Cellular therapy in bone-tendon interface regeneration

Abstract

The intrasynovial bone-tendon interface is a gradual transition from soft tissue to bone, with two intervening zones of uncalcified and calcified fibrocartilage. Following injury, the native anatomy is not restored, resulting in inferior mechanical properties and an increased risk of re-injury. Recent in vivo studies provide evidence of improved healing when surgical repair of the bone-tendon interface is augmented with cells capable of undergoing chondrogenesis. In particular, cellular therapy in bone-tendon healing can promote fibrocartilage formation and associated improvements in mechanical properties. Despite these promising results in animal models, cellular therapy in human patients remains largely unexplored. This review highlights the development and structure-function relationship of normal bone-tendon insertions. The natural healing response to injury is discussed, with subsequent review of recent research on cellular approaches for improved healing. Finally, opportunities for translating in vivo findings into clinical practice are identified.

Keywords: :

• bone-tendon healing
• enthesis
• fibrocartilage
• mesenchymal stem cells
• insertion

Introduction

Now in the third decade following its introduction by YC Fung and the subsequent publication of the seminal paper of Langer and Vacanti,¹ the field of tissue engineering still offers tremendous promise in allaying or eliminating multiple diseases. Perhaps no organ system is closer to broad application of tissue engineering strategies than that of the musculoskeletal system,² due in part to the relatively well-characterized structure and function of the these tissues, coupled with the prevalence of musculoskeletal disease. Annually, there are 32 million musculoskeletal injuries in the United States, of which 45% involve tendons or ligaments (T/L).³ Multiple researchers have already demonstrated the ability to engineer T/L that possess similar structure, function, and cellular behavior as normal tissue,⁴ but the in vivo integration of these engineered constructs with the surrounding normal tissues is a persistent challenge. In particular, the apposition of T/L against their bone insertion site fails to restore the complex anatomy found at the native bone-ligament or bone-tendon interface, commonly known as the enthesis. As a result, the repaired tissue possesses inferior mechanical and biochemical properties, increasing the risk of re-injury and continued disability. Therefore, restoration of the complex structure and function of the enthesis is the next step in bringing the promise of T/L tissue engineering to fruition. While several approaches have been explored, including the application of growth factors, scaffolds, bone cements, and inhibitors of matrix metalloproteinases,⁵ recapitulation of the fibrocartilaginous layer interposed between bone and tendon in the native enthesis will require guided chondrogenesis of a cell source, whether exogenously-applied or endogenously-derived.

Given the central role of cells in enthesis repair, this review examines the in vivo evidence to support cellular therapy in bone-tendon interface regeneration. We first describe the complex anatomical and biomechanical characteristics of the bone-tendon interface, and then present the embryonic development and postnatal maturation of the enthesis, with special emphasis on the cellular origin of interfacial tissues and the biophysical cues that coordinate their differentiation. We next discuss the pathophysiology of bone-tendon injury and the subsequent intrinsic healing cascade, and how its insufficiency in restoring native structure and function has led to the development of current tissue engineering strategies, specifically cellular therapies. In closing, the regulatory and scientific hurdles impeding the adoption of cellular therapy in the clinical care of bone-tendon injuries are presented and potential solutions are offered.

Structure-Function Relationship of the Bone-Tendon Interface

While tendons and ligaments connect muscle-to-bone and bone-to-bone, respectively, they share a similar hierarchical structure of collagen fibers and resulting mechanical properties. Furthermore, the interface between T/L and bone are otherwise indistinguishable, although a large degree of variability exists depending on the location of the attachment site. Therefore, tendons and ligaments will be considered interchangeable for the purpose of this review. In general, T/L that attach to the bone diaphysis display fibers inserting directly into cortical bone or periosteum, and are known as fibrous entheses.⁶ Conversely, T/L that insert into bone epiphyses, often within or around a synovial joint, possess a gradual transition from connective tissue to bone with two intervening layers of fibrocartilage.⁷ Interfaces that possess this four-zone structure are known as fibrocartilaginous entheses.⁶ As the vast majority of injuries of the bone-tendon interface involve fibrocartilaginous entheses, with surgical repair failing to restore this native structure, no further discussion on the anatomy and healing potential of fibrous entheses will be performed. For a detailed description on the latter type of bone-tendon insertion, please refer to references 7 and 8.

