Immune response and innervation signatures in aseptic hip implant loosening

Centro Hospitalar São João ethics committee approved this study and all patients consented to the use of their tissue and blood for research purposes. The followed procedures were in accordance with the Helsinki Declaration of 1975, as revised in 2000. Samples from synovial membrane-like interface tissue/aseptic interface membrane were collected from twenty patients during hip revision surgeries due to AL of hip prostheses. All revised hips had a metal-on-polyethylene (MoP) coupling, eleven out twenty cemented and bone defects were classified according to the Paprosky classification [23]. Relevant clinical information is summarized in Table 1. Infection, recurrent dislocation and periprosthetic fracture were considered exclusion criteria in the aseptic loosening group. Osteoarthritic synovial tissues were collected from 15 patients undergoing primary hip replacement surgeries for primary OA. OA patients were classified for OA severity according Tönnis OA grade [24] and presented scores from moderate (2) to severe (3). The clinical information of these patients is summarized in Table 2. For both OA and AL groups, the same orthopedic team performed the collection of tissue samples. Previously to surgery, blood was collected and leukograms and plain radiographies were registered.

Immediately after excision, both OA synovial tissues and aseptic interface tissues were split. Half of the tissue was immersed in formalin for further histological analysis, while dry ice was used to freeze the remaining tissue until storage at −80 °C. No more than 4 h elapsed between tissue collection and storage.

Half of collected tissues were formaldehyde-fixed paraffin embedded and cross-sections of 3 μm thickness were cut. Contiguous sections were stained with hematoxylin & eosin (H&E) and Masson’s trichrome (MT). For immunohistochemistry, tissue sections were deparaffinized and rehydrated before heat induced antigen retrieval (98 °C, 10 mM citrate buffer, pH 6.0). Endogenous peroxidases were blocked using 3 % H₂O₂ and non-specific binding sites were blocked using Background Block (Cell Marque, USA), previously to the incubation with antibody diluent 1 % BSA (negative control) or with primary antibodies: anti-CD20 (clone L26, dilution 1:100, Cell Marque, USA), anti-CD163 (clone MRQ-26, dilution 1:150, Cell Marque, USA), anti-CD68 (clone 514H12, dilution 1:100, Novocastra, UK), anti-HLA-DR (clone TAL1B5, dilution 1:5000, Abcam, USA) and anti-CD3 (clone PS1, dilution 1:100, Biocare Medical, USA). After primary antibody incubation, tissue sections were incubated with BrightVision Poly-HRP-Anti Mouse/Rabbit/Rat IgG (Immunologic, the Netherlands) and then revealed using DAB Plus Substrate System (Thermo Scientific, USA), before hematoxylin counterstaining. The specificity of immunostainings of CD20, CD163, CD68, HLA-DR and CD3 was confirmed using human spleen as positive control. For immunofluorescence studies, tissue section were deparaffinized and antigen retrieval of rehydrated sections was performed using Proteinase K (0.2 mg/mL in PBS) for NF200 immunostaining and incubated for 20 min at 98 °C in Citrate buffer (pH 6.0), or TE Buffer (pH 9.0) for SP and CGRP staining, respectively. After quenching endogenous fluorescence with 0.1 % sodium borohydride and 100 mM NH₄Cl, sections were incubated with blocking buffer (10 % FBS, 1 % BSA, 0.2 % Triton X-100). Primary antibody rabbit anti-human neurofilament heavy subunit (NF200) (dilution 1:1000, Abcam, USA), anti-Substance P (dilution 1: 1000, Millipore, USA), anti-TH (dilution 1:100, Millipore, USA), anti-CGRP (dilution 1:4000, Sigma-Aldrich, USA) or blocking buffer (negative control) was applied overnight at 4 °C. For signal detection, tissue sections were incubated with anti-rabbit Alexa Fluor 568 antibody (dilution 1:1000, Life Technologies, USA), incubated with DAPI and then mounted with Fluoroshield Mounting Medium (Abcam, USA). The specificity of immunostainings for NF200, Substance P, TH and CGRP was confirmed using a specimen of human Morton’s neuroma as positive control.

In order to characterize synovial microenvironment, tissues were semi-quantified considering the total immunoreactive area and scored into four categories: absent (0), present (1), frequent (2) or abundant (3), according to the histological grading system illustrated in the Additional file 1: Figure S1 (tissue reaction), Additional file 2: Figure S2 (prosthetic debris accumulation) and Additional file 3: Figure S3 (immune cells distribution), similarly to the methodology followed by others [25, 26].

Structural changes, fibrosis and necrosis in OA synovial tissues and aseptic interface tissues were evaluated after H&E and MT staining. The thickening/hyperplasia and increased villi of synovial membrane were assessed in OA patients. The accumulation of polymeric, ceramic and metallic particles was assessed in synovial membrane-like interface tissues. A polarized filter was used to detect polymeric particles. Ceramic and metallic particles were studied through scanning electron microscopy (SEM) and their elemental composition assessed by energy dispersive spectroscopy (EDS). Additionally, a phase contrast filter (Ph3) was used with an optical microscope to ease the detection of clusters of ceramic or metallic particles in histological slices, allowing a deep study of the interaction between particles and cells. The infiltration of target immune cell populations, identified by immunohistochemistry, was determined regarding the presence of macrophages (CD68⁺ cells), T cells (CD3⁺ cells) and B cells (CD20⁺ cells). The semi-quantification of polymorphonucleated cells (PMNs) was performed after PMNs identification as these cells have a lobed nucleus while multinucleated giant cells (GC) were detected due to their bigger size, multiples nucleus and CD68⁺ labeling.

