*2.5. Correlations Between Histological Data and Morphometric Analysis*

The age of overall cohort correlated positively with the adipocyte area (r = 0.591, *p* < 0.0001), and negatively with the adipocyte number (r = −0.412, *p* = 0.003) (Figure S3a,b, Supplementary Materials). However, considering the two groups separately, both correlations were not maintained. The adipocytes area was positively correlated with BMI of overall cohort (r = 0.621, *p* < 0.0001) (Figure S3c, Supplementary Materials) while, separating the two groups, this correlation was present only in end-stage OA IFP (r = 0.413, *p* = 0.040) (Figure S3e,f, Supplementary Materials). The number of adipocytes of the overall cohort was negatively correlated with BMI (r = −0.427, *p* = 0.003) (Figure S3d, Supplementary Materials) and this correlation was maintained only in the OA group (r = −0.514, *p* = 0.009) (Figure S3g,h, Supplementary Materials).

#### *2.6. Influence of BMI and Age*

Since well-known risk factors for OA, such as BMI and age, were not equally distributed between the two patient groups, we applied linear models to control their influence (Tables S4–S6, Supplementary Materials). Vascularity did not fit a normal distribution, and so it was log-transformed before being modeled. Lymphocytic infiltration, IL-6 immunohistochemical grading, vascularity, and adipocyte numbers were all independent from BMI and age. Furthermore, BMI and age did not have any significant effect on the response variables, whether analyzing OA and ACL rupture separately or combined. Only the adipocyte area was almost associated with BMI in the OA group (*p* = 0.057) and significantly associated when the two patient groups were considered together (*p* = 0.033).

#### **3. Discussion**

This is the first study investigating the histological, morphometric, and molecular characteristics of end-stage OA IFP compared to ACL IFP. In particular, we have explored not only IFP inflammation but also adipocytes and ECM changes in these two groups. The studies published so far mostly focalized on IFP inflammation and utilized the subcutaneous adipose tissue of the knee region as "healthy" control, which might not be an optimal "healthy" control because of the differences in physiological functions and biomechanical characteristics of these fat depots [21]. Furthermore, OA IFP characterization is of great importance to unravel its role in OA pathology as occurred for other tissues such as synovial membrane and cartilage [22,23].

The important novelty of this study is the evidence that, beyond inflammation, also adipocytes and ECM characteristics of OA patients are different compared to IFP derived from ACL patients. In particular, these data suggest that OA pathology induces molecular changes in IFP, affecting the cells and ECM composition in addition to the increase of inflammation.

Regarding the adipocyte characteristics, we observed that the adipocyte area was 1.7-fold higher in end-stage OA compared to ACL IFP, while the adipocyte number was lower in end-stage OA compared to ACL IFP. Interestingly, the adipocyte area of IFP was positively correlated with BMI, while adipocyte number was negatively correlated with BMI, but only the adipocyte area was confirmed to be associated with the BMI at linear regression model in the overall population and in the end-stage OA group. The absence of the correlation in the ACL IFP group between BMI and adipocyte area could be explained, considering that all these subjects were lean.

Previous studies evaluated the adipocyte area in OA IFP subgrouping the patients according to BMI, observing that the area was smaller in lean OA patients (BMI < 25 kg/m2) compared to severely obese OA patients (BMI <sup>≥</sup> 35 kg/m2) [24]. In contrast, de Jon et al. did not observe any difference in adipocyte size of OA IFP related to patient BMI, reporting that adipocyte volume and size of OA IFP were smaller compared to subcutaneous adipose tissue used as control [25]. The influence of BMI on the adipocyte area was not confirmed in other studies, both in animals and humans. Barboza et al. showed an increase of IFP volume in mice with high-fat diet-induced obesity compared to controls. In contrast, no difference was reported in the adipocyte area, suggesting that obesity does not influence

this IFP feature [26]. In addition, other authors demonstrated that BMI did not exhibit any effect on adipocyte area in humans [27].

In general, the adipocyte area has been studied from all adipose tissue anatomical locations and in both sexes. The adipocyte area increases in size along with adiposity level, reaching a plateau in massively obese subjects [28]. Here, we showed that age was positively associated with the adipocyte area and negatively with the adipocyte number of IFP, although these findings were not confirmed by the linear regression model. No specific studies have been published so far evaluating the effects of age on IFP adipocyte characteristics. However, no significant association was observed between age and adipocyte size distribution parameters in omental and subcutaneous adipose tissue [29].

