Functional MRI Mapping of Human Meniscus Functionality and its Relation to Degeneration

Meniscus pathology may promote early osteoarthritis. This study assessed human meniscus functionality (i.e. its response to loading) ex vivo based on quantitative T1, T1ρ, and T2 mapping as a function of histological degeneration and loading. Forty-five meniscus samples of variable degeneration were harvested from the lateral meniscus body region of 45 patients during total knee arthroplasties. Samples underwent serial mapping on a 3.0-T MRI scanner (Achieva, Philips) using a force-controlled and torque-inducing compressive loading device. Samples were measured at three loading positions, i.e. unloaded, loaded to 2 bar (compression force 37 N) and 4 bar (69 N). Histology (Pauli classification) and biomechanics (Elastic Modulus) served as references. Based on histology, samples were trichotomized as grossly intact (n = 14), mildly degenerative (n = 16), and moderate-to-severely degenerative (n = 15) and analyzed using appropriate parametric and non-parametric tests. For T1, we found loading-induced decreases in all samples, irrespective of degeneration. For T1ρ, zonal increases in intact (apex) and decreases in degenerative samples (base) were found, while for T2, changes were ambiguous. In conclusion, force-controlled loading and serial MR imaging reveal response-to-loading patterns in meniscus. Zonal T1ρ response-to-loading patterns are most promising in differentiating degeneration, while T1 and T2 aren’t clearly related to degeneration.and may provide an imaging-based indication of functional tissue properties.

. Histologically, this sample demonstrated slight fraying of the meniscus apex zone but otherwise smooth surfaces without fraying or fibrillation (score 1), normal cell distribution (score 0), diffuse foci of degenerated extracellular matrix (score 1), and moderate staining intensity for proteoglycans (score 2). Pauli sum score 4 (Pauli grade-I). Of note, focal calcifications of the tibial meniscus surface are only secondary histological findings and not relevant to the histological scoring.

Results
General findings. All 45 samples underwent complete MRI measurements and subsequent histological and biomechanical reference characterization. In all samples, sufficient tissue thickness was available for these analyses.
Elastic Modulus at strains 20% and 80% was not significantly different between the subgroups: strain 20%: 15 Unloaded, we did not find significant degeneration-dependent differences in the various regions-of-interest (ROIs) for T1 or T1ρ. For T2, however, we determined significant differences in the apical zone with significantly lower T2 medians in Pauli grade-I than in Pauli grades-II (and higher) samples (p = 0.004). Overall, qMRI parameter values tended to be highest in Pauli Grades ≥ III samples and in the apical zone. Unloaded absolute qMRI parameters are given in Table 1.
Qualitative changes in response to loading. Qualitatively, loading-induced changes were variable across the samples. Meniscus samples were subject to substantial compression and deformation under loading. Consequently, the physiological wedge shape underwent considerable flattening with increased loading.
In Pauli Grade-I samples, only the apical and intermediate zones underwent slight changes, i.e. decreases in T1, and increases in T1ρ and T2, while the base zone was largely unaltered (Fig. 2). In Pauli Grade-II and ≥ III samples, the apical zone underwent largest changes. In some samples, signal characteristics became more homogeneous so that pre-existent signal hyperintensities gradually disappeared (Fig. 3), while in other samples, focal signal hyperintensities that were not well appreciable in the unloaded configuration became more prominent under loading, in particular at the apical zone (Fig. 4). All samples, irrespective of degeneration, were characterized by substantial loading-induced decreases in T1 throughout the entire cross-sectional area, whereas changes in T1ρ or T2 characteristics were less uniform. In Figs Quantitative changes in response to loading. Quantitatively, loading-induced adaptations as detailed above were reflected by several changes: First, mean pixel numbers decreased significantly from 801 ± 237 (unloaded, δ 0 ) to 718 ± 263 (loading position 1, δ 1 ), and 604 ± 236 (loading position 2, δ 2 ) (p < 0.001, entire sample). Second, we found distinct changes in T1, T1ρ and T2 in response to loading. Table 1 summarizes absolute  qMRI parameter values as a function of increasing loading and histological Pauli Grades, while Table 2  Pauli-grade III ( †) and are printed in bold type. Significant differences are indicated by "*" and non-significant differences by "ns". δ 0 refers to the unloaded configuration, while δ 1 and δ 2 refer to loaded configurations at 2 bar (compression force 37.1 N) and 4 bar (69.1 N). www.nature.com/scientificreports www.nature.com/scientificreports/ larger at the meniscus base and related to loading intensity, i.e. larger loading induced larger decreases (e.g. all samples, entire ROI: Δ 1 : −15.1 ± 7.0%, Δ 2 : −20.5 ± 8.5%, p = 0.001). Significantly larger relative changes were found for the intermediate zone and the entire sample.  