The “other” 15–40%: The Role of Non-collagenous Extracellular Matrix Proteins and Minor Collagens in Tendon

Abstract

Extracellular matrix (ECM) determines the physiological function of all tissues, including musculoskeletal tissues. In tendon, ECM provides overall tissue architecture, which is tailored to match the biomechanical requirements of their physiological function, that is, force transmission from muscle to bone. Tendon ECM also constitutes the microenvironment that allows tendon- resident cells to maintain their phenotype and that transmits biomechanical forces from the macro-level to the micro-level. The structure and function of adult tendons is largely determined by the hierarchical organization of collagen type I fibrils. However, non-collagenous ECM proteins such as small leucine-rich proteoglycans (SLRPs), ADAMTS proteases, and cross-linking enzymes play critical roles in collagen fibrillogenesis and guide the hierarchical bundling of collagen fibrils into tendon fascicles. Other non-collagenous ECM proteins such as the less abundant collagens, fibrillins, or elastin, contribute to tendon formation or determine some of their biomechanical properties. The interfascicular matrix or endotenon and the outer layer of tendons, the epi- and paratenon, includes collagens and non-collagenous ECM proteins, but their function is less well understood. The ECM proteins in the epi- and paratenon may provide the appropriate microenvironment to maintain the identity of distinct tendon cell populations that are thought to play a role during repair processes after injury. The aim of this review is to provide an overview of the role of non-collagenous ECM proteins and less abundant collagens in tendon development and homeostasis.

Tendon contains 55–70% water and 30–45% dry mass, which is mostly comprised of extracellular matrix (ECM) (Fig. 1).¹ 60-85% of the dry mass is attributed to collagens, predominantly collagen type Consequently, the “other” 15–40% of tendon dry mass is attributed to less abundant collagens and non-collagenous ECM proteins.^1,2 These ECM proteins are involved in several physiological processes, such as collagen fibril formation and tenocyte homeostasis that contribute to overall tendon function and health. Most studies on the function of these “other” ECM proteins in tendon focused on a subset of small leucine-rich proteoglycans (SLRPS) and their role in collagen fibril formation.⁵ However, recent knockout studies in mice made it clear that ECM proteins not directly involved in collagen fibrillogenesis may play important roles in tendon physiology. For example, such ECM proteins can provide unique microenvironments to maintain the identity of the different types of tendon-resident cells.^6,7 In addition, ECM molecules or their bioactive peptides, generated by proteases, can coordinate cellular responses to microdamage and macrodamage of tendon tissue. In this review, we will illustrate the complexity of tendon tissue by providing an overview of the function of the “other” 15–40% of tendon ECM proteins. Genetic disorders associated with these ECM proteins are listed in Supplementary Table S1.

Figure 1.

Tendon composition. Relative quantities of major tendon extracellular matrix (ECM) proteins. Numbers are based on Kjaer, Elliott, Connizzo et al., and Thorpe et al. and references therein.^1–4 n.d., not determined.

ECM PROTEINS INVOLVED IN COLLAGEN FIBRILLOGENESIS AND FIBRIL BUNDLING

The correct formation of collagen type I fibrils and their subsequent hierarchical bundling into larger functional units is paramount for the proper function of tendons.⁸ Collagen fibril formation requires the coordinated activity of multiple proteins in the tenocyte pericellular matrix (PCM) and the ECM. The size and arrangement of individual collagen fibrils within the collagen fiber is determined mainly by SLRPs and collagen fibrils are stabilized by enzymatic cross-linking.⁹

