low-grade synovitis group (p = 0.3745). In contrast, no such trend was observed in the no-synovitis and high-grade synovitis groups (p = 0.6535 and p = 0.6535, respectively) (Fig. 4).

Overall, TGF-β expression ranged from weak to moderate. In the lining layer, there was a tendency toward higher TGF-β expression in the high-grade synovitis group compared to the low-grade and no-synovitis groups, though the difference was subtle. TGF-β expression in the sublining layer was consistent across all groups. There was no significant difference in TGF-β expression in the lining layer compared to the sublining layer in the no-synovitis, low-grade synovitis and high-grade synovitis groups (p = 0.8385, p = 0.8385, and p = 0.8385, respectively) (Fig. 4).

In the no-synovitis group, the synovial membrane demonstrated a single layer of cells in the lining, with occasional synoviocytes displaying NF-κB and TGF-β immunopositivity (Fig. 5). No increase in total cellularity or inflammatory infiltrates was observed in the sublining layer, though some perivascular NF-κB immunopositive cells and diffusely localized TGFβ immunopositive cells were present. Additionally, a few cells in the vascular walls of the sublining layer demonstrated TGF-β expression. Overall, TNF-α immunopositivity was weak in both layers.

Fig. 5.

Representative microphotographs of synovial tissue stained for NF-κB, TNF-α, and TGF-β in patient groups classified by synovitis grade. Panels depict immunostaining results in synovium from patients without synovitis (1), with low-grade synovitis (2), and with high-grade synovitis (3) (scale bars 100 μm). Abbreviations: NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta.

In the low-grade synovitis group, the lining layer was mildly thickened with abundant NFκB immunolabeled synoviocytes, along with fewer TGF-β and some TNF-α positive synoviocytes (Fig. 5). A slight increase in cellularity was noted in the sublining layer, with numerous diffusely localized NF-κB positive cells and some cells showing TNFα and TGF-β expression. A slight to moderate increase in vascular beds within the sublining layer was observed, with occasional TGF-β expression.

In the high-grade synovitis group, the lining layer was moderately to greatly thickened and contained numerous NF-κB, TNF-α, and some TGF-β positive synoviocytes (Fig. 5). The sublining layer showed moderate increases in cellularity, with inflammatory infiltrates and follicle-like aggregates. Abundant NF-κB immunopositive cells were diffusely distributed and present within the follicle-like aggregates in the sublining layer. Numerous diffusely localized TNF-α-positive cells were also observed, along with moderately increased vascularization.

Correlation analysis of NF-κB, TNF-α, and TGF-β expression in synovial membrane

The comparison of the three study groups revealed common correlations among them (Fig. 6). In all groups, NF-κB expression in the lining layer synoviocytes demonstrated moderate-to-strong positive correlation with NF-κB positive cells located diffusely in the sublining. A similar correlation pattern was observed for TGF-β expression in the lining layer synoviocytes and TGF-β positive cells located diffusely in the sublining.

Fig. 6.

Correlation cluster matrices of the main markers of all three groups. Graphs visualize the co-expression of immunohistochemically determined markers of synovial membrane tissue, where the direction and level of association of the variables are presented according to the color scale from blue to red. The tables show the variables, the correlation coefficients of their most significant associations with the corresponding p-values. Abbreviations: NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta, lin–lining layer, sublin–sublining layer, ρ–Spearman’s correlation, p–p-value.

Distinct correlations were found that specifically characterize the low- and high-grade synovitis groups. In the low-grade and high-grade synovitis groups, compared to the no-synovitis group, NF-κB-positive cells located diffusely in the sublining showed a moderate-to-strong positive correlation with NF-κB-positive cells in the walls of blood vessels. A similar correlation was observed for TNF-α expression in the perivascular region of the sublining layer and TNF-α positive cells in the vascular bed.

