OP0223 DISTINGUISHING OSTEOARTHRITIS AND RHEUMATOID ARTHRITIS SYNOVIUM WITH MACHINE LEARNING USING AUTOMATED CELL DENSITY AND PATHOLOGIST SCORES

Background

Joint damage in the knee can be severe in both rheumatoid arthritis (RA) and osteoarthritis (OA) such that total knee replacement (TKR) is often the only management option. Pathological assessment of the extent or type of synovial tissue inflammation from joint explants or biopsies can be useful. However, an ongoing challenge in using semi-quantitative assessments of synovitis is the disagreement between human pathologist scores of the same sample. We previously developed and validated a computer vision algorithm to automatically count each cell nucleus in an H&E-stained synovial whole slide image and yield a value of cell density, defined as mean nuclei count per mm2 of tissue1.

Objectives

We sought to develop methods to distinguish OA from RA based on machine learning analysis of histologic features on H&E-stained synovial tissue samples.

Methods

We measured 14 pathologist-scored histology features (137 RA and 152 OA patients) and computer vision quantified cell density (60 RA and 147 OA patients) in H&E stained synovial tissue samples from total knee replacement arthroplasty explants. A random forest model was trained using disease state (OA vs RA) as classifier and histology features and/or cell density as inputs, and feature importance scores for the model were calculated.

Results

Synovium from patients with RA exhibited increased lymphocytic inflammation, lining hyperplasia, neutrophils, detritus, plasma cells, Russell bodies, binucleate plasma cells, sub-lining giant cells, synovial lining giant cells, and fibrin (all p<0.001), while synovium from patients with OA had increased mast cells and fibrosis (both p<0.001). Fourteen pathologist-scored features allowed for discrimination between RA and OA samples, producing a macro-averaged area under the receiver operating curve (AUC) of 0.85. This discriminatory ability was comparable to that of the computer vision score of cell density alone (AUC = 0.88). Combining the pathologist scores with the cell density metric improved the discriminatory power of the model (AUC = 0.91). The three most important features in this combined model were mast cells followed by cell density and fibrosis (Figure 1). AUC values for each individual feature are provided in Table 1. The optimal cell density threshold to distinguish RA from OA synovium was 3,400 cells per mm2, which yielded a sensitivity of 0.82 and specificity of 0.82.

Table 1.

Area under receiver operating characteristic curves (AUC) of the synovial features in distinguishing RA and OA patients

FeatureAUCAutomated Cell Density0.88Fibrosis0.84Mast cells0.80Lining hyperplasia0.78Lymphocytic inflammation0.69Fibrin0.68Plasma cells0.66Detritus0.64Binucleate plasma cells0.60Neutrophils0.60Synovial giant cells0.58Sub-lining giant cells0.57Russell bodies0.56Germinal centers0.51Mucoid change0.50Figure 1.

Importance of synovial features in distinguishing RA and OA synovium

Feature importance scores for supervised machine learning model including all 14 pathology scores and the computer vision-generated cell density.

Conclusion

H&E-stained images of RA and OA TKR explant synovium are distinct. We identified cell density, mast cells and fibrosis as the three most important features for making this distinction, with RA being characterized by increased cell density, low mast cells, and low fibrosis. Cell density greater than 3400 per mm2 of tissue yields a sensitivity of 0.82 and a specificity of 0.82 for distinguishing RA from OA. In the future, this can have clinical and research applications as this technique removes the requirement for subjective selection of a certain field of interest, is reproducible, and is scalable as it does not require technical expertise of a pathologist.

References

[1]Guan S, Mehta B…Orange DE. Rheumatoid Arthritis Synovial Inflammation Quantification Using Computer Vision. ACR Open Rheumatology. 2022 Jan 10;acr2.11381.

Acknowledgements

This work was supported by the C. Ronald MacKenzie Young Scientist Endowment Award, the Leon Lowenstein Foundation, and the Kellen Scholar Award supported by the Anna Marie and Stephen Kellen Foundation Total Knee Improvement Program.