We observed an increase of lymphocytic infiltration as well as of vascularity in end-stage OA IFP compared to ACL IFP, confirming our previous data obtained comparing end-stage OA IFP with that of cadavers considered as healthy controls [30]. Increased lymphocytic infiltration and vascularity were independent of age and BMI in both groups in the linear models, suggesting that these inflammatory features play an important role in OA pathology. Moreover, in agreement with the histological IFP scores, we observed an increase of *IL-6* and *VEGF* mRNA expression levels in OA compared to ACL IFP. The increase of IL-6 expression was also confirmed by IHC in agreement with our previous results [30]. Surprisingly, we did not observe increased MCP-1 expression levels in OA IFP compared to ACL IFP, at variance with what we observed in a previous study using IHC [30]. The expression of all classical adipose tissue markers was increased in OA IFP compared to ACL IFP, in agreement with previously published data showing an increase of adipogenesis genes in end-stage OA IFP compared to early OA IFP [31].

The expression of other genes putatively involved in adipose tissue ECM organization and cell differentiation, such as *SERPIN2, GPNMB,* and *ITH5* considered as novel adipokines, was investigated only in OA IFP patients without any comparison with controls [32]. Differences in *GPNMB* and *ITH5* expression were observed between the OA and ACL IFP, with no differences for *SERPIN2*. Several studies highlighted the presence of fibrosis in OA IFP compared to non-OA tissues and subcutaneous adipose tissue [13,30]. GPNMB is a transmembrane protein involved in adipose tissue-derived inflammation in a mouse model of obesity [33]. We showed an increase of *GPNMB* in OA compared with ACL IFP with the increase of the inflammatory pattern in OA IFP. ITIs proteins are directly involved in the stabilization of ECM forming complexes binding hyaluronic acid molecules [34]. ITIH5 is highly expressed in white adipose tissue, and the expression seems to be increased in obese subjects [35]. Here, we observed an increase in its expression in OA IFP, in agreement with the increase of fibrosis. However, we cannot exclude that the differences observed in GPNMB and ITIH5 expression could be influenced by BMI, an important risk factor for OA.

The expression of other genes involved in the ECM composition was also investigated. ECM of adipose tissue is composed of several types of collagen and is particularly rich in COLVI that is positively correlated with BMI [36]. We evaluated *COLI, III,* and *COLVI,* showing that *COLI* and *COLIII* levels were decreased in OA compared to ACL IFP, while no differences were detected in *COLVI* expression. The differences in COLI and COLIII were also confirmed at the protein level, supporting the hypothesis of a change in ECM composition in OA, leading to the fibrotic phenotype of OA IFP [24,30]. This could also explain the reason why OA IFP has a biomechanical behavior different from that of healthy IFP [20].

Our study pointed out for the first time that OA pathology has a direct impact on IFP ECM and, in particular, on the expression of collagens and adipokines involved in the fibrotic process. These findings might open the possibility that fibrosis could be the target of a novel therapeutic strategy to counteract the OA progression and related pain in OA patients. This is also supported by a recently published paper by Onuma et al. that established a novel inflammation-induced knee pain model in rats and showed that the inflammation-induced fibrotic changes in the IFP were associated with the prolongation of joint pain [37]. The main limitation of our work is the age and BMI differences between the two groups of subjects, due to their specific categorization, given the fact that ACL rupture

usually occurs in young, lean athletes, while OA occurs mainly in aging females. However, we showed that our analysis was not influenced by these confounders by general linear models except for the adipocyte area in the OA group and the overall cohort. In particular, inflammatory features such as lymphocytic infiltration, vascularization, and IL-6 protein expression as evaluated by IHC were increased in OA IFP compared to controls independently from BMI, suggesting an important role of IFP inflammation in OA pathogenesis. Unfortunately, it was not possible to apply the linear models to the gene expression analysis due to the small sample size. In addition, we cannot exclude that the acute trauma occurring during ACL rupture influenced our analysis [38]. However, in our study, we have enrolled patients who underwent ACLR at least 6 months after the injury in order to avoid the inflammatory phase occurring after the trauma. Bigoni et al. analyzed cytokines levels in synovial fluid of patients with ACL tears divided in the study population into 4 groups according to the time elapsed from the injury. Those authors demonstrated that there was an increased level of pro-inflammatory cytokines in the acute phase of inflammation, followed by a decrease with a minimum of three months after injury [39]. Current studies on patients with ACLR are mainly focused on the evaluation of inflammatory markers in the synovial fluid. Heilmeier et al. correlated the synovial fluid inflammatory markers with the IFP/synovial membrane abnormalities detected by MRI [40]. They found that the degree of IFP abnormality correlated positively with the synovial fluid levels of the inflammatory cytokine markers and with chondro-destructive markers as MMP-1 and −3. Since these Authors did not analyze IFP and synovial membrane inflammatory cytokines expression, it is not possible to discern the contribution of each joint tissue in the secretion of these inflammatory markers [40].