In the typical fibrocartilaginous enthesis, there are four discernible zones—dense connective tissue (T/L), uncalcified fibrocartilage (UF), calcified fibrocartilage (CF), and bone (Fig. 1).⁶ The first zone (T/L) is similar to the midsubstance of these tissues in terms of architecture and extracellular matrix protein composition, with aligned collagen type I fibers and small amounts of the proteoglycans decorin and biglycan.^8,9 The UF zone possesses columns of oval chondrocytes embedded in a matrix rich in collagen types II and III, with small amounts of collagen types I, IX, and X, and proteoglycans aggrecan and decorin.^7,8 The CF zone consists predominantly of collagen type II, with significant amounts of collagen type X and aggrecan.⁹ Collagen type X is produced by the hypertrophic chondrocytes found in this third zone (Fig. 1B). Both fibrocartilaginous zones stain positive for glycosaminoglycans (Fig. 1A). Lastly, the bony region displays the cellular and biochemical composition of typical bone, with a predominance of type I collagen and a high mineral content.⁸

Figure 1. Histological features of the fibrocartilaginous bone-tendon interface stained with (A) Safranin-O, (B) H&E, and viewed with (C) polarized light microscopy. Note the four zones of the native enthesis. Magnification 20×. B, bone; CFC, calcified fibrocartilage; UFC, uncalcified fibrocartilage; T, tendon. Reproduced with permission from reference 122.

Much like the osteochondral interface of articular surfaces, a basophilic tidemark seen on histological sections separates the UF from CF and was thought to represent a mineralization front (Fig. 1B and C). In support, Moffat et al.¹⁰ detected mineral only within the CF and bone regions. Conversely, the application of Raman spectroscopy revealed a mineral-to-collagen ratio that increased linearly from tendon to bone, rather than abruptly, with a concomitant increase in crystalline order when moving from soft to hard tissue.¹¹ Therefore, it seems increasingly unlikely that four zones of distinct biomechanical, structural, and composition properties exist. Rather, there is a gradual transition from the compliant soft tissue (tensile modulus ~200 MPa) to rigid bone (~20 GPa). This serves to minimize stress concentrations and allow the efficient transmission of muscular forces to the skeleton.⁸

In further support, Thomopoulos et al.¹² developed a two-dimensional finite element model of an insertion based upon collagen fiber orientation, as determined by polarized light microscopy. In comparing the stress concentrations in the idealized model to those of three comparison models, it was concluded that the microstructure of the insertion serves to (1) reduce stress concentrations and material mass, and (2) shield the insertion’s outward splay from higher stresses. In subsequent work, it was shown that an increasing mineral concentration coupled with decreasing collagen fiber orientation as one moves from the tendon toward bone, provides a functionally graded material composition that can explain the variation of stiffness over the length of the insertion.¹³ Likewise, Moffat et al.¹⁰ previously showed that the calcified fibrocartilage region possessed a higher Young’s modulus than the adjacent uncalcified fibrocartilage. Taken together, the transitional composition of the enthesis microstructure creates a gradient of increasing material stiffness, thereby allowing efficient joint mobilization while minimizing the risk of injury. Nevertheless, damage to the bone-tendon interface does occur, and the natural healing response fails to restore the complex anatomy of the native insertion. To recapitulate the embryonic events that give rise to the normal microstructure, thereby promoting bone-tendon regeneration, a comprehensive understanding of enthesis development is needed.

Enthesis Development

Embryonic origin of the bone-tendon interface

An understanding of both the cellular origin of interfacial tissues and the molecular mechanisms that mediate their development is in the nascent stage, due in large part to the only recent discovery of a selective marker of tendon and ligament cell-lineage. Scleraxis (Scx), a basic helix-loop-helix transcription factor, was first recognized at day 9.5 post coitum (p.c.) in the sclerotome of the somites and the mesenchymal cells in the body wall and limb buds of a mouse embryo.¹⁴ Subsequent expression within mesenchymal precursors of the axial and appendicular skeleton in advance of chondrogenesis suggested a role in the developing skeleton. However, at later time-points in development, Scx expression is a highly specific marker for all the connective tissues that mediate attachment of muscle to bone.¹⁵ Brent et al.¹⁶ identified a Scx-positive region interposed between the sclerotome and myotome, two previously-formed somitic compartments that develop into the axial skeletal system and body wall muscles, respectively. Induction and differentiation of this interposing somitic compartment, the syndetome, is dependent on signals from the adjoining sclerotome and myotome, although the syndetomal precursors themselves are derived from the sclerotome.¹⁶ Conversely, progenitor cells that ultimately constitute tendons of the limb develop independently of muscle, although subsequent tendon maturation requires reciprocal signaling between these two mesenchymal progenitors.¹⁷