www.nature.com/scientificreports www.nature.com/scientificreports/ For T1ρ, we made numerous distinctly different observations. Even though mean relative changes were not significantly different for the entire samples (all samples, entire ROI: Δ 1 : 0.9 ± 13.1%, Δ 2 : 3.6 ± 17.6%, p = 0.425), they were distinctly different in the meniscus zones. First, we found -by trend-loading-induced decreases at the base zone. These were significant in Pauli Grades ≥ III samples (base ROI: For T2, we found ambiguous zonal loading responses, even though overall, decreases were more prevalent than increases. Even though in all samples, significant decreases were found with increasing loading (all samples, entire ROI: δ 0 : 24.7 (22.8-27.7) ms, δ 1 : 24.2 (22.5-26.0) ms, δ 2 : 23.0 (21.9-24.5) ms, p = 0.001), further zonal analysis revealed that these significant changes were exclusive to the apical zone with differences being significant in all samples (p < 0.001), Pauli Grade-II samples (p = 0.002) and Pauli Grades ≥ III samples (p < 0.001).For the other zones, however, changes were not significant. By trend, relative changes tended to be larger at higher loading intensity, even though significant differences were only found for the apical zone of all samples (all samples, apex ROI: Δ 1 : 0.9 ± 13.5%, Δ 2 : −9.1 ± 16.3%, p = 0.002).

Discussion
The most important findings of this study are 1) lateral human meniscus body samples undergo distinctly different zonal changes in T1, T1ρ and T2 in response to loading and 2) these changes are associated with histological degeneration but not with biomechanical properties, i.e. Elastic Modulus. More specifically, in histologically grossly intact meniscus, the physiological response to loading is defined by decreases in T1, divergent changes . Data are given as mean ± standard deviation. Δ 1 (or Δ 2 ) were calculated by relating the absolute qMRI parameter values at loading position δ 1 (or δ 2 , respectively) to the unloaded configuration δ 0 . Histological Pauli grade-related differences for the distinct regions-of-interest at the successive loading positions were assessed using one-way ANOVA tests ( †), while differences between Δ 1 and Δ 2 were assessed using unpaired Student's t-tests ( §). Statistically significant are printed in bold type. Please refer to www.nature.com/scientificreports www.nature.com/scientificreports/ in T1ρ with increases at the apex and decreases at the base, and ambiguous changes in T2. In severely degenerative meniscus however, the pathological response to loading is characterized by homogenous decreases in T1ρ throughout the tissue, while for T1 and T2, similar changes as for the intact samples were found. Across all samples and histological subgroups, substantial inter-individual variability could be determined. Even though we employed statistical measures and tests that decrease sensitivity to outliers, the substantial variability certainly prevented clearer statistical inferences and biological conclusions. Except for T2 in the apical zone, we could not find any significant degeneration-related differences for any ROI in the unloaded configuration. In all samples, we found that the apical zone (by trend) had the highest and the intermediate zone the lowest T1, T1ρ, and T2 values. This finding is well in line with recent in-vivo and in-vitro studies that reported regional and zonal variability in human meniscus due to tissue heterogeneity in structure, composition and biomechanical properties 4,13,[22][23][24] .
In our study, loading induced morphological deformation and flattening of the wedge shape and, correspondingly, significant decreases in the compressed samples' pixel numbers, which is indicative of sufficient sample pressurization. Within the confines of our study's in-vitro design, the loading intensity may be considered grossly representative of the in-vivo situation. The applied compressive forces of 38 N (δ 1 ) and 69 N (δ 2 ) were actuated on meniscus samples of standard length (15 mm) and variable width (mean: 8.4 mm) 25 , resulting in a mean sample surface area of 126 mm² and mean pressure levels of 0.30 MPa (δ 1 ) and 0.55 MPa (δ 2 ), respectively. For the lateral meniscus body region, such pressures are well in line with those experienced during level walking (ca. 0.5 MPa) 26 , even though peak pressures may be considerably higher during unlevel walking, running or stair climbing. Thus, our loading regime is representative of static loading configurations as in prolonged one-or two-leg standing.