Disclosure of Interests

Bella Mehta Paid instructor for: Novartis, Susan Goodman Consultant of: UCB, Grant/research support from: Novartis, Edward DiCarlo: None declared, Deanna Jannat-Khah Shareholder of: AstraZeneca, Cytodyn, and Walgreens, J. Alex Gibbons: None declared, Miguel Otero Consultant of: Regeneron Pharmaceuticals, Grant/research support from: Tissue Genesis, Laura Donlin Speakers bureau: Stryker, Consultant of: Stryker, Grant/research support from: Karius, Inc, Tania Pannellini: None declared, William Robinson: None declared, Peter Sculco Consultant of: Intellijoint Surgical, DePuy Synthes, Lima Corporate, Zimmer Biomet, and EOS Imaging, Grant/research support from: Intellijoint Surgical and Zimmer Biomet, Mark Figgie Shareholder of: HS2, Mekanika, and Wishbone, Consultant of: Lima and Wishbone, Jose Rodriguez Consultant of: ConforMIS, Medacta, Exactech, Inc, and Smith & Nephew, Grant/research support from: DePuy, Exactech, Inc, and Smith & Nephew, Jessica Kirschmann: None declared, James Thompson: None declared, David Slater: None declared, Damon Frezza: None declared, Zhenxing Xu: None declared, Fei Wang: None declared, Dana Orange: None declared

Distinct inflammatory macrophage populations sequentially infiltrate bone‐to‐tendon interface tissue after ACL reconstruction surgery in mice

Macrophages are important for repair of injured tissues, but their role in healing after surgical repair of musculoskeletal tissues is not well understood. We used single‐cell RNA sequencing (RNA‐seq), flow cytometry, and transcriptomics to characterize functional phenotypes of macrophages in a mouse anterior cruciate ligament reconstruction (ACLR) model that involves bone injury followed by a healing phase of bone and fibrovascular interface tissue formation that results in bone‐to‐tendon attachment. We identified a novel “surgery‐induced” highly inflammatory CD9+ IL1+ macrophage population that expresses neutrophil‐related genes, peaks 1 day after surgery, and slowly resolves while transitioning to a more homeostatic phenotype. In contrast, CX3CR1+ CCR2+ macrophages accumulated more slowly and unexpectedly expressed an interferon signature, which can suppress bone formation. Deletion of Ccr2 resulted in an increased amount of bone in the surgical bone tunnel at the tendon interface, suggestive of improved healing. The “surgery‐induced macrophages” identify a new cell type in the early phase of inflammation related to bone injury, which in other tissues is dominated by blood‐derived neutrophils. The complex patterns of macrophage and inflammatory pathway activation after ACLR set the stage for developing therapeutic strategies to target specific cell populations and inflammatory pathways to improve surgical outcomes. © 2022 The Authors. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.

Risk factors for multi-joint disease in patients with glucocorticoid-induced osteonecrosis

This study investigated risk factors for osteonecrosis involving multiple joints (MJON) among glucocorticoid-treated patients. The best predictor of MJON was cumulative oral glucocorticoid dose. Risk of MJON was 12-fold higher in patients who had a second risk factor for osteonecrosis. Further research is needed into strategies for prevention of MJON.

INTRODUCTION

Osteonecrosis (ON) is a debilitating musculoskeletal condition in which bone cell death can lead to mechanical failure. When multiple joints are affected, pain and disability are compounded. Glucocorticoid treatment is one of the most common predisposing factors for ON. This study investigated risk factors for ON involving multiple joints (MJON) among glucocorticoid-treated patients.

METHODS

Fifty-five adults with glucocorticoid-induced ON were prospectively enrolled. MJON was defined as ON in ≥ three joints. Route, dose, duration, and timing of glucocorticoid treatment were assessed.

