Shear Mechanics of Articular Cartilage and Cartilage-Inspired Materials

Abstract

Articular cartilage is a load-bearing, hierarchically organized tissue composed of a network of type II collagen embedded in an aggrecan-rich polyelectrolyte gel. Its ability to resist deformation and dissipate energy arises from spatially varying matrix composition and architecture. Here, we review experimental and theoretical advances that elucidate the mechanistic basis of cartilage shear mechanics. Recent studies have shown that the tissue operates near a rigidity transition, in which small changes in collagen density, cross-linking, or osmotic stress can produce large, nonlinear changes in shear stiffness. We discuss how this behavior is captured by models rooted in rigidity percolation, continuum elasticity, and micromechanics, and how these frameworks connect depth-dependent composition to macroscale mechanical response. Throughout, we emphasize physical principles that describe observations across native, degraded, and engineered tissues, and we highlight emerging strategies for designing cartilage-inspired materials with tunable, anisotropic mechanics, with applications in soft robotics, synthetic gels, and load-bearing biomaterials.

1. INTRODUCTION

“How did you go broke?” Bill Gorton asks in The Sun Also Rises. “Two ways,” replies Mike Campbell. “Gradually and then suddenly.” The same can be said of articular cartilage (AC) failure during osteoarthritis. AC may degrade over years of wear or rupture from a single acute injury. Yet for most of its life, it performs exceptionally well.

AC is a sparsely cellular and avascular load-bearing soft tissue that enables mammalian joints to support decades of repetitive mechanical loading—often exceeding ten times our body weight—without undergoing structural failure. Although it is only a few millimeters thick, its resistance to fracture and damage remains unmatched among synthetic analogs. These mechanical attributes arise from a multicomponent extracellular matrix (ECM) composed primarily of type II collagen and the charged PG aggrecan, embedded within a water-rich gel. This composite architecture enables AC to function as a self-assembled, long-lived material that is simultaneously stiff in compression, tough in shear, and capable of resisting crack propagation.

The structural and mechanical organization of cartilage suggests an underlying set of physical principles that remain only partially elucidated. In particular, recent studies indicate that the collagen network in AC operates near a mechanical phase transition. Small changes in composition or prestress can result in large variations in stiffness. This proximity to a rigidity threshold implies that AC’s macroscale properties may be governed by microscale parameters such as fiber density, cross-linking, and osmotic stress—features that are readily perturbed during disease progression or tissue engineering.

In this review, we discuss the mechanistic basis of cartilage function, drawing on results from both experiments and theoretical modeling. We summarize key findings that establish structure–function relationships under shear and compression, and we highlight the role of depth-dependent composition in governing tissue response. In parallel, we discuss theoretical frameworks, including poroelastic and rigidity percolation theories, that have successfully captured these behaviors and provided predictive insight into the mechanics of both native and engineered tissues. We emphasize approaches that connect microscale composition and architecture to macroscale mechanical function, with the goal of identifying principles that can guide both the diagnosis of cartilage degeneration and the design of next-generation biomimetic materials.

2. ARTICULAR CARTILAGE COMPOSITION AND STRUCTURE

AC is a specialized hyaline cartilage that forms a low-friction, load-distributing interface at the synovial joints (1). While in infancy AC is dense with cells, by adulthood the tissue consists primarily of ECM. The main components of cartilage ECM are collagen, primarily type II, proteoglycans (PGs), and water. Specialized cells known as chondrocytes are sparsely distributed throughout this matrix and decrease in number during adulthood. Although infant and adolescent cartilage is nourished by canals that extend from the bone to which it is attached, adulthood AC is avascular and aneural. Adult cartilage also has a very slow collagen turnover, with a half-life on the order of 100 years; thus, the tissue has very little capacity to regenerate after childhood.

The remarkable mechanical properties of AC arise from pronounced spatial and material heterogeneity. The tissue is broadly divided into three zones: a superficial surface zone, a mid zone, and a deep zone in which the tissue anchors to the bone (Figure 1). From the superficial zone to the deep zone, the components of the ECM generally become denser, and the chondrocytes and the collagen network exhibit a distinct spatial structure at each level. Near the surface, chondrocytes have a flat, discoid morphology with their long axes oriented horizontally. In the mid zone, chondrocytes have a more oblong morphology and are more uniformly distributed. In the deep zone, near the bone, the chondrocytes are the largest, have a roughly spherical shape, and are organized into vertical columns. Similarly, collagen fibers are oriented approximately parallel to the tissue surface in the superficial zone, isotropically in the mid zone, and vertically in the deep zone. As we discuss below, for small strains collagen concentration has been observed to be the most important predictor of tissue shear modulus, whereas orientation appears to play a more important role for larger strains.

Figure 1.

Schematic of a bovine knee joint highlighting the patellofemoral groove harvesting site for articular cartilage. The layered structure of cartilage is shown, comprising the superficial (tangential), mid, and deep zones. Each zone exhibits distinct collagen and aggrecan concentrations and fiber orientations. The depth-wise coordinate system is indicated, with $z=0$ defined at the articular surface and increasing toward the subchondral bone. Figure adapted with permission from Reference 30.

After collagen, the second most prominent ECM component in cartilage is the large aggregating PG known as aggrecan. Aggrecan is composed of an ~200 kDa core protein decorated with side chains of chondroitin sulfate and keratan sulfate glycosaminoglycans (GAGs). These sulfate groups are ionized at physiologic pH, endowing cartilage with a high negative fixed charge density. GAG chains account for 90% of the mass of the aggrecan molecule, which has a molar mass of up to 2 MDa (2). Through the action of link proteins, aggrecan associates with hyaluronic acid, effectively forming a polyelectrolyte hydrogel interpenetrating with the collagen network. Electrostatic repulsion between charges on the aggrecan network provides up to half of the tissue’s compressive stiffness (3). Furthermore, the negative fixed charge density attracts an excess of mobile cations in the interstitial fluid, attracting water through osmotic pressure via Donnan equilibrium (4).

3. MECHANICAL STRUCTURE-PROPERTY RELATIONSHIPS: EXPERIMENTS

AC exhibits remarkable mechanical performance due to its hierarchical structure and complex composition (5). A range of experimental studies have quantified how the tissue’s microstructure (collagen fiber network, PG content, etc.) gives rise to its unique mechanical properties under various loading modes. Below, we review key experimental findings with an emphasis on shear mechanics, organized by loading modality and experimental design. We begin with a brief historical perspective on compression studies and then turn to explants and engineered tissues subjected to shear in both linear and nonlinear regimes, highlight the depth and rate dependence of mechanical response, and discuss the role of osmotic swelling.