Collagen Propeptidases

The basic units of collagen fibrils are tropocollagen triple helices. The formation of these triple helices requires the proteolytic removal of the N- and the C-terminal propeptide from the procollagen molecule by secreted ECM proteases. The N-terminal collagen propeptide is removed by a disintegrin-like and metalloprotease with thrombospondin type I motives (ADAMTS)2, −3, or −14.¹⁰ ADAMTS2 is expressed at high levels in tendon and other tissues rich in collagen type I.^11–13 Mutations in ADAMTS2 cause dermatosparaxis which presents with extreme skin fragility and joint laxity in humans and cattle (Supplementary Table S1).¹⁴ Adamts2 knockout mice develop a similar skin phenotype, but show less pronounced joint laxity.¹⁵ In ADAMTS2-deficient skin, collagen is deposited in abnormal ribbon-like sheets due to the continued presence of the N-terminal propetide.¹⁶ Tendon was not analyzed in these studies but the abnormal collagen type I deposition in skin and joint laxity in humans and mice suggest compromised tendon function in the absence of ADAMTS2. Interestingly, in arthrochalasia Ehlers-Danlos syndrome, where the N-terminal collagen propeptide processing site itself is mutated, patients present with joint laxity, but the skin is less affected.^17,18 The C-terminal collagen propeptide is removed by BMP1/tolloid-like proteases and the Bmp1 knockout mice show abnormal collagen fibril formation.^19,20 In contrast to ADAMTS proteases, where proteolytic activity is negatively regulated by endogenous inhibitors such as tissue-inhibitors of matrix metalloproteases (TIMPs), the activity of BMP1/tolloid-like proteases is positively regulated by procollagen C-proteinase enhancers (PCPE, PCOLCE), fibronectin, or periostin.^21–24 Both, the N- and C-terminal procollagen processing proteases have additional substrates, such as decorin, biglycan, and lysyl oxidase (LOX) and thus may be involved in regulating additional steps during collagen type I fibril formation.^20,25,26 Recently, a third group of proteases, meprin α and meprin β, have been implicated in procollagen processing.²⁷

Minor Collagens

Besides collagen type I, several other types of fibrillar and non-fibrillar collagens, such as collagen type III, V, VI, XI, XII, and XIV, are present in tendon and are expressed predominantly during tendon development and in the postnatal growth phase.^28,29 While collagen type III, V, and XIV expression decreases during later stages of development in chick and/or mouse tendon, collagen type VI and XI continue to be expressed during the postnatal growth phase of tendon and beyond. The knockout mouse models for collagen type V, VI, XI, and XIV collectively suggest a role for these collagens in the assembly and maturation of collagen fibrils. For example, disruption of collagen type V resulted in irregular and disorganized collagen fibrils.³⁰ Knockout of collagen type VI resulted in an increased number of smaller diameter collagen fibrils, while knockout of collagen type XI and XIV resulted in an increase of larger diameter collagen fibrils.^31,32 Collagen type XII and XIV are bridging collagen type I fibrils with each other or with neighboring ECM molecules and thus are important in maintaining integrity of collagen type I fibers.^33–36 However, collagen type XII and XIV are differentially distributed in tendon and it was proposed that they might play independent roles in the formation of larger tendon compartments, such as tendon fascicles, by organizing or maintaining tissue architecture (collagen type XII) or by modulating collagen fibrillogenesis (collagen type XIV).^37–39 Together these studies suggest that these minor collagens play major roles in the formation and organization of collagen fibrils in tendon and that they might be involved in higher-order tendon tissue organization by contributing to collagen fiber bundling or tendon fascicle formation. A comprehensive network-level understanding how all these different collagens work together and how they collaborate with other ECM proteins in collagen fibrillogenesis and in establishing larger tendon compartments remains to be established.

Collagen Cross-Linking

Intramolecular and intermolecular cross-linking of collagen fibrils is an enzyme-mediated process that is required to register collagen molecules during the formation of the tropocollagen triple helix and to stabilize collagen fibrils in tendons. One family of enzymes which mediates the formation of covalent collagen crosslinks are the LOXs comprising LOX itself, and LOX-like 1–4.⁴⁰ LOX activation requires proteolytic processing by BMP1, which also processes the C-terminal collagen propeptide.²² In addition to BMP1, periostin appears to be required for LOX activity, since periostin-deficient tendon showed reduced LOX-mediated collagen cross-linking.²² LOX requires binding to the tropocollagen triple helix to initiate the cross-linking reaction.^41,42 During embryonic chick tendon development, LOX upregulation correlated with the maturation of the biomechanical properties of tendons.⁴³ When LOX activity was inhibited pharmacologically, tendons showed reduced tensile strength, despite maintaining their collagen organization and overall collagen content.^43,44 Fibulin-4, an ECM glycol protein, was identified as yet another modulator of LOX-mediated collagen cross-linking.⁴⁵ Mutations in the human fibulin-4 gene can cause severe vascular abnormalities, cutis laxa, arachnodactyly, and joint laxity (Supplementary Table S1).⁴⁶ A mouse model for fibulin-4 deficiency showed more abundant collagen fibrils in tendon cross-sections and these alterations were attributed to impaired LOX-mediated collagen cross-linking.⁴⁵ An independent fibulin-4 knockout mouse showed forelimb contractures and hypermobile joints.^47,48 During development, fibulin-4 deficiency disrupted collagen fibril packing and the number of collagen fibril bundles was reduced. These data suggest that LOX may need several accessory ECM proteins to modulate its activity or to serve as a scaffold. Together, the role of LOX underscores that tendon tissue formation not only requires the right amount and composition of ECM, but also requires appropriate covalent posttranslational modifications, such as crosslinks to elaborate fully its biomechanical properties. The role of other cross-linking enzymes in tendon biology, such as transglutaminases, should also be considered.⁴⁹