RESULTS

Mean age of enrolled subjects was 44 years, 58% were women. Half had underlying conditions associated with increased ON risk: systemic lupus erythematosus (29%), acute lymphoblastic leukemia (11%), HIV (9%), and alcohol use (4%). Mean daily oral dose of glucocorticoids was 29 mg. Average cumulative oral dose was 30 g over 5 years. The best predictor of MJON was cumulative oral glucocorticoid dose. For each increase of 1,000 mg, risk of MJON increased by 3.2% (95% CI 1.03, 1.67). Glucocorticoid exposure in the first 6 months of therapy, peak dose (oral or IV), and mean daily dose did not independently increase risk of MJON. The risk of MJON was 12-fold in patients who had a second risk factor (95% CI 3.2, 44.4).

CONCLUSIONS

Among patients with glucocorticoid-induced ON, cumulative oral dose was the best predictor of multi-joint disease; initial doses of IV and oral glucocorticoids did not independently increase risk. Further research is needed to better define optimal strategies for prevention and treatment of MJON.

MEF2C regulates osteoclastogenesis and pathologic bone resorption via c-FOS

Osteoporosis is a metabolic bone disease with dysregulated coupling between bone resorption and bone formation, which results in decreased bone mineral density. The MEF2C locus, which encodes the transcription factor MADS box transcription enhancer factor 2, polypeptide C (MEF2C), is strongly associated with adult osteoporosis and osteoporotic fractures. Although the role of MEF2C in bone and cartilage formation by osteoblasts, osteocytes, and chondrocytes has been studied, the role of MEF2C in osteoclasts, which mediate bone resorption, remains unclear. In this study, we identified MEF2C as a positive regulator of human and mouse osteoclast differentiation. While decreased MEF2C expression resulted in diminished osteoclastogenesis, ectopic expression of MEF2C enhanced osteoclast generation. Using transcriptomic and bioinformatic approaches, we found that MEF2C promotes the RANKL-mediated induction of the transcription factors c-FOS and NFATc1, which play a key role in osteoclastogenesis. Mechanistically, MEF2C binds to FOS regulatory regions to induce c-FOS expression, leading to the activation of NFATC1 and downstream osteoclastogenesis. Inducible deletion of Mef2c in mice resulted in increased bone mass under physiological conditions and protected mice from bone erosion by diminishing osteoclast formation in K/BxN serum induced arthritis, a murine model of inflammatory arthritis. Our findings reveal direct regulation of osteoclasts by MEF2C, thus adding osteoclasts as a cell type in which altered MEF2C expression or function can contribute to pathological bone remodeling.

FRI0403 CLINICAL FEATURES OF PROSTHETIC JOINT INFECTIONS DIFFER IN PATIENTS WITH INFLAMMATORY ARTHRITIS AND OSTEOARTHRITIS

Background:

Inflammatory arthritis (IA) patients are at increased risk for prosthetic joint infections (PJI). However, because active IA patients without infections can have elevated inflammatory markers that mimic joint infection, PJI diagnosis is challenging in this population.

Objectives:

We used an institutional PJI registry to identify and compare the clinical, microbiologic, and histopathologic features of culture positive (CP) and culture negative (CN) total hip and knee PJI in IA and OA patients. We also evaluated the relationship between culture positivity, IA, and clinical outcomes.

Methods:

A retrospective cohort of THA/TKA PJIs, from 2009 to 2016, were identified by ICD codes, and confirmed by chart review. IA diagnosis was also confirmed by use of IA-specific medications. CN cases were defined as PJIs with no evidence of microbial growth in intraoperative cultures and CP PJI cases were defined by positive microbial growth in intraoperative cultures. Treatment failure was defined as subsequent surgical treatment for infection after the initial infection surgery. H&E slides of OA and IA PJI cases matched by age (+/-5) sex, and culture status were reviewed by a pathologist for evidence of the histopathologic features listed in Table 2. Fisher’s exact test, chi-square test, and Kaplan-Meier estimates were used.

TABLE 1.

Patient characteristics in IA and OA PJIs

IAOAN%/SDN%/SDp-valueTotal36771Age58.511.466.812<.001BMI30.26.7306.70.861Female2877.833243.1<.001CCI2.81.71.72.10.002Smoking411.18611.20.792Glucorticoids1027.8395.1<.001Culture Negative1027.810914.10.024Treatment Success at 2 years1952.8509660.146

IA- inflammatory arthritis; OA – osteoarthritis; PJI -prosthetic joint infection; CCI – Charlson Comorbidity Index

TABLE 2.

