Gait data from a knee OA patient (male; age = 52 years; weight = 73 kg; height = 1.73 m; Body-Mass-Index (BMI) = 24.4 kg/m²) were used to estimate the loading conditions (Caravaggi et al., 2021). Knee joint flexion/extension angles and adduction/abduction moments (Leardini et al., 2007) were recorded in five comfortable walking-speed trials at 100 Hz and time-normalized over stance phase duration. The average knee contact force was obtained by using the linear relationship between knee adduction moment and contact forces reported by Kutzner et al. (Kutzner et al., 2013). Average knee pressure at the medial compartment was estimated by dividing the knee contact force by the contact area reported by Liu et al. (Liu et al., 2010). The estimated average peak pressure of 3 MPa, which is consistent also with what reported by Gilbert et al. (Gilbert et al., 2014) as contact stress at the knee in walking, set the maximum pressure for the sinusoidal pressure/time waveform used as loading condition for the cartilage samples (Caravaggi et al., 2021). A waveform frequency of 1 Hz was chosen to simulate a typical knee joint compression/relaxation pattern during normal walking. Each sample was subjected to cyclic compression for 45 min. This time was chosen to simulate a moderate stimulation of the knee, slightly above the minimum recommended exercise time of around 30 min per day (Lundebjerg, 2001) (Loew et al., 2012).

For study purposes, 19 mono-compartmental knee OA cases (Female/Male: 10/9; age±SD: 71 ± 9 years; mean BMI±SD: 27.3 ± 4.1 kg/m²; 13 medial compartment OA due to varus knee, six lateral compartment OA due to valgus knee) undergoing total knee replacement surgery were recruited. Femur’s mechanical axis alignments were 171.5 ± 4.8 deg and 174.6 ± 3.6 deg for varus and valgus cases, respectively. In mono-compartmental knee OA (with predominant biomechanical aetiology), the medial or lateral compartment in varus and valgus knee, respectively, is mainly degenerated compared to the contralateral one (Felson et al., 2013; Sharma et al., 2010, PMID 20511608). This allowed to compare articular compartments showing different cartilage degeneration within the same articular milieu, thus representing different stages of OA progression with minimum variability (paired samples). Accordingly, cartilage from the same knee was separated in affected areas, i.e., where tissue histology showed the worst score, and contralateral areas showing lower degeneration.

The study was approved by the local Ethic Committee (CE-AVEC Prot. Kneeload N. EM603-2018_89/2015/Sper/IOR_EM1) and written patient consent was recorded from all participants. Medial and lateral femoral condyles were collected at time of knee replacement surgery as described in Caravaggi et al., 2021 (Caravaggi et al., 2021). Medial and lateral condyles were divided in four topographical areas, corresponding to the original anatomical location and to different in-vivo loading conditions: medial anterior, medial posterior, lateral anterior and lateral posterior (Figure 1A). Differentiating the condylar region, as medial and lateral compartments, and within the same compartment, between anterior and posterior areas, is critical to apply the correct biomechanical conditions and to obtain more in-depth information on the biological modulations of cartilage samples which are consistent in terms of long-term wear and mechanical loadings in-vivo (Caravaggi et al., 2021).

A macroscopic score was assigned by the surgeon to all the identified areas, according to the Collins grading system of disease severity (from 0 to 4, with 0 corresponding to apparently normal cartilage, and 4 corresponding to completely degenerated tissue) (Pritzker et al., 2006; Caravaggi et al., 2021). Each area could appear homogeneous (with only one attributed macroscopic score) or further zoned in sub-areas in case of heterogeneous tissue degeneration. In the latter case, subzones from the same area were separately recovered and further analysed.