SLRPs

The SLRPs are ubiquitously expressed and play key roles in cell signaling, ECM assembly, and ECM turnover.^5,50 SLRPS consist of a core protein and covalently attached glycosaminoglycan (GAG) chains. SLRPs can interact with many ECM proteins, such as collagen type I, II, III, V, VI, and XI, and fibronectin and they can bind directly to growth factors, such as trans-forming growth factor β (TGFβ) or epidermal growth factor (EGF), modulating their activity.^51,52 In addition, SLRPs contribute to the overall stability of collagen fibrils by protecting them from proteolysis by collagenase.⁵³ The two pairs of SLRPs that are relevant for tendon biology, decorin/biglycan, and fibromodulin/lumican, are discussed below.^54–58

Decorin

Decorin was named for its ability to “decorate” the collagen fibril network in the ECM.^59,60 High-affinity binding to collagen fibrils is mediated by the decorin core protein and is independent of the GAG side chains.⁶¹ The decorin-binding domains on collagen fibrils are identical to the biglycan binding sites and both SLRPs can compete for binding to collagen fibrils. Decorin has multiple binding sites for collagen fibrils, suggesting that decorin can non-covalently cross-link several collagen fibrils during growth and maturation and as such contribute to the proper arrangement of collagen fibrils.⁶² In decorin-deficient tendons, the diameter of collagen type I fibrils was increased and showed a higher variability.^63,64 Despite the altered collagen type I fibril diameters, alterations of the bio-mechanical properties of decorin-deficient tendons seemed to depend on the specific tendon type under investigation and the age of the tissue. For example, knockout of decorin had no effect on the biomechanical properties of tendon fascicles from mouse tails, but decorin-deficient patellar tendon showed a higher modulus and increased stress relaxation when uniaxial tension was applied.⁶⁵ Biglycan may compensate in tail tendon fascicles, but in mouse patellar tendon biglycan was absent, similar to human patellar tendon, and thus may not be able to compensate for the loss of decorin.⁶⁶

Biglycan

Biglycan is highly expressed during postnatal growth in tendon, bone, and cartilage. Biglycan interacts with collagen type I, VI, and several other ECM proteins.^67–69 The absence of biglycan in mice resulted in abnormal collagen fibril formation in tail tendon, where collagen fibril diameters were smaller and the diameter distribution was less regular.⁷⁰ In biglycan-deficient patellar tendon however, no alteration in collagen fibril diameter was observed, despite alterations in the viscoelastic properties of the tendon.⁷¹ Biglycan expression was increased after tendon injury, in contrast to decorin, which was decreased.^72,73 Moreover, biglycan contributes to the tendon stem/progenitor cell (TSPC) niche, where biglycan induced tenogenic differentiation of TSPCs while simultaneously inhibiting the chondrogenic and osteogenic differentiation program.^6,74

Lumican

Lumican interacts with collagen type I and regulates collagen fibril assembly.⁷⁵ Addition of purified lumican to cells limits collagen fibril growth in vitro, resulting in thinner fibrils.^61,76 Lumican and fibromodulin share the same binding site on collagen fibrils, but the appearance of the collagen type I fibrils in the single knockout and double knockout mice are distinct, suggesting only partial compensation of lumican function by fibromodulin in collagen fibrillogenesis.^77–79 Lumican knockout mice show collagen fibrils with a very smooth surface, but an irregular diameter profile with a shift towards smaller diameters.⁷⁹ Biomechanically, the absence of lumican resulted in tendons with a significant reduction in tensile strength.⁷⁵