3.1. Mechanical Response in Compression

Under compressive loading, AC behaves as a poroelastic soft material with contributions from a fluid phase and a solid matrix. Classic experiments by Mow et al. (5) demonstrated that when cartilage is compressed under radial confinement, interstitial fluid pressurization provides significant initial load support, leading to a high instantaneous stiffness, followed by gradual stress relaxation as fluid flows out of the tissue (6). This response is captured by continuum models including both mixture theory (i.e., biphasic and triphasic theory) (5, 6) and poroelastic models coupled with electrokinetics (7). The equilibrium compressive modulus, after fluid exudation, reflects the stiffness of the collagen-PG solid matrix. Notably, half of the stiffness at equilibrium is due to electrostatic effects from the PG network (8). AC is also routinely subject to time-varying compressive loads, necessitating a dynamic mechanical response involving the interplay of poroelastic fluid flow and intrinsic matrix viscoelasticity, which act on distinct timescales and length scales. At low frequencies, stress-relaxation and creep experiments reveal that load decay is dominated by fluid exudation through the porous matrix (9, 10). During dynamic compression at comparatively higher frequencies, such as that experienced during walking or running, cartilage is pressurized due to there being insufficient time for fluid to drain from the tissue, making cartilage up to ten times stiffer than under static loading (11).

The ability of the tissue to pressurize so effectively is tied to the PG network (12) owing to the fact that spacing between GAG chains, and hence tissue pore size, is less than 10 nm (13). Mechanical tests have shown strong correlations between cartilage’s biochemical composition and its compressive properties. Julkunen et al. (14) measured stress-relaxation in human cartilage under unconfined compression and found that variations in PG and collagen content (and collagen orientation) predict the mechanical response. Similarly, infrared spectroscopic analysis by Rieppo et al. (15) demonstrated that the compressive stiffness of cartilage can be quantitatively predicted from its molecular composition: samples with higher PG content and collagen cross-linking showed greater equilibrium moduli. These studies underscore that PGs, which create swelling pressure and restrict fluid movement, and the collagen network, which provides tensile reinforcement, together determine compressive stiffness. The spatial distribution of GAG content is highly correlated with local compressive modulus, which is lower near the articular surface and higher in the deep zone (16).

Perturbing the matrix structure leads to predictable changes in compressive mechanics. For example, enzymatic depletion of GAGs causes a marked drop in the tissue’s compressive modulus and alters its swelling behavior. In a classic study, Bonassar et al. (17) treated cartilage with stromelysin (MMP-3) and reported an increase in tissue hydration due to damage to collagen network cross-linking as well as decreased compressive stiffness due to degradation of PGs. These results indicate that PGs are critical for maintaining compressive resistance by generating osmotic pressure and maintaining solid matrix integrity. In contrast, removal of collagen (e.g., with collagenase) virtually eliminates the load-bearing framework, leading to mechanical failure under compression (18). Overall, compression experiments confirm that the composition of cartilage, particularly PG concentration and collagen network integrity, is directly linked to its ability to bear compressive loads.

3.2. Mechanical Response in Shear

In pure shear deformation, the mechanical response of cartilage is governed almost entirely by the elastic ECM, because the interstitial fluid does not support the shear stress. As a result, the shear modulus is orders of magnitude lower than the instantaneous compressive modulus and is critically dependent on the organization of the collagen network. The shear behavior of cartilage has been studied for over a half century. Early studies of the shear behavior of cartilage demonstrated important nonlinearities, notably time-dependent (i.e., viscoelastic) behavior (19, 20), as well as stress-stiffening at higher levels of loading (21). Zhu et al. (22) expanded on this by using torsional shear experiments to show that the dynamic shear modulus is frequency dependent and modulated by PG content and compressive strain, implicating osmotic swelling in the modulation of shear stiffness. This behavior has been described reasonably well using empirical or phenomenological models (23) such as spring-dashpot systems or relaxation functions. Nevertheless, there remains a lack of mechanistic models to describe the structural and compositional basis of cartilage behavior in shear.

The zonal structure of cartilage is known to produce a different mechanical behavior that changes with distance from the surface. This zonal dependence was initially studied by cutting samples from different regions of the tissue, but such approaches are limited by the fact that they do not preserve the connectivity between layers and are further limited by the spatial resolution of the sectioning technique and the ability to grip and test very thin samples (24). Elliott et al. (25) inferred depth-dependent shear moduli by measuring tensile modulus and Poisson’s ratio across $~500\text{-}\mu \text{m}$-thick cartilage sections, providing early evidence of zonal variation, although this study limited by assumptions of isotropy and coarse spatial resolution. High-resolution confocal studies by Guilak et al. (26) and Schinagl et al. (27) introduced experimental methods to quantify spatial variations in strain and chondrocyte deformation. More recently, Buckley et al. (28) used confocal elastography techniques to more accurately map cartilage shear properties and to identify important length scales associated with these heterogeneities. Using such techniques, the local shear modulus $G$ has been measured through the depth of bovine cartilage, revealing a dramatic depth-dependent variation (up to a hundredfold difference) across the tissue thickness. In particular, the shear modulus is lowest in the transition zone (approximately $50–250\mu \text{m}$ below the articular surface) and increases toward a plateau in the deeper regions (28).

These depth-dependent trends appear to be conserved across anatomical locations. Silverberg et al. (29, 30) found that explants from different joint regions of neonatal bovine cartilage displayed similar qualitative profiles but differed in absolute modulus values and zone thicknesses (29, 30). To investigate the mechanistic basis of this variation, mechanical measurements were integrated with polarized light microscopy and Fourier transform infrared spectroscopy (FTIR) imaging to assess modulus, fiber orientation, and matrix composition (Figure 2). Variations in shear modulus (about hundredfold) were found to be more strongly correlated with modest differences in collagen concentration (about twofold) than with fiber orientation. This sharp dependence of the modulus on the density of the matrix is consistent with the rigidity percolation models (31–35), as discussed in Section 4.

Figure 2.

(a) Micrograph of cartilage illustrating two FTIR-I measurement regions. The two spectra were decomposed to a linear combination of pure type II collagen and pure aggrecan spectra. (b) Plotting the wet volume fraction of aggrecan $v_{\text{a}}(z)$ and type II collagen $v_{\text{c}}(z)$ reveals depth-dependent variations similar to those in the shear modulus. Plotting $\left|G^{*}(z)\right|$ against (c) $v_{\text{a}}(z)$ and (d) $v_{\text{c}}(z)$ reveals a correlation between modulus and matrix density. The inset in panel $c$ shows a linear relationship between $v_{\text{a}}$ and $v_{\text{c}}$. Figure adapted with permission from Reference 30.

