The main load-bearing component of the tendon ECM is type I collagen (Col1), which composes 80–90% of tendon dry weight (see Supplementary Figure S2) (Kannus, 2000; Wang et al., 2012). Collagen type I is a heterotrimeric molecule composed of two α1 chains and one α2 chain forming a triple helix. Many collagen types (V, IX, X, XI, and XII) are found in smaller amounts within the tendon, though they still provide important functions. For example, collagen fibril diameter is partially regulated by collagen type V (Birk and Mayne, 1997; Wang et al., 2012). Other components involved in ECM modulation include type III collagen (Col3) which is important for quick cross-linking and stability of injury sites (Liu et al., 1995), integrin transmembrane proteins involved in sensing loads, and proteoglycans (decorin, aggrecan, etc., make up 1–5% of tendon dry weight) that hydrate the tendon and contribute to fibrillar slippage and resisting compression, respectively (Fessel and Snedeker, 2009; Wang et al., 2012; Screen et al., 2015). Matrix metalloproteinases (MMPs, mainly 1, 3, and 12) and ADAMTS-1, -4, -5, -8, -9, -15, and -20 (disintegrin and metalloproteinase with thrombospondin motifs, or aggrecanases) enzymatically breakdown various extracellular matrix components (Buono et al., 2013). Glycoproteins such as fibronectin, elastin (1–2% of tendon dry weight) (Kannus, 2000), and tenascin-C contribute to tendon repair and the elastic properties of the tendon (Wang et al., 2012). Intermolecular enzymatic cross-linking is an interaction mainly between thick Col1 fibrils, thinner Col3 fibrils, elastin, tenascin-C (Midwood et al., 2016), and proteoglycan components that occurs during maturation and adaptation to mechanical loads (Eyre et al., 1984; Kubow et al., 2015; Zitnay and Weiss, 2018). Consequently, the architectural arrangement of cross-linked components influences the load-bearing ECM-derived functional properties (Kjaer, 2004).

Transforming growth factor beta (TGF-β) is highly correlated to mechanical stress intensity and thus the tendon’s mechanoresponse via the Smad2/3 pathway (Figure 2A) (Dahlgren et alk., 2005; Maeda et al., 2011; Munger and Sheppard, 2011; Mendias et al., 2012). Downstream of the Smad2/3 pathway, Scleraxis (Scx), Mohawk (Mhk), and Tenomodulin (Tnmd) are tenocyte-specific markers that regulate proteoglycan production and collagen synthesis (Berthet et al., 2013; Subramanian et al., 2017). In developing mouse tendons, Tnmd expression is downstream of Scx, though they both likely play a role in Col1 transcription and adaptation to mechanical loading as well as tenocyte proliferation (Léjard et al., 2007). These markers have been identified as significant factors involved in tendon homeostasis and degeneration. They are heavily studied in model systems though exact mechanisms have not been clearly identified in all loading scenarios.

Current knowledge of the interaction between integrin proteins in cells and the ECM during loading is limited. It is unclear whether or not the cells respond to loads primarily via an integrin-mediated pathway or via stretch-activated ion channels that are caused by actual deformation of the cytoskeleton and nucleus in the cell. Decoupling of the mechanosensing process during maturation can be achieved using in vitro models of stem cell differentiation and ECM maturation in response to mechanical loads (Cosgrove et al., 2016). In other words, defining the cell-matrix interactions versus cell-cell interactions is necessary using simplified models.

Inflammation is vital to healing of acute injuries, yet the tendon has low regeneration potential once past early post-natal stages (Ansorge et al., 2012; Howell et al., 2017; Stauber et al., 2020). Thus, it is necessary for inflammatory pathways to be resolved to limit tissue damage and prevent continual degradation. The timeline for tendon healing is estimated (Figure 3) where pro-inflammatory signaling resolves between about week 1 and 6 post-injury, yet the underlying mechanisms for resolving inflammation (or not resolving inflammation) is unclear. Endogenous TPCs are capable of regenerating tenocytes, but this ability becomes limited due to aging and their susceptibility to injury (Xu and Liu, 2018). Furthermore, TPCs are implicated in fatty infiltration, fibrosis, and calcification in tendinopathy through TPC differentiation into adipocytes and chondrocytes (Figure 1), signaled by high levels of the pro-inflammatory cytokine prostaglandin E2 (PGE2) (Zhang and Wang, 2010, 2014; Agarwal et al., 2017).