As tendon progenitors of both the axial and appendicular musculoskeletal system upregulate Scx early in development and maintain its expression following maturation into tenocytes, Scx is widely considered to be the most selective marker of tendon and ligament differentiation. However, despite the identification of Scx expression in cells comprising T/L, its complete functional role remains unknown. Scx heterodimerizes with bHLH protein E47 to selectively bind cis-acting elements that control collagen type I expression,¹⁸ suggesting an important role in tenocyte-mediated fibrillogenesis.¹⁹ However, Scx expression does not appear to be absolutely necessary for tendon formation.²⁰ Indeed, the molecular mechanisms underlying tenogenic and ligamentogenic differentiation are complex and multifactorial, surpassing the scope of this review. For a more detailed account of tendon development, the reader is referred to ref. 18. Nevertheless, the association of Scx expression with tendon development has proven important in elucidating the origin of interface tissues.

More specifically, Asou et al.²¹ found that Scx expression was closely associated with, but distinct from, formation of the skeletal primordia, which expressed the known chondrogenic transcription factor Sox9. However, this study relied on whole mount in situ hybridization to determine the spatiotemporal pattern of mRNA transcript expression, thereby limiting the resolution to probe the heterotypic interactions of the primordial tendon-skeletal interface. Armed with a genetic mouse model capable of inducing a Sox9 expression reporter during development, Soeda et al.²² demonstrated that at least some of the cells of tendons and ligaments originate from Sox9-expressing mesenchymal cells of the cartilage primordial, suggesting their contribution to the formation of tendon. Most recently, it was shown that a Scx⁺/Sox9⁺ progenitor pool contributes to the establishment of the interface between cartilage and T/L in the developing murine embryo.²³ Of particular interest, the gradual transition of Scx and Sox9 expression resembles the functional transition seen in mature entheses (Fig. 2). Namely, Scx^-/Sox9⁺ mesenchymal cells were localized to the developing cartilaginous primordium, which subsequently develops into the bony skeleton by endochondral ossification. Scx⁺/Sox9^- cells were found in the developing tendon midsubstance, while the closer the tendon was to the cartilaginous primordium, the greater the number of Scx⁺/Sox9⁺ progenitors. Furthermore, conditional inactivation of Sox9 expression in Scx⁺/Sox9⁺ progenitors results in defective formation of cartilage-T/L attachment sites²³ and, more specifically, the loss of the bony eminence normally seen at insertion sites.²⁴ Therefore, the Scx⁺/Sox9⁺ progenitor pool is a multipotent mesenchymal population capable of differentiating into the tenocytes, ligamentocytes, and chondrocytes that ultimately establish the enthesis.

Figure 2. Embryonic development of the bone-tendon interface. (A) The degree of Scx and Sox9 expression determines cellular phenotype. Scx⁻/Sox9⁺ chondroprogenitors (CP) become chondrocytes of the skeletal anlagen while Scx⁺/Sox9⁻ tenoprogenitors (TP) become tenocytes of the tendon midsubstance. Varying levels of Scx and Sox9 expression are seen in teno-/ligamento-/chondro-progenitors (TLCP), which give rise to cells of the bone-tendon interface. (B) Scx⁺/Sox9⁺ progenitors give rise to the primordial chondro-tendinous/ligamentous junction (CTJ/CLJ), which forms the osteo-tendinous/ligamentous junction (OTJ/OLJ) following birth. Reproduced with permission from reference 23.

Signaling pathways active in enthesis formation

As the enthesis consists of multiple cellular phenotypes embedded in a complex extracellular matrix spanning a length less than 1 mm, the molecular mechanisms underlying its formation and maintenance are equally intricate. Much like the growth plate of bone,²⁵ the enthesis expresses Indian hedgehog (Ihh), parathyroid hormone-related peptide receptor (PTHrPR), collagen type II α1 (Col2α1), and collagen type X α1 (Col10α1).⁹ In the growth plate, an elegant paracrine loop exists where Ihh secretion from hypertrophic chondrocytes upregulates PTHrP expression in a zone of proliferating chondrocytes, thereby inhibiting their hypertrophic differentiation and allowing continued longitudinal bone growth.²⁵ The presence of such a loop in the developing enthesis has not been established, but as discussed in Thomopoulos et al.,⁸ there is growing evidence to support that a similar mechanism exists to guide bone ridge formation and subsequent mineralization in the fibrocartilaginous enthesis.²⁶ In addition, bone morphogenetic protein-4 (BMP-4), a member of the Transforming Growth Factor-β (TGF-β) superfamily, also plays a role in bone ridge patterning. Specifically, Scx promotes secretion of BMP-4 from nascent tenocytes, which in turn initiates formation of the bony ridge.²⁷ Inhibition of BMP-4 expression in Scx-positive cells leads to the failed formation of many bony ridges in the embryo, suggesting the pivotal role of this signaling pathway in early enthesis formation.