Within the premise of this study's in-vitro setting, we consider the inclusion of two different loading regimens (alongside the unloaded configuration) to provide a thorough understanding of the loading-induced intra-tissue adaptations and to indicate if (and to what extent) variations in loading intensities are related to variations in tissue functionality as assessed by qMRI parameters.
The observed changes in the T1, T1ρ, and T2 maps are grossly reflective of the morphological changes. Overall, the samples' responses to loading were variable. For T1, we found consistent loading-induced decreases throughout all ROIs of all histological Pauli Grades. For cartilage, T1 relaxation has been linked to tissue hydration 27 , while for meniscus, the exact structural and/or compositional correlate is yet unknown. The loading-induced decreases are most likely due to water redistribution within the tissue, its partial loss and relative increases in the solid matrix constituents. Decreases in T1 were larger with more intense loading, in particular in the intermediate zone, which may be explained by the fact that this zone is biomechanically most compliant 28 . Mean overall decreases in T1 tended to be larger at the meniscus base, which is plausible, too, given the higher water content 4 and the abundant amount of loose vascularized connective tissue 29 . Under loading, T1 values decreased in all samples, irrespective of degeneration, which indicates the dominant contribution of water to T1 relaxation that most likely overpowers the contribution of the biochemical variations in collagen and/or proteoglycans 4,30 . Also, MRI measurements were performed after extended sample storage in medium which may have increased tissue hydration beyond the constitutively high water content.
Loading-induced changes in T1ρ were less homogeneous. While we found decreases at the meniscus base in all samples, changes at the meniscus apex and -to a lesser extent-the intermediate zone were significantly different in the distinct Pauli Grade subgroups. In Pauli Grade-I samples, T1ρ values were increased, while in Pauli Grades ≥ III samples, they were decreased. Similar to T1, the consistent decreases in T1ρ at the meniscus base are most likely secondary to the fact that T1ρ relaxation is dominated by extracellular water, especially in advanced degeneration 4 . Consequently, degeneration-related changes in the a-priori heterogeneous nature of the tissue are responsible for the discrepant changes at the apical (and intermediate) zones. Under loading, the tight collagen network is compressed and condensed. Intuitively, one would speculate that T1ρ values are decreased because of more restricted spin motion. In fact, T1ρ values were increased in Pauli Grade-I samples, which is in line with previous in-vivo data 23 . As T1ρ probes the low-frequency interactions of the tissue's macromolecules and bulk water 31 , physiological adaptations of the structurally intact matrix, i.e. changes in collagen fibre anisotropy, restrictions in water redistribution and upheld swelling pressure, may have led to these increases. Nonetheless, the significantly different loading responses may indicate sufficient load transmission (in Pauli Grade-I samples) and its failure (in Pauli Grades ≥ III samples), thereby corroborating earlier findings 22,23 , even though the exact biophysical correlates of these changes remain unclear.
For T2, the response-to-loading patterns were ambiguous and not related to histological degeneration. In meniscus, T2 is widely considered to be sensitive to water interactions and related to content, orientation and anisotropy of the collagen fibers 19 . Loading-induced decreases were more prevalent than increases, which is primarily indicative of changes in water distribution and content 4 . Consequently, stronger loading was associated with larger changes in T2, at least for the apical zone. Here, loading induced significant decreases in T2 in all samples except for Pauli grade-I samples, which is an indication of more pronounced degeneration-dependent changes in the apical zone. Due to its lack of intrinsic ability for self-repair, the inner zone is considered particularly prone to permanent tissue damage 32 , and consequently, degenerative tears usually extend from the inner zone to the periphery 33 . This clinical observation was confirmed in our study as tissue fraying and tearing in more degenerative samples usually originated from the apical zone.