Mechanical micro-ruptures and metabolic byproducts during repetitive or hyper-physiological loading stresses the tendon cells. Ultimately, this triggers an inflammatory cascade, leading to ECM degradation and further pathological characteristics (Figure 2B). Other factors highly involved in these pathways include pro-inflammatory cytokines interleukin-1 (IL-1) and tumor necrosis factor (TNF)-α (John et al., 2010), glutamate, vascular endothelial growth factor (VEGF), and substance P (Abate et al., 2009; Schulze-Tanzil et al., 2011). These signals activate angiogenesis and nerve ingrowth, leading to pain. In addition, they contribute to fibrosis formation due to macrophage signaling and localization and thus fibroblast activation (Dakin et al., 2012, 2017). Inflammatory pathways lead to ECM degradation and mechanical weakening of ECM via MMP upregulation and downregulation of tissue inhibitors of MMPs (Abate et al., 2009; Buono et al., 2013). Prolonged presence of macrophages further contributes to degradation of the ECM as they also secrete MMPs. Overall, micro-ruptures may also contribute to the onset of tendinopathy through the inability to properly heal or resolve inflammation before reoccurrence.

Studies in ex vivo models have shown that cell apoptosis and upregulation of inflammatory pathways occurs at 9–20% strain loading values (Maeda et al., 2010; Thorpe et al., 2015; Wang et al., 2015). Tendon rupture also leads to massive tenocyte cell death through high expression of TGF-β (Maeda et al., 2011). Meanwhile, underloading-induced tendon degeneration is driven primarily by enzymatic digestion which affects the tendon mechanics (Maeda et al., 2011; Wang and Chen, 2018). For example, unloaded rabbit patellar tendons in the presence of collagenase have an 80% decrease in elongation to failure, max failure force, and linear stiffness as compared to tendons loaded at 4% static strain (Nabeshima et al., 1996). Only physiological dynamic loading, which is generally between 4–8% strain at 0.25–1 Hz in ex vivo models, leads to increased collagen production and regenerative properties (Legerlotz et al., 2013; Wang et al., 2013, 2015). Highlighted tendon mechanobiology ex vivo work is summarized in Table 1 for recent findings from years 2015–2021. In addition, a general schematic of the combined culture and loading systems (bioreactors) used for ex vivo studies included in this review are illustrated in Figure 5A.

Many of the previously described ex vivo studies were limited in analyzing longer-term inflammatory effects, especially in overloading of explants. Only Maeda et al., 2011, measured cytokine factors after 7 days of culture with no difference found between unloaded and physiologically loaded groups. Despite this finding, measuring pro-inflammatory factors in hyper-physiologically loaded tissues over at least a 7-day period may help elucidate pathomechanisms associated with tendon matrix modulation by removing other tissue effects. Downstream of TGF-β signaling, hyper-physiological loading is expected to increase IL-1β expression (Figure 2B). It would be helpful to characterize temporal IL-1β expression in hyper-physiological loading scenarios to identify potential feedback loops within the tissue.

Inflammatory cascades within tendon can originate from other sources (Arvind and Huang, 2021). In a rotator cuff muscle, bone, and tendon explant culture model, 7 days in culture results in upregulation of TNF-α and IL-6, atrophy, and tenocyte cell death as compared to the tendon only explant culture (Connizzo and Grodzinsky, 2018). These findings correlate with upregulation of TNF-α, interleukin-6 (IL-6), and IL-1 in bone injuries (Connizzo and Grodzinsky, 2018) and reinforce that unloaded tendon explants alone may produce limited quantities of pro-inflammatory factors (Maeda et al., 2011). This indicates that the muscle and bone likely play a role in the inflammatory signaling and ultimate degeneration of tendon after injury. Culture conditions, specifically serum-supplemented media, likely play a role in directing tenocytes to a pro-inflammatory phenotype as well in both whole tendon explants and 2D culture (van Vijven et al., 2021). Underlying presence of pro-inflammatory factors is generally unknown or varies in serum batches, which may contribute to the pro-inflammatory phenotypes. A primary limitation with these studies is that the tissues were unloaded once explanted; however, these results highlight the need to understand the role of other tissues and culture conditions in inflammatory signaling.