In addition to Ihh and its effects on PTHrP, TGF-β plays a central role in orchestrating tendon and cartilage differentiation at the developing interface. When examining the developing rat supraspinatus enthesis, Galatz et al.²⁸ could distinguish the rotator cuff as early as 13.5 d post coitum (d.p.c.). Although four insertion site zones (zone 1, tendon; zone 2, UF; zone 3, CF; zone 4, bone) were not distinct until 7 d after birth, TGF-β3 expression in zone 1 was discernible 13.5 d.p.c through 15.5 d.p.c, with TGF-β1 expression becoming upregulated at 15.5 d.p.c before diminishing after 18.5 d.p.c. In a related study, Lorda-Diez et al.²⁹ showed a differential effect of TGF-β signaling on chondrogenesis and fibrogenesis when comparing an in vitro to in vivo experimental model. Namely, TGF-β supplementation of a high density micromass culture suppressed chondrogenic markers Sox9 and Aggrecan, while upregulating fibrogenic markers Scx and Tenomodulin. Conversely, TGF-β upregulated Sox9 in vivo, thereby inducing ectopic chondrogenesis. The discrepancy between model systems may be explained by transcriptional repressors of TGF-β, including TGF-interacting factor Tgif1 and SKI-like oncogene SnoN. Regions of the developing digit expressing these transcription repressors went on to become soft connective tissues such as ligament, joint capsule, or tendon, while TGF-β regions devoid of SnoN and Tgif1 adopted a chondrogenic lineage. Furthermore, limb mesodermal cells cultured in a TGF-β-supplemented medium showed downregulation of Sox9 and Aggrecan expression when transfected with Tgif1.²⁹

Just as this orchestrated spatiotemporal expression pattern of signaling molecules results in regional differences in molecular markers that are indicative of differentiating cellular phenotypes, the expression of extracellular matrix proteins also varies by location. While the insertion site of the developing rat supraspinatus does not display four zones until 7 d following birth, the primordial tendon region stains positive for collagen type I beginning at 13.5 d.p.c.²⁸ Likewise collagen type II is expressed in the cartilaginous anlagen from 13.5 d.p.c. until postnatal day 21, when endochondral ossification replaces cartilage with bone. A fibrocartilaginous region interposed between bone and tendon is distinguishable as early as 7 d following birth, with collagen type II expression found in both zones 2 and 3. Collagen type X, indicative of chondrocyte hypertrophy, is first expressed in zone 3 at postnatal day 14, with continued expression into adulthood.²⁸ In the developing mouse patella tendon, collagen type I, the proteoglycan fibromodulin, and the glycoprotein tenomodulin (TNMD), were expressed evenly along the tendon at embryonic day 17.5 (E17.5), but TNMD expression was significantly reduced at the insertion site starting at postnatal day 7.³⁰ Conversely, glycoprotein tenascin C expression was found only at the insertion at E17.5, but increased expression in the midsubstance was noted after birth. Likewise, both biglycan and cartilage oligomeric protein were concentrated at the insertion site.³⁰

The role of mechanical loading on enthesis development

The mature enthesis, as with other musculoskeletal tissues, is responsive to mechanical stimulation.⁷ However, the role of muscle loading in the formation of a functional bone-tendon interface has only recently been explored. As the fibrocartilaginous zones of the developing rat supraspinatus enthesis are not visible until postnatal day 7,²⁸ it was hypothesized that muscle loading following birth contributes to the formation of the graded transition seen in adult interfaces. Thomopoulos et al.³¹ injected the left supraspinatus muscle of a rat with botulinum toxin A (BtxA) at birth, effectively inducing paralysis of the left shoulder. At day 14, evidence of transitional fibrocartilage was seen forming between the supraspinatus tendon and its humeral head insertion, but by day 21 the fibrocartilage had regressed. Instead, there was a layer of disorganized mesenchymal-like cells and hypertrophic chondrocytes. By day 56, the tendon-bone interface in the control shoulders consisted of fully developed transitional fibrocartilage organized in the classical four-zone structure. The shoulders paralyzed by BtxA injection, however, showed fibrochondrocytes embedded in a disorganized matrix lacking any indication of a zonal structure.³¹

During normal enthesis development of the rat supraspinatus, a gradient in the mineral-to-bone ratio is detectable at the leading edge of the hard-soft tissue interface as early as postnatal day 7.³² While the length of this mineralized region (~20 μm) remains constant through day 28,³² the amount of mineralized bone increases over time due to endochondral ossification.³¹ Mechanical loading is necessary for this continued maturation, as BtxA-induced paralysis results in significant reductions in bone volume, as compared with saline-injected controls.³¹ Likewise, the BtxA-unloaded shoulders displayed reduced collagen fiber organization and crystallographic atomic order of the hydroxylapatite phase, a measure of crystallinity. These microstructural and compositional changes resulted in significantly reduced strength, modulus, and toughness in the paralyzed shoulders.³³ Taken together, these findings emphasize the importance of appropriate muscle loading in enthesis formation, and support the need to consider the mechanical microenvironment of the healing bone-tendon interface following injury.