Perturbative experiments targeting specific zones further validate the functional significance of these gradients. Griffin et al. (18) demonstrated that localized enzymatic degradation of cartilage matrix components—collagen via collagenase and PGs via trypsin—leads to dramatic, depth-dependent changes in shear mechanics near the articular surface. Using high-resolution confocal strain mapping, they showed that collagen degradation reduces the local shear modulus by up to fiftyfold and increases energy dissipation by up to a hundredfold, whereas PG removal produces more modest changes and preserves the characteristic depth-dependent profile. These findings highlight the extreme sensitivity of cartilage shear properties to matrix composition within the top few hundred microns of tissue and support the view that the collagen network operates near a mechanical threshold in which small structural perturbations can drive large mechanical consequences.

Together, these results point to a mechanically stratified system in which each depth zone plays a distinct but interdependent role: the superficial zone dissipates energy and resists shear, the mid zone buffers deformation, and the deep zone provides compressive support and anchoring. This organization arises during development, adapts to the anatomic context (29), and is disrupted in early degeneration. Engineered constructs continue to fall short of replicating native tissue function, largely because reproducing the depth-dependent composition and organization remains difficult.

AC also exhibits pronounced nonlinear shear mechanical properties governed not simply by strain amplitude but by fiber density and network connectivity. At low collagen concentrations, the network is floppy and supports little load. Near the rigidity threshold, finite fiber-bending stiffness suppresses floppy modes, producing a bending-dominated, approximately linear mechanical response. As fiber density increases, the system transitions to a regime in which tension-bearing fibers drive a sharp increase in stiffness. This fiber density-dependent transition reflects collective behavior near a mechanical critical point (29, 30). At larger strains, fiber recruitment and alignment create anisotropic responses with directional, multistage onset of rigidity (36). Normal compressive loading further modulates these behaviors by promoting fiber buckling and suppressing alignment, allowing adaptive tuning under multiaxial loading (28).

3.3. Role of Osmotic Stress

The osmotic swelling pressure, generated by the high fixed charge density of PGs, is a fundamental driver of cartilage mechanics. Negatively charged GAG side chains attract counterions and water, producing a Donnan osmotic pressure that prestresses the collagen network in tension. This internal prestress alters the tissue’s mechanical baseline and modulates its response to external loading. As such, osmotic forces are intrinsic to the load-bearing and energy-dissipating properties of cartilage.

Triphasic theory, an extension of poroelastic formulations, incorporates ionic concentration, fluid flow, and solid matrix deformation to model osmotic swelling and electrochemical interactions (37). Experimental manipulations of osmotic conditions validate this framework. Immersion of cartilage in hypoosmotic solutions induces water uptake and swelling, reducing compressive stiffness. In contrast, hyperosmotic environments drive deswelling and compaction, which can transiently increase stiffness but also weaken the collagen network, altering its ability to resist deformation (10).

Quantitatively, changes in ionic strength shift both tissue thickness and equilibrium moduli in predictable ways. Chremos et al. (38) used synthetic hydrogels to model a prestressed fiber-reinforced matrix, showing that osmotic swelling induces a preloaded state that significantly raises the effective modulus and alters the strain-stiffening threshold. In native cartilage, this prestressed state is critical: The collagen network resists osmotic expansion, storing elastic energy and enhancing tissue turgor. The loss of this balance, via PG degradation, leads to both mechanical and morphological changes.

Experimentally, PG depletion (e.g., by stromelysin or trypsin treatment) lowers the fixed charge density and reduces the pressure of internal swelling, resulting in a decrease in equilibrium modulus and an increase in tissue deformability (17). Interestingly, loss of osmotic prestress can delay the onset of nonlinearity during shear loading, as collagen fibers remain unengaged until larger strains are applied (28). Thus, osmotic stress influences not only the compressive response but also the recruitment and mechanical phase of the collagen network.

Finite element simulations incorporating osmotic prestrain reproduce observed changes in both modulus and strain distribution (39). Models that omit the osmotic loading consistently underestimate stiffness and fail to capture strain gradients, especially in the superficial zone. These results underscore the role of osmotic stress as a boundary condition that sets the mechanical operating point of the tissue.

In total, osmotic swelling pressure is not simply a passive consequence of matrix composition; it is an active regulator of tissue mechanics. Through its interplay with collagen architecture, it establishes a prestressed, nonlinearly tunable material capable of sustaining physiological loads. Disruption of this coupling, through enzymatic degradation, disease, or aging, compromises the mechanical performance of cartilage on both the local and tissue-wide scales.

3.4. Tissue Engineering Contexts

Efforts to replicate the mechanical function of native AC in engineered tissues require recapitulation of its multicomponent composition, depth-dependent structure, and emergent mechanical behavior. Although cell-seeded hydrogels and scaffold-based constructs have made progress in mimicking the biochemical profile of cartilage, reproducing its mechanical performance, particularly under shear and dynamic loading, remains a central challenge.

Engineered constructs often exhibit spatially homogeneous architecture, with limited development of the zonal organization critical for depth-specific mechanics. As a result, their equilibrium modulus, shear resistance, and energy dissipation capabilities fall short of those in native tissue. Longitudinal studies of in vivo implants have shown that though the PG and collagen content may gradually increase over time, the restoration of functional gradients and fiber organization is incomplete. In equine models, matrix-induced autologous chondrocyte implantation grafts remained mechanically inferior to adjacent native cartilage even after 12 months, particularly in the superficial region where collagen alignment was poorly restored (40).

This mechanical deficit reflects the absence of a robust collagen network and the failure to reestablish a zonally differentiated prestressed matrix. Attempts to address this have employed biomimetic scaffolds designed to guide fiber alignment and promote zonal stratification. For example, nanofiber-based scaffolds that mimic collagen architecture have been shown to enhance matrix deposition and improve mechanical properties in vitro and in vivo (6). Constructs incorporating spatially patterned biochemical cues or subjected to mechanical loading regimes have demonstrated improved alignment and zonal differentiation, but often lack sufficient collagen cross-linking to support physiological loads.

Multiscale mechanical assessments have identified critical compositional thresholds for function. Regions within constructs with locally low PG or collagen concentrations exhibit strain concentrations, suggesting that even small-scale heterogeneity can compromise global performance (41). These findings underscore the importance of establishing not only bulk composition but also appropriate spatial distribution and mechanical integration with host tissue.

4. MECHANICAL STRUCTURE-PROPERTY RELATIONSHIPS: THEORY AND MODELING

Theoretical frameworks play a central role in deciphering how the composition and architecture of AC give rise to its emergent mechanical behavior. Given the complex, depth-dependent, and multiphasic nature of the tissue, no single theory can capture all aspects of its response. Instead, different modeling paradigms have been developed to interrogate specific regimes—poroelastic theory (42–47) for fluid–solid interactions under compression, and rigidity percolation (33, 34, 48–50) for understanding the linear and nonlinear, composition-sensitive response to shear. These theories complement experimental studies by providing a mechanistic language for interpreting data and generating predictive insights into how small-scale perturbations cascade into large-scale mechanical outcomes.