Mechanical loading to a specific strain value is less physiologically relevant as compared to stress-specific loading. This is due to the viscoelastic nature of the tendon, where repeated loading to a certain strain level on the tendon leads to decreasing absolute stress over time. Variations in cross-sectional area lead to differential load distribution across samples when stretched to the same strain as well. An ex vivo bioreactor was designed for this issue specifically, where researchers could characterize the link between tendon explant fatigue loading with tendon degeneration using cyclic loads and stresses instead of strains (Pedaprolu and Szczesny, 2021).

Physiological mechanical loading is necessary for tissue homeostasis though may also have therapeutic effects following injury. In a study by Wang et al., researchers induced early-stage tendinopathy of the rabbit Achilles tendon by unloading the tendon for 6 or 12 days, and then they reversed the effect in both treatment groups via external mechanical stimulation in a previously developed bioreactor (Wang et al., 2015). They stimulated the tendon under 6% strain, 0.25 Hz, and 8 h per day for 6 days based on their previous work estimating loads of 5–6% strain during hopping (schematic illustrated in Table 1). After the 6 days, results indicated reversal of the biochemical, structural, and mechanical changes that were induced by unloading. Significantly, they measured increased Col1 production and decreased cell apoptosis and ECM degradation (Wang et al., 2015). Their work highlights the need for further research on the regenerative capacity of tendon.

Overall, few studies have focused on gliding tendons and the role that compressive and shear forces play in tendon pathologies such as carpal tunnel syndrome (Tse and Keir, 2020). The pathomechanism is widely unknown for such cases. A pilot study by Tohidnezhad et al. measured tenocyte-specific marker expression and ECM remodeling components (MMP-1 and 13 and proteoglycans) in both the traction and gliding, with chondrocyte-like tenocytes, areas of the tendon after 48 hours of uniaxial cyclical loading (refer to Table 1 for the loading schematic). They found that the gliding area of the tendon that is exposed to greater shear and compressive forces had greater upregulation of matrix degradation markers compared to the traction area (Tohidnezhad et al., 2020). This may correlate to clinical observations that tendon ruptures frequently occur in the gliding area of tendons. However, limitations of this study include the short stimulation period (48 h) and limited analysis of the many factors involved in matrix remodeling.

Previous work indicates that loading type (biaxial versus uniaxial) and structural environment (2D versus 3D) affect cellular responses in vitro. Loading strains of greater than 10–12% induce expected pathways, such as MMP upregulation, inflammatory factors, cell apoptosis, and angiogenic factors in 2D (Arnoczky et al., 2002) and upregulation of inflammatory factors and MMP with decreased tenogenic factors in 3D cultures (Pentzold and Wildemann, 2022). Similarly, underloading led to activation of cellular inflammation and matrix degradation pathways (Arnoczky et al., 2002;; Tsuzaki et al., 2003a; Chen et al., 2009). Physiological loading ranges vary depending if the platform is 2D or 3D as well as 3D construct material (Patel et al., 2017). In general, local strains to the cells in 3D culture are dependent on the scaffold material and crosslinking density of hydrogels (Bryant et al., 2004). Fibrous scaffolds also may lead to higher variability than bulk 3D hydrogels that can more uniformly transfer loads to all cells within the construct (Kloxin et al., 2010). Highlighted tendon mechanobiology in vitro work is summarized in Table 2 for recent findings from years 2015–2020. Bioreactor setups used for in vitro work are outlined in Figures 5B,C.

For in vitro studies, loading type and frequency affect catabolic and anabolic pathways. In 2D, a frequency of 1 Hz (similar to walking) compared to 2 Hz (similar to sprinting) lead to greater cell proliferation of 2D cultures, greater cell viability, greater Col 1 expression, reduced expression of MMP-1 and MMP-13, and decreased expression of VEGF (Kubo et al., 2020). The loading schematic is illustrated in Table 2 for Kubo et al., 2020. In 3D cell constructs, dynamic loading type affects TPC differentiation and tendon formation via altering pathway activation. With biaxial loading, tenocytes experience both longitudinal and radial or lateral loading. Uniaxial loading provides a more physiologically relevant stimulation of the 3D cell culture niche that is necessary for tenogenic differentiation (Wang and Thien, 2018). Uniaxial loading in 3D cell culture led to the discovery that the transmembrane integrin protein is activated via the PI3K/AKT pathway (Manning and Cantley, 2007; Paterno et al., 2011). Meanwhile, tissue constructs aim to mimic the collagen fibers loading scheme with the correct ratio of higher magnitude tensile to lower magnitude shear forces (Patel et al., 2017; Sawadkar et al., 2020). In native tendon, whole tissue strain varies from the local fiber strains, with fiber extension reaching only about 40% of the whole tissue (Thorpe et al., 2013; Shepherd et al., 2014). One particular 3D hydrogel was developed to mimic the dynamics of collagen fibers in the native tendon and therefore forces, such that the sliding of fibers produced shear forces and fiber extension as tension (Patel et al., 2017).