Bone-Tendon and Bone-Ligament Injury and Natural Healing

Incidence of bone-tendon injury and the natural healing response

Injuries to the bone-tendon or bone-ligament interface are common and are traditionally classified as either acute ruptures or chronic degenerative changes.⁶ However, such a strict dichotomy is unlikely to exist. While the microstructure of the degenerative enthesis (enthesopathy) has not been directly compared with that of the acutely ruptured interface, it is probable that degenerative changes precede acute tears, as is seen in the injuries of the tendon midsubstance. Kannus and Josza³⁴ reported that nearly all (97%) of spontaneously ruptured tendons showed histopathological changes indicative of chronic tendon degeneration. Likewise, acutely ruptured tendons are significantly more degenerated than tendinopathic tendons.³⁵

Enthesopathies are defined as pathological changes in the enthesis and are frequently seen in athletic populations, in addition to those suffering from rheumatological conditions, such as rheumatoid arthritis, spondylarthropathy, calcium pyrophosphate skeletal hyperostosis (CPPD), and diffuse idiopathic skeletal hyperostosis (DISH).⁷ Given such a broad array of etiologies, enthesopathies will not be covered in this review. Further description is provided in references 7–8. In terms of acute damage to the bone-tendon interface, the two most common injuries involve tearing of the rotator cuff tendons of the shoulder and the cruciate ligaments of the knee. The rotator cuff is comprised of the interdigitating tendons of four muscles—subscapularis, supraspinatus, infraspinatus, and teres minor—that attach on the lateral aspect of the humeral head, serving important roles in both stabilizing and mobilizing the shoulder. Rotator cuff tears are a common cause of debilitating pain, reduced shoulder function, and weakness, affecting more than 40% of patients older than 60 y of age and resulting in 30 000 to 75 000 repairs performed annually in the United States.³⁶ Despite advances in surgical techniques and in the understanding of shoulder pathology, chronic tears fail to heal in 20–95% of cases.^37,38 Likewise, damage to ligaments of the knee are very common, with an estimated incidence of 2/1000 people per year in the general population.³⁹ Ninety percent of these injuries involve the anterior cruciate ligament (ACL) and medial collateral ligament (MCL).³⁹ Depending on the site of injury of the MCL, conservative therapy can often be employed with satisfactory results. However, the ACL does not heal spontaneously and requires reconstructive surgery to restore knee stability.⁴⁰ Although the majority of ACL ruptures occur in the midsubstance, successful reconstruction is dependent upon initial fixation and subsequent osseointegration of a graft into a bone tunnel.⁴⁰ Traditionally, the bone-patella tendon-bone (BPTB) autograft has been the preferred autologous graft due to superior bone-to-bone healing, as opposed to autografts derived from soft tissues, most often hamstring tendons. The improved osseointegration of bone blocks in bone tunnels results in enhanced mechanical properties at early time points of healing.⁴¹ However, BPTB autografts are associated with higher morbidity rates,⁴⁰ leading to increased use of tendon grafts. Furthermore, rotator cuff repair involves the apposition of the debrided tendon against the humeral head.⁴² Therefore, complete restoration of the native bone-tendon interface could reduce the risk of re-rupture, while allowing earlier and more aggressive rehabilitation protocols.

While some studies have shown histological evidence of fibrocartilage formation following surgical repair of the bone-tendon interface,⁴³ the majority have failed to demonstrate restoration of the native enthesis structure. Using a flexor tendon autograft for ACL reconstruction in a dog model, Tomita et al.⁴⁴ found no fibrocartilage formation between the graft and the bone tunnel up to 12 wk, resulting in an ultimate strength that was 42% of the normal ACL. In a similar study, Newsham-West et al.⁴⁵ examined the long-term healing of the bone-tendon interface in a sheep model following surgical reattachment of the patellar tendon to the patella. At 2 y, macroscopic observation revealed a tendon that merged seamlessly with the bone tissue. However, microscopic evaluation showed residual hypercellularity, extracellular matrix disorganization, and no fibrocartilage layer. Therefore, it is generally recognized that current surgical approaches are unable to recapitulate the four zones of the native fibrocartilaginous enthesis, resulting in reduced mechanical properties and an increased risk of re-injury. Consequently, several researchers have explored the possible mechanisms to explain the failed regeneration.