Cylinders of cartilage tissue were harvested with an 8 gauge-diameter corer (internal diameter: 2.5 mm) after subchondral bone removal and cultured overnight in D-MEM (SIGMA, Sigma Aldrich, St. Louis, United States) in serum starvation conditions. Ex-vivo mechanical compression was applied with the FlexCell FX-4000C stage presser apparatus (Flexcell International Corporation, United States), allowing to homogenously apply uniform loading to different specimens. Fresh cartilage cylinders were weighed and positioned into compression plate wells (consisting of a chamber bonded to the bottom of a flexible silicon rubber membrane) with 2 ml D-MEM without FCS. The cylinders were carefully oriented, all placed with the upper cartilage layer facing up. A stationary platen was added to each well and screwed down to the top of the tissue. Loading uniformity was guaranteed by the positioning of only one cartilage cylinder per well, therefore each compression chamber was set at the specific height of the sample. Air pressure applied to the bottom caused the chamber to rise and apply an unconfined compression to explants. All the experiments were performed at 37°C, 5% CO₂. A 3 MPa amplitude 1 Hz frequency sinusoidal compressive load was applied for 45 min. Paired control explants were maintained at 37°C, 5% CO₂ in unloaded conditions. For each donor, two conditions were tested: basal and stimulated with 2 ng/ml Interleukin (IL)-1β (pro-inflammatory stimulus; rhIL-1β, R&D Systems, Minneapolis, United States). For each experimental condition, two/three explants (depending on available cartilage) were used. After compression, cartilage samples and culture supernatants were immediately recovered and frozen (liquid nitrogen and −80°C, respectively) for molecular analyses and soluble factor determination.

A series of markers of cartilage turnover was identified to evaluate the effect of ex vivo loading. Some markers were quantified as mRNAs and others as proteins released in culture supernatants. Table 1 is reporting all the analyzed variables and their main function in cartilage metabolism.

Before compression, one cartilage sample from each sub-area was used for microscopic cartilage scoring to confirm the previously attributed macroscopic score. Samples were freshly collected, formalin fixed, and paraffin embedded until use. A representative number of cartilage samples from 16 out of 19 OA cases was scored after conventional Safranin O-staining. Briefly, 5 μm thick sections of FFPE samples were rehydrated and stained with Hematoxylin/Eosin, 0.1% Safranin-O and 0.02% Fast Green (SIGMA-ALDRICH, Munich, Germany). One slice from each sub-area was independently analyzed by two experienced biologists for histopathology grading score attribution following the OARSI criteria (Pritzker et al., 2006). Scores attributed ranged from 0 (intact cartilage), to 4 (complete cartilage degradation). Grades 5 and 6 were never attributed since the osteochondral bone was excluded from the analysis. The entire section was evaluated, and the analysis performed with an Eclipse 90i microscope and NIS elements software (NIKON CORPORATION, Tokyo, Japan).

Depending on the available material, one or two cartilage cylinders (20–80 mg) for each experimental condition (and for each donor sample) were dedicated to gene expression analysis, as described in Caravaggi et al., 2021 (Caravaggi et al., 2021). Briefly, liquid nitrogen frozen samples were pulverized using the Mikro-Dismembrator S (Sartorius Stedim Italy SpA, Italy) grinding mill in 5 ml PFTE shaking flasks with a stainless-steel grinding ball at a shaking frequency of 2000/min, for 45 s. Total cellular RNA was extracted using RNApure (EUROCLONE, Milan, Italy) and reverse transcribed using SuperScript VILO cDNA Synthesis kit (INVITROGEN, Life Technologies, NY, United States) by random hexamer priming, following manufacturer’s instructions.

Messenger RNA expression was evaluated by semi-quantitative Real-Time RT-PCR in a Light Cycler Instrument (ROCHE Molecular Biochemicals, Mannheim, Germany) using the SYBR Premix Ex Taq (TAKARA Biomedicals, Tokyo, Japan) with the following protocol: 95°C for 10 s, 45 cycles at 95°C for 5 s and 60°C for 20 s. Analyzed genes and primers sequences are shown in Supplementary Table S1.