Fibromodulin

Fibromodulin is highly expressed in connective tissues including tendons and ligaments.⁸⁰ Fibromodulin affects the maturation steps of collagen fibrils through modulation of LOX activity to prevent premature cross-linking.^79,81,82 Therefore, fibromodulin-deficient tendons showed a higher proportion of larger diameter collagen fibrils.⁷⁹ The double knockout of fibromodulin and lumican resulted in joint laxity and gait abnormalities, suggesting a deficiency in tendon strength, specifically a decrease in tendon stiffness.^78,79 When fibromodulin was deleted in biglycan-deficient mice, collagen fibrils were smaller and heterotopic tendon ossification was observed.⁸³ The authors speculate that the decrease in tendon stiffness due to smaller collagen fibril diameter may be compensated by ossifying parts of the tendon in an effort to (re)establish the bio-mechanical properties of the tendon to match its functional requirements. More recently, the double knockout of the ADAMTS7 and ADAMTS12 proteases also resulted in heterotopic ossification in the Achilles and patellar tendon concomitant with a reduction in biglycan, fibromodulin, and decorin.⁸⁴ In these mice, the profile of collagen fibril diameters was skewed towards larger and more irregular fibrils. If ADAMTS7 and ADAMTS12 modulate the turnover of biglycan and fibromodulin in tendon ECM or if they work as proteoglycans independent of their protease activity remains to be established. Finally, biglycan and fibromodulin were identified as part of the TSPC niche. TSPCs isolated from biglycan and fibromodulin double knockout mice proliferated faster and formed more colonies.⁶ However, gene expression of tendon markers such as scleraxis or collagen type I was reduced, suggesting that these TSPCs proliferated at the expense of their tenocytic differentiation potential or that they maintained their stemness.

Tenascin-C and Periostin

Tenascin-C is expressed in mechanically loaded tissues, such as tendon, but is virtually absent in non-loaded tissues.^85,86 However, a tendon phenotype for the tenascin-C knockout mouse was not reported.⁸⁷ Single nucleotide polymorphisms (SNPs) in tenascin-C are associated with rotator cuff degeneration but a mechanistic role for tenascin-C in tendon development or homeostasis remains to be established.^88,89 Periostin is expressed in collagen-rich tissues, including tendon, and was shown to bind directly to collagen type I.⁹⁰ In skin from periostin knockout mice, collagen fibrils had a smaller diameter and the skin had decreased stiffness. Collagen cross-linking was reduced in the absence of periostin, suggesting that the major role of periostin is to promote LOX activity in collaboration with BMP1.²² Periostin interacts with tenascin-C and is required for the incorporation of tenascin-C in the ECM.⁹¹ The tendon phenotype was not analyzed in periostin knockout mice. In a cell culture study using tenogenic differentiation of mesenchymal progenitor cells, it was shown that periostin is upregulated and that overexpression of periostin in human mesenchymal stem cells resulted in ectopic tendon formation when these modified cells were injected into nude mice.⁹² This suggests that periostin might play a role in tenocyte differentiation and may be relevant to provide the tenocyte microenvironment together with tenascin-C rather than contributing directly to the bio-mechanical properties of tendon.

The Sliding Proteoglycan Filament Model

On the basis of electron microscopy imaging, computer modeling, and biophysical and biochemical studies, a role for SLRPs and other GAGs and proteoglycans was proposed as interfibrillar collagen bridges to convert compression into tensile strain and thus allow for the recovery of tendon shape after physiological stretching.⁹³ It was further suggested that GAG bridges between collagen fibrils would facilitate sliding of collagen fibrils against each other and thus allow for elastic stretching of tendon tissue beyond strains of >10%, at which individual collagen fibrils are irreversibly damaged. GAG bridges between collagen fibrils were observed using electron microscopy.⁹⁴ In addition, the X-ray diffraction patterns of tendon tissue under tension are sharper compared with non-stretched tendon, suggesting higher supramolecular order.⁹⁵ Computational modeling identified GAG content and the area fraction of the collagen fibrils as strong predictors of mechanical behavior, but not collagen fibril diameter itself.⁹⁶ However, enzymatic or genetic GAG depletion in tendon tissue resulted in surprisingly small changes in mechanical tensile behavior.^97–100 Therefore, other biomechanical roles for the GAG-containing proteins in tendon have been proposed, such as the distribution of transverse loads, and contribution to ultimate failure strength or creep/fatigue recovery. In support of biomechanical roles for GAGs and proteoglycans in tendon other than stretch recovery, more recent studies showed, that tendon from decorin knockout mice had decreased tendon viscoelasticity without affecting the elastic properties of tendon.^101,102 The exact role for each of the SLRPs and other GAGs and glycoproteins and their cooperative interactions in tendon development and tendon function needs to be further investigated.

