Lack of high BMI-related features in adipocytes and inflammatory cells in the infrapatellar fat pad (IFP)

In total, 155 patients with knee OA were recruited in the study. The patients were 68% women, with mean age 65 years and mean (SD) BMI 29.9 (5.7) kg/m². The distribution of the patients among BMI categories were: underweight (BMI below 18.5), none; normal (BMI 18.5–24.9), 28 patients; overweight (25.0–29.9), 61 patients; and obese (30.0 and above), 66 patients. Characteristics of the patients represented in Figs. 1, 2, 3 and 4 are shown in Additional file 1: Table S1. A total of 102 patients were part of the GEneration of Models, Mechanisms & Markers for STratification of OsteoArthritis patieNts (geMstoan) study, an observational study in patients with knee OA to find new biomarkers for OA progression. The patients were included between 2008 and 2014, had symptomatic knee OA, following the American College of Rheumatology (ACR) criteria [22], and attended the orthopaedic department of the Leiden University Medical Centre (LUMC) or the orthopaedic department of the Alrijne Ziekenhuis in Leiden. Written informed consent is available from all geMstoan patients. Of these 102 patients who participated in the geMstoan study, a total of 79 patients underwent magnetic resonance imaging (MRI), 30 of whom underwent joint-replacement surgery and 49 of whom underwent arthroscopy. IFPs and subcutaneous adipose tissue (SCAT) were obtained from 53 patients participating in geMstoan, who were undergoing joint replacement surgery after giving informed consent. Leftover IFP and SCAT were obtained during arthroplasty from an additional 23 patients from the LUMC, and 30 patients from Erasmus Medical Centre (MC). SCAT was obtained from the thigh, next to the incision made for the total knee-replacement surgery. Diagnosis, age, gender and BMI were available for the latter patients. The study was approved by the local medical ethical committee. Consent was given in accordance with the guidelines of the Federation of Biomedical Scientific Societies (http://​www.​federa.​org) after approval by the local ethical committee (MEC 2008-181 and MEC 2012-267). Not all experiments could be performed with each tissue sample as the sample size was limited.

We used a 3 T Philips Achieva MR system (Philips Healthcare, Best, The Netherlands) with an 8-channel dedicated knee coil. Sagittal proton density (PD) fast spin-echo (FSE)-driven equilibrium images were obtained with a field of view (FOV) of 150 × 150 mm, an acquisition matrix of 304 × 240, slice thickness of 3 mm, repetition time (TR) of 3000 ms and echo time (TE) of 34 ms. Subsequently, contrast enhanced (CE) MR images were obtained following injection of gadoterate meglumine (0.2 ml/kg) (Dotarem; Guerbet) into the cubital vein using a power injector (Medrad) with a rate of 2 ml/s followed by a 40-ml saline flush also at a rate of 2 ml/s. We subsequently obtained frequency selective fat-suppressed T1-weighted, FSE with TR of 655 ms, and TE of 20 ms, in both the axial and sagittal planes.

IFP volume was measured by manual segmentation of IFP boundaries on section-by-section sagittal PD FSE images, using the software program OsiriX. T1-weighted CE images were used to distinguish and compare between IFP and non-IFP structures (Additional file 1: Figure S1). The software program OsiriX measured IFP volume by making a 3D model of the drawn contours. Two independent observers measured IFP volume on all MRI scans. The intraclass correlation (ICC) was 0.957 for intra-observer reliability (measured in all images). The ICC was 0.909 for inter-observer reliability (measured in all images).

Adipocytes were isolated as previously described [13]. Briefly, adipose tissue was digested with collagenase type 1A (Sigma-Aldrich) for 1 h and the tissue was filtered through a 250-μm nylon mesh. Adipocytes were washed three times with medium (DMEM/F12 supplemented with 0.5% free fatty acid (FFA) free bovine serum albumin (BSA), 15 mM Hepes, 2 mM glutamax and 100 U/ml penicillin/streptomycin) by allowing them to float to the surface, followed by removal of medium and addition of fresh medium. The diameter of 100 adipocytes was determined with light microscopy with an ocular micrometer and the mean volume was calculated, based on the formula:

V = πd³/6,

where d is the diameter of the adipocyte, and the mean volume of adipocytes was then calculated. Explants of the IFP and SCAT were also cryosectioned and stained with haematoxylin and eosin (H&E) and imaged using an Olympus SC30 camera (Olympus, Zoeterwoude, The Netherlands). The cross-sectional area of the imaged adipocytes was calculated using Fiji Is Just ImageJ software with the additional Adiposoft plugin. Three separate sections, with a minimum of 25 adipocytes in each section were measured per donor. The Adiposoft application was calibrated to identify cells with a diameter between 30 and 130 μm. A 0.33 μm/pixel measuring-scale was also used by the application to determine the cross-sectional area (size) of each adipocyte identified in the images. A manual inspection of output data was performed to confirm the consistency of the measurements.