We also found correlations that are distinct for the low- and high-grade synovitis groups. In the low-grade synovitis group, a moderate positive correlation between NF-κB and TNFα expression in the lining synoviocytes was observed. In the high-grade synovitis group, a strong positive correlation between NF-κB and TNF-α-positive cells located diffusely in the sublining was identified. Furthermore, in the same group, a strong positive correlation was found between TNF-α positive cells in the perivascular region and those located diffusely in the sublining.

Analysis of NF-κB, TNF-α, and TGF-β levels in synovial fluid across all study groups

Assessment of NF-κB and TNF-α expression in synovial fluid revealed detectable levels in all patients across the study groups. However, measurable levels of TGF-β in synovial fluid were found in only 30% of patients in the no-synovitis group (n = 3), 25% in the low-grade synovitis group (n = 3), and 44% in the high-grade synovitis group (n = 4). Subsequent analyses of biomarker expression in synovial fluid included all patients for NFκB and TNF-α levels, while only those with detectable TGF-β levels were included in the TGFβ analysis, in accordance with the detection range (Fig. 7).

Fig. 7.

Distribution of NF-κB, TNF-α and TGF-β levels in synovial fluid in groups without synovitis, with low-grade synovitis and high-grade synovitis; presented as a bar plots. Abbreviations: NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β– transforming growth factor beta.

In the no-synovitis group, the median levels of synovial fluid biomarkers were 2.7285 ng/mL (IQR 2.4470–4.6662) for NF-κB, 3.6160 pg/mL (IQR 3.5234–3.7395) for TNFα, and 297.7500 pg/mL (IQR 207.9000–460.0850) for TGF-β.

In the low-grade synovitis group, median biomarker levels showed an increase, with NF-κB at 3.3005 ng/mL (IQR 2.6618–4.6448) and TNF-α at 4.0667 pg/mL (IQR 3.5675–4.3192), while TGF-β levels were considerably lower, at 134.6200 pg/mL (IQR 80.0350–151.0400).

The high-grade synovitis group presented median synovial fluid levels of 2.7820 ng/mL (IQR 1.2680–3.3920) for NFκB, 3.7925 pg/mL (IQR 3.3694–4.0931) for TNF-α, and 1902.10 pg/mL (IQR 1320.7125–2241.8975) for TGF-β. TGF-β levels in this group showed a substantial increase compared to both the no-synovitis and low-grade synovitis groups, consistent with the higher synovitis grade.

Bayesian correlation networks for comparative analysis of models including pain scores and key biomarker expression in synovial fluid and membrane across study groups

Bayesian correlation network analysis was used to compare the relationships between pain assessmentsand the expression of NF-κB, TNF-α, and TGF-β in both synovial fluid and synovial membrane across study groups.

Bayesian model for correlating pain assessments and key signaling molecules detected in synovial fluid

In the no-synovitis group, Bayesian network analysis identified a strong probabilistic dependency between VAS pain score and both the WOMAC rest pain and WOMAC movement pain indexes, although only the latter showed a direct conditional relationship with NF-κB levels in the synovial fluid (Fig. 8).

Fig. 8.

Comparative Bayesian correlation networks were constructed for no-synovitis (1), low-grade synovitis (2), and high-grade (3) synovitis groups, focusing on two variable categories: pain assessments and levels of key biomarkers detected in synovial fluid via ELISA. The network visually represents the presence of relationships (indicated by lines), with blue lines showing positive associations and red lines showing negative associations. The thickness of the lines reflects the strength of the relationship between variables. Number of non-zero edges represents the count of edges in the network that have a value greater than zero, indicating the presence of a connection or relationship between nodes. Sparsity highlights key relationships in the network, with higher values indicating a greater focus on these by reducing the number of active connections. Abbreviations: VAS–visual analogue scale, WOMAC–Western Ontario and McMaster Universities Osteoarthritis Index, NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta.

