Microscale investigation of the anisotropic swelling of cartilage tissue and cells in response to hypo‐osmotic challenges

Tissue swelling represents an early sign of osteoarthritis, reflecting osmolarity changes from iso‐ to hypo‐osmotic in the diseased joints. Increased tissue hydration may drive cell swelling. The opposing cartilages in a joint may swell differently, thereby predisposing the more swollen cartilage and cells to mechanical injuries. However, our understanding of the tissue–cell interdependence in osmotically loaded joints is limited as tissue and cell swellings have been studied separately. Here, we measured tissue and cell responses of opposing patellar (PAT) and femoral groove (FG) cartilages in lapine knees exposed to an extreme hypo‐osmotic challenge. We found that the tissue matrix and most cells swelled during the hypo‐osmotic challenge, but to a different extent (tissue: <3%, cells: 11%–15%). Swelling‐induced tissue strains were anisotropic, showing 2%–4% stretch and 1%–2% compression along the first and third principal directions, respectively. These strains were amplified by 5–8 times in the cells. Interestingly, the first principal strains of tissue and cells occurred in different directions (60–61° for tissue vs. 8–13° for cells), suggesting different mechanisms causing volume expansion in the tissue and the cells. Instead of the continuous swelling observed in the tissue matrix, >88% of cells underwent regulatory volume decrease to return to their pre‐osmotic challenge volumes. Cell shapes changed in the early phase of swelling but stayed constant thereafter. Kinematic changes to tissue and cells were larger for PAT cartilage than for FG cartilage. We conclude that the swelling‐induced deformation of tissue and cells is anisotropic. Cells actively restored volume independent of the surrounding tissues and seemed to prioritize volume restoration over shape restoration. Our findings shed light on tissue–cell interdependence in changing osmotic environments that is crucial for cell mechano‐transduction in swollen/diseased tissues.

RGPIN-2023-05127; Canada Research Chair Programme; Killam Foundation osmotic environments that is crucial for cell mechano-transduction in swollen/ diseased tissues.

K E Y W O R D S
cell shape, cell volume, multiphoton laser microscopy, osteoarthritis, tissue swelling