Pieces of IFP were fixed in 4% formalin overnight followed by storage in EtOH, before embedding in paraffin. Sections of 4 μm at different depths in the tissue were deparaffinised in xylene (Merck, Germany) and EtOH. Endogenous antigens were peroxidised in a MetOH/H₂O₂ solution, and antigen retrieval was performed with EDTA (DAKO, USA) at 96 °C for 30 min. After cooling, two consecutive coupes of three different layers were stained for mouse anti-human CD68 (clone KP-1 1:800, DAKO, USA) or the isotype control (mIgG1 (DAKO, USA)) for 1 h at room temperature (RT). CD68⁺ cells were then visualized using DAKO EnVision and DAB Ni kits, according to manufacturers’ instructions (DAKO, USA; Vector Laboratories, Canada) before counterstaining with haematoxylin (Klinipath, The Netherlands). Slides were then embedded in pertex (Histolab Products, Sweden) and analysed using a Leica microscope and Leica software. Two to three slides per patient were used and a total of 11–37 high power field (HPF) pictures were taken depending on the size of the tissue. Two independent scorers (AJJ and SNA) quantified the amount of CLS (the ICC for inter-observer reliability was 0.985) in each HPF and an average of both scorers was generated for the amount of CLS per HPF.

The stromal vascular fraction (SVF) was isolated from the IFP and SCAT as previously described [13]. SVF cells were counted by light microscopy, surface staining was performed and the remaining cells were plated overnight in a 6-well plate at a density of maximal 5 × 10⁶ cells/well in medium (DMEM/F12 supplemented with 0.5% FFA free BSA, 15 mM Hepes, 2 mM glutamax and 100 U/ml penicillin/streptomycin) supplemented with 50 IU/ml IL-2 (Peprotech, USA) and 3 μg/ml Brefeldin A (Sigma Aldrich, Germany). Thereafter, cells were harvested using a cell scraper and surface and intracellular staining was performed. The SVF of the IFP from two patients was used to isolate CD14⁺ cells using magnetic-labelled anti-CD14 beads (Miltenyi Biotec) according to the manufacturer’s instructions. CD14⁺ cells were plated in a 96-well plate and supernatant was harvested after 2 days of culture.

Approximately 100,000 freshly isolated SVF cells were stained for 30 min at 4 °C with surface antibody (Ab) solutions containing mixes of the following Abs: PE-conjugated CD1d, CD3, DC-SIGN, CD206 and CD16; FITC-conjugated HLA-DR, CD11c and CD45; APC-conjugated CD8 and CD163; PE-Cy-7-conjugated CD14, PB-conjugated CD4 and Pe-Cy5-conjugated CD206 (all Abs from BD biosciences, except Ab to CD163 and CD206, which were from Biolegend). When specified, approximately 400,000 SVF cells harvested after overnight incubation with brefeldin A were stained for 30 min at 4 °C for surface markers and intracellular cytokines using the BD intracellular cytokine fixation/permeabilization solution kit (BD Biosciences) according to the manufacturer’s instructions. The following Abs were used: PE-Cy7-conjugated CD14; APC-conjugated CD163; PE-conjugated IL-10, TNFα and IL-6 (all Abs from BD biosciences, except Ab to IL-6 which was from eBioscience). Exclusion of dead cells was performed in all experiments using the Dead Cell discrimination kit (Miltenyi Biotec, Germany). Cells were fixed with 1% paraformaldehyde and analysed with a LSR II flow cytometer using Diva 6 software (BD biosciences).

Fat-conditioned and adipocyte-conditioned medium (FCM and ACM, respectively) were obtained as previously described [13]. Briefly, FCM was obtained by culturing 100 mg/ml of small pieces of IFP or SCAT in 6-well plates in medium (DMEM/F12 supplemented with 0.5% FFA free BSA, 15 mM Hepes, 2 mM glutamax and 100 U/ml penicillin/streptomycin). Medium was refreshed after 2 h and supernatant was collected after 24 h. For ACM, 100 μl/ml of adipocytes were cultured in 6-well plates in medium for 24 h. Supernatants were collected and stored at −80 °C until use.