In the second and third groups, the network structure revealed dependencies between WOMAC rest and movement pain, as well as between VAS and both NF-κB and TNF-α levels in the synovial fluid (Fig. 8). Compared to the first group, the probabilistic relationship between VAS and WOMAC rest pain was weaker in the second group and absent in the third. Conversely, the dependency between WOMAC movement pain and TNFα levels in the synovial fluid strengthened in the third group. Interestingly, a negative conditional relationship between NF-κB and TNF-α in the synovial fluid was observed in the third group. Unlike the first group, the Bayesian network analysis did not reveal any dependency between VAS and WOMAC movement pain in either the second or third groups.

Bayesian model for correlating pain scoring systems and key signaling molecules detected in synovial fluid and synovial membrane

In the no-synovitis group, Bayesian analysis incorporating multiple parameters revealed a direct probabilistic dependency between WOMAC rest pain and elevated TNF-α levels in the synovial fluid (Fig. 9). In this group, no relationship was observed between total expression of NF-κB in the synovial membrane and NF-κB in the synovial fluid, a connection that was present in both the second and third groups.

Fig. 9.

Comparative Bayesian correlation networks were constructed for the no-synovitis (1), low-grade synovitis (2), and high-grade synovitis (3) groups, focusing on three variable categories: pain characteristics; total immunopositivity of NF-κB, TGF-β, and TNF-α in the synovial membrane; and the levels of these key biomarkers detected in synovial fluid via ELISA. The network visually represents the presence of relationships (indicated by lines), with blue lines showing positive associations and red lines showing negative associations. The thickness of the lines reflects the strength of the relationship between variables. Number of non-zero edges represents the count of edges in the network that have a value greater than zero, indicating the presence of a connection or relationship between nodes. Sparsity highlights key relationships in the network, with higher values indicating a greater focus on these by reducing the number of active connections. Abbreviations: VAS–visual analogue scale, WOMAC–Western Ontario and McMaster Universities Osteoarthritis Index, NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta.

In the second group, the Bayesian network analysis identified a continued dependency between VAS and WOMAC rest pain, which was absent in the third group. However, the dependency between VAS and WOMAC movement pain was absent in both the second and third groups. A dependency between WOMAC rest pain and WOMAC movement pain was observed in both groups. Additionally, in the second group, the network model revealed a conditional relationship between elevated NF-κB and TNF-α levels and subjective pain scores measured by VAS, as increased inflammation appeared to intensify pain perception. NF-κB levels were also conditionally associated with both total expression of NF-κB and TGFβ in the synovial membrane (Fig. 9).

In the third group, dependencies were similar to those in the second. Additionally, a negative conditional relationship between TNF-α and NF-κB levels in the synovial fluid was detected, while the association between NF-κB levels in the synovial fluid and total expression of NF-κB in the synovial tissues disappeared entirely (Fig. 9).

Characterization of Late-Stage osteoarthritis profiles through cluster analysis of study participants

Hierarchical clustering analysis of the summarized clinical data, morphological data, key biomarker expression detected in the synovial membrane, and their levels in the synovial fluid identified three patient clusters, one of which revealed two sub-clusters (Fig. 10). Based on these findings, four distinct late-stage OA patient subgroups were identified. The synovitis grades were not uniform across three clusters, showing homogeneity only the high-grade synovitis group.

Fig. 10.

A dendrogram visually represents hierarchical clustering, illustrating relationships within data sets. It consists of stacked branches (clades) that progressively divide into smaller branches. At the lowest level, individual elements are shown, and as one moves upward, these elements are grouped based on attributes, forming clusters that become progressively fewer. The terminal points of each clade, referred to as leaves, represent the actual data points. In this context, the data sets encompass 24 parameters analyzed from 31 patients. Variables are color-scaled from blue to red, representing values ranging from the lowest to the highest. Patients are categorized into four subgroups based on differences in clinical measurements, pain assessments, immunostaining of synovial tissues, and synovial fluid analysis. The dendrogram’s constellation plot, located in the upper-right part of the figure, highlights these four major data clusters. Abbreviations: BMI–body mass index, CRP–C-reactive protein, VAS–visual analogue scale, WOMAC–Western Ontario and McMaster Universities Osteoarthritis Index, NF-κB–nuclear factor kappa B, TNF-α–tumor necrosis factor alpha, TGF-β–transforming growth factor beta, SF–synovial fluid, SM–synovial membrane.

