Introduction
Osteoarthritis (OA) of the knee affects over one-third of the United States population aged 60 years or greater [
1]. The incidence of knee OA is expected to increase over the coming decades owing to the aging population and increases in obesity [
2]. Pain and functional limitations from knee OA have major impacts on quality of life for persons living with arthritis [
3], and surgical management of lower extremity arthritis now comprises the largest procedural expenditure in the Medicare budget [
4]. Knee replacement surgery is associated with improvements in pain and function [
5], but 15–20% of patients have continued pain and dissatisfaction following surgery [
6,
7].
Osteoarthritis is characterized macroscopically by loss of articular cartilage and changes to subchondral bone, involving changes to all tissues in the joint through a complex interplay of inflammatory molecules [
8,
9]. Synovial inflammation and hypertrophy occur in joints affected by OA [
10,
11], and severity of synovitis positively correlates with symptoms [
12,
13]. Synovial inflammation present in OA is mediated by inflammatory cytokines [
14], but disease mechanisms of inflammatory OA are incompletely understood. Owing to the growing disease burden of knee OA and the cost and imperfect outcomes associated with its treatment, there remains a critical need to better understand the pathophysiology of synovial inflammation both before and after knee replacement.
Microbial products derived from the microbiome have been proposed to contribute to joint inflammation in OA [
15], and there is increasing evidence that circulating microbial debris contributes to OA [
16]. Peptidoglycan (PG), a structural component of the bacterial cell wall and a highly conserved pathogen-associated molecular pattern (PAMP), has been shown to trigger inflammatory responses in both Lyme arthritis (LA) and rheumatoid arthritis (RA) [
17‐
19]. PG is recognized by innate immune cells via pattern recognition receptors [
18,
20,
21]. Bacterial DNA and bacterial debris, including PG, have also been reported in synovium of limited cohorts of patients with OA [
22‐
24], but the prevalence and potential impact of PG in synovium in patients with advanced knee arthritis remain incompletely understood.
The aim of this study was to characterize the prevalence of PG in synovial fluid and tissue samples at time of total knee arthroplasty performed for advanced OA, and to define its association with synovitis, inflammatory cytokines, and patient outcomes.
Methods
All bacteria were grown to mid-exponential phase, harvested at 4,000 x g for 15 min, and then washed twice with PBS. For peptidoglycan purification, bacterial pellets were resuspended in PBS and added dropwise into boiling SDS (5% w/v, final concentration) and boiled for 1 h as previously described [
17]. All Gram-positive bacteria were bead-beat (BeadBug, Benchmark Scientific) prior to SDS boiling for 3 cycles of 60 s on, 60 s on ice. After boiling, all samples were cooled to 30ºC, and the pellets washed with autoclaved H
2O four times using ultracentrifugation at 283,346 x g for 60 min at 30ºC. The pellets were then resuspended in H
2O and treated with lipase (1 mg/ml) for 3 h, benzonase nuclease (4 µl/ml) for 2 h, and overnight with chymotrypsin (0.3 mg/ml), all with shaking at 37ºC. The next day 0.5% SDS was added to each pellet and heated to 80ºC for 30 min. The pellets were washed 3 times with autoclaved H
2O at the same centrifugation conditions listed above. The Gram-positive samples were treated with a final concentration of 1 M HCl while continuously rotating at 4ºC for 48 h and centrifuged/washed 3 times, as described above. The dry weight was measured to quantify the amount of PG purified. To create the anti-peptidoglycan antibody, 5 BALB/cJ mice purchased from Jackson Laboratories were injected subcutaneously with 200 µg total of peptidoglycan from the bacteria listed above and mixed with equal parts of Freund’s Complete adjuvant (Thermo Scientific Ref: 77,140) (2 mg/ml final of PG). After 3 weeks all mice received a 265 µg booster injection of the same PG mixture. The mice were euthanized 2.5 weeks later and blood was collected. The blood was incubated at room temperature for 30 min prior to spinning at 1,500 x g for 10 min at 4ºC. The serum was then removed, pooled together, and frozen at -20ºC. The specificity of the antibody was tested using immunofluorescence and was found to bind
S. mutans, D. radiodurans,
S. aureus, and
E. coli PG (data not shown) using methods described elsewhere [
25,
26].
Discussion
This study provides evidence that peptidoglycan (PG), a bacterial cell wall component, is present in the synovial tissue of over half of patients undergoing primary total knee arthroplasty for degenerative osteoarthritis. Furthermore, our results suggest that PG may play a role in the symptomatology and disease progression of osteoarthritis, as levels of PG positively correlated with synovitis and pro-inflammatory cytokine levels as well as younger age at time of arthroplasty. These results indicate that PG, likely derived from the microbiome, is involved in pathogenesis of synovial inflammation in advanced knee OA for at least a subset of patients undergoing TKA. PG is a pathogen-associated molecular pattern (PAMP) that is recognized by several immune receptors that yield a pro-inflammatory response [
18,
20,
21]. Synovitis has been linked to clinical progression of OA [
12,
13,
27,
28].