Cytokines were measured in supernatants of CD14⁺ cells isolated from SVF and in FCM and ACM using the Milliplex Human Cytokine/Chemokine kit (Millipore), the Bio-Plex array reader and Bio-Plex software in accordance with the manufacturer’s instructions.

Associations between IFP volume measured by MRI and patient characteristics were determined by linear regression analysis after establishing that the assumptions underlying linear regression analysis were met. Both univariate and multivariate linear regression analyses were performed to determine which patient characteristics were associated with IFP volume. The unpaired or paired Student’s t test (for normally distributed variables) or the Mann-Whitney test or Wilcoxon’s matched-pairs signed rank test (for non-Gaussian distributions) was used to compare differences between groups. Correlation was tested by calculating the Pearson correlation coefficient (for normally distributed variables) or by using Spearman’s ranked correlation test (for non-Gaussian distributions), as specified. For Pearson correlation, the trend line with 95% confidence interval was depicted. Correction for multiple testing was performed using Bonferroni’s correction. A P value ≤0.05 was considered statistically significant.

To obtain insight into whether the volume of the IFP is associated with obesity-related features or other patient characteristics, we made use of a clinically well-characterized population of patients participating in the geMstoan study (N = 79). Patient characteristics are shown in Table 1. The mean (SD) IFP volume was 23.6 (5.4) mm³. Univariate linear regression analysis indicated that BMI is not associated with IFP volume (Table 2). Other obesity-related features such as waist-to-hip ratio (R ² = 0.09) and fat percentage (R ² = 0.19) were associated with IFP volume, while waist circumference was not. Moreover, IFP volume was associated with male gender (R ² = 0.37), height (R ² = 0.47) and weight (R ² = 0.07), but not with age, nor with radiographic damage (Kellgren-Lawrence (KL) scores) (Table 2). Multivariate linear regression analysis in which all factors that were significantly associated with IFP volume were included indicated that only height remained independently associated with IFP volume (Table 2). The association with gender was still present, but was no longer significant. To understand how the factors included in the multivariate analysis vary with height and gender, we included height or gender as covariates in the linear regression analysis. These analyses indicate that gender is a confounder for most of the observed associations and that only height remained significantly associated with IFP volume upon correction for gender (Table 2). Stratification based on gender indicated that height is associated with IFP volume in women but not in men (data not shown). Of all patients, 17.9% had cardiovascular comorbidities, indicating the presence of metabolic complications. There was no association between IFP volume and BMI upon stratification for the presence of these comorbidities (data not shown).

Next, we investigated adipocytes and SVF cells in the IFP. The average volume of adipocytes was 271 pl (Fig. 1a), and the average size was 1933 μm (Fig. 1b). Neither the volume nor the size of adipocytes correlated with BMI. Furthermore, the distribution of adipocyte sizes was not different between lean and obese individuals. Macrophage staining indicated that CLS (Fig. 1c) were present in small numbers in the IFP (mean = 0.18 CLS/HPF) and did not correlate with BMI (Fig. 1d). Likewise, the number of SVF cells in the IFP did not correlate with BMI (Fig. 1e). In contrast, we did observe BMI-related differences in adipocyte volume (Additional file 1: Figure S2a) and size (Additional file 1: Figure S2b-c), and the number of SVF cells (Additional file 1: Figure S2d) in SCAT, a control adipose tissue. Direct comparison between the IFP and SCAT revealed that IFP adipocytes are smaller compared to SCAT adipocytes (Additional file 1: Figure S2e-f), but IFP has more SVF cells compared to SCAT (Additional file 1: Figure S2g).

Next, we investigated the type of immune cells in the IFP. Therefore, we determined the percentage of macrophages and T cells in the SVF from the IFP, since these cell types are the most abundant immune cells in adipose tissue. Our data indicated that there was no correlation between CD3⁺ T cells (Fig. 2a), CD4⁺ T cells (Fig. 2b), CD8⁺ T cells (Fig. 2c) or macrophages (Fig. 2d) and BMI, suggesting no BMI-related differences in the types of immune cells in the IFP.