In the first group, the smallest sub-cluster (blue cluster) included only male patients with an operated hip joint and a short duration of symptoms. All patients had a severe K/L grade, normal CRP levels, and higher VAS pain scores. NF-κB expression and levels were generally elevated in both the synovial membrane and fluid, despite a low to absent synovitis grade and a low density of inflammatory infiltrates in tissues. TGF-β expression in the synovial membrane was high, although fluid levels of TGF-β remained low. The second sub-cluster of the first group (green cluster) primarily comprised female patients with synovitis scores ranging from no-grade to high-grade. In this sub-cluster, increased self-reported pain scores were associated with a severe K/L grade, higher BMI, and a tendency toward diabetes, hyperlipidemia, and elevated serum CRP. On the cellular level, more pronounced hyperplasia, increased vascularization, and greater cellularity were observed. Some patients also showed elevated TNF-α and NF-κB levels in the synovial fluid.

The next patient group (red cluster) predominantly included those without synovitis and had a relatively higher proportion of individuals with hyperlipidemia and diabetes. This group was characterized by the shortest duration of symptoms, lower K/L grade, and lower pain assessment scores. Some patients had slightly elevated TGF-β levels in the synovial fluid without corresponding elevated expression in the tissue.

The final group (orange cluster) included patients with high-grade synovitis and notably altered morphological parameters in the synovial tissues, such as increased overall cellularity in both lining and sublining layers, increased vascularization, a higher density of inflammatory infiltrates, and a greater presence of giant cells. This group was distinguished by the oldest age, lower BMI, a tend toward a higher K/L grade, and elevated pain scores, especially WOMAC-assessed rest and movement pain. Notably, no patients in this group had diabetes, and there were fewer individuals with hyperlipidemia. This group also had slightly elevated TGF-β expression in the tissue, with markedly increased TGF-β and relatively high TNF-α levels in the synovial fluid, as well as NF-κB expression in the synovial membrane for some patients.

Discussion

This study highlights the dynamic role of inflammation-associated markers and their effect on pain intensity in late-stage OA across three patient groups: no-synovitis, low-grade synovitis, and high-grade synovitis. First, clinical, morphological, and molecular parameters were assessed using bivariate statistics. Second, model-based analyses were conducted for each parameter, incorporating co-factor effects to simulate the natural course of OA. Bayesian analysis highlighted the relationships between inflammation-associated biomarker expression in the synovium, their levels in synovial fluid, and pain perception across synovitis grades, while hierarchical clustering identified distinct late-stage OA profiles.