As with ex vivo models, there is a need to characterize inflammatory effects in these models of tendon degeneration. Studies found that IL-1β expression is upregulated in tendon cells by exogenous IL-1β (M Tsuzaki et al., 2003b), which triggers the cytokine-induced MMP matrix degradation pathways due to fibroblast activation. In addition, IL-1β alters metabolic pathways and inhibits TPC differentiation potential irreversibly through downregulation of Col1, Col3, Scx, Tnmd, and biglycan (Zhang et al., 2015). Another pilot study by Stolk et al. highlighted that pro-inflammatory pre-stimulated human tenocytes co-cultured with macrophages lead to altered surface markers and increased expression of inflammatory cytokines, such as IL-6 and IL-8, as well as influenced macrophage polarization to a pro-inflammatory state after 3 days (Stolk et al., 2017). Results from this work corroborates cytokine signals originating from other tissues might play a role in tendon degenerative pathways as well as contribute to a positive-feedback loop in tendon cells.

Although 3D cultures facilitate manipulation of the loading protocols, the current limitation of this work is that limited inflammatory analysis is performed on tissue constructs. More specifically, combinatory effects of loading and culture conditions on inflammatory responses needs to be explored. Determining inflammatory effects in vitro is also further complicated by tenocyte response variations associated with the model species used (Oreff et al., 2021), which indicates the importance of selecting a model that most appropriately matches the research question.

Live animal models are commonly used in the study of tendinopathy as they are the most physiologically relevant. Overloading of the Achilles tendons is often performed in rodents as they have homologous anatomy and physiology to humans. They have shown that intensive treadmill running induces degenerative tendon pathologies, including TPC differentiation into chondrocyte-like cells that lead to non-tendinous tissue signaling (Zhang et al., 2020). Meanwhile, more moderate treadmill running promotes TPC differentiation into tenocytes that achieve homeostasis (Zhang et al., 2020). In a recent equine model, researchers found that biomechanical and compositional adaptation that occur from mechanical loading are localized to the non-collagenous extracellular matrix (Zamboulis et al., 2020). However, discrepancies may exist between in vivo and in vitro models. A pilot study illustrated differences between expressions of ECM components for in vitro mouse Achilles tenocytes versus in vivo mouse Achilles tendons. Specifically, in vitro loading led to a higher expression of collagen type III versus in vivo (Fleischhacker et al., 2020). Though successful at modeling tendinopathy, in vivo models are plagued by confounding variables compounded by the difficulty in identifying the underlying mechanisms.

Nevertheless, in vivo models highlight important findings that could shift focus for future investigations, as presented by Zamboulis et al., 2020 with the finding of non-collagenous matrix adaptations. A piezo-bioelectric device used with in vitro and in vivo systems of tendon regeneration suggest the need to further investigate the role of mechanical stimulation in modulating tenogenic-specific phenotypes (Fernandez-Yague et al., 2021). Fernandez-Yague et al. suggests that mechanical stimulation may induce pro-regenerative pathways via modulation of ion channel sensitivity. Furthermore, findings from Passini et al. suggest that shear stress is sensed through the specific PIEZO1 calcium ion channel in rodents and leads to upregulation of collagen cross-linking factors related to tendon stiffness (Passini et al., 2021). The direct effect of various mechanical stimuli in modulating these mechanosensing pathways is important to uncover in future work.

Overall, live models of tendon degeneration and regeneration show age-dependent mechanistic effects (Freedman et al., 2022; Kinitz et al., 2022). Specifically, mouse Achilles tendon following transection recruits Scx-lineage tenocytes to the repair site in neonatal development, but not in adult mice (Howell et al., 2017). The neonatal healing strategy ultimately results in functional restoration, whereas the adult suffers from dysfunction due to fibrotic scarring. Despite this new finding, the underlying mechanism of Scx-lineage tenocyte recruitment is unclear.