Cycle Threshold (C_T) values were determined for each sample. Specificity of the amplicons was checked by gel electrophoresis and confirmed at each run by melting curve analysis. Amplification efficiency (above 90% for each primer couple) was evaluated using 10-fold serial dilutions of positive control cDNAs and calculated from the slopes of log input amounts plotted versus crossing point values. Expression levels were quantified in respect to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) housekeeping gene following formula (1 + E)^ΔC _T, where E represents the reaction efficiency and ΔC_T the difference between the GAPDH crossing point and the specific crossing point for each sample.

Matrix MetalloProteinase 13 (MMP13), a Disintegrin And Metalloproteinase with Thrombospondin motif 5 or Aggrecanase-2 (ADAMTS-5), Aggrecan (Aggr), Cartilage Oligomeric Matrix Protein (COMP), Procollagen II C-Terminal Propeptide (PIICP), Chondroitin Sulfate (CS), and Collagen Type 2 Cleavage (COL-2CAV) were measured by ELISA in culture supernatants recovered after compression, according to manufacturer’s instructions (Cloud-Clone Corp, United States for MMP13, ADAMTS-5, Aggr, COMP, PIICP and CS kits; MyBioSource, United States for COL-2CAV) using Infinite^® M Plex multimode microplate reader equipped with Magellan Software (Tecan Trading AG, Switzerland) and a programmable microplate washer (Hydro flex Tecan Trading AG, Switzerland). Results were normalized to mg of tissue of the corresponding explant. Due to bioreactor technical requirements, each cartilage cylinder was cultured in 2 ml medium, resulting in high dilution of released factors in supernatants. For this reason, supernatants were vacuum concentrated before ELISA test.

Data are presented as medians, 25^th and 75^th percentiles, minimum and maximum values; percentages, means ± standard deviations, as appropriate.

The main analysis was the Generalized Linear Mixed Models (GLMM) analysis with cartilage markers as dependent variables; articular tissue sample position and grading, knee OA compartment, IL-1β stimulus and compression as fixed effects. The analysis performed resulted in estimated means with 95% CI; the comparisons between the subgroups were performed using the Sidak multiple comparisons test. Results in graphs and tables are resumed with medians and 25^th–75^th percentiles to facilitate reading of data that were highly asymmetrical.

A flowchart of the statistical analysis is presented in Table 2. Briefly, to find the suitable distribution and link function for the model, the Shapiro test was used to assess normality of the distributions of the cartilage markers, which were all not normally distributed. Subsequently, logarithmic transformation was applied only to those markers with all positive values and the Shapiro test was used to assess log-normality of these latter markers (stage 1). If cartilage markers presented zero-values, Tweedie distribution with log-link function was used in the model; in case of absence of zero-values and not log-normal distribution the gamma distribution with log-link function was used (stage 3).

The random effects to be used in the model were chosen between age, BMI, OA grade and gender using the appropriate univariate analysis (Spearman’s Rank correlation for age and BMI; Kendall’s tau correlation OA for grade; and Mann Whitney U-test for gender) (stage 2).

To choose the right model, the complete model was firstly applied with all the interactions, and then run without the not significant interactions, possibly with only the fixed main effects (stage 4).

The Sidak test was then used to compare the results of fixed effects.

The level of statistical significance was set at p < 0.05. All statistical analysis was performed using SPSS v.19.0 (IBM Corp, Armonk, NY, United States).

In Figure 1, representative examples of surgical samples with attribution of macroscopic score, together with explant collection via tissue coring, are shown. Analyses were performed separately in these areas. Patients’ characteristics are summarized in Table 3.

Samples from mono-compartmental knee OA cases, corresponding to varus (medial condylar compartment prevalently affected) or valgus knee (lateral condylar compartment prevalently affected) allowed to compare, in the same donor sample, the behaviour of areas with distinct pathology grade. Accordingly, a correspondence between the affected articular compartment and the macroscopic score was observed: varus knee cases showed higher OA scores in medial compartments (median medial value 2.50; median lateral value 1.00, p = 0.003), whereas valgus knee cases showed higher OA scores in lateral compartments, albeit without reaching statistical significance (median medial value 1.50; median lateral value 2.00, p = ns) (Figure 2).