Potential mechanisms to explain the failed healing response

The role of intrinsic T/L or bone cells in healing at the enthesis is currently not understood. As the fibrocartilaginous region is avascular,⁶ endogenous progenitor cells that might reconstitute the enthesis must be derived from the surgically apposed tendon, as transdifferentiating tenocytes or tendon-derived stem cells, or from pericytes located in the granulation tissue or the subchondral bone. In seems unlikely that tendon-derived cells, found either in an autograft or the torn tendon, contribute significantly to bone-tendon healing, as shown in a recent study.⁴⁶ Conversely, bone marrow-derived cells may contribute to enthesis healing. In a mouse model that expresses green fluorescent protein (GFP) only in the bone marrow- and circulation-derived cells, Kida et al.⁴⁷ surgically repaired transected supraspinatus tendons. Drilling into the bone marrow was performed in the greater tuberosity of the right shoulder prior to repair, while the left shoulder was repaired without drilling. GFP-positive mesenchymal cells were present in the repaired rotator cuffs of both groups, with a higher concentration seen in the drilling group at 2, 4, and 8 wk. Furthermore, the ultimate load was significantly higher in the drilling group at both 4 and 8 wk.⁴⁷

While this study demonstrates the possibility of bone marrow-derived cells improving bone-tendon healing, the phenotype of these cells was not explored. It is unknown if these cells differentiated into resident fibrochondrocytes of the healing enthesis, or if they served as mediators of repair by paracrine factors and/or direct cell-cell contact. In vitro, Wang et al.⁴⁸ showed that osteoblast-fibroblast co-cultures slightly upregulated markers of chondrogenesis, including collagen type II and aggrecan. However, whether the osteoblasts or fibroblasts were principally responsible for this change in gene expression was not investigated. In a related study, a trilineage coculture system (osteoblasts–BMSC–fibroblasts) on a hybrid silk scaffold was created.⁴⁹ Both RT-PCR and immunohistochemistry demonstrated that the BMSCs differentiated into fibrochondrocytes, with a calcified region forming between the BMSCs and osteoblasts, but not the BMSCs and fibroblasts. Therefore, it appears that osteoblasts and fibroblasts are capable of supporting fibrocartilage differentiation of MSCs when interposed between the two. Why this does not consistently occur during the natural healing of the bone-tendon interface remains unknown.

One possible explanation is that there is an insufficient cell concentration to reconstitute the native fibrocartilage at the injury site. Given the avascular nature of the fibrocartilaginous region, this is an attractive hypothesis; it is has been the central basis for cellular therapies in bone-tendon healing, as discussed in detail below. A second, mutually compatible, explanation involves the biochemical and biomechanical microenvironment of the lesion. The synovial fluid of joints is known to inhibit tendon-to-bone healing,⁵⁰ ostensibly due to elevated pro-inflammatory cytokines and metalloproteases found in association with chronic tendon degeneration or acute tears. Additionally, synovial fluid contains the anti-adhesion protein lubricin, which is found on the end of ruptured intrasynovial tendons.⁵¹ Regardless of the identity of the inhibitory factors, the healing of the bone-tendon interface is worse when exposed to an intra-articular environment.⁵² Consequently, surgeons have developed repair strategies that minimize permeation of synovial fluid into the healing bone-tendon interface.⁵³ While these surgical techniques increase initial fixation strength, improve contact areas, and decrease gap formation at the healing enthesis, there is insufficient clinical evidence at this time to definitively confirm superior structural healing or functional outcomes in patients.⁴²

Mechanical loading of the post-operative bone-tendon interface is also known to affect healing. Using a rat model of a transected supraspinatus tendon, Thomopoulos et al.⁵⁴ examined the effect of post-operative activity levels on the biomechanical, structural, and compositional properties of the healing shoulder. Of the three activity levels—plaster-cast immobilization, normal cage activity, and daily exercise—shoulders that were immobilized possessed the best quasilinear viscoelastic properties. Likewise, the immobilized shoulders demonstrated significantly higher collagen orientation and expressed extracellular matrix genes in a pattern similar to uninjured insertions.⁵⁴ In contrast, complete removal of load by injecting BtxA into the supraspinatus was detrimental to rotator cuff healing, especially when combined with immobilization.⁵⁵ In a similar study examining the effect of early and delayed mechanical loading on tendon-to-bone healing after ACL reconstruction, it was found that delayed loading improved mechanical and biological parameters to a greater extent than immediate loading or prolonged post-operative immobilization. Therefore controlled mobilization of the healing bone-T/L interface is required to optimize the restoration of native anatomical structure and, in turn, the functional outcome.⁵⁶