THE PCM OF TENDON-RESIDENT CELLS

The PCM represents the cell surface-associated ECM in the immediate vicinity of the cell membrane. The PCM serves as the interface between tissue resident cells and the more distal territorial and interterritorial ECM networks.¹⁰³ The functional importance of the PCM is well established in cartilage.¹⁰⁴ The properties of the PCM can be modulated by the activity of cell-surface-associated proteases or by altering the gene expression of PCM components in response to physiological and pathological stimuli.^105,106 Recently, proteomics studies identified more than 100 proteins at the cell surface and the PCM of tenocytes, illuminating the complexity of this ECM compartment.^103,107 Some of these PCM components, for which functions have been established, are discussed below.

Tenomodulin

Tenomodulin localizes predominantly to the tenocyte PCM around the more mature, elongated tenocytes, but was less prominent in the PCM of immature, more roundish tenocytes in developing chick tendon.^108,109 Tenomodulin is furin-processed at the cell surface and the C-terminal domain, which mediates the physiological functions of tenomodulin, is released into the PCM.^110,111 Tendons of tenomodulin knockout mice showed larger and more variable collagen fibril diameters.¹¹¹ In addition, tenocyte proliferation was decreased in the absence of tenomodulin, resulting in a lower cell density in tendon. Tenomodulin-deficient TSPCs showed reduced self-renewal and cell proliferation capacity, which could be rescued by over-expressing the bioactive C-terminal domain of tenomodulin.¹¹² No binding partners for tenomodulin are known and the molecular mechanisms by which tenomodulin contributes to tendon homeostasis remain to be discovered.

Versican and Aggrecan

Versican and aggrecan are large proteoglycans which are present in many tissues and which are anchored in the PCM via their interactions with hyaluronan and the cell surface receptor CD44.¹¹³ The versican and aggrecan PCM facilitates cell shape changes, cell proliferation, and cell differentiation processes.¹¹⁴ The dimension of the versican and aggrecan PCM can be modulated by versicanases or aggrecanases, such as ADAMTS1, −4, or −5.¹¹⁵ In tendon, versican localizes primarily to the tendon midsubstance whereas aggrecan is found closer to the fibrocartilage of the enthesis and in tendon regions that are adjacent to bone, suggesting a role in buffering biomechanical compression.^116–120 Versican accumulation was associated with increased ECM volume in patellar tendinosis.¹²¹ By knocking out ADAMTS5, the amount of aggrecan, but not versican, was increased in the tenocyte PCM and tendons showed a more heterogeneous distribution of collagen fibril diameters.¹²²