The implementations of theories of cartilage mechanics can be broadly grouped into continuum-level and micromechanical models. Continuum models are typically used to describe macroscopic deformations using constitutive relations derived from continuum elasticity or phenomenological principles (9, 51–54). In contrast, micromechanical models resolve structural heterogeneity explicitly and are well suited to exploring the impact of fiber architecture, network topology, compositional variations, and statistical mechanics (31–34, 36, 55–64) on tissue-scale mechanical response. Both classes of models are essential for connecting experimental measurements to mechanistic understanding.

4.1. Poroelasticity Theory

Poroelasticity provides the foundational framework for describing the time-dependent compressive response of cartilage as a fluid-saturated porous medium. This continuum theory treats the tissue as a composite of a viscoelastic solid matrix and an interstitial fluid phase. Building upon Biot’s theory of porous media (42–44), biphasic formulations developed by Mow et al. (9, 10) demonstrated that compressive loads are initially supported by fluid pressurization, which relaxes over time as fluid flows through the matrix. The characteristic timescale for this relaxation, $t_{\text{c}}~L^2/D$, depends on sample thickness $L$ and hydraulic diffusivity $D$. The equilibrium response is governed by the mechanical stiffness of the solid matrix, whereas the transient response reflects the dynamic coupling between matrix deformation and solvent redistribution. Extensions to the biphasic theory, including triphasic formulations, incorporate ionic species and fixed charges to model swelling pressure and electrochemical gradients (37). These models capture key features of cartilage mechanics, including creep, stress relaxation, osmotic swelling, and electromechanical coupling.

4.2. Rigidity Percolation Theory

Although poroelastic models capture the essential features of time-dependent compression, they do not account for the observed sensitivity of the shear modulus to microscale structure. Rigidity percolation theory addresses this by modeling the collagen network as a disordered assembly of semiflexible fibers. In this statistical mechanical framework, the system undergoes a rigidity transition when the average coordination number of fibers exceeds a critical threshold. Below this point, the network is floppy and incapable of bearing shear stress; above it, a rigid spanning cluster forms that supports mechanical loads (30, 33, 34, 48, 59, 62).

Experimental observations reveal that cartilage near the articular surface lies near this critical point (30). The shear modulus varies by orders of magnitude over submillimeter-depth scales, correlating more strongly with collagen volume fraction than with fiber orientation. This extreme sensitivity is indicative of a system poised near a rigidity threshold. Simulations based on kagome-lattice models with depth-varying connectivity reproduce the observed modulus gradients and predict that small reductions in collagen density or cross-linking—such as those caused by enzymatic degradation—can push the tissue below the threshold, dramatically softening its shear response (30, 62). Conversely, the presence of a reinforcing aggrecan gel matrix can lower the rigidity threshold, stabilize the network, and suppress nonaffine deformations. In this view, the mechanical phase of cartilage is set by the interplay of collagen connectivity, fiber stiffness, and gel reinforcement—a parameter-sensitive regime in which subtle compositional changes produce large mechanical consequences.

This model is supported by depth-resolved mechanical measurements that show twofold changes in collagen concentration can produce one- to two-orders-of-magnitude differences in shear modulus. Such scaling behavior is not easily explained by alignment or bulk material density but is consistent with a mechanical phase transition governed by critical-like behavior. The transitional zone, which exhibits the lowest collagen volume fraction and highest compliance, appears to operate closest to this rigidity threshold (29).

Simulations of disordered filament networks, based on bond-diluted kagome lattices, explicitly model fiber mechanics by assigning stretching and bending stiffnesses to discrete collagen segments. These networks exhibit a rigidity phase transition, with the shear modulus following a power law $G~{\left(p-p_{\text{c}}\right)}^{\beta }$ as the system approaches the rigidity threshold from above, where $p$ is the bond occupation probability and $p_c$ is the critical value. In the linear regime near the threshold, the response is largely bending-dominated, producing pronounced nonaffine deformations. Stretch-dominated behavior emerges only once sufficiently rigid, percolating load paths form, which can subsequently lead to strain stiffening. The aggrecan-rich gel reinforces the collagen network by adding an elastic energy term that penalizes transverse fiber displacements, thereby increasing mechanical constraints and lowering the rigidity percolation threshold. In composite models, the shear response reflects contributions from both the fiber network and gel, enabling the mechanical response to be tuned via volume fractions and coupling strength.

These theoretical predictions align with experimental perturbation studies. Enzymatic depletion of aggrecan raises the percolation threshold and reduces stiffness, especially in zones near criticality. Conversely, degrading collagen reduces connectivity to the point that the shear modulus approaches that of the gel matrix alone (18). The nonlinear stiffening observed under shear strain can also be understood as a strain-controlled percolation: At large strains, fibers align with the direction of applied strain and transition from slack to taut, shifting the network from bending-to stretching-dominated mechanics.

Altogether, rigidity percolation theory provides a mechanistic explanation for the emergent and highly tunable shear behavior of cartilage. It connects microscale architecture to tissue-scale function and explains how minor changes in matrix composition or organization—due to aging, disease, or injury—can result in large changes in mechanical performance. The theory further offers predictive power for engineered tissues, emphasizing the importance of reaching sufficient collagen connectivity and coupling to recapitulate native-like mechanics.

Together, poroelastic and rigidity percolation theories frame the multiscale mechanics of cartilage across distinct but complementary regimes. Poroelasticity governs compression and time-dependent deformation through fluid-solid coupling, whereas rigidity percolation governs shear and nonlinear mechanics through compositional and architectural thresholds. These theories, when used in concert, offer a powerful lens for interpreting experimental data and guiding the design of engineered tissues that aim to replicate the complex mechanical environment of native cartilage.

4.3. Continuum Models

In continuum elasticity formulations (9, 23, 53), cartilage is typically treated as a deformable solid in the Lagrangian frame, with material points labeled in the undeformed configuration and displacements tracked through a deformation map such that each material point $\overrightarrow{X}$ in a reference state is displaced by some amount $\overrightarrow{u}(\overrightarrow{X})$ to a deformed position $\overrightarrow{x}(\overrightarrow{X})=\overrightarrow{X}+\overrightarrow{u}(\overrightarrow{X})$. As soft biological tissues can often support macroscopic and reversible deformations, it is common to formulate continuum theories of tissues using finite strain elasticity theory. For clarity of exposition, we briefly review some important concepts here. Further details may be found, for instance, in Reference 65. In $d$ dimensions, the deformation gradient tensor $\mathbf{F}$ is defined such that $F_{iJ}=\frac{\partial x_i}{\partial X_J}$, $1\le i,J\le d$, where lower case indices correspond to deformed coordinates and upper case indices correspond to coordinates in the reference state. A deformation is characterized by local eigen vectors of $\mathbf{F}$ and the accompanying eigenvalues ${\lambda }_i,1\le i\le d$. The ${\lambda }_i$ are termed the principal stretches and are bounded from below by 0, with a value less than 1 indicating compression, a value greater than 1 indicating extension, and a value of 1 indicating neither extension nor compression along a given eigen direction. The nonlinear Cauchy–Green deformation tensor is then defined as $\mathbf{C}={\mathbf{F}}^T\mathbf{F}$, and the Green-Lagrange strain tensor as $\mathbf{E}=\frac{1}{2}\left({\mathbf{F}}^T\mathbf{F}-\mathbb{I}\right)$.