In the no-synovitis group, the synovium demonstrated no significant morphological changes. However, TGF-β and NF-κB expression in the lining correlated with diffuse positive cells in the sublining, suggesting coordinated expression of these markers within non-inflamed synovium. Studies indicate that local expression of TGF-β and NF-κB may result from excessive mechanical loading and the release of damage-associated molecular patterns (DAMPs), which stimulate synoviocytes to express these markers even in the absence of overt inflammation^32–35. Given TGF-β’s diverse functions, its localized expression may facilitate an adaptive mechanism to tissue injury²⁷. An interesting finding was the dissociation between NF-κB expression in the synovium and its levels in the synovial fluid. It could reflect local low-level inflammatory signaling driven by tissue stress response, possibly due to an imbalance between reactive oxygen species and antioxidants or altered cellular metabolism^35–37. However, it is important to note that NF-κB is not an extracellularly secreted molecule, and to the best of our knowledge, there is limited evidence-based data regarding the direct detection of NF-κB in synovial fluid using ELISA. Instead, it is mainly detected in tissue homogenates³⁸. Nonetheless, we hypothesize that the positive ELISA signal may reflect cell-free NF-κB1 released either via extracellular vesicles remaining in the supernatant post-centrifugation or from apoptotic or damaged cells within the inflamed joint environment. NF-κB is a key transcription factor, which is involved in the regulation of inflammatory processes. Upon activation, NF-κB dimers, the most common p65/p50, translocate to the nucleus and activate pro-inflammatory genes. Therefore, when studying inflammatory activation in synovial tissues, it is most appropriate to focus on p65, which contains the transactivation domain and thus directly reflects the active signaling pathway^18,36. In contrast, NF-κB1 lacks a transactivation domain and cannot initiate transcription independently³⁹. Simultaneously, p65 is less stable outside the cell and may be difficult to detect in biological fluid samples, including synovial fluid. Therefore, ELISA kits are more commonly developed against NF-κB1, as it is more abundant and represents a more reproducible biomarker in biological fluids. In our study, the NF-κB p65 subunit was analyzed in synovial tissues as a marker of the transcriptionally active form, indicating activation of inflammatory signaling within cell nuclei. Meanwhile, the levels of NF-κB1 in synovial fluid potentially reflect a combination of regulatory buffering, degradation and passive cellular release, thus working as an indirect indicator of inflammatory activity^35,40. We also observed weak TNF-α immunopositivity in both layers of the synovial membrane, which agrees with our finding of a slight local inflammatory state³. The lack of observable morphological synovitis suggests that the moderate subjective pain may originate from factors beyond inflammatory infiltrates in the synovium. This is consistent with studies highlighting the multifactorial nature of OA pain, including sensitization, mechanical alterations, psychological factors, biochemical changes, and individual variability in pain thresholds^14,41,42. Interestingly, a link was found between NF-κB in synovial fluid and movement pain, probably induced by mechanical stress. Furthermore, a probabilistic dependency was observed between rest pain and TNF-α in synovial fluid, despite the overall low local inflammation (Fig. 9). It aligns with previous research identifying TNF-α as a rest pain amplifier in OA. It suggests that TNF-α may play a role in pain mechanisms, including co-induction of fibrosis, sensitization of pain receptors, and effects on cartilage and subchondral bone remodeling^15,17,21,23.

Patients with low-grade synovitis showed notable hyperplasia of lining synoviocytes compared to the no-synovitis group, along with mild increases in cellularity and a higher number of perivascular immune cells in the sublining. Despite these morphological changes, we didn’t find an association between the density of inflammatory infiltrates and rest pain. This suggests that rest pain in patients with low-grade synovitis may not depend solely on the inflammatory microenvironment but may also be associated with the proliferation of synoviocytes, which promotes fibrosis and thereby intensifies rest pain^14,43. In contrast, movement pain intensified with advanced synovitis, increased follicle-like aggregates, and elevated TNF-α in synovial fluid, highlighting the role of inflammation in exacerbating movement-related pain. This association was also observed in the high-grade synovitis group. Hattori et al. similarly found that movement pain is linked to OA-related synovitis, basing their findings on MRI-detected joint inflammation¹². NF-κB was the predominant marker in this group, consistent with the report by Ostojic et al., who emphasized NF-κB’s central role in local immune response at this stage of synovitis in patients with both early and late-stage OA¹⁸. NF-κB showed a positive correlation between diffuse cells in the sublining and those in the vascular walls. This also points to the involvement of NF-κB in vascular remodeling^18,44. Furthermore, NF-κB and TNF-α expression in lining synoviocytes showed positive correlation, indicating a pronounced interaction between these two pro-inflammatory markers in low-grade synovitis. It is worth mentioning that in high-grade synovitis, NF-κB and TNF-α-positive cells demonstrated a positive correlation in the sublining. This aligns with the opinion that the coupled action of these markers leads to a robust immune response as inflammation progresses²¹. Bayesian modeling revealed a conditional relationship between elevated levels of NF-κB and TNF-α in synovial fluid and subjective pain, with the link’s strengthening as synovitis progressed. This supports the current view that inflammation amplifies overall pain perception⁴⁵. Overall, these findings suggest that although morphological changes and specific pro-inflammatory markers play crucial roles in low-grade synovitis, their contributions to rest, movement, and overall subjective pain intensity differ^15,32.