| INTRODUCTION
Articular cartilage consists of collagens, proteoglycans (PGs), water, ions, and cells.Cartilage contains only one cell type, chondrocytes.Chondrocytes are surrounded by a thin layer of pericellular matrix and rely on mechanical signals transduced from joint loading of cartilage to maintain a healthy cartilage metabolism. 1,24][5][6] Changes in cartilage also extend to the cells.Chondrocytes in degenerated cartilage are swollen, and the amount of swelling has been shown to correlate with tissue degeneration. 7Chondrocyte volume and shape are indicators of their well-being.Specifically, chondrocytes are known to regulate their volume to stay at a resting set point so as to maintain metabolic homeostasis 8 and prevent membrane rupture from excessive swelling. 9,10Cell shape is insofar important as chondrocyte shape determines cell phenotype and the type of cartilage matrix (hyaline or fibrocartilage) that the cells produce. 11,12Interestingly, given that chondrocytes can actively regulate their osmolarity to match the extracellular osmolarity to preserve volume, 10,13 it is unclear why chondrocytes are swollen in degenerated cartilage, 7 and if the changing physical constraints associated with matrix swelling play a role in determining cell volume in degenerated cartilage.
Simultaneous measurement of the swelling behavior of tissue and cells may provide insights into the questions above.However, tissue and cell swelling are typically studied separately.Classically, tissue swelling has been studied in terms of bulk volumetric swelling by measuring the water content of sliced tissue samples of hundreds of micrometer thickness. 4,14,15Also, kinematic techniques have been used to study the two-dimensional (2D) swelling-induced tissue strains. 168][19] However, cells behave differently when removed from their surrounding tissue matrices. 20Even under in situ conditions, cellular responses to hypo-osmotic challenges are affected by the structural integrity of the extracellular matrix (ECM). 21,22Analysis using numerical modeling suggests that cell swelling depends crucially on the tissue collagenous structure. 23However, this suggestion has never been proven experimentally, probably due to the technical challenge of studying three-dimensional (3D) volumetric swelling of tissue and cells at the same time.
5][26] Structural and compositional differences are thought to play a role in the unequal rate of degeneration observed for these two cartilages. 27Mechanically, PAT cartilage is softer than FG cartilage. 28,29As the synovial environment of OA joints is often hypoosmotic, 30 the two opposing cartilages in the joint may swell to a different extent, making the more swollen cartilage more vulnerable to mechanical injury.Furthermore, considering that the cell morphology and the surrounding ECM are closely linked, 31,32 the cells in PAT and FG cartilages may also swell differently in a hypo-osmotic environment, causing altered cartilage metabolism.However, our understanding of the relationship between tissue and cell swelling in cartilage is limited.
Therefore, the purpose of this study was to determine the tissue and cell responses of lapine PAT and FG cartilages when exposed to a hypo-osmotic challenge.The hypo-osmotic loading was provided through deionized distilled water (ddH 2 O).We used multi-photon laser microscopy to measure the 3D kinematics of the tissues and cells during the period of hypo-osmotic loading.We hypothesized that cartilages experience tensile strains in all directions in response to hypo-osmotic loading, and that PAT cartilage swells more than FG cartilage.Correspondingly, the cells in the loaded tissues also expand in all directions, and cells in the PAT cartilage swell more than cells in the FG cartilage.| 55 specimen fixation.The fixed specimens were then equilibrated in an isotonic saline (315 mOsm) for at least 30 min.
A three-dimensional (3D) grid pattern with lines evenly spaced at 10 µm in axial (i.e., along objective axis) and transverse directions was imprinted onto a volume of 110 × 110 × 50 µm 3 in the superficial cartilage zone by photobleaching of the green-fluorescing ECM using a laser intensity of ~35 mW. 33The grid-marked tissue region was then imaged over time to measure ECM deformation.As the fluorescence emitted by the cells within the grid region was severely affected by grid photobleaching, cells located at 200 µm to the right of the grid region were imaged instead for cell deformation measurement (Figure 1).
Since the same cells were imaged nine times over a duration of 76 min, the effect of photobleaching on cell volume determination 34 was accounted for by correcting cell volumes during osmotic loading using measurements of cell volumes before osmotic loading obtained in an identical manner for nine consecutive times from a separate region of interest (ROI) located 200 µm to the left of the grid region (Figure 1).
The signals emitted by ECM and cells were collected simultaneously in the backward (epi-) direction and filtered by band pass filters at wavelengths of 516 nm (FF01 518/45, Semrock Inc.) and 590 nm (FF01 610/70, Semrock Inc.), respectively.127 × 127 µm 2 planar images were serially acquired along the objective axis at 1 µm intervals from 25 µm above the articular surface to a depth of 60 µm by a high-speed resonant scanner (pixel size: 0.249 µm; bit-depth: 10; dwell time: 0.067 µs, scan direction: roundtrip).Acquisition time for each image stack was approximately 13 s.Reference image stacks of tissue and cells were first acquired from the two ROI described above for isotonic conditions.The bathing solution was then changed from isotonic to the hypotonic ddH 2 O, after which the image stacks of tissue and cells were taken six times (at 1, 3, 5, 25, 40, 75 min; Figure 1).Cells were imaged 1 min after every ECM image acquisition to allow for a two-way planar translation between the tissue ROI and the cell ROI.All osmotic experiments were conducted at room temperature (21°C).

| Image-based kinematic analyses of ECM and cells
Before any kinematic analysis, the surface orientation was determined from the through-thickness view of the reconstructed 3D image stacks, and it was used to rotate the tissue and cell images for alignment with the horizontal.

| Cartilage ECM
The ECM of paired cartilage samples harvested from the FG and PAT of six knee joints were studied.The grid-marked ECM images (Figure 1) were processed with the custom open-source software 'lsmgridtrack'. 33,35e deformed image stacks of the ECM acquired following the osmotic loading were registered to the reference image stack by determining an optimal third-order basis spline (B-spline) transform using a two-step procedure. 36First, lists of corresponding points at the corner nodes of each transverse grid plane in the reference and deformed images were provided by the user to calculate an initial guess for the B-spline transform.Then, the Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton algorithm 37 was employed to further optimize this initial guess.