Surprisingly, we did not find any significant differences in the biomechanical properties of the histological subgroups. As earlier reports on the relationship of meniscal degeneration and biomechanical properties are conflicting 34,35 , this finding is only partially in line with the pertinent literature. Potential reasons for this discrepancy relate to differences in study design, reference evaluation, sample numbers, degeneration severity, harvesting locations and biomechanical testing protocols. In their comparative evaluation of medial and lateral menisci, Katsuragawa  www.nature.com/scientificreports www.nature.com/scientificreports/ standardized cylindrical specimens were prepared from the anterior horn and subjected to confined compression testing to determine aggregate modulus and permeability (as measures of the solid and fluid components' contributions, respectively). Although histological work-up was performed, degeneration severity was only assessed qualitatively. The authors did not find any significant differences in the biomechanical parameters for lateral menisci from OA versus non-OA knees, while for medial menisci, they determined large differences indicating considerable softening in degenerative medial menisci. In contrast, Fischenich et al. found progressive decreases in biomechanical properties with increasing tissue degeneration 35 . The authors included 24 medial and lateral meniscus pairs from TKAs, sectioned them into anterior and posterior regions and performed i) indentation relaxation tests to determine the instantaneous and equilibrium compressive moduli and ii) tension testing to determine tensile modulus. Meniscal degeneration was assessed macroscopically based on inspection of gross tissue morphology, while histology was not performed. While tensile modulus was largely retained throughout degeneration, significant degeneration-dependent decreases in instantaneous and equilibrium compressive moduli were found. Whatever the exact relation of meniscal degeneration and biomechanical properties, fluid flow seems to predominate over matrix composition in determining the compressive properties of meniscus 7 , which suggests that even (histologically) moderate-to-severely degenerative meniscus upholds viscoelastic capacities to withstand compressive loading, at least when quantified by qMRI mapping techniques. This is in line with our findings that suggest that the structural and/or compositional correlates of T1 and T2 (and their loading-induced changes) are not significantly different across the different grades of histological degeneration. In contrast, zonally discrepant changes in T1ρ, i.e. significant increases at the apex in grossly intact samples and significant decreases at the base in moderate-to-severely degenerative samples, indicate that the correlates of T1ρ undergo degenerative processes that may be detected by functional imaging techniques. Moreover, the underlying tissue model/theory used to determine the mechanical properties of the meniscus samples and their assumptions are highly relevant. In the present study, we used a traditional two-parameter exponential model 36 that is common in modelling approaches of nonlinear properties of fibril-reinforced tissues in tension such as meniscus 35,[37][38][39] . Nonetheless, alternative models such as the linear biphasic theory 40 have also been applied before to quantify mechanical properties of meniscus 34,41 .
This study has several limitations. First, our study's ex-vivo setting limits overall transferability to the in-vivo setting. Alongside differences in tissue hydration due to sample storage, additional factors such as temperature, culture conditions, preparation procedure, and sample position and orientation as well as the fact that we performed the MRI measurements in standard saline solution may alter relaxation characteristics and consecutive response-to-loading patterns 42 . Nonetheless, in a clinically relevant basic research context, this study comprehensively assessed degeneration-related response-to-loading patterns in human meniscus. For once, measurements were performed on a clinical MRI scanner (and clinically applicable sequence parameters) and on samples from a clinical population representative of the entire continuum of health and disease. Meanwhile, the increasing interest in the association of joint imaging and joint mechanics (with its traditional focus on articular cartilage) has produced a number of MRI-compatible loading devices that are intended to realize the clinical translation of stress MRI techniques. Loading of patients' knee joints (and menisci) is based on suspending weights via dedicated pulley systems [43][44][45][46][47] or direct compressive loading along the leg axis [48][49][50] . In practical terms, further in-situ and in-vivo studies assessing meniscus functionality on the basis of such loading devices and quantitative MRI techniques are necessary to corroborate our study's findings of the partial relatedness of loading-induced qMRI parameter changes and (histological) degeneration.