A continuum mechanical model demands a constitutive relation that maps a deformation onto the energy per unit volume required to produce that deformation, $W(\mathbf{F})$. Many polymeric materials are presumed to be elastically isotropic, so that $W$ should transform as a true scalar under rotation of the local displacement field. Any such function of $\mathbf{F}$ can be written as a function of three principal invariants, $I_1,I_2$, and $I_3$, defined as

$$\begin{gathered} I_1\equiv \text{tr}\left(F^{T}F\right), \\ {I}_2\equiv \frac{1}{2}\left[I_1^2-\text{tr}\left(C^2\right)\right],\text{and} \\ {I}_3\equiv \text{det}(F{)}^2. \end{gathered}$$ 1.

It is also convenient to define $J\equiv \text{det}(\mathbf{F})=\sqrt{I_3}$. As $J$ is simply the Jacobian of the deformed coordinates relative to the undeformed coordinates, a deformation is contractile if $J<1$, dilatational if $J>1$, and volume-conserving or isochoric if $J=1$. Once $W(\mathbf{F})$ has been identified, the nominal stress, $\mathbf{s}$, is defined as $s_{iJ}=\frac{\partial W}{\partial F_{iJ}}$. The Cauchy stress, $\mathit{\sigma }$, may in turn be computed using the relation ${\sigma }_{ij}=\frac{s_{iK}{F}_{jK}}{J}$, where summation over repeated indices is assumed.

In gel theories using the above formalism, care must be taken in identifying the reference state. In some approaches, the reference state is the dry or preparation state, prior to the polymer network imbibing solvent. In other cases, the reference state is taken to be the swollen state, after contributions to the deformation energy due to polymer entropy, polymer elasticity, osmotic stress, and screened electrostatic interactions have been balanced. Theories that explicitly track ion concentrations involve chemical potentials ${\mu }_i:{\mu }_i=q_i\varphi +\frac{\partial W}{\partial C_i}$, where $\varphi$ is the electrostatic potential, and $q_i$ and $C_i$ are the charge per ion and the concentration of the $i$th ionic species, respectively. In modeling dynamic processes, gradients in chemical potentials are related to currents via mobilities, whereas in equilibrium, chemical potential gradients, as well as chemical potential differences across permeable membranes, should vanish. See, for instance, References 66–69 for additional discussion of the modeling of polyelectrolyte gels.

In interpreting the shear response of a model ECM containing reconstituted collagen and GAGs, Chen et al. (70), extending upon work in Reference 71, identified three contributions to the deformation energy density: GAGs, collagen not aligned with the shear direction, and collagen aligned with the shear direction. The GAG and unaligned collagen contributions are described with energy densities of the form

$$W\propto \frac{\mu }{2}\left(\frac{\text{tr}\left({\mathbf{F}}^T\mathbf{F}\right)}{J^{2/3}}-3\right)+K(J-1{)}^2,$$ 2.

where $K$ is the bulk modulus of the material. The first term punishes shear, and the second term punishes expansion and contraction. In the case of GAG, Equation 2 was scaled by the ratio of the current to the reference value of the GAG concentration, whereas in the case of unaligned collagen fibers, Equation 2 was scaled by $J^{\alpha }$ for $J>1$, with $\alpha$ being a strain-hardening exponent. Contributions to aligned fibers were accounted for by functions of the eigenvalues of the shearing or deviatoric part of the strain gradient tensor.

In modeling the growth of regenerated tissue within a polyethylene glycol (PEG) hydrogel, Sridhar et al. (72) found it sufficient to linearize the elastic stress of the hydrogel and cultured ECM phases; i.e.,

$${\sigma }_{ij}=\lambda \text{tr}\left(\mathit{\epsilon }\right){\delta }_{ij}+2\mu {\epsilon }_{ij}.$$ 3.

Here, $\mathit{\epsilon }$ is the linearized strain tensor given by $\frac{1}{2}\left(\mathbf{F}+{\mathbf{F}}^T\right)$, and $\mu$ and $\lambda$ are the Lamè coefficients (73). In the hydrogel phase, $\lambda$ and $\mu$ are related by the assumption that the hydrogel is nearly incompressible, and the shear modulus diminishes as the cultured ECM replaces the initial PEG scaffold and cross-links are dissolved.

4.4. Micromechanical Network Models

Although continuum models are useful for capturing bulk mechanical behavior and providing computationally tractable descriptions at the tissue scale, resolving the mechanics at microscopic length scales is essential for understanding emergent phenomena such as strain localization, nonaffine deformation, and failure. In particular, local rearrangements and the onset of rigidity in sparse, disordered networks are not well described by continuum theories alone. Collagen networks have been studied extensively in the biophysics literature, both in isolation and in combination with surrounding matrix components (74–77). Here, we describe modeling approaches designed to capture the elastic response that arises from the large bending stiffness and dense cross-linking characteristic of type II collagen fibers in extracellular matrices. Collagen has a hierarchical structure: Fibrils with nanoscale radii (~50–500 nm), formed by the self-assembly of tropocollagen, bundle into fibers with microscale radii. Most mathematical micromechanical models treat fibers as the elastic units.

Collagen fibrils are relatively stiff, with reported persistence lengths ranging from tens to approximately 100 nm at physiological temperature (78). The exact value depends on fibril radius and local biochemical environment. This persistence length is comparable with typical cross-linking distances measured in histological studies (79), placing the network in the semiflexible regime, where mechanical energy associated with bending and stretching dominates over entropic contributions. This behavior contrasts with materials such as vulcanized rubber, where configurational entropy drives elasticity (80, 81).