The objective function was used to minimize the Mattes Mutual
Information Metric, which performs robustly on low-quality images. 38,39e optimal B-spline transform was then applied to the vertices, X X X X = ( , , ) 1 2 3 , of a rectilinear grid of 5 µm edge length overlaid on the grid-marked image region in the undeformed reference configuration to obtain the deformed vertex positions, x x x x = ( , , ) 1 2 3 .The material points, X, were mapped into the spatial deformed grid positions, x, by a configuration map, χ, that is assumed to be affine, that is, in Cartesian coordinates, where u is the vertex displacement from current to reference configuration, calculated in components as and δ iK is a generalized Kronecker delta (rigorously speaking, a shifter; see, e.g., Eringen 40 ), valued one when the indices i and K take the same value among 1, 2, and 3, and zero otherwise.
The deformation gradient, F, at the centroid of every grid element, was determined by employing the isoparametric formulation for linear eight-node hexahedrons 33 : where N a are the trilinear Lagrangian shape functions of a linear hexahedron.
The Green-Lagrange strain tensor, E, was calculated at each grid element centroid as where I is the identity tensor and F T is the transpose of the deformation gradient.
The eigenvalues of E were calculated and numbered in decreasing order: 3 .The corresponding eigenvectors were chosen so that the eigenvector triad was positively oriented, that is, Finally, the maximum shear strain, and the volumetric strain, where det is the determinant, were calculated.Through the novel three-dimensional (3D) grid lines imprinted non-invasively on the cartilage ECMs, 3D tissue deformation (e.g., volumetric strains) could be measured at the micrometer scale.Image segmentation of the cells selected in (C) yielded robust and high-fidelity quantification of cellular geometries.

| Cartilage cells
In total, 115 and 99 cells were selected for kinematic analysis from the top 50 µm of FG and PAT cartilages, respectively (see Supporting Information: S1 for sample distribution and selection criteria).The cell images (Figure 1) were processed by a customwritten open-source software 'resonant_lsm'. 35Briefly, a list of seed points corresponding to the approximate centroids of the selected cells in pixel coordinates was inputted by the user.The image contrast was first enhanced by adaptive histogram equalization. 41Then, an edge-free active contour model 42  The segmented binary image was further smoothed by an iterative anti-aliasing algorithm 43 before the generation of a triangulated cell surface with the flying edges algorithm. 44r the calculation of cell strains, the simplified case of uniform cellular deformation was assumed (i.e., the apparent cell shape changes were measured).Specifically, the optimal affine transforms, which mapped the triangulated cell surface vertices from the reference to the deformed cells, were determined by the iterative 'closest point optimization' algorithm.The deformation gradient, F, was determined from the linear map contained in these affine transforms.
The Green-Lagrange strain tensors, principal strains, and maximum shear strains were calculated from the deformation gradients as outlined in Section 2.3.1.The volume and surface area of the cells were calculated from the segmented surfaces by using the computational geometry algorithms available in the Visual Toolkit. 45e volumetric strain and surface area strains were defined as the percentage difference of volume and surface area between the deformed and reference cells, respectively.
To account for possible cell shrinkage artifacts caused by photo-bleaching, cell images acquired nine consecutive times from a separate ROI (Figure 1B) were analyzed for the volume, surface area, first principal, third principal, and maximum shear strains.The lines best-fitted to each of these kinematic variables as a function of scan repeat were used to correct for the photo-bleaching effect on the measured cell dimensions.
Cells were considered to swell if they gained volume by ≥2% at the beginning of the hypo-osmotic loading.The time taken by each chondrocyte to reach peak volume was recorded.Swelling cells were considered to undergo a regulatory volume decrease (RVD) if they lost more than 2% of volume following peak volume occurrence.The rate of cell RVD was calculated as the linear approximation of the time rate of volume loss from the instant of peak volume occurrence until 41 min following peak volume occurrence for every cell.

| Statistical analysis
Commercial statistical software (SPSS version 27, IBM Corp.) was used for all tests.The volumetric, first principal (maximum tensile strain if positive), third principal (maximum compressive strain if negative), and maximum shear strains of ECM and cells, as well as the surface area and the ratio of surface area to volume of the cells, were analyzed for the effects of time (0-76 min following osmotic loading) and joint region (patella vs. FG) using a generalized estimating equation (GEE, under Genlin Mixed procedures in SPSS) to take into account the correlated nature (FG and patella were from the same animals) of the observations and the unbalanced design (unequal number of cells for each animal) of the study.In addition, using the same statistical procedure, the average orientation of the first principal strains was also compared between the tissue and the cells and between the patella and the FG samples at 25 and 2 min, respectively, which were the time points at which maximal volume changes occurred for the tissue and cells, respectively (see Results and Figure 2).Moreover, since the strain tensor is symmetric by definition, the Spectral Theorem 46 guarantees that its eigenvectors (principal directions of strain) are mutually orthogonal and its eigenvalues (principal strains) are all real.Since the first and third principal strain directions are orthogonal by virtue of the Spectral Theorem, only the orientation of the first principal direction was reported.Two-tailed distribution testing of type I error probability (α = 0.05) was assumed.Data are presented as estimated marginal means (EMM) ± 1 standard error (SE), unless otherwise stated.