Second, the compressive loading device does only partially emulate physiological loading. While the experimental setup provides indications of the intra-tissue changes secondary to compressive loading that goes through the meniscus, its reflection of the complex biomechanical environment during loadbearing in vivo (that involves tensile and shear stresses 2,26,51 ) is limited. Secondarily, the meniscus undergoes complex motion and adaptations, thereby deforming and displacing the perimeniscal soft tissues 51,52 . Another related aspect involves dynamic contact mechanics across the joint that change during activities of daily living, e.g. gait and stair climbing, and has considerable effects on resultant peak contact stresses of meniscus and cartilage 26 . Despite these limitations, our study's implementation of standardized loading of meniscus samples in vitro is a basic requirement to realize reproducible and comparable loading regimes across sample cohorts at optimized image resolution and signal-to-noise ratio. Nonetheless, for more complex biomechanical evaluations, in-situ and in-vivo configurations are critical. Moreover, we deliberately chose an elastic polyvinylsiloxane inlay to circumferentially confine the meniscus samples. Under loading, the inlay was deformed as a sign of effective load transmission. This, however, increases the complexity of the subsequent biomechanical analyses, e.g. in finite element simulations. Third, even though sample size allowed for sound statistical analyses and valid inferences, it was limited, not least due to the study's single-institution design. At this stage, it remains speculative -in particular when considering the qMRI parameters' substantial statistical variability-if larger sample sizes might have brought about less equivocal findings. Fourth, as meniscus properties and their imaging correlates demonstrate considerable zonal and regional variability 4,13,22,23 , samples were harvested from the lateral meniscus body region only for the sake of topoanatomic consistency. Accordingly, our findings are only reflective of this region and may be different for the anterior and posterior regions of the lateral meniscus, let alone the medial meniscus. Fifth, samples were only harvested from total knee arthroplasties, which raises the question of our sample cohort's representativity when it comes to truly healthy samples. In this regard, future studies should include alternative sample sources with a focus on young non-OA donors to be more representative of the entire spectrum of health and disease. Another related aspect concerns the semiquantitative histological scoring used as reference standard. Histological signs of degeneration across one sample may be variable, thereby rendering mean (subcomponent) scores potentially unrepresentative of focal alterations in matrix structure and composition that, however, may substantially affect qMRI characteristics. Sixth, even though our findings indicate an exciting scientific approach to further differentiate meniscus in health and disease, statistically significant differences between histological subgroups do not www.nature.com/scientificreports www.nature.com/scientificreports/ imply the ability to diagnose on qMRI parameters alone. With reproducibility of research being an increasingly important cornerstone of modern biomedical research, additional translational studies need to be conducted to confirm that the statistically significant differences observed in this study are of scientific and diagnostic relevance. In this regard, the apical meniscus zone should be primarily considered in efforts to differentiate the tissue's functional status. Seventh, for biomechanical evaluation in terms of unconfined compression testing, an invasive preparation procedure was chosen that was characterized by resection of the top and bottom of the specimen cylinder. Even though biomechanical properties were certainly altered, the methodology is scientifically sound and biomechanically valid because the innermost third of the specimen, which largely determines compressive properties 53 , was entirely included in all specimens. Eighth, the use of a standard 3.0 T MRI device and the choice of sequence parameters was intended to facilitate eventual translation of our findings into clinical practice. While a scanner with higher field strength (e.g. 7.0 T or 9.4 T) would have provided better signal-to-noise ratio, such scanners have not (yet) seen widespread clinical application and relaxation times are different from those at 3.0 T 54 . We deliberately chose sequence parameters similar to those in earlier studies [55][56][57][58][59] to render generalizability and transferability of the findings less related to actual sequence parameter settings.
In conclusion, this study is the first to systematically assess human meniscus functionality based on advanced quantitative MRI techniques and in relation to histological and biomechanical reference measures. Under standardized compressive loading, we determined different response-to-loading patterns across the meniscus zones that are related to histological degeneration. Changes in T1ρ relaxation and in the apical zone seem to indicate distinctly different loading-induced intra-tissue adaptations and may help to identify potential load transmission failure in degenerative joint disease. However, tissue hydration likely overpowered the contribution of biochemical alterations in this in-vitro study, which warrants further in-vivo studies. We used dedicated software for statistical power analyses (G*Power freeware from Heinrich-Heine-University Düsseldorf, v3.1.9.4, URL: http://www.psychologie.hhu.de/arbeitsgruppen/allgemeinepsychologie-und-arbeitspsychologie/gpower.html) 61 to determine minimum sample sizes. Based on earlier studies 22,23 , we estimated a minimum sample size of 39 for a statistical power of 0.9, a type-I error of 0.05, an effect size of 0.5, and three groups undergoing three measurements.