A simple yet powerful modeling framework treats individual fibers as Eulerian elasticas, in which the fibers are described as curves $\overrightarrow{r}(s)$, where $s$ is an arc length coordinate ranging from 0 to $L$, the length of the fiber (see, for instance, 59, 73, 82). It is also useful to define the local direction vector, $\hat{t}$, of the fiber $\hat{t}=\frac{\frac{\text{d}\overrightarrow{r}}{\text{d}s}}{\left|\frac{\text{d}\overrightarrow{r}}{\text{d}s}\right|}$. The work to deform the elastica consists of a bending contribution, governed by the bending modulus, $B$ (with dimensions of energy times length), and a stretching contribution, governed by the stretching modulus, ${\mu }_{\text{s}}$ (with dimensions of energy per unit length). If each point along the initial curve $\overrightarrow{r}(s)$ is displaced by some vector $\overrightarrow{u}(s)$, and the fiber has an initially straight conformation, the position of a point parameterized by arc length coordinate $s$ is now ${\overrightarrow{r}}^{\prime }(s)=\overrightarrow{r}(s)+\overrightarrow{u}(s)$, with associated direction vector ${\hat{t}}^{\prime }(s)$. The work of deformation is given by

$$U=\frac{B}{2}{\int }_0^L{\left|\frac{\text{d}{\hat{t}}^{\prime }}{\text{d}s}\right|}^2\text{d}s+\frac{{\mu }_{\text{s}}}{2}{\int }_0^L{\left|\hat{t}\cdot \frac{\text{d}\overrightarrow{u}}{\text{d}s}\right|}^2\text{d}s\approx \frac{B}{2}{\int }_0^L{\left|\frac{d^2{\overrightarrow{u}}_{\perp }}{\text{d}{s}^2}\right|}^2\text{d}s+\frac{{\mu }_{\text{s}}}{2}{\int }_0^L{\left|\frac{\text{d}{\overrightarrow{u}}_{\Vert }}{\text{d}s}\right|}^2\text{d}s,$$ 4.

where ${\overrightarrow{u}}_{\Vert }\equiv \overrightarrow{u}\cdot \hat{t}$ and ${\overrightarrow{u}}_{\perp }\equiv \overrightarrow{u}-{\overrightarrow{u}}_{\Vert }$.

In micromechanical modeling of fiber networks, it is commonplace to discretize fibers into a set of $N$ segments of length $l_0$. If consecutive segments $m$ and $m+1$ have direction vectors ${\hat{t}}_m$ and ${\hat{t}}_{m+1}$, and ${\hat{t}}_{m+1}$ is rotated by an angle $\Delta {\theta }_m$ relative to ${\hat{t}}_m$, then the magnitude of the local curvature over these segments is approximated as

$$\left|\frac{\text{d}\hat{t}}{\text{d}s}\right|\approx \frac{\left|{\hat{t}}_m-{\hat{t}}_{m+1}\right|}{2l_0}\approx \frac{\sqrt{1-\text{cos}\left(\Delta {\theta }_m\right)}}{l_0}\approx \frac{\Delta {\theta }_m}{l_0\sqrt{2}}.$$ 5.

If segment $m$ currently has a length $l_m$, it is supposed that this segment is subjected to a uniform strain ${ϵ}_m=l_m/l_0-1$. With these approximations, the continuum result in Equation 4 is replaced by

$$U_{\text{discrete}}=\frac{B}{2l_0}\sum _{m=1}^{N-1}\Delta {\theta }_m^2+\frac{{\mu }_{\text{s}}}{2}\sum _{m=1}^N\frac{{\left(l_m-l_0\right)}^2}{l_0}=\frac{\kappa }{2}\sum _{m=1}^{N-1}\Delta {\theta }_m^2+\frac{\alpha }{2}\sum _{m=1}^N{\left(l_m-l_0\right)}^2,$$ 6.

where we define $\kappa \equiv \frac{B}{l_0}$ and $\alpha \equiv \frac{{\mu }_{\text{s}}}{l_0}$. If a fiber is divided into segments of varying length (as is the case for Mikado networks, discussed below), then the preceding expression for $\kappa$ for the joint between segments $m$ and $m+1$ is replaced by $\kappa \equiv \frac{2B}{l_{0,m}+l_{0,m+1}}$, where $l_{0,m}$ and $l_{0,m+1}$ are the rest lengths of segments $m$ and $m+1$, respectively.

Model networks of discretized filamentous proteins fall into two broad categories: lattice-based and off-lattice networks. Lattice-based networks are constructed by placing fibers spanning the breadth of the simulation volume such that fibers intersect at points belonging to a Bravais lattice or a lattice with a basis. Off-lattice networks are constructed by placing fibers with random positions and orientations. Lattice-based networks have the advantages of ease of construction and the availability of analytic predictions for elastic moduli, whereas off-lattice models ensure elastic allotropy is not introduced as an artifact of the network creation process and allow decoupling of filament density and network topology. One can further adjust the fiber concentration and introduce disorder by randomly removing connections between adjacent cross-links in the network, such that fiber segments between adjacent cross-links are retained with a bond occupation probability $p$. At a critical concentration, $p_c$, elastic fiber networks exhibit a remarkable phenomenon: a many-decade increase in shear modulus over a narrow range of $p$. This transition has many features in common with thermodynamic phase transitions and is known as rigidity percolation. In general, a cluster of vertices connected by elastic bonds is considered rigid if there is no way to vary the distance between any two vertices in the cluster without an energy cost.

Once the network has been defined, the energy of deformation associated with a displacement field $\overrightarrow{u}$ is (34, 57)

$$U_{\text{net}}=\sum _{⟨ij⟩}\frac{\alpha }{2}{\left(l_{ij}-l_{0,ij}\right)}^2{p}_{ij}+\sum _{⟨ijk⟩}\frac{\kappa }{2}\Delta {\theta }_{ijk}^2{p}_{ij}{p}_{jk}.$$ 7.

The first summation is over all pairs of vertices $⟨ij⟩$ joined by a bond, whereas the second sum is over all triples of consecutive points $⟨ijk⟩$ along a fiber. We further define the following:

$$p_{ij}\equiv \left\{\begin{array}{ll}1,& \text{bond}{p}_{ij}\text{is retained}\\ 0,& \text{otherwise}.\end{array}\right.$$

For initially straight fibers subjected to a small displacement field, Equation 7 is often approximated by its leading order expansion:

$${\left.U_{\text{net}}\approx \frac{1}{2}{\overrightarrow{u}}^T\frac{{\partial }^2{U}_{\text{net}}}{\partial {\overrightarrow{u}}^2}\right|}_{\overrightarrow{u}=0}\overrightarrow{u}=\sum _{⟨ij⟩}\frac{\alpha }{2}{\left({\overrightarrow{u}}_{ij}\cdot {\hat{r}}_{ij}\right)}^2{p}_{ij}+\sum _{⟨ijk⟩}\frac{\kappa }{2}{\left|\left({\overrightarrow{u}}_{ij}+{\overrightarrow{u}}_{jk}\right)\times {\hat{n}}_{ijk}\right|}^2{p}_{ij}{p}_{jk},$$ 8.

where ${\overrightarrow{u}}_{ij}$ is the difference in the displacements of vertices $i$ and $j,{\hat{r}}_{ij}$ is a unit vector pointing from vertex $j$ to vertex $i$ in the undeformed state, and ${\hat{n}}_{ijk}$ is the axis about which bond $ij$ rotates toward bond $jk$.