| ECM response
Following the change of the bathing solution from isotonic PBS to hypotonic ddH 2 O, the superficial tissue of the FG and PAT cartilage samples swelled continuously for ~25 min before reaching an approximate steady state.The increase in local tissue volume was small at <3%, and different between joint regions (Figure 2A).FG cartilage experienced a smaller volume increase (0.8%) than PAT cartilage (2.3%) at 75 min.Notably, tissue swelling was not isotropic.
At 75 min, the first principal strains were 0.016 and 0.044 (Figure 2B), while the third principal strains were negative at −0.009 and −0.015 for FG and PAT cartilages, respectively (Figure 2C).The first principal strain direction of FG and PAT cartilages were similar at 61 ± 0.4°and 60 ± 0.4°relative to the horizontal (p > 0.05).Maximum shear strains of the FG cartilage were smaller at 0.025 for the FG than the PAT cartilage (0.055) at the end of the testing period (Figure 2D).

| Cell shrinkage artifacts caused by photo-bleaching
Cell volume decreased by 0.4% per scan while cell surface area decreased by 0.3% per scan.The first principal strains decreased by ~0.1% per scan, while the third principal strains decreased by 0.3-0.4% per scan.The measured rate of change of these kinematic variables caused by photobleaching artifacts was included in the Supporting Information S2 and were used to correct for the cell swelling response.

| Cell response
FG (654 ± 13 µm 3 ) and PAT (636 ± 14 µm 3 ) cell volumes were similar in the isotonic reference condition.Most of the cells stayed alive throughout the experimental period (see Supporting Information: S3).
Eighty-six percent of the FG cells and 96% of the PAT cells responded to the hypo-osmotic challenge by swelling, while the remaining cells experienced a reduction in volume from the beginning of the loading period (Supporting Information: Table S1, Supplementary Material S1).From the cells that swelled, 95% and 88% experienced RVD following peak volume occurrence for the FG and PAT cells, respectively.Interestingly, for those that experienced RVD, 54% of the FG cells and 24% of the PAT cells lost volume relative to their initial reference volumes by the end of the experiment (Supporting Information and Table S2, Supplementary Material S4).The following analyses focused only on the cells that swelled and showed RVD during hypo-osmotic loading (FG: 84 cells, PAT: 94 cells).
The peak ECM volume change occurred at 15-20 min (Figure 2A) and the peak cell swelling occurred within the first 2-6 min of the osmotic challenge (Figure 2E).The ECM stayed swollen throughout the hypo-osmotic loading period, while the majority of the cells exhibited RVD.The time history of the initial swelling and the ensuing RVD differed slightly between FG and PAT chondrocytes.FG chondrocytes reached peak volume gain in 3.3 ± 0.3 min, which was significantly faster than the PAT chondrocytes (6.0 ± 0.9 min).The average FG chondrocyte volume increased by 11%, recovered the initial volume by 41 min, and had a 4% decreased volume at the end of the testing period.The PAT chondrocytes gained a maximum volume of 14%-15% and remained at a 5% increased volume to the end of the 76 min hypotonic test condition (Figure 2E).The rate of RVD calculated from individual cells was similar for FG and PAT chondrocytes (FG: 0.25 ± 0.03%/min, PAT: 0.27 ± 0.04%/min).