Thrombospondins

The thrombospondin (TSP) protein family of matricellular proteins comprises TSP1–4 and cartilage oligomeric matrix protein (COMP, TSP5).¹²³ As matricellular proteins, TSPs regulate diverse cellular processes such as inflammation, angiogenesis, cell migration, and cell proliferation through participation in cell–ECM and ECM–ECM interactions.¹²⁴ The TSP2 knockout mouse has lax tail tendons and the tail can be tied into a knot.¹²⁵ On an ultrastructural level, TSP2-deficient tendons showed a higher percentage of larger collagen fibrils.¹²⁶ In addition, tendon fascicle formation was abnormal in the absence of TSP2. TSP4 was found in the PCM of tenocytes and to a lesser extend between collagen fibrils. In triceps tendon, TSP3 and TSP4 showed distinct localization and COMP co-localized with TSP4.¹²⁷ In TSP4-deficient patellar tendons, the distribution of collagen fibril diameters was more irregular with larger diameter collagen fibrils being observed.¹²⁷ COMP is about tenfold more abundant in weight bearing tendon compared with positional tendons suggesting a role in elaborating the biomechanical properties of these tendons.^128,129 A role for load in regulating COMP expression is further supported by its upregulation after birth when load is applied to tendons. For example, COMP is almost absent in horse tendons at birth, but can constitute up to 3% of dry weight in the superficial digital flexor tendon by the age of 2 years.¹²⁹ Mechanistically, COMP binds to the gap region of fibrillar collagens and thus reinforces collagen fibrils.¹³⁰ In humans, mutations in COMP cause pseudoachondrodysplasia or multiple epiphyseal dysplasia, which present with dwarfism and lax joints, suggesting a weakening of tendon and ligaments.¹²³ In a mouse model for pseudoachondrodysplasia, joint laxity was observed concomitant with a reduced size of the patellar tendon.¹³¹ In a knock-in of a human pseudoachondrodysplasia mutation, collagen fibrils with a larger diameter were more abundant in the presence of mutant COMP in the Achilles tendon in these mice.¹³² On a tissue level, the Achilles tendon were thinner without a reduction of the collagen fiber occupied area, the ECM between the collagen fibrils, or proteoglycan content. In biomechanical testing, COMP mutant tendon could sustain more stretch and stress.¹³² However, when undergoing cyclic mechanical testing, COMP mutant tendons became more lax over time when compared to wild type tendon.

Fibrillins and Elastin

Fibrillins form microfibrils in the ECM that serve as a scaffold for several ECM proteins, including tropoelastin, the precursor of elastin.¹³³ Depending on the location, mutations in fibrillin-1 (FBN1) can cause Marfan syndrome or acromelic dysplasias, presenting with either joint laxity or joint contractures, respectively (Supplementary Table S1).^134–136 Mutations in fibrillin-2 (FBN2) are associated with congenital contractural arachnodactyly, that is, joint contractures.^137,138 Elastin plays a prominent role in sustaining the interaction between tenocytes and ECM in tendon.¹³⁹ Elastin is found in the PCM of tenocytes and can account for up to 2% of the total dry weight of tendon (Fig. 1).^140,141 Some of the stretch and recoil properties of tendon are attributed to elastin.¹⁴² Elastic fibers are also found in the interfascicular matrix (IFM), specifically in energy storing tendons and here, elastic fibers become disorganized during aging.^143,144. Elastin haploinsufficiency in mice resulted in larger collagen fibrils in the Achilles tendon but not in the supraspinatus tendon.^145,146 Surprisingly, elastin-deficient tendons showed no alterations in biomechanical properties, with the exception of a slight increase in linear stiffness.¹⁴⁵ FBN1 and FBN2 were found in bovine flexor tendon and in canine flexor digitorum profundus tendon where FBN2 was localized in the PCM of the central tenocytes, while FBN1 localized to the PCM of the peripheral tenocyte layers.^143,147 The PCM around the tenocytes is rich in FBN2, which co-localized with versican, where it forms a sturdy ECM that allows for the proteolytic and mechanical isolation of intact tenocyte arrays.¹³⁹ The FBN2 microfibrils originated from tenocytes and traversed through the PCM finally emerging out from the tendon cell arrays.¹³⁹ So far, the tendon phenotype of the Fbn1 knockout mice has not been described. Fbn2 knockout mice showed temporary forelimb contractures and decreased collagen fibril cross-linking in the digital flexor tendons.^148,149 We recently described a role for ADAMTSL2 in modulating the composition of the PCM around tenocytes in the Achilles tendon.¹⁵⁰ Specifically, we found increased FBN1 deposition in the absence of ADAMTSL2. Remarkably, tenocyte organization in linear arrays was distorted and Achilles tendons were shorter and wider. Mutations in ADAMTSL2 cause geleophysic dysplasia, which can also be caused by mutations in FBN1.^151,152 Patients with geleophysic dysplasia present with joint contractures and tiptoe walking, and other clinical findings, suggesting dysfunctional tendons.