Tissue Engineering Approaches to Bone-T/L Healing

Tissue engineering of the bone-T/L interface, like any organ system, combines cells, scaffolds, bioactive molecules, and biophysical stimulation, in an effort to restore structure and function. However, unlike certain elements of the musculoskeletal system such as the T/L midsubstance, biological interfaces present a more complex engineering challenge due the multiple cellular phenotypes and graded mechanical properties existing in highly organized, but spatially small (<1 mm), extracellular matrix. Recent efforts to fabricate graded scaffolds capable of supporting zone-appropriate cell types by incorporating biochemical components⁵⁷ or viral vectors⁵⁸ have shown promise. Others have designed triphasic scaffolds, with controlled matrix topography and mechanical properties, thereby supporting cell phenotypes germane to the enthesis.^59,60 Efforts to enhance enthesis healing by scaffolding have been extensive, with several FDA-approved products available for clinical use (see ref. 36 for a thorough review of scaffolds designed to improve bone-T/L healing).

Given the role of growth factors⁶¹ and metalloproteinases⁶² in enthesis development and healing, many investigators have used bioactive agents to augment surgical repair in animal models. BMP-2, a known regulator of osteogenesis and tenogenesis,⁶³ has been applied with variable success to multiple injury models, enhancing the mechanical properties⁶⁴ and osteoid formation⁶⁵ at the healing enthesis. Under a similar rationale, bone cement has also been applied to the surgically repaired interface. Gulotta et al.⁶⁶ augmented tibial tunnel fixation of a tendon autograft in ACL reconstruction by applying a magnesium-based bone cement. While there no was effect at 3 wk, the augmented group showed increased fibrocartilage and bone formation, as well as superior ultimate load, at 6 wk.⁶⁶ As a final example, several investigators have delivered TGF-β isoforms to the surgical site. TGF-β plays a significant role in interface development and healing, as discussed above. Unfortunately, exogenous application of this growth factor has shown little effect on bone-tendon healing.^5,67 For a more comprehensive review of interface tissue engineering approaches using bioactive agents, see reference 5.

Cell therapy in bone-T/L interface regeneration—animal models

During surgical repair of the bone-tendon interface, the ruptured end of the tendon is debrided and the bony insertion site is decorticated, destroying any remaining fibrocartilage.⁴³ As a result, restoration on the native zonal enthesis structure requires transdifferentiation of osteoblasts or tenocytes into fibrochondrocytes, or the chondrogenesis of endogenous progenitor cells. As discussed above, it appears unlikely that the tendon contributes a substantive cell source to recapitulate the normal fibrocartilage zones. While bone marrow-derived progenitor cells can infiltrate into the healing interface,⁴⁷ this intrinsic response is insufficient to restore native anatomy. Instead, a fibrous enthesis often forms, with an immediate transition from soft connective tissue to stiff cortical bone. In an effort to promote fibrocartilage formation between the native bone and T/L, investigators have applied an interposing layer of cells that possess chondrogenic ability (Fig. 3). Four classes of cells/tissues have been explored to date—ACL-derived stem cells, chondrocytes, MSCs, and periosteum. A detailed description of the experimental design and results of these studies is provided in Table 1, and is summarized below.

Figure 3. Cell sources for augmenting bone-tendon healing. Animal studies exploring the use of MSCs to improve bone-tendon healing have derived these cells primarily from the bone marrow, although other tissues (e.g., adipose) also harbor MSCs (not shown). ACL-derived MSCs can be obtained from the ruptured ligament, while chondrocytes/cartilage plugs can be taken from non-weightbearing articular surfaces. Lastly, the periosteum is typically harvested from the anteromedial tibia, given the ease of access in the absence of overlying soft tissues.