F I G U R E 2 Contrasting kinematic responses of cartilage ECM (A-D) and cells (E-H, yellow shaded
) to the hypo-osmotic challenge.Note that only cells showing a swelling response to hypo-osmotic loading were included in this analysis (see Results for further details).Volumetric strains (A, E), first principal strains (B, F), third principal strains (C, G), and maximum shear strains (D, H) of the ECM and cells were compared.The swelling responses of the ECM and cells of the patella (red circles) were different from that of the opposing femoral groove (green triangles) despite their interdependence for mechanical loading in a knee joint during locomotion.The results for cell deformation (bottom row) were corrected for photo-bleaching effects.For the tissue response, *, $, and & indicate statistical difference between the response at the marked time point with the response at 0 (reference), 15, and 25 min within a joint, respectively, whereas for the cellular response, * and # indicate statistical difference between the response at the marked time point with the response at 0 and 2 min within a joint, respectively.† denotes significant difference between patella and femoral groove at the corresponding time points.
Similar to tissue strains, cell strains were also direction dependent.However, the direction of the first principal strains of the cells was significantly different from that of the tissue (p < 0.01), with direction lying near the transverse plane (13 ± 2°for FG cells, 8 ± 2°f or PAT cells, p > 0.05).First principal strains of FG cells increased to 0.14 in 2 min and then decreased to 0.08 at 76 min (Figure 2F).First principal strains for PAT cells increased to 0.15-0.16from 2 to 6 min, and decreased to 0.12 at 76 min.The third principal strains for FG and PAT cells were approximately −0.08 at 2 min, then decreased further to −0.11 at 76 min (Figure 2G).Maximum shear strains stayed constant at 0.19 and 0.23 after the first 2 min for FG and PAT cells, respectively (Figure 2H).

The surface area of cells from FG and PAT cells increased by 10%
in the first 2 min.At 76 min, FG cells recovered their initial surface area, and PAT cell surface areas remained increased by 5% (Figure 3A,C).The surface area to volume ratio of the FG cells continuously increased under hypotonic conditions and was increased by 6% by the end of the testing period.Conversely, the area to volume ratio of PAT cells decreased by 3% in the first 2 min, recovered to the initial value by 40 min, and then decreased again by 3% at 76 min (Figure 3B,D).