Perlecan

The heparan sulfate proteoglycan perlecan is yet another PCM component for which a tendon and collagen fibril phenotype was described.¹⁵³ In a mouse model that lacks the perlecan heparan sulfate chains, smaller collagen fibrils were observed in tail tendon, similar to the phenotype observed in collagen type VI deficient tendon.^153,154 The altered collagen fibril structure resulted in increased tensile strength. Perlecan can form a network together with fibrillin microfibrils, elastic fibers, and collagen type VI in the PCM/ECM.^155,156 In addition, perlecan can interact with several ECM proteins and growth factors and can anchor fibrillin microfibrils into the basement membrane in skin.¹⁵⁷ As such, perlecan may assist in organizing the PCM around tenocytes and connect the tenocyte PCM to the more distal ECM networks.

IFM AND PARATENON

The hierarchical organization of collagen fibrils into fibers and tendon fascicles and the bundling of tendon fascicles into functional tendon gives rise to several distinct connective tissue compartments within the fibers and fascicles, in between tendon fascicles (endotenon or IFM), and surrounding bundles of tendon fascicle (epitenon/paratenon) (Fig. 2).¹⁵⁸ In a proteomics study, 130 proteins were identified in the IFM and 92 in the fascicular matrix, several being unique to either one of these ECM compartments.³ Among the SLRPs, immunostaining for lumican, biglycan, and lubricin was more prominent in the IFM in different types of horse tendon, whereas fibromodulin was enriched in the fascicular matrix.¹⁵⁹ Decorin was equally distributed between the IFM and the fascicular matrix. In addition, elastic fibers were prominent in the IFM in the superficial digital flexor tendon in horse, the equivalent of the Achilles tendon, suggesting a role for elastic fibers in mediating recoil of energy storing tendons.¹⁴³ Lubricin is a glycoprotein that was identified in the IFM where it may facilitate the gliding between tendon fascicles due to its anti-adhesive properties.^160–162 This role was supported in pullout tests of lubricin knockout mouse-tail tendon fascicles, which showed increased gliding resistance.¹⁶³ In mouse and dog models of tendon injury, a thickening of the epitenon was observed due to cell proliferation and a similar response was seen in an injury model of horse tendon.^164,165 The proliferating cells were characterized by stronger expression of fibronectin, which is an ECM master organizer and one of the earliest ECM proteins deposited during wound healing.^166,167 In rat, the epitenon/paratenon contained tenascin-C, collagen type I, and fibromodulin.¹⁶⁴ In a patellar injury model in rat, paratenon-derived cells bridged the initial gap space and initiated collagen fibril formation in a circumferential orientation.¹⁶⁴ One of the better characterized of the less abundant collagens in the IFM is collagen type III. Collagen type III content is positively correlated with tissues requiring greater compliance and thus is higher in energy storing tendons, compared with positional tendons.³ Collagen type III is more abundant in the IFM compared with the fascicular matrix and is specifically upregulated in the IFM in injured equine tendon as part of the repair response.¹⁶⁸ Collagen type III binds to several cell surface receptors, such as the discoidin domain receptors (DDR1, DDR2), PCOLCE, or platelet-derived growth factor (PDGF).^169–171 Taken together, the IFM and the tissues surrounding the tendon proper are important compartments that contribute to the biomechanical properties of tendons and support cell-mediated tendon healing processes. Since overall much less is known about the formation and function of the IFM compared with the fascicular matrix and collagen fibers, elucidating the formation and function of the IFM and its adaption to load will be an important area for future research.

Figure 2.

Hierarchical organization of tendon and its extracellular matrix (ECM) compartments. Proteins that play a major role in each of the ECM/pericellular matrix (PCM) compartments in tendon are listed. However, most of the proteins are found in several ECM/PCM compartments and their expression and localization is dynamically regulated during development, growth, and homeostasis and in response to chronic or acute injury. Images on the bottom show (from left to right): Hematoxylin/eosin-stained longitudinal section through mouse Achilles tendon; PCM of murine tenocytes stained for fibronectin (FN, red); electron microscopy image of cross-section through mouse Achilles tendon showing a tenocyte (T) and collagen fibrils (CF).