Histopathology and clinical presentation in IA and OA PJIs

OA (N=57)IA (N= 31)CP-IA (N=23)CN-IA (N=8)N (%)p-valueN (%)p-valuePathology Review>10 PMN per HPF42 (74)22 (71)0.80620 (87)2 (25)0.003Chronic Inflammation13 (23)23 (74)0.00118 (78)5 (63)0.393Necrosis17 (30)9 (29)18 (35)1 (13)0.38Clinical PresentationMSIS50 (88)26 (84)0.74722 (96)4 (50)0.009Sinus Tract7 (12)7 (23)0.2335 (22)2 (25)1Elevated ESR or CRP41 (72)24 (77)0.62217 (74)7 (88)1Elevated Synovial WBC33 (58)19 (61)0.82313 (57)6 (75)1Elevated Synovial %PMN31 (54)20 (65)0.37714 (61)6 (75)0.333

OA – osteoarthritis; IA – inflammatory arthritis; CP – culture positive; CN – culture negative; MSIS – meets Musculoskeletal Infection Society diagnostic criteria

Results:

807 PJI cases were identified including 36 IA (33 RA and 3 SLE) and 771 OA. A higher proportion of IA PJI were CN (N=10, 27%) vs. OA PJI (N=109, 14%, p=0.02). IA-PJI were younger, female, on glucocorticoids, and with more comorbidities. Type of surgical treatment did not differ significantly between IA and OA groups. Comparing CN-IA vs. CP-IA, no difference was observed in age, smoking, diabetes, surgical treatment, IA-specific meds or Charlson comorbidities. One-year survivorship of CN-IA and CN-OA were 66% and 87% (p>0.05). Across all CP cases, 57% were staphylococcal, with no differences between groups. Treatment failure was more frequent for CP-IA (42%) compared to CP-OA (30%), (p=0.2).

Histopathology of 88 PJIs (31 IA and 57 OA) was reviewed. The IA cohort presented with more chronic inflammation (p=0.001) than the OA cohort. Within the IA cohort, a higher proportion of CP-IA had >10PMN per HPF (p= 0.003) and met MSIS criteria (p=0.009). Comparing CP-OA and CN-OA, there were no significant differences in histopathology findings or number of patients meeting MSIS criteria.

Conclusion:

IA PJIs are more likely to be culture negative than OA PJIs. Although our analysis was limited by our cohort size, our findings including differences in histopathology, and better clinical outcomes suggest the presence of biologic differences between CN and CP PJI that require further study.

Disclosure of Interests:

Milan Kapadia: None declared, Tania Pannellini: None declared, Carine Moezinia: None declared, Andy Miller: None declared, Mark Figgie: None declared, Peter Sculco: None declared, Michael Cross: None declared, Michael Henry: None declared, Linda Russell: None declared, Laura Donlin Consultant of: Consultant – Genentech/Roche, Allina Nocon: None declared, Susan Goodman Shareholder of: Reginosine- Investment, Grant/research support from: Novartis, Horizon, Consultant of: Novartis, Celgene, UCB

Monomeric allergoid intragastric administration induces local and systemic tolerogenic response involving IL-10-producing CD4(+)CD25(+) T regulatory cells in mice