| DISCUSSION
Tissue swelling is one of the early signs of OA. 4 Tissue swelling and associated increase in hydration have been shown to stiffen articular cartilage 47,48 and cause swelling of chondrocytes, which makes them more susceptible to injury, 49,50 probably through a size reduction of the so-called membrane reservoir in swollen cells. 51,52In contrast, a reduction in cell volume by water efflux is thought to offer protection against mechanical injury to chondrocytes. 49,50,53Cell swelling leads F I G U R E 3 Surface area (A) and surface area-to-volume ratio (B) as a function of time following hypotonic osmotic loading of cells residing in the femoral groove (green triangles) and patella (red circles).Normalized results relative to the initial reference state of the cells before osmotic loading in the form of (C) surface area strain, and (D) percent change of surface-area-to-volume ratio.Note that only cells showing a swelling response to the hypo-osmotic loading were included in this analysis (see Results for further details).Cell deformations were corrected for photobleaching artifacts.* and # indicate statistical difference between the response at the marked time point with the response at 0 and 2 min within a joint, respectively.† denotes significant difference between patella and femoral groove at the corresponding time points.
to a change of intracellular composition, and negatively affects cell metabolism. 8In this study, we used ddH 2 O with zero osmolarity to investigate the resilience of articular cartilage and chondrocytes when subjected to extreme hypo-osmotic swelling.We measured, for the first time, the swelling response of tissue and cells in the same cartilage samples and compared these responses for directly opposing PAT and FG cartilage regions in patellofemoral joints.Specifically, tissue swelling responses were measured non-invasively at micrometer resolution for the top 60 µm superficial layer of cartilage, a region that is most susceptible to osteoarthritic degeneration. 54Such high-resolution measurements were possible because of a novel 3D grid imaging technique developed on a multiphoton laser microscopy platform. 33We found that: (i) hypoosmotically loaded ECMs were stretched by 2%-4% in the first principal strain direction, which was oriented at 61°from the articular surface, but compressed by 1%-2% along the third principal direction, which lead to an average increase in tissue volume by 1-2.5% (Figure 2A-D); (ii) under hypo-osmotic conditions, most cells increased in volume to a greater extent than the corresponding local ECM.Instead of expanding equally in all directions, as seen in isolated cells, 9,17 the swelling strains of the in situ cells were anisotropic; that is, compressive in the third principal direction, but tensile in the first principal direction (Figure 2E-H).Interestingly, the cell expansion, as defined by the first principal strain and associated direction, occurred near the transverse plane that is parallel to the articular surface (Figure 4), and differed from the direction of tissue expansion by 48-53°(see Results); (iii) 88%-95% of the in situ cells that gained volume in the initial phase of exposure to hypo-osmotic loading underwent RVD to regain their initial reference volume, and 24%-54% of them had less volume at 76 min than at the initial pre-intervention state; and (iv) the volume increase, and the associated deformation at the tissue and cell level, were larger for the PAT cartilage than for the FG cartilage (Figure 2).At the tissue level, the local ECM, on average, gained volume in the hypo-osmotic bathing solution due to the Donnan effect caused by the PGs' fixed charged density (FCD). 55However, the tissue volume increase remained below 3%, and the tissue strains below 4%, likely because of the tension-resistant fibrous network in healthy cartilages (Figure 2A-C).Similar swelling-induced tissue strains were found previously for canine knee cartilage. 16Narmoneva et al. 16 measured the 2D swelling-induced strains across the tissue thickness and found that 80% of the tissue experienced uni-directional tensile strains along the tissue depth.However, they were not able to determine volume changes and strains in the top 100 µm of the cartilage due to the limited resolution of their approach, yet this thin layer of superficial cartilage is crucial for the overall tissue swelling behavior. 56In contrast, we measured the swelling response for the most superficial cartilage, and found that the swelling-induced deformation was more profound than the one described by Narmoneva et al. 16 Specifically, the third principal strains were compressive whereas the first principal strains were tensile (Figure 2B-C).We speculate that the observed tissue deformations are governed by the structure of the local collagenous network. 57,58e collagenous network of mature cartilage is known to have the socalled Benninghoff arcade-like structure with collagen fibrils aligned tangentially to the articular surface in the superficial tissue, 59,60 making the tissue mechanics highly anisotropic.The structural and mechanical anisotropy likely caused the observed directional swelling behavior of the superficial cartilage.
At the cell level, we found that not all chondrocytes gained volume following the hypo-osmotic loading.Fourteen percent and 4% of the analyzed FG and PAT cells, respectively, experienced a volume loss at all measured time points.It is possible that these cells gained volume within the first 2 min before the first imaging time point (Figure 1A), and then lost volume rapidly thereafter.Local tissue deformation and non-uniformities in local tissue structure may play a key role in causing the cell volume loss for this small group of cells.
Nevertheless, this hypothesis should be investigated further.For simplification, these non-swelling cells were excluded from our analysis.For the remaining cells that gained volume early in the hypoosmotic loading exposure phase, the volume increase was 11%-15% (Figure 2E), which was much greater compared with the tissue volumetric increase (<3%).This difference in volume increase between cells and local tissue may be due to the fact that chondrocytes are surrounded by the pericellular matrix, which is rich in type VI collagen and mechanically softer than the distant territorial matrix. 61Importantly, the swelling-induced strains on the cells were not always tensile (Figure 4).Rather, the cells had tensile first principal strains and compressive third principal strains (Figure 2F-G).
When compared with the adjacent ECM deformation, the swellinginduced deformation of the cells was not only larger in magnitude (0.03 vs. 0.15 and −0.01 vs. −0.08 for the first and third principal strains of the tissue and the cells, respectively), but also along different direction (61°vs.11°relative to the articular surface for the first principal strain direction of the tissue and the cells, respectively).
It has been suggested that cell deformation resulting from hypoosmotic loading is governed by the local collagenous architecture, which is aligned parallel to the articular surface. 23Considering that the collagenous network in the superficial cartilage is disrupted in early OA, 3,54 cell swelling behavior in early OA may also differ from that observed here in healthy cartilage samples.The significance of the directional expansion of cells for hypo-osmotic conditions will be investigated in detail in the near future.