NON-COLLAGENOUS ECM PROTEINS IN TENDON TISSUE ENGINEERING

From the diverse functions of the non-collagenous ECM proteins and the minor collagens, it becomes clear that tendon represents a very complex and dynamic tissue, especially during tendon development, growth, und in repair processes upon injury (Fig. 3). To engineer tendon or augment tendon healing after injury, it is likely that some aspects of normal tendon development and growth need to be recapitulated to generate fully functional tendon tissue. Can some of the knowledge on the individual proteins or protein networks in the tendon ECM be harnessed for novel tendon tissue engineering approaches? One such approach utilizes ECM from decellularized tendon tissue as a scaffold for cells.¹⁷⁴ The idea is that decellularized tendon ECM has already the ideal biomechanical properties and contains the appropriate biochemical cues to successfully home tenocytes or allow for the differentiation of TSPCs.¹⁷⁵ The challenges with this approach are: (i) To identify a source for tendon tissue, taking into account availability, interspecies compatibility, and immunogenicity; (ii) To develop an efficient decellularization protocol that preserves the composition and structure of native tendon ECM as much as possible; (iii) To identify sources for TSPCs to be reseeded on the scaffold under the right culture conditions to allow for differentiation into mature tenocytes. Despite these challenges, decellularized tendon tissue allografts have been evaluated in small and large animal models and showed some promise in augmenting tendon repair after injury.^176–179 In addition, overexpression or exogenous addition of individual ECM proteins in combination with a diverse set of growth factors have been shown to enhance tenogenic differentiation of stem cells and augment repair processes in injury models, respectively, and as such could represent an integral part of a comprehensive tendon tissue engineering and repair strategy.^74,180–182 An understanding of the dynamic tendon ECM and tenocyte PCM will provide useful guidance to design future therapeutic approaches to augment tendon healing and to invigorate tendon tissue engineering.

Figure 3.

Complexity of the extracellular matrix (ECM)/pericellular matrix (PCM) protein network in tendon tissue. The network of biochemical or proposed functional interactions between ECM/PCM proteins was generated using the String database (https://string-db.org/) which is a curated database that draws information from multiple experimental databases and by text mining.¹⁷² The network was visualized using Cytoscape (https://cytoscape.org/).¹⁷³

CONCLUSION AND OUTLOOK

Tendons are known to withstand substantial mechanical loads, yet various daily activities can cause chronic microdamage or acute injury to tendons.^183,184 The healing response, especially of adult tendon, is insufficient to regenerate the tissue in situ as seen by scar formation at the site of injury that can result in chronic functional tendon deficiency.⁷ Understanding the contribution of tendon-resident cells, their PCM, and the ECM that they construct during tendon development and homeostasis has important implications for tendon regeneration. While much knowledge has been gained from knockout studies of individual genes, providing a comprehensive understanding of the temporal and spatial organization of the PCM and ECM network(s) and of the identity and lineage of the various cell types in tendon is challenging. For example, individual knockout of many tendon ECM proteins resulted in the alteration of collagen fibril diameters either skewing the size distribution towards larger, smaller, or more irregular diameters with very little compensation between related ECM proteins. How do all these proteins work together? What are the (bio-mechanical) parameters that determine the correct size of collagen fibrils? Could the absence of some of these ECM components compromise the biomechanical properties of tendon independent of collagen fibril alterations and could the observed alteration in collagen fibril diameters represent a generic physiological response to (re)adjust the biomechanical properties of tendon? Temporally controlled conditional deletions of these ECM components would address at which stage these proteins are required for tendon homeostasis and such studies already revealed distinct postnatal functions for biglycan and decorin.¹⁰¹ Lineage tracing experiments over the past several years made it abundantly clear that tendon tissue is host to more than one cell type.^7,164,185 The questions that arise here are: (i) What components of the tendon ECM do individual cell types synthesize? (ii) Does each of these cell types require its own specific microenvironment, that is, PCM/ECM, to maintain their phenotype; (iii) What is the structure of these microenvironments and how are they created; (iv) How do these microenvironments change during natural tendon aging and in chronic tendinopathies? On a macroscopic level, very little mechanistic information is available on postnatal tendon growth and the formation of larger tendon compartments such as tendon fascicles, or on how tendon fascicles are bundled to form the tendon proper.¹⁸⁶ It is conceivable that the ECM plays a major role as a structural glue or in the initial registration of adjacent tendon fibers and fascicles. With these questions in mind, we are looking into a bright future in tendon research specifically trying to understand the complexity and diversity of tendon tissues on a molecular, cellular, and tendon-specific level, reaching far beyond collagen type I fibrils and “the” tenocyte.