Recent advancements in multiscale models have enabled researchers to estimate in vivo stress and strain values (Pizzolato et al., 2020; Devaprakash et al., 2022) as well as investigate the role of mechanical loading (both quasi-static and dynamic) and tendon healing on multiscale mechanical, structural, and compositional properties (Freedman et al., 2018). Patient-specific finite element (FE) models of the human Achilles tendon and an OpenSim neuromusckuloskeletal model may be combined for measuring real-time stress and strain values during various activities (Pizzolato et al., 2020). In another FE analysis, macroscale tendon strain stiffening is reduced with higher magnitude and longer duration loading as well as increased laxity and delayed fiber re-alignment with applied strain (Freedman et al., 2018). Microscale properties during early healing were found to differ greatly from uninjured tendon. Predicted deficits in ECM stress transmission were found following fatigue loading and during healing in FE analysis (Freedman et al., 2018). In general, FE analysis allows for rapid testing that can facilitate experimental designs without the use of live animals, but it also may aid in patient-specific tendon rehabilitation protocols.

Various model systems are utilized to study tendon mechanobiology. As shown in Table 3, key features of each type of model system (ex vivo, 2D in vitro, 3D in vitro, and in vivo) are semi-quantitatively rated on their overall ability to replicate seven desired features when performing assessments of tendon mechanobiology. These key features include “biological relevance”, “biological controllability”, “biomechanical relevance”, “biomechanical controllability”, “accessibility”, “usability”, and “translatability”. Only mammalians are included in this assessment as they share similar immunological characteristics to humans. Key feature rating values were assigned to the model system types on a relative scale of one to four due to the four types of model systems that are being evaluated. For instance, a value of “one” for “biological relevance” for 2D in vitro systems was assigned as it least mimics the physiological environment as compared to 3D in vitro (second lowest = 2), ex vivo (3), and in vivo (highest = 4).

Each key feature has particular aspects that were considered in the rating assessment of the model systems though some of their ratings are similar. For “biological relevance”, the presence of other physiological cues was considered, such as immune and cell signaling crosstalk (2D lowest, in vivo highest). Because cells in 2D culture are isolated from other tissues and extended cell culture may alter cell phenotype, 2D in vitro systems rank the lowest for “biological relevance”. For “biological controllability”, the ability to control crosstalk factors was considered and the opposite rating trend was assigned (2D highest, in vivo lowest). Cultures in 2D have the greatest controllability as crosstalk can be isolated to specific cell types or factors using co-culturing techniques. Then cell-matrix interactions can be captured and manipulated with 3D systems, which is why 3D systems rank the second highest for “biological controllability”, followed by ex vivo (2) and in vivo (1) systems. Similarly, “biomechanical relevance” is related to the ability of the investigator to recapitulate the in vivo biomechanical environment (2D lowest, in vivo highest), while “biomechanical controllability” relates to the ability to control the mechanical loading environment (2D highest, in vivo lowest). Explanted tissues, or ex vivo models, maintain the native extracellular matrix, which affects mechanosensing pathways, and this contributes to its higher rating than both 2D and 3D in vitro systems for both biological and biomechanical relevance. In vivo models rank highest in biomechanical relevance because natural variation in loading parameters, gait, and load sharing with other tissues are not recapitulated in ex vivo systems.

Aspects for the other three key features are straightforward. The “accessibility” of a model system relates to the ability to obtain the desired tissue or animal sources for that system. Because primary cell lines are often the most accessible, 2D in vitro systems scored highest (4), followed by 3D in vitro (3), and then ex vivo and in vivo systems share a rating of 1 or 2 as access to whole tissue for ex vivo work or live animals for in vivo work may depend on the institution and regulations. Usability relates to the technical feasibility of developing the model system for dynamic loading experiments (2D highest, followed by 3D, in vivo, and then ex vivo systems). Development of the model system includes aspects such as accruing a bioreactor, troubleshooting, and experimental timelines. Again, these rankings might be institution- or resource-dependent. Last, translatability relates to how well these model systems translate to clinical work, where 2D in vitro has the lowest level of translation while in vivo systems are the most translatable.

In this assessment, each key feature was equally weighted to minimize subjective evaluation; however, this weighting could change based on the research question. For example, “biomechanical controllability” would likely be weighted more heavily than “biomechanical relevance” if an investigator was aiming to determine underlying cell signaling factors for varied loading conditions.