Previously, we found that TNFα secretion by FCM but not ACM correlated with BMI [13]. Using a different cohort of patients, we now showed that FCM derived from the IFP of patients with a high BMI (BMI >30) had higher levels of TNFα compared to FCM derived from the IFP of patients with a low BMI (BMI ≤25), confirming our previous data (Additional file 1: Table S2). In line with our previous findings, levels of TNFα in ACM did not differ between these high and low BMI groups (Additional file 1: Table S2). We further expanded our analyses to a broad range of cytokines. These studies indicated that several cytokines, such as interferon (IFN)α, IL-2, IL-3, IL-4, IL-5, IL-9, IL-13, IL-15, IL-17, platelet-derived growth factor (PDGF)-ABBB, sCD40L, transforming growth factor (TGF)α and TNFβ were undetectable in most FCM and ACM samples. Moreover, FCM had higher levels of granulocyte macrophage-colony stimulating factor (GM-CSF), IL-8 and monocyte chemotactic protein 3 (MCP3) in the group with BMI >30 compared to the group with BMI ≤25 (Additional file 1: Table S3). Likewise, although ACM contained very low levels of TGFα, these seemed to be lower in the group with BMI >30 compared to the group with BMI ≤25 (Additional file 1: Table S4). These differences, however, were no longer significant after correction for multiple testing.

Because macrophages are abundant in the adipose tissue and their phenotype has been shown to change with obesity, we investigated the phenotype of macrophages in the IFP. Virtually all macrophages expressed human leukocyte antigen (HLA)-DR, while CD1c, CD1d, CD11c and DC-SIGN were present on a low percentage of cells (Fig. 3a). Furthermore, among the surface molecules associated with a pro-inflammatory macrophage phenotype, CD16 was the most abundantly expressed, being present on approx. 20% of macrophages, while 80% of the macrophages expressed the mannose receptor CD206, a surface molecule associated with an anti-inflammatory phenotype. Furthermore, less than half of the macrophages expressed the scavenger receptor CD163. None of the surface markers correlated with BMI.

Intracellular cytokine staining revealed that macrophages from the IFP produced mainly IL-6 and TNFα, but little IL-10 directly ex vivo (Fig. 3b), indicating a predominantly pro-inflammatory phenotype. No correlation with BMI was observed for any of these cytokines (data not shown). The intracellular cytokine findings were confirmed in culture supernatants of isolated macrophages from two patients (Fig. 3d). Furthermore, these analyses revealed that CD14⁺ cells from the IFP were also capable of secreting eotaxin, fibroblast growth factor-2 (FGF-2), Flt3-ligand, fractalkine, growth-regulated oncogene (GRO), granulocyte colony stimulating factor (G-CSF), GM-CSF, IFNα2, IFNγ, IL-1α, IL-1β, IL-1RA, IL-7, IL-8, IL-12p40, IP-10, MCP-1, MCP-3, MDC, macrophage inflammatory protein (MIP)1α, MIP1β, RANTES, sCD40L, sIL-2Rα, TNFβ and vascular endothelial growth factor (VEGF) (data not shown).

Surface marker expression indicated an anti-inflammatory phenotype of IFP macrophages, while the cytokine production profile suggests that pro-inflammatory macrophages are predominant. However, a restricted percentage of macrophages secrete IL-10, indicating them as anti-inflammatory. The scavenger receptor CD163 has been associated with the resolution of inflammation and tissue regeneration [23‐25], but also with inflammation [26‐31]. This indicates that the phenotype of CD163⁺ macrophages is unclear. Therefore, we set out to investigate whether the CD163⁺ macrophages comprise the anti-inflammatory macrophage population and compared them to their CD163^- counterparts. Flow cytometric characterization indicated that CD163⁺ macrophages are bigger than their CD163^- counterparts (Fig. 4a). Furthermore, we found that the percentages of cells positive for CD16, CD206, DC-SIGN and HLA-DR were higher, while the percentage of cells positive for CD1d was lower in CD163⁺ macrophages compared to their CD163^- counterparts (Fig. 4b). Additionally, CD163⁺ macrophages were more often positive for IL-6 and TNFα than for IL-10 and seemed to be more positive for TNFα than CD163^- macrophages (Fig. 4c). These data indicate that although CD163 is regarded as an anti-inflammatory macrophage marker, CD163⁺ macrophages from the IFP display a pro-inflammatory phenotype.