This linearization enables effective medium theory (EMT) calculations (33, 34, 48, 50, 57), in which the disordered network is approximated as a homogeneous medium that, on average, deforms affinely under simple boundary conditions such as simple shear or uniaxial extension or compression. These approximations yield analytic estimates of modulus under shear, extension, or compression. EMTs agree well with numerical simulations when $p>p_{\text{c}}$, but they show considerable deviation from computer simulations near the percolation threshold, where nonaffine deformations dominate.

In the limiting case in which $\alpha \gg \kappa /l^2$, with $l$ being the characteristic spacing between the cross-links, the elastic response of the network is dominated by the resistance of the bonds to stretching and compression. In this case, termed the central force limit, the rigidity percolation threshold agrees well with a classic criterion due to J.C. Maxwell (83), which gives $z_{\text{CF}}\ge 2d$, where $z$ is the mean number of nearest neighbors to which a cross-link is joined by a fiber segment, and $d$ is the dimensionality. This requirement can be relaxed with the application of prestress, a phenomenon discovered heuristically by Fuller & Applewhite (84) and formally established by Calladine (85). Let sections of fiber between cross-links be subjected to prestresses, such that fiber segment $i$ is under a force ${\tau }_i$, and let a negative sign of ${\tau }_i$ correspond to compression and a positive sign correspond to extension. Then a compatibility matrix can be defined as $\mathbf{C}\in {\mathbb{R}}^{d{N}_{\text{c}}\times N_{\text{B}}}$, with $N_{\text{c}}$ being the number of cross-links, and $N_{\text{B}}$ being the number of bonds, such that the resultant forces $\overrightarrow{f}\in {\mathbb{R}}^{d{N}_{\text{c}}\times 1}$ on the cross-links are $\overrightarrow{f}=\mathbf{C}\overrightarrow{\tau }$.

Null eigenvectors of $\mathbf{C}$ correspond to states of self stress: configurations of internal tension that produce no net force on any node. Such prestresses can stabilize otherwise floppy networks. This concept is especially relevant to cartilage, in which osmotic swelling and fiber entanglement generate internal tension. Experiments show that modulating the ionic environment reduces swelling pressure and tissue stiffness (86), whereas mechanical compression has similar effects on modulus (28).

Numerous procedures exist to estimate the elastic moduli of fiber networks, depending on the loading conditions. In the quasistatic regime, the procedure involves applying prescribed deformations and relaxing the network to its minimum energy configuration, which is subject to boundary conditions on the displacement field. Given a sufficiently small shear or dilatational strain $ϵ$, the deformation energy $U$ is related to the corresponding elastic modulus $C$ as $U=\frac{C{ϵ}^{2}V}{2}$, where $V$ is the generalized volume of the sample. If a strain-dependent modulus is desired, or the network is subject to multiple types of strain (such as uniaxial compression followed by shear), the differential modulus, $K_{\text{d}}$, is given by

$$K_{\text{d}}=\frac{1}{V}\frac{{\partial }^{2}U}{\partial {ϵ}^2}.$$ 9.

In simulations, a common practice is to evaluate the energy at a number of closely spaced strains and approximating Equation 9 using a centered difference approximation (see, for instance, 62). Biological tissues have long been known to exhibit pronounced strain stiffening, and network models presuming individual elements that are governed by the constitutive relation Equation 4 have successfully accounted for the observed nonlinear elasticity of in vitro collagen gels (60, 87). Sharma et al. (74) found a model of the form Equation 7 that could quantitatively reproduce the scaling of the shear modulus with strain for networks initially too poorly coordinated to bear appreciable shear stress.

Although rigidity percolation has been successful in reproducing the shear mechanics of in vitro collagen networks, explicit modeling of the elasticity of the background polyelectrolyte gel is important for capturing the mechanics of AC. Silverberg et al. (30) extended Equation 8 to include a coupling term between the network and the surrounding ECM to give an overall energy,

$${𝒰}_{\text{tissue}}=\sum _{⟨ij⟩}\frac{\alpha }{2}{\left({\overrightarrow{u}}_{ij}\cdot {\hat{r}}_{ij}\right)}^2{p}_{ij}+\sum _{⟨ijk⟩}\frac{\kappa }{2}{\left|\left({\overrightarrow{u}}_{ij}+{\overrightarrow{u}}_{jk}\right)\times {\hat{n}}_{ijk}\right|}^2{p}_{ij}{p}_{jk}+\frac{\mu }{2}\sum _{⟨ij⟩}{\left|{\overrightarrow{u}}_{ij}-\left({\overrightarrow{u}}_{ij}\cdot {\hat{r}}_{ij}\right){\hat{r}}_{ij}\right|}^2.$$ 10.

The final term penalizes transverse displacements relative to the bond axis and captures the elastic support from the aggrecan-rich matrix. This formulation models the mechanics of composite systems and reproduces experimentally observed shear modulus profiles with spatial gradients in fiber density. This model captures the remarkable observation that a modest (about twofold) variation in collagen volume fraction can give rise to a dramatic (about hundredfold) change in shear modulus by treating the collagen matrix as a disordered fiber network operating near the rigidity threshold (30) (Figure 3).

Figure 3.

(a) A spatially heterogeneous network generated from experimentally measured collagen concentrations $v_{\text{f}}(z)$ at a given tissue depth $z$. Zoomed insets show the region near the articular surface is below the percolation threshold, whereas the region at greater $z$ is more well connected. For each inset, the largest percolating cluster is colored red, whereas the remaining network is colored black. (b) For specific model parameters, a comparison between experiment (light gray points) and simulation (dark gray points) shows reasonable agreement. The shear modulus is decomposed into contributions from the fiber network ($G_{\text{fiber}}\left(v_{\text{f}},\mu /\alpha \right)$, dashed blue) and the background medium (${\mu }_0\left(1-v_{\text{f}}\right)$, solid blue). (c) The spatially homogeneous $v_{\text{f}}$ is replaced with experimentally measured $v_{\text{c}}(z)$ to generate a depth-dependent $G(z)$ profile (red), which when superimposed on experimental data (light gray lines) shows qualitatively similar behavior. Figure adapted with permission from Reference 30.

Recent extensions of rigidity percolation theory have incorporated double-network models composed of a primary stiff network coupled to a mechanically distinct, more compliant secondary network. In such systems, the interaction between the two networks imposes additional mechanical constraints that lower the rigidity percolation threshold and modify the mechanical response of the system (61). When the primary network lies near its percolation threshold, the secondary network plays a dual role: It can either relax internal stresses through nonaffine modes or reinforce the network to support larger loads. Increasing the connectivity of the secondary network lowers the effective percolation threshold, increases the shear modulus, and alters failure characteristics including peak stress, strain at failure, and energy dissipation. These results highlight how double networks introduce tunable structure-function relationships that are not accessible in single network systems. In the context of AC, these models provide a mechanistic framework to understand how local compositional heterogeneity modulates phase behavior, nonlinear mechanics, and fracture resistance.