Tissue samples retrieved from five out thirty-five patients (4/15 OA and 1/20 AL) were excluded from histopathological evaluation because they do not correspond to OA synovial tissue or aseptic interface tissue. H&E, MT and immunohistochemistry slices were analyzed using a light microscope Olympus CX31 while immunofluorescence, polarized light and phase contrast Ph3 images were captured on Carls Zeiss Axiovert 200 inverted microscope.

Synovial tissues were homogenized in liquid nitrogen using a mortar and pestle to preserve RNA integrity. RNA was extracted and purified using TRIzol (Invitrogen, UK) and Direct-zol™ RNA MiniPrep (ZYMO Research, USA), according to the manufacturers’ instructions. The amount and quality of extracted RNA were evaluated by Nanodrop ND-1000 (Thermo Fisher Scientific, USA) and running RNA samples in a 2 % agarose gel. The transcriptional levels of pro-inflammatory cytokines (TNF-α, IL-1β, IL-6, IL-12a and iNOS), anti-inflammatory cytokine (IL-10), osteoclastic factor (RANKL) and bone remodeling factor (TGF-β1) were evaluated by quantitative real time PCR (qRT-PCR) in PCR iQ™5 system (Bio-rad, USA). All used primers, listed in Table 3, were optimized and melting curves of PCR products were evaluated to guarantee primers specificity. β2 microglobulin (B2 M) and β-actin were used as reference genes. Experiments were performed in triplicated. Relative gene expression levels were calculated using the quantification cycle (C_q) method, according to MIQE guidelines [27]. Fourteen out thirty-six cases (5/15 OA and 9/20 aseptic loosening) were excluded from gene expression analysis when average Cq for reference genes was above 26, to avoid bias in the evaluation of genes with later expression.

Serum transforming growth factor β1 (TGF-β1) concentrations were measured using Quantikine^® ELISA Kit for human TGF-β1 (R&D Systems, USA), according to the manufacturer’s protocol. Cytokine concentration was calculated against a standard curve.

In this study, 15 patients underwent primary hip replacement due to OA (Fig. 1a) and twenty hip implants were revised and their acetabular defects classified by X-ray imaging (Fig. 1b). Macroscopically, the collected synovial tissues at primary and revision surgeries were morphologically distinct. Synovial surface was white and with diffuse papillary architecture (dashed black line, Fig. 1c, d) whereas synovial membrane-like interface tissues were highly fibrotic (Fig. 1e) or with greyish appearance (Fig. 1f), in case of metallosis.

Semi-quantitative analysis of tissue organization and immune cell distribution in OA synovial tissue and aseptic interface tissues showed distinct local profiles. OA synovial membranes displayed a bicellular lining layer (LL) and a sublining layer (SLL) (Fig. 2a) enriched in blood vessels (black arrows) and collagen (Fig. 2b), as well as structural changes induced by synovial inflammation, such as synovial hyperplasia (Fig. 2c, d) and increased number of villi (Fig. 2e, f). OA patients presented at least one sign of synovial inflammation (thickening or villi) but their magnitude was variable among patients (Fig. 2c, e). Synovial membrane-like interface tissues presented fibrotic stroma with increased collagen deposition (Fig. 2g, h) than OA synovial tissues (p = 0.001) and with some necrotic regions (Fig. 2i).

The accumulation of prosthetic debris in synovial membrane-like interface tissues was semi-quantified. Polymeric particles were just detected in 11/20 analyzed aseptic interface tissues (Fig. 3a) after tissue analysis under polarized light (Fig. 3b, c). No significant difference was found between AL patients with cemented and uncemented MoP bearings regarding the amount of polymeric particles entrapped in synovial membrane-like tissues. Zirconia particles (ZrO₂) were observed in almost all synovial membrane-like interface tissues retrieved from patients with loose cemented prostheses (Fig. 3d–f) and mainly phagocytized by macrophages or multinucleated giant cells (Fig. 3g). Ph3 contrast filter eased the detection of ZrO₂ particles (Fig. 3e, f). White color particles under Ph3 filter were confirmed to be ZrO₂ and to be organized in clusters of nanoparticles by SEM/EDS (Fig. 3h, i). Intense deposition of metallic particles was just found in the four cases of metallosis that were patients with uncemented metal-back acetabular cups (Fig. 3j). Aseptic interface membranes with metallosis presented high deposition of metallic particles with concomitant macrophage infiltration (Fig. 3k), tissue fibrosis (Fig. 3l) and necrosis (Fig. 3m). Under Ph3 filter, metallic nanoparticles and haemoglobin presented violet and red color respectively, which allow distinguishing them from ZrO₂ particles (Fig. 3n, o).