It is worth pointing out that none of the cells burst by "overswelling," as seen in isolated cells, 9,10 despite using ddH 2 O as the bathing solution.We hypothesize that in situ cells are protected from bursting because they are surrounded by PGs, which elevate the extracellular osmolarity through the local FCD by approximately 150-170 mOsm. 18,62In situ chondrocytes only behave as a perfect osmometer for extracellular osmolarity ranging from 200 to 600 mOsm.Further reduction of the extracellular osmolarity leads to little additional swelling. 17 In agreement with our hypothesis, we observed a significantly greater swelling-induced deformation of tissue and cells for the PAT than the FG cartilage (Figure 2).This finding can be explained by the difference in structure and composition for these two cartilages, specifically differences in the top 10% superficial zone tissue.
Compositionally, the FCD in FG cartilage is ~10% higher than that in PAT cartilage (168 vs. 153 mEq/L), resulting in a relatively lower extracellular osmolarity in PAT than FG cartilage for increased swelling. 62Structurally, the orientation angle of the collagenous network relative to the articular surface (0°: parallel, 90°: perpendicular) was ~19% smaller in the FG cartilage than in the PAT cartilage (43°vs.51°). 62Mechanically, the dynamic modulus, which is mainly caused by the stiffness of the collagenous network and the fluid pressure, has been shown to be 75% greater in the FG than the PAT cartilage (7.8 vs. 4.4 MPa). 63These structural, compositional, and mechanical differences likely caused the larger swelling-induced deformation in the PAT tissue and cells compared to the FG tissue and cells.In addition, we speculate that the structural differences in FG and PAT cartilages may also be contributing to the longer swelling duration observed for PAT than FG chondrocytes (6 vs. 3.3 min).
We found that 88%-95% of the analyzed cells exhibited RVD following the initial phase of volume gain.In previous studies, a similar proportion of in situ cells showing RVD has been reported. 13D is thought to be an active, temperature-dependent, regulatory process that is initiated by the cells to restore the osmolarity balance with the extracellular environment through transmembrane ion fluxes, hence restoring cell volumes to a resting set point.Cells undergoing RVD become stiffer because of the contraction of their cytoskeleton. 64In the current study, despite being swollen to varying degrees, FG and PAT chondrocytes had similar rates of RVD at F I G U R E 4 (A) Exemplary three-dimensional (3D) images of cells in patellar cartilage before (left) and 4 min (i.e, near peak volume increase, right) following the start of osmotic loading, showing clear expansion of cells in the transverse (xy-) plane of the tissue.The associated cell volume change caused by hypo-osmotic loading is also shown.(B) Cell shape changes during hypo-osmotic loading, calculated based on the measured principal strains and expressed in a two-dimensional aspect ratio (see Discussion, and Supporting Information S5).(C) Schematic illustration of the difference in swelling and the corresponding regulatory volume decrease during hypo-osmotic loading for isolated and in situ chondrocytes, drawn based on realistic cell shapes.Note that the isolated cells maintain the spherical shape throughout the hypo-osmotic loading, whereas the in situ cells change shape in the early swelling phase of the hypo-osmotic loading, before maintaining a constant shape afterwards.approximately 0.25%/min (see Results, Figure 2).However, this rate of RVD is much slower than previously reported values of about 15%/min, 13 despite the fact that both studies were conducted at room temperature (21°C).This difference in RVD may be due to structural and compositional differences of cartilages from different animal species (lapine vs. bovine).The ECM integrity of the tissue may also have played a role.The cells were imaged through the intact articular surface in the current study, but were imaged through a cut surface of tissue slices in Bush and Hall. 13It has been shown that tissues swelled by >10% after being sliced, 8 and that cells adjacent to cut surfaces had altered mechanical and swelling behavior than cells surrounded by intact tissue. 22,65Cartilage explants deprived of the subchondral bone also showed drastically different tissue swelling response, with tensile strain of 2%-9% in all directions. 66,67Nevertheless, further investigations are required to clarify the discrepancies in RVD rates reported in the literature, and the 'over-correction' of volume by the RVD process seen in 24%-54% of the analyzed cells (see Results).
We also investigated the relationship between RVD and cell shape changes.It seems that cells minimize the first principal strain, while driving the third principal strain towards a more negative value during RVD (Figure 2F-G).We note that while the cells recovered their volume and surface area to the initial resting state during RVD (Figures 2-3), the volume and surface area restoration occurred at different rates.For FG chondrocytes, the rate of surface area recovery was slower than the rate of volume recovery, thereby leading to an overall increase in the surface-area-to-volume ratio (Figure 3B-D).To help visualize the cell shape changes, we took the 2D cell aspect ratios measured previously for PAT and FG chondrocytes 28 and assuming that the measured first and third principal strains occurred along the minor and major principal axes of an ellipse (Figure 4, see also Supporting Information: S5 for more details).Interestingly, while the cells became more flattened during the initial volume gain phase, this new cell shape was maintained during the RVD (Figure 4A).Cell shape is closely related to the cytoskeletal organization, 11 while cell volume is tightly coupled to the membrane tension, 10 both of which are crucial for cellular mechano-transduction.Our results suggest that the cell volume recovery preceded cell shape recovery (Figure 2E vs. Figure 4A).The mechanisms behind the sequencing of these two regulatory events should be studied in the future.
There are limitations that should be considered while interpreting the current results.First, due to the limited penetration of laser microscopy for effective grid-line imaging, 33 only the top 60 µm superficial zone cartilage could be investigated.Second, the intracellular fluorescence was diluted by the water exchange over time, thereby limiting the measurement duration to the first 76 min of the osmotic loading.Third, the rapid exposure to the hypotonic bathing solution is not physiologically relevant. 19The current testing conditions were used to measure the upper limit of tissue and cell responses to hypo-osmotic loading.Future studies should investigate the swelling-induced deformation of tissue and cells as a function of tissue depth.Narmoneva et al. 16 found a compressive swelling-induced strain in the deep tissue adjacent to the cartilage-bone interface.It would be interesting to investigate this tissue region with the current 3D grid-line imaging technique.As tissue stiffens and cells swell in hypo-osmotic loaded cartilages, 21,22,47,48 whereas tissue softens and cells shrink in hyper-osmotic loaded cartilages, 53,68 the changing mechanics of tissue and cells to osmotic loading and the corresponding effects on joint health and cell metabolism should be further elucidated.Hypo-osmotic loading with an osmolarity range (180-250 mOsm) that is more physiologically relevant 30 should be used and applied gradually 19 to mimic diseased joint conditions.Finally, multi-scale finite element modeling, particularly models that consider cartilage as a fiber-reinforced material 6,57,69,70 and cells as being surrounded by a pericellular matrix with structurally accurate thickness distribution, 71 should be developed to determine the cause of the anisotropic swelling behavior of cartilage and cells observed in the current study, and the effect of osmotic loading rate on cell mechanics. 72