The high-grade synovitis group exhibited numerous giant cells, moderately increased vascularization, and dense inflammatory infiltrates forming follicle-like aggregates. A tendency toward higher rest pain was observed in this group compared to the other two groups, supporting our aforementioned finding that a more nuanced relationship exists between rest pain and inflammation. Histologically, this group exhibited an abundance of NF-κB, TNF-α, and some TGF-β-positive synoviocytes in a markedly thickened lining, while the moderately hypercellular sublining demonstrated diffuse NF-κB and TNF-α positivity, particularly within follicle-like aggregates, indicating a robust inflammatory response. Interestingly, a significantly higher NF-κB expression in synovial sublining layer compared to the lining layer was observed. The finding consistent with a previous study revealing elevated NF-κB expression in the synovial sublining layer in late-stage OA¹⁸. In comparison to the previous groups, increased TGF-β expression in the synovium was accompanied by elevated marker levels in the fluid. Consistent with the previous findings, TGF-β expression contributes to fibrosis, chronic synovial thickening, and pain exacerbation under conditions of pronounced synovitis^24,26,27. Interestingly, although high levels of NF-κB were detected in the synovial fluid, a negative conditional relationship between TNF-α and NFκB levels was identified. This finding might be attributable to prolonged, chronic inflammation, the upregulation of regulatory mechanisms, and the expression of specific proteins that interfere with its further activation, as studied by Yun et al.⁴⁶. Additionally, it may imply distinct compartmental origins of the two markers^23,32,40. Our findings highlight potential differences in the regulation of inflammatory pathways as synovitis progresses, along with distinct activation and interaction mechanisms of studied markers across tissue compartments and variations in immune cell dynamics^32,47.

The hierarchical clustering analysis identified four distinct subgroups of patients, which did not correspond to the three-group classification based on synovitis grades described earlier. Our results suggest that OA affects the entire organism, and that synovitis is not a uniform condition across all subgroups. Likely, it represents a spectrum with varying molecular mechanisms and clinical outcomes.

The first cluster revealed two sub-clusters with different clinical and molecular characteristics. The smaller sub-cluster (first subgroup) pooled exclusively male patients with hip joint arthroplasty and a relatively short duration of disease manifestations. Despite the recent appearance of symptoms, these patients showed advanced K/L grade and reported higher subjective pain. This may be explained by prolonged compensatory anabolic pathways in man that lead to the belated seek for healthcare⁴⁸. Elevated NFκB expression was observed in both the synovium and fluid, even in the absence of notable synovitis. This finding may represent NF-κB involvement in destructive processes that matches its alternative function beyond inflammation – promoting the release of catabolic factors^17,32.

In contrast, the second sub-cluster (second subgroup) included mainly female patients with a synovitis severity ranging from no-grade to high-grade. This variability in synovitis was coupled with increased pain intensity, severe K/L grade, greater BMI, hyperlipidemia, diabetes, and higher CRP levels in their blood. The increased proportion of metabolic conditions in these patients links metabolic status with higher synovitis grade⁴⁹. Increased CRP levels are consistent with findings that connect higher BMI with worse OA-related pain⁵⁰. Elevated TNF-α levels in synovial fluid were observed. It aligns with the findings of Agrawal et al., which reported TNF-α induction by leptin, an adipokine produced by adipose tissues that is elevated in chronic inflammatory conditions⁵¹. This also aligns with previous studies highlighting the influence of systemic metabolic disorders on changes in the cellular composition and structure of the synovium^37,49,52. The observed increase in pain intensity suggests that, for these patients, it may be driven by metabolic disorders and systemic and local inflammation⁵³. This subgroup emphasizes the link between local and systemic inflammatory responses in modulating synovitis, as well as the need to consider systemic metabolic factors when addressing pain management in OA.