Immune cell distribution was studied in tissues from AL and OA patients. In AL, macrophages (CD68⁺ cells) were more abundant than in OA (p = 0.007; Fig. 4a). While macrophages were often confined to lining layer in OA patients (Fig. 4b), signifficant macrophage infiltration was found in synovial membrane-like interface tissues (Fig. 4c). In these tissues, co-localization between polymeric particles and macrophages was often found (Fig. 4d) and a similar pattern was observed in AL patients with cemented and uncemented MoP bearings. In AL patients, macrophages were the most prevalent immune cell population, even in metallosis cases, and in overall highly express M1 (HLA-DR; Fig. 3e) and M2 (CD163; Fig. 3f) markers. Multinucleated giant cells in foreign body reaction setting were mostly found in synovial membrane-like interface tissues (p = 0.037, Fig. 4g) surrounding big polymeric particles, likely PMMA, with ZrO₂ particles entrapped inside (Fig. 4h). T cells (CD3⁺ cells) (Fig. 4j), and B cells (CD20⁺ cells; Fig. 4l) in lower number, could be detected in the majority of tissues collected from AL and OA patients (Fig. 4i, k), including AL patients with uncemented implants or signs of metallosis. Moreover, the number of PMNs was also low in both aseptic interface membranes and OA synovial tissues. Overall, the immune cell populations addressed, namely macrophages, T cells, B cells and PMNs, in this study were most located in the vicinity of synovial membrane in OA patients, while in AL patients these cells were identified throughout the aseptic interface membranes.

Myelinated nerve fibers identified by NF200 immunoreactivity were detected in both synovial membrane-like interface tissues and OA synovial tissues. They were presented as single fibers (Fig. 5a, c) and were preferentially arranged around blood vessels (Fig. 5c) particularly in tissue regions of reactive vascularization induced by immune responses underlying AL and OA. Alternatively, nerve fibers were also found in neurome-like structures (Fig. 5b, d) in both AL and OA patients. Sensory and sympathetic innervation, as seen by immunohistochemical markers of sensory nerve-associated peptides (SP and CGRP) and catecholaminergic marker of sympathetic neurons (TH), showed different pattern between OA patients and AL patients. TH immunoreactive nerve fibers were observed in OA synovial tissues (Fig. 5f) but not in synovial membrane-like interface tissues (Fig. 5g, h). SP and CGRP were found both in OA synovial tissues and synovial membrane-like interface tissues but showed different pattern. In OA synovial tissues, nerve fibers immunoreactive to SP and SGRP were observed mainly around blood vessels in the vicinity of synovial membrane (Fig. 5j, n). In addition, a high number of cells in the OA synovial membrane also stained for TH, SP and CGRP (Fig. 5e, i, m). In synovial membrane-like interface tissues, the expression of SP and CGRP was also found in both nerve fibers and cells but with a broad distribution throughout the tissue (Fig. 5k, l, o, p).

The expression levels of pro-inflammatory cytokines tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β) and IL-6 were similar in aseptic interface tissues and OA synovial tissues (Fig. 6a–c). Inducible nitric oxide synthase (iNOS) and IL-12a expression levels were found to be low when compared with the TNF-α, IL-1β and IL-6 mRNA levels detected in AL and OA groups (Fig. 6d, e). Interestingly, the anti-inflammatory cytokine IL-10 presented a tendency (p = 0.084) to be higher expressed in synovial membrane-like tissues than in OA synovial tissues (Fig. 6f). Two genes involved in bone remodeling were evaluated: TGF-β1 and receptor activator of nuclear factor kappa-B ligand (RANKL). The mRNA levels of TGF-β1 were significantly reduced (p = 0.038) in aseptic interface tissues when compared to OA synovial tissues (Fig. 6g). However, no differences between AL and OA patients were verified for RANKL (Fig. 6h) and no correlation was found between mRNA levels of RANKL and the bone defect type.

In both AL and OA patients, TGF-β1 was detected in blood vessels endothelium cells (Fig. 7a), macrophages (Fig. 7b) and fibroblasts (Fig. 7c). Interestingly, differences in the pattern of TGF-β1 expression were found between AL patients and OA. Aseptic interface tissues presented a trend (p = 0.1672) toward increased number of regions expressing TGF-β1 (Fig. 7d). In OA synovial tissues, TGF-β1 was present in the sublining layer of synovial membrane (Fig. 7e), in aggregates of lymphocytes (Fig. 7F) and blood vessels. In synovial membrane-like interface tissues, the distribution of TGF-β1-positive regions was heterogeneous (Fig. 7g, h).

The preoperative leukograms were analyzed and the concentration of TGF-β1 determined in plasma of both AL and OA patients. The percentage of circulating monocytes in both groups tended to be similar (Fig. 8a). Interestingly, the percentage of lymphocytes seemed to be low in AL patients but within the reference interval in OA group (Fig. 8b). Neutrophils did not present significant alterations compared with the reference values (Fig. 8c). Remarkably, the levels of TGF-β1 in serum were similar in both groups (Fig. 8d).