Methods
After surgery, we prepared the specimens according to standard. First, we identified the meniscus body region of each surgical specimen and removed any adherent capsular soft tissues (Fig. 5a 1 ). Second, we cut the specimen to standard anteroposterior diameter (i.e. length) of 15 mm using a cutting device that contained notches of variable depth (Fig. 5a 2 ). Specimen width was not altered (Fig. 5a 3 ). Tissue-marking dye (Polysciences, Warrington, US) was used to define the mediolateral imaging plane (from meniscus apex to base) for future reference. Samples were stored in sterile DMEM medium (Gibco-BRL, Gaithersburg, US) containing 100 U/ml penicillin, 100 μg/ml gentamycin and 1.25 U/ml amphotericin B (all from Gibco-BRL).
force-controlled compressive loading device. We applied standardized compressive loading on the meniscus samples using an MRI-compatible compressive loading device, which loads meniscus samples by means of torque 62 . In practical terms, we placed individual meniscus samples into the dedicated lever device within the sample box attached to the pneumatically driven sample tray (Fig. 5a 4 ). By control of set pressure levels, the sample tray was displaced upwards towards a half-sphere attached to the vertically adjustable cover screw within the device's upper frame (Fig. 5b). The upward movement was actuated by a pneumatic mechanism and resulted in a defined torque moment on the compression lever that conformed well to the sample's wedge shape. Confining inlays (polyvinylsiloxane, Wirosil, Bego, Germany) circumferentially contained and supported the samples to prevent loading-induced sample displacement. We selected appropriately sized confining inlays to contain half of the meniscus base height. Additionally, polymethylmethacrylate (PMMA) plates of corresponding width were placed between inlay and sample box outlines. We loaded samples by control of pressure via an electronically actuated valve with high precision (±0.01 bar, Type: VPPM-6L-L-1-G18-0L6H-V1P-S1C1, Festo, Esslingen, Germany), a digital-to-analogue converter (Multifunction I/O USB-6001, National Instruments Corporation, Austin, USA) and customized software routines implemented in LabVIEW software (v2017, National Instruments Corporation, URL: https://www.ni.com/de-de/shop/labview.html). To this end, we connected the pneumatic mechanism via pressure lines (Festo) to the standard hospital pressure outlet providing pressure levels of up to 4.69 bar. The control components were located outside the MRI scanning room. During MRI measurements we set target pressures www.nature.com/scientificreports www.nature.com/scientificreports/ to three different levels, i.e. p 0 = 0 bar (δ 0 , i.e. unloaded), p 1 = 2 bar (δ 1 ), and p 2 = 4 bar (δ 2 ). Earlier pressure-force calibration studies based on digital sensors (K-Scan 4000, Tekscan, Boston, USA) indicated that set pressure levels resulted in highly reproducible compressive forces of f 1 = 37.1 ± 0.8 N (δ 1 ) and f 2 = 69.1 ± 1.1 N (δ 2 ) on the half-spherical piston and resultant torques of t 1 = 0.67 Nm (δ 1 ) and t 2 = 1.24 Nm (δ 2 ) on the compression lever 62 .

MRi measurements.
Within 24 h after preparation, the meniscus samples underwent standardized MRI measurements in standard saline solution using a clinical 3.0 T MRI system (Achieva, Philips, The Netherlands) and the MRI-compatible compressive loading device. We selected appropriately sized polyvinylsiloxane inlays and PMMA plates to confine the sample circumferentially (Fig. 5). Next, we adjusted the height of the half-sphere (using the cover screw) to be in loose contact with the lever and the femoral meniscus surface. Imaging was performed using a modified single-channel (receive-only) prostate coil (BPX-30 endorectal coil, Medrad/Bayer, Germany) stripped off its inflatable balloon that circumferentially comprised the sample box. Radiofrequency pulses were applied via the scanner's body coil and scout views were acquired to guide the imaging sections along the mediolateral plane for each sample and loading position individually. The imaging protocol consisted of Proton Density-weighted (PDw) sequences in the axial, coronal and sagittal orientation and T2, T1ρ, and T1 maps in the coronal orientation (Table 3). We imaged the samples at the reference configuration, i.e. δ 0 , to be followed by two consecutive pressure levels, i.e. δ 1 and δ 2 . After each change in pressure level and before initiating the MRI measurements, we observed an equilibration period of 5 min. This time period was based on an earlier study that investigated stress relaxation as a function of physiological strains. Chia and Hull found equilibration in human meniscus to be complete after three to four minutes 39 . In other studies, considerably longer equilibration periods of up to 20 min were observed 35 . Loading was brought about by pneumatically controlled