Historical paradigms held that the joint space was free of microbes and microbial debris in the absence of clinical infection, yet data has suggested that immune responses mediated by microbial byproducts may play a role in arthritis. The concept of microbial debris as a mediator of joint inflammation first emerged regarding inflammatory arthritis [
24,
29‐
32]. Newer data indicates that microbial debris, including PG as well as bacterial DNA fragments, is present in joint tissue in degenerative arthritis [
17,
22,
23,
33]. Supporting evidence for the role of microbial debris as a mediator of synovitis includes a study showing positive correlation between the PAMP lipopolysaccharide and knee OA severity [
34]. PG in particular has been shown in animal models to be strongly arthritogenic [
17,
35,
36] and may be exploited as a potential therapeutic target [
37]. Our study is the first of this size to quantify PG in a cohort of patients with advanced knee OA and to characterize PG’s association with synovitis and inflammation. Together, our data and previous studies strongly support the premise that microbial debris derived from the host microbiome can act as a driver of synovitis in knee osteoarthritis.
There are several plausible mechanisms by which bacteria or bacterial byproducts from the host microbiome could travel to the knee joint hematogenously. PG has been identified in the blood of healthy individuals without clinical infection [
38,
39]. Potential sources of PG include gastrointestinal (GI), oral, and skin flora. Translocation of bacteria from the gastrointestinal tract through a permeable gut barrier has been postulated as a driver of surgical site infections [
40]. Gut microbiota, intestinal permeability, GI inflammation, and other gut-associated factors are proposed to contribute to OA, so this phenomenon could also occur in the absence of clinical infection and could include bacterial byproducts [
16,
41,
42]. Boer et al. found that gut dysbiosis is associated with joint pain and inflammation [
43]. Obesity, known to be strongly associated with OA, is linked to alterations in the gut microbiome that promotes increased absorption of bacterial byproducts [
44,
45], although further studies are required to identify the source of PG in synovial tissue. Bacteria or bacterial byproducts may travel directly through the gut barrier or could travel inside white blood cells [
46,
47]. An OA microbiome from Goswami et al. suggested that the skin microbiome may also contribute to contamination of microbes within arthritic joints of some patients [
48]. Moentadj et al. described the ability of PG-polysaccharide polymers from oral streptococcal species to induce arthritis in mice [
35].
We found PG staining of synovium colocalized with both macrophages and synovial fibroblasts. In seeming contradiction, Schrijver et al. [
24]. previously showed localization of PG staining from synovial samples only within cells expressing markers of antigen presentation (HLA-DR, CD40, CD80, CD86)
in situ. However, we and others have subsequently shown that synovia from Lyme arthritis [
17] and rheumatoid arthritis [
49] contain distinct populations of HLA-DR + CD90 + synovial fibroblasts, particularly within the synovial sub-lining and perivascular regions. The CD90 + cells with fibroblast morphology containing bacterial PG in this study display phenotypically similar characteristics. Our in vitro results further demonstrate that PG induces an inflammatory and fibrotic response in synovial fibroblasts, similar to senescent fibroblasts in other chronic inflammatory and fibrotic diseases [
50]. This is further supported by previous
ex-vivo findings in PG-infected synovial cells [
24]. These data support a dual role for synovial fibroblasts, and likely other tissue-resident immune cells, as mediators of the pathogenic response to PG in synovium via upregulation of pro-inflammatory and pro-fibrotic cytokines.
We found no associations between PG and patient reported outcome measures following surgery. While the possibility of type II error cannot be excluded, we do not detect a strong signal that PG present at time of surgery is prohibitive of good outcome following knee replacement surgery. Nonetheless, further investigation of a possible role for PG to adversely affect post-TKA outcomes is warranted. There were no occurrences of periprosthetic joint infection in our elective TKA cohort out to 1 year following surgery. This indicates that the PG identified at time of surgery was not indicative of active clinical infection, but instead represented prior intrusion of these PAMPs into the joint space.
There are several limitations with this study. First, the use of polyclonal antisera limited us to identification of PG antigens restricted to the immunogens used to generate the mouse antisera. Although we selected bacterial species with different PG structures for this study, it is possible that there exists PG in the ‘negative’ patients that may be detected by more sensitive techniques. Furthermore, this study only examined a small portion of discarded synovial tissue from each patient, and there are likely other PG ‘negative’ patients that may have PG in other regions of the synovium. Another limitation is that the assay used was unable to distinguish between different types of PG in patients’ synovia, and it is unknown whether we were detecting PG from intact bacteria, or PG fragments alone. Other studies have found a distinct shift in microbial DNA from gram-positive bacteria to gram-negative bacteria in OA and distinct microbial signatures based on hospital of origin and prior intraarticular steroid injection [
23,
48]. The last observation also indicates that skin microbiota may be contributing to PG contamination into the joint environment. Furthermore, the type of.
PG may have a marked impact on cellular phenotypes of nearby cells within the synovial microenvironment. Spatial imaging approaches, analysis of microbial DNA, and use of PG antibodies specific for distinct PG types may be needed to resolve some of these outstanding questions. Another limitation is the types of primary fibroblasts available for this study. We were not able to obtain synovial tissue from healthy patients, which would be an ideal control. We were able to obtain primary fibroblasts from two patients with knee trauma who underwent TKA. Differences in cytokine levels between stimulated vs. unstimulated cells from trauma patients did not achieve statistical significance (Supplemental Figure
S2), but interpretation of these data are difficult because of the low sample number in this subgroup. Further experiments using more sensitive techniques and reagents will be needed to resolve these outstanding questions.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.