4.5. Data-Driven Rigidity Percolation Models

To test and calibrate rigidity percolation models against experimental data, recent work has integrated spatially matched measurements of ECM composition and shear mechanics in healthy and degraded cartilage tissue (Figure 4) (62). Using FTIR imaging to quantify local collagen and aggrecan concentrations, and confocal elastography to measure the depth-dependent shear modulus in corresponding samples, this approach provides a direct means of evaluating composition–function relationships in situ.

Figure 4.

(a) A schematic of the primary extracellular constituents of cartilage, collagen, and aggrecan. (b) The case in which aggrecan was digested away with an enzyme. (c) Rigidity percolation model based on a network with energetic contributions from stretching, bending, and movement through a background gel. This model was shown to accurately model healthy tissue and tissue in which aggrecan was digested with trypsin (d) as shown by predicted surface that captures experimental data points (solid lines). Panels $c$ and $d$ adapted from Reference 62 (CC BY 4.0).

These data reveal that the sensitivity of the shear modulus to aggrecan concentration strongly depends on the local collagen content. In regions near the cartilage surface, where the collagen network is sparse and close to its percolation threshold, small changes in aggrecan lead to large changes in modulus. In contrast, in the deep zone, where the collagen network is well connected, aggrecan plays a comparatively minor mechanical role. This behavior is consistent with a phase-sensitive regime in which the contribution of aggrecan is amplified when the primary network is marginally rigid.

To model these observations, a rigidity percolation framework was constructed using a kagome lattice representing the collagen network embedded in an elastic background gel representing aggrecan and other matrix components. The total energy includes terms for fiber stretching, bending, gel elasticity, and a coupling term that penalizes transverse displacements between the network and the gel. This coupling lowers the effective rigidity threshold and is essential for reproducing the experimentally observed shear modulus profiles.

Simulations spanning a range of collagen and aggrecan concentrations yielded shear modulus predictions that quantitatively match experimental measurements across both healthy and degraded tissue. In particular, exclusion of the gel–network coupling term leads to substantial misfit, particularly at low collagen concentrations, underscoring the importance of mechanical interactions between networks. These results show that data-driven modeling anchored in compositional measurements enables predictive mapping of cartilage mechanics and provides a mechanistic framework to assess disease progression and scaffold design (62).

5. CONCLUDING REMARKS AND FUTURE OUTLOOK

AC is a hierarchically structured composite material whose mechanical behavior arises from a finely tuned interplay among composition, architecture, and multiscale mechanical response. In this review, we highlight how structure-function relationships govern cartilage depth-dependent mechanical properties, particularly when the tissue is subject to shear stresses or strains. Across experimental and theoretical studies, a consistent picture emerges: Cartilage behaves as a mechanically heterogeneous composite operating near a rigidity transition, in which small changes in collagen or aggrecan concentration can produce large changes in stiffness and load-bearing capacity.

Despite these advances, key challenges remain. Experimentally, measurements of local mechanics and composition in vivo are still limited in spatial and temporal resolution. Techniques such as FTIR imaging and confocal elastography now enable microscale mapping of composition and shear modulus, but broader integration into in vivo or dynamic studies are needed. Cartilage exhibits significant heterogeneity across joint regions and with depth; capturing how this heterogeneity modulates strain, energy dissipation, and failure remains an open question. Developing minimally invasive tools to probe tissue-scale mechanics in living joints would enable direct tests of percolation-based hypotheses and provide critical data for personalized modeling.

On the theory and modeling front, treating cartilage as a disordered fiber network embedded in a prestressed gel has provided important insights, especially near the percolation threshold. However, open questions remain regarding how degradation, remodeling, and repair alter the topology of these networks over time. For example, how does chondrocyte-mediated matrix turnover or damage accumulation affect the system’s proximity to rigidity transitions? Can compositional gradients be tuned to prevent failure propagation or to prolong functionality under cyclic loading? Extending current models to include damage evolution, dynamic prestress, and feedback between mechanics and matrix synthesis will enable more predictive frameworks for tissue function, degeneration, and repair.

Emerging data-driven and machine learning approaches offer new tools to address this complexity. These methods can analyze multiscale mechanical and compositional data sets across scales, identify precursors to failure or dysfunction, and accelerate the discovery of phase boundaries or mechanical tipping points. Hybrid models that combine physics-informed mechanistic simulations with data-driven calibration have already demonstrated success in mapping local concentration profiles to mechanical outcomes. As these tools mature, they can allow real-time prediction of disease progression, individualized tissue diagnostics, and optimization of repair strategies.

Insights from cartilage mechanics are also guiding the design of synthetic and engineered materials. Double-network architectures—consisting of a marginally rigid fiber network embedded in a soft, prestressed matrix—capture many features of cartilage toughness and tunability. When tuned near a rigidity threshold, these systems exhibit enhanced stiffness and fracture resistance. Translating these principles into biomimetic hydrogels, soft robotic components, and tissue-engineered constructs holds promise for creating materials with cartilage-like resilience and mechanical performance. Spatial gradients, prestress, and marginal connectivity observed in native cartilage offer a blueprint for durable integration and functionality in synthetic systems.

Looking ahead, artificial intelligence-driven platforms combined with physics-based modeling could accelerate the discovery of optimal architectures, compositions, and fabrication protocols. Advances in 3D bioprinting, stem-cell-based engineering, and programmable materials will enable fabrication of constructs that replicate cartilage’s zonal, anisotropic, and prestressed structure. Engineered cells may not only deposit appropriate matrix components but also actively remodel the tissue in response to mechanical cues. In parallel, optogenetic tools and light-responsive materials will allow precise spatiotemporal control over matrix assembly, cross-linking, and contractility, opening new possibilities for directing tissue development and adaptation.

In conclusion, cartilage illustrates how biological systems exploit proximity to mechanical phase transitions to achieve functional adaptability. Its ability to straddle the boundary between fluid- and solid-like response, remaining compliant under low loads yet rigid enough to bear weight, underscores the relevance of soft matter concepts such as percolation, jamming, and nonaffine deformation in living systems. Future progress will require multiscale models that integrate composition, prestress, structural anisotropies, clustering, and remodeling; experimental techniques to map local properties in vivo and over time; and data-driven methods to predict and encode structure-function relationships and personalize predictions of tissue behavior. By addressing these challenges, the field stands to advance both fundamental understanding and translational impact. This progress will not only improve our ability to diagnose and treat joint disease but also guide the design of materials and constructs that match, or exceed, the mechanical performance of native cartilage.