upward displacement of the sample box (2) containing the lever device (3) and meniscus sample (4). The sample box was screwed onto the pneumatic piston (5) that had been laid out to generate (measured) forces in the range of 0-76N 62 once the pressure chamber (6) was filled with air via the pressure connection port (7) and pressure lines (8). A half-sphere attached to the vertically adjustable cover screw contained within the device's upper frame (9) then induced defined torque loading on the samples. Samples were confined by appropriately sized polyvinylsiloxane inlays (10) and PMMA plates (11) to prevent their lateral displacement upon loading. A dedicated modified prostate single-channel coil (12) that circumferentially comprised the sample box was used for imaging. c) Loading device in operation in the clinical 3.0-T MRI scanner. the manual segmentations of the meniscus samples. Using the morphological PDw images obtained along the mediolateral plane as references, segmentation masks of sample outlines were generated while excluding boundary pixels to eliminate partial volume effects. SN (clinical MSK radiologist, 6 years of experience) validated these masks against the processed T1, T1ρ, and T2 images. In addition to the entire sample outlines, further zonal ROIs were defined by automatic subdivision of segmented outlines using a custom-made routine implemented in MATLAB (MatlabR2018b, Natick, USA). First, this routine determined the sample's maximum mediolateral diameter by projecting the segmented outline's most peripheral pixels onto a horizontal line parallel to the lever device's bottom. Second, after dividing this line into thirds, the division was back-projected onto the segmented outline to obtain the apical zone (I), intermediate zone (II), and outer zone (III) (consistent with the white-white, red-white, and white-white zones as defined histologically) 2 . Zonal ROIs were thus defined by these vertical thirds (as inner borders) and the segmented sample outline (as outer borders) (Fig. 1b).
Biomechanical and histological reference evaluation. For biomechanical assessment, we harvested two cylindrical samples of 4 mm diameter from the red-red to the red-white transition, bilaterally adjacent to the mediolateral plane, using a 4mm-diameter biopsy punch (pfm-medical, Cologne, Germany) that was oriented perpendicular to the tibial surface. After resection of the femoral and tibial meniscus surfaces, we cut the punched-out core cylinders to standard height of 3 mm and carried out unconfined compression testing on a materials testing machine (Z2.5, Zwick/Roell, Ulm, Germany) equipped with a 20 mm wide rigid and impermeable piston. A tare load of 0.2 N was applied to maintain standardized interaction of the interfaces and mechanical equilibrium throughout the measurements. No additional preconditioning loads were applied. As in previous studies 63, 64 , we applied a displacement rate of 0.0083 mm/sec and a strain of 100%.
The resultant stress-strain data obtained were fitted to a two-parameter exponential model 36,37 . Here, the strain energy Ψ is given by 2 where b scales the stress response and c gives the degree of nonlinearity, while ε is strain. The resulting expression for the Elastic Modulus EM is given by the second derivative with respect to ε, hence For each sample, non-linear optimizations were run in MATLAB to specifically fit the material parameters. Eventually, we determined the Elastic Modulus EM for the strains of 20% and 80% to obtain representative data from high and low strains and to limit the amount of data for subsequent multiple comparisons. Representative samples' measured stress-strain curves and modelling predictions are included as supplementary material (Supplementary Figure 1).
For histological assessment, samples were fixed in paraformaldehyde (4%) for 7 days, embedded in paraffin, sectioned along the mediolateral plane, cut to 5-µm thick slices and stained with hematoxylin-eosin and Safranin O 13 . We used a digital light microscope (BZ-9000, Keyence, Osaka, Japan) to visualize the samples and merged six individual micrographs into one image per sample. Two investigators (SN [fellowship trained, 10 years of experience in histopathology]; LD [2 years of experience in histopathology]) graded the tissue status semi-quantitatively according to the Pauli classification 30 in consensus. On a per-sample basis, final scores were discussed until consensus was reached. No measures of inter-or intra-rater variability were calculated. The Pauli classification assesses surface integrity for the femoral, tibial and inner surfaces (score 0-3 each, i.e. 0-9), cellularity (score 0-3), collagen organization and alignment (score 0-3), and matrix staining intensity (score 0-3).

Data availability
The datasets generated and analyzed in this study are available from the corresponding author on reasonable request.