The third cluster grouped patients without synovitis but with varying levels of pro-inflammatory markers as well as a predominance of metabolic disorders. This subgroup also showed the lowest K/L and pain scores among all subgroups, despite having the longest duration of symptoms. Overall, this group more likely represents an indolent form of late-stage OA, characterized by a mix of presumably metabolically driven, mild inflammatory responses without local synovitis, along with structural and degenerative alterations in the joint. This group may fit in the biomarker-based, low-tissue-turnover OA endotype reported by Angelini et al.⁵⁴.

High-grade synovitis represented the fourth subgroup, displaying the most pronounced inflammatory alterations. The expression patterns of markers in this group were similar to those observed in the second cluster. Despite this, patients had a lower BMI and no evidence of metabolic comorbidities. These findings suggest a different underlying mechanism driving inflammation compared to the aforementioned late-stage OA patient clusters^54,55. Patients also experienced intense pain, likely due to high local inflammation grades accompanied by concomitant expression of NF-κB, TNF-α, and TGF-β^15,18,26. Interestingly, the group was characterized by the highest mean age, consistent with the findings that aging is associated with synovial inflammation and altered pain-related behavior⁵⁶.

This study has some limitations, primarily related to the relatively small sample size. However, this was addressed by optimizing the study design and group allocation based on pilot data, ensuring sufficient power to detect clinically relevant differences. NF-κB levels detected in synovial fluid by ELISA reflect the joint’s inflammatory status, though this method does not capture its transcriptional or functional activity. Rather, it provides an indirect indication of extracellular signaling or cellular turnover associated with inflammation. Additionally, in the present study, the identification of cellular constituents within the synovial membrane was based on their histological features and spatial distribution, with a particular focus on the quantitative assessment of NF-κB, TNF-α, and TGF-β expression within anatomically distinct layers of the synovial membrane. While our results do not allow for direct attribution of marker expression to individual cell types, the distinct layer-specific expression patterns, together with the spatial arrangement of cells, provide important insights into the inflammatory microenvironment of the synovial membrane in OA.

Conclusions

Here, we reported whether and how inflammatory markers – specifically, NF-κB, TNFα, and TGF-β – play a role in subjective pain intensity associated with movement in late-stage OA patients at various stages of synovitis. The movement pain showed the most direct link with inflammation and associated markers. NF-κB plays a substantial role in movement-related pain, while TNF-α becomes more relevant in advanced stages, indicating changes in the inflammatory responses. Rest pain showed more nuanced relationships with inflammation, meaning that more complex mechanisms underlie this pain type. The study presents four unique late-stage OA patient subgroups with distinct clinical and molecular traits, which offers a more comprehensive understanding of late-stage OA heterogeneity, including: (1) male patients with high radiographic severity, elevated NF-κB in both the synovium and synovial fluid, and no synovitis; (2) predominantly female patients with metabolic comorbidities, variable synovitis severity, and elevated TNF-α levels in synovial fluid; (3) patients with low subjective pain scores and the absence of synovial inflammation; and (4) patients with high-grade synovitis, increased NFκB, TNF-α, TGF-β expression in synovial membrane, but without metabolic comorbidities. These findings demonstrate the potential for personalized therapeutic strategies catered to the unique features of each late-stage OA profile. More study is required to create precision medicine approaches that address the distinct inflammatory, structural, and metabolic drivers of OA to improve patient clinical outcomes.

Abbreviations

OA

Osteoarthritis

ELISA

Enzyme-linked immunosorbent assay

NF-κB

Nuclear factor kappa B

TNF-α

Tumor necrosis factor alpha

TGF-β

Transforming growth factor beta

BMI

Body mass index

WOMAC

The Western Ontario and McMaster Universities Arthritis Index

VAS

Visual Analogue Scale

K/L

Kellgren–Lawrence grade

CRP

Serum C-reactive protein

Author contributions

Conceptualization – S.Sk., V.G.; formal analysis – S.Sem., L.S., S.Sv., S.Sk.; data curation– S.Sem., L.S., P.S.; statistical analysis – S.Sv., visualization – S.Sk., S.Sv., original draft preparation – S.Sem., S.Sk.; review and editing, S.Sem., V.G., S.Sk. All authors have read and agreed to the published version of the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.