| CONCLUSION
We conclude that the swelling-induced deformation of in situ cells is anisotropic, which is distinctly different from the isotropic swelling of isolated cells (Figure 4B).Although the deformation of the ECM surrounding the cells was also anisotropic, its expansion was distinctly different from that of the cells, with their first principal strain directions differing by 48-53°.Also, unlike the ECM, cells actively restored their volume towards the initial resting set point following the hypotonic exposure, and cell volume restoration precedes cell shape restoration.Our findings may provide novel insights into the interdependence of the kinematic responses of tissue and cells to changes in the osmotic environment, which may be important for the study of cell mechano-transduction.
All aspects of animal care and experimental protocol were approved by the Life and Environmental Sciences Animal Care Committee of the University of Calgary.Knee joints were harvested from the right hindlimbs of six 10-12-month-old skeletally mature New Zealand white rabbits.Immediately after opening the knee joints, rectangular osteochondral blocks of 5 × 5 mm 2 were sawn from the weight-bearing surfaces of the medial FG.The opposing PATs were harvested intact.The osteochondral specimens were incubated in a phosphate-buffered saline (PBS) solution containing fluorescent dyes for staining the ECM and live cells.The cartilage ECM was stained by 16 µM 5-(4,6dichlorotriazinyl) aminofluorescein (5-DTAF, Ex: 492 nm, Em: 516 nm, Thermo Fisher Scientific Inc.) while live cells were labeled by 3 µMCalcein Red/Orange AM (Ex: 577 nm, Em: 590 nm, Thermo Fisher Scientific Inc.).After staining for 1 h, the excess dye was washed away from the specimens by three rounds of 5 min gentle shaking in dyefree saline.The stained specimens were then attached to a specimen holder using dental cement in a dark room.The medial articular surface of PAT and FG cartilages was carefully aligned horizontally during MOO ET AL.

F
I G U R E 1 (A) Time points for image acquisition of the extracellular matrix (ECM) and cells and (B) the respective locations of the measured volumes in the cartilage samples.(C) Example images of the acquired ECM and cell fluorescent signals with user-indicated seed points overlaid in yellow.(D) The acquired image stacks were used for kinematic analyses of the ECM and cells.
was used to determine a levelset function defining the image region containing the cell within a local image sub-volume of 25 × 25 × 30 µm 3 centered at each seed point.This local region of interest was larger than the typical volume of a cartilage cell in the superficial tissue to ensure segmentation of the entire cell.The levelset function was smoothed by a median filter with a two-voxel radius.Cells were defined as the domain containing positive levelset function values.
Furthermore, chondrocytes are physically bound to the ECM, which likely helps reduce the amount and rate of swelling.