The efficacy of sublingual immunotherapy, at present one of the treatments of choice for respiratory allergy, relies on the tolerance induced by oral mucosa-associated immune system; however, the gut-associated lymphoid tissue (GALT: Peyers patches and isolated lymphoid follicles) and mesenteric lymph nodes could also be involved, being stimulated by the ingested part of the allergen extract. The aim of the present study is to assess whether the exposure of the allergen exclusively to the GALT induces a tolerogenic response. For this purpose, mice were sensitized with ovalbumin or Par j 1 allergens. The corresponding gastric-resistant monomeric allergoids were then administered via orogastric gavage. After treatment, all mice were tested for: serum IgE, in vitro Th1 and Th2 cytokine release by allergen-stimulated peripheral blood lymphocytes, CD4(+)CD25(+) and CD4(+)CD25(+)IL-10(+) T cells in Peyers patches, mesenteric lymph nodes and spleen. Compared to the control, sensitized groups showed higher levels of serum IgE, lower frequency of CD4+CD25+IL-10+ T cells, at all sites, and higher amounts of in vitroreleased IL-4, IL-6 and TNF-alpha. Compared to the sensitized groups, higher frequency of CD4(+)CD25(+)IL-10(+) T cells was observed in the spleen of both Par-j 1 and OVA sensitized/treated groups and, only for ovalbumin-treated mice, in the Peyers patches and mesenteric lymph nodes, IgE and in vitro cytokines were significantly lower and equivalent to the control group. The results give the first evidence that the intragastric-restricted administration of gastric-resistant allergens restores local and peripheral tolerance in allergen-sensitized mice.

WIP null mice display a progressive immunological disorder that resembles Wiskott-Aldrich syndrome

The Wiskott-Aldrich syndrome (WAS) is an X-linked immunodeficiency syndrome caused by mutations in the WAS protein (WASP). This participates in signalling and cytoskeletal homoeostasis, and some of its activities are regulated by its binding to the WASP interacting protein (WIP). WIP deficiency, however, has not yet been shown to be of pathological significance in humans. Here we show that, in WIP null (WIP(-/-)) mice, it produces haematological alterations and anatomical abnormalities in several organs, most probably as a consequence of autoimmune attacks. Granulocytosis and severe lymphopenia are associated with a proportional increase in segmented cells and fewer bone marrow erythrocytes and lymphocytes. Splenomegaly is accompanied by an increase of haematopoietic tissue and red pulp, reduction of the white pulp, and fewer B (B220(+)) lymphocytes (also apparent in the lymph nodes and Peyer's patches). Ulcerative colitis, interstitial pneumonitis, glomerular nephropathy with IgA deposits, autoantibodies, and joint inflammation are also evident. These progressive immunological disorders closely mimic those seen in WAS. WIP deficiency may thus be implicated in some cases in which mutations in the gene encoding WASP are not detected.

The expression of LEC/CCL16, a powerful inflammatory chemokine, is upregulated in ulcerative colitis

Ulcerative colitis (UC) is a chronic inflammatory disease of unknown aetiology and pathogenesis. The presence in the colonic mucosa of reactive cells expressing proinflammatory cytokines and chemokines is associated with high levels of IL-10, an anti-inflammatory cytokine. Our aim was to investigate the role of IL-10 and the beta chemokine LEC/CCL16 selectively up-regulated by IL-10 in inflammatory cell recruitment and cytokine and chemokine production during UC. We studied histologically, immunohistochemically and ultrastructurally colonic biopsies from 20 active UC patients and 10 control specimens taken far from any macroscopically detectable lesion in age and sex-matched patients with noninflammatory bowel disease. In active UC, immature dendritic cells (DCs) in the LP are associated with IL-10 in the T cell rich area. Furthermore, most of the LP-infiltrating macrophages strongly expressed LEC/CCL16, a chemokine upregulated by IL-10. To evaluate if LEC/CCL16 plays a role in the inflammatory reaction present in UC, we performed morphological studies in mice injected s.c. with syngeneic tumor cells engineered to produce LEC/CCL16. We found that the LEC protein locally released by LEC-gene-transfected tumor cells is a potent proinflammatory chemokine that induces the recruitment of a reactive infiltrate, and an angiogenic process mirroring that in human UC.

Neutrophils in anti-cancer immunological strategies: old players in new games

This review highlights the new "immunological identity" of neutrophils within the cytokine network and their role in biology of diseases, particularly in tumor biology. The latest preclinical evidence of their involvement in anti-cancer immunotherapeutic and prophylactic strategies will be discussed with particular reference to the real possibilities of transferring experimental results to a clinical setting.