A histopathologic score was attributed to a total of 37 previously identified cartilage areas from 16 donors. These areas carried a macroscopic score attribution from 0 to 3 (no areas with macroscopic score 4 were included in correlation analysis since this score corresponded to completely degenerated cartilage and no histology was possible). In Figure 3A, a representative example of histological evaluation in areas with different macroscopic score, is shown.

No sample (even those with macroscopic score = 0) showed histological score corresponding to 0, thus confirming that even apparently intact cartilage from an OA joint carries tissue alteration (Figure 3B). Spearman correlation analysis highlighted a positive correlation between the two scores (r = 0.99 between macroscopic scores and mean microscopic scores; Figure 3B), in accordance with previous results (Caravaggi et al., 2021). This confirms the reliability of the macroscopic score for sample partition in areas of different pathology level before compression.

Tissue response to mechanical stimulation was analyzed as the effect of compression both in relationship to the anatomical original position and to the level of cartilage degeneration. To this aim, samples from areas differently affected were compared. Based on the correlation between macroscopic score and cartilage position (Figure 2), cartilage originating from varus knee medial compartment and cartilage from valgus knee lateral compartment were considered as “affected compartment”, whereas cartilage from varus knee lateral compartment and valgus knee medial compartment were considered as “contralateral compartment”.

SOX9 gene expression showed no significant differences before and after compression, even in IL-1β stimulated samples; COL2A1 was reduced in contralateral compartment compressed unstimulated samples (p = 0.0142) compared to the affected compartment and aggrecan was reduced in uncompressed stimulated samples of the contralateral compartment compared to the affected compartment (p = 0.04) (Figure 4).

The analysis of markers involved in inflammation and innate immunity as well as soluble factors, was performed in a sub-group of IL-1β stimulated samples (n = 60). Interestingly, we observed a downmodulation of IL-6 and IL-8 expression after compression. This only happened in contralateral compartment cartilage samples (p = 0.0126 and p = 0.0327, respectively), whereas compression failed in downmodulating IL-6 and IL-8 expression in the affected compartment (Figure 5). Moreover, if considering the effect of compression in relationship to the anatomical position, a prevalent effect of compression is evident in the anterior contralateral compartments compared to posterior. In fact, the downmodulation of both IL-6 and IL-8 gene expression appears in both compartments, but it reaches statistical significance only in the anterior one (p = 0.02 for IL-6 and p = 0.02 for IL-8) (Figure 5).

Compression produced no effect on IL-4Rα, C3, and CFB expression (Figure 6). Even for these markers, differences in gene expression emerged if comparing anterior and posterior areas: IL4-Rα anterior areas of the affected compartment presented, after compression, significantly lower gene expression values than posterior areas of the same compartment and of the contralateral anterior compartment, again highlighting position-dependent differences (Figure 6A).

We noticed that soluble factors were the most affected by compression among the molecules analyzed. Indeed, COL-2CAV release was significantly downmodulated by compression only in contralateral compartment cartilage samples (p = 0.0367); moreover, for this marker, anterior and posterior areas showed the same trend (Figure 7A).

ADAMTS5 expression resulted, on the contrary, upregulated by compression in the contralateral compartment (p = 0.030; Figure 7B); the same trend, even without reaching statistical significance, was observed in the affected compartment and both in the anterior and posterior areas (Figure 7B).

PIICP and COMP release was increased after compression in the affected compartment (p = 0.013 and p = 0.018, respectively); the same trend, even without reaching statistical significance, was observed in the contralateral compartment and both in the anterior and posterior areas (Figures 7C,D).

Finally, CS release was again stimulated by compression, both in the affected and in the contralateral compartments (p = 0.006 affected; p = 0.006 contralateral). The trend was the same if considering sample position, with a statistically significant increase in the posterior affected compartment and in the anterior contralateral compartment (Figure 7E).