Table 1. In vivo studies of cellular therapy in bone-tendon interface healing

Cell Source	Site of Repair	Animal Model	Results	Ref.
Anterior Cruciate Ligament (ACL) Tissue/Cells
Autologous ruptured ACL	Anterior cruciate ligament	• Beagle dogs • Bilateral ACL reconstruction with ipsilateral flexor digitorum superficialis tendon autograft • “Tissue-treated” group augmented with resected native ACL tissue in tibial bone tunnel	• Tissue-treated group showed endochondral ossification-like integration with enhanced angiogenesis and osteogenesis starting at 2 wk • At week 4, Tissue-treated group had smaller tibial tunnel cross-sectional area as compared with contralateral control • At week 4, ultimate load in tissue-treated group was over twice that of controls 80% of controls failed at bone tunnel; Only 20% of tissue-treated samples failed at tunnel	69
ACL-derived stem cells	Anterior cruciate ligament	• Nude rats • ACL reconstruction with intracapsular injection of 5x10^5 human stem cells obtained from ruptured ACL • Four experimental groups – (1) CD34+ cells, (2) Nonsorted cells, (3) CD34- cells, (4) no cells	• Human stem cells honed to healing bone-tendon interface • Collagen fiber deposition and collagen 2-expressing cells seen at 2 wk • Human stem cells induced angiogenesis and osteogenesis in rat cells, and differentiated into these lineages themselves • Bone volume/total tissue volume increased at 4 wk in CD34+ and NS groups • **In all above cases, greatest effects seen in CD34+ > NS > CD34- cells • At week 8, ultimate load greatest in CD34+ > NS > CD34- cells	70
ACL-derived stem cell sheets	Anterior cruciate ligament	• Nude rats • ACL reconstruction with flexor digitorum longus autograft • Three experimental groups – (1) autograft wrapped with human CD34+ cell sheet, (2) Intra-articular injection of CD34^+ cells, (3) no cells	• Transplanted cells observed directly in tendon graft and at bone tunnel; Greater number for cell sheet vs intra-articular injection • Increased vascularization and collagen type 3 expression in tendon graft • At two weeks, evidence of intrinsic osteogenesis and neovascularization around bone tunnel; Human cell markers also co-localized with endothelial and osteogenic markers • ** In all above cases, greatest effects seen in sheet > injection > no cells • At week 8, ultimate load greatest in sheet > injection > no cells	71
Chondrocytes/Cartilage
Allogeneic chondrocyte pellet	Partial patellectomy	• NZW rabbit • Distal 1/3 patella resection • Placed chondrocyte pellet between remaining patella and ligament • Immobilized 4 wk long cast	• Chondrocyte pellet placed between 1/3 distal-resected patella and patella ligament showed improved integration and formation of fibrocartilge-like zone at weeks 12 and 16. No integration seen without pellet • No mechanical testing performed • Collagen/mineral composition of interface not performed	72
Articular cartilage plug	Partial patellectomy	• Chinese goats • Distal 1/3 patella resected and patella tendon sutured to proximal 2/3 patella • Repair alone vs. cartilage wafer (from autologous hyaline cartilage on resected 1/3 patella • Casted 6 wk, weight-bearing immediately	• Autologous articular cartilage interposition resulted in more fibrocartilage transition zone formation than direct repair at all times • No group differences in new bone formation, but increased with time in both groups • Ultimate load increased for both groups, while ultimate stress did not change. No group differences at any time	73
MSCs or Chondrocytes in fibrin clot	Achilles	• Male Wistar rat • Achilles enthesis destroyed with transection and burr • Repair with suture through bone tunnels • Three experimental groups (1) repair only, (2) 4x10^6 chondrocytes in 100 ul fibrin clot, (3) 4x10^6 MSCs in 100 ul fibrin	• Both chondrocytes and MSCs produced fibrocartilage (as demonstrated by Col2 and GAG staining) starting at day 15. No fibrocartilage in untreated group up to 45 d • On day 45, only MSC recapitulated columnar structure of hypertrophic chondrocytes • Failure to load improved over time for all groups, but chondrocyte and MSC significantly higher than controls at day 45 (and equal to native controls)	74
Enthesis Cells
Tendon-to-bone interface cells	Supraspinatus	• Lewis Rats • 2x2 mm of supraspinatus excised and enthesis destroyed Five groups – (1) native control, (2) no repair, (3) suture repair, (4) suture repair with gelfoam, (5) suture repair with gelfoam + 1.5 x 10^5 cells	• At week 3, gelfoam + cell group (V) displayed increased cellularity, vascularity, inflammation, and collagen disorganization vs. other groups • At week 12, gelfoam + cell group (V) had better collagen organization than other treatment groups (II-IV) • Cells retained at healing site up to 12 wk No biomechanical testing performed	75
Mesenchymal Stem Cells (MSCs)
BMSCs in fibrin clot	Supraspinatus	• Lewis rats • Supraspinatus transection with acute repair augmented with nothing, fibrin clot, or fibrin + allogeneic BMSCs • Not immobilized	• Ad-LacZ-transduced MSCs demonstrated cell viability at 2 and 4 wk (but decreasing) • No difference in amount of new cartilage or collagen fiber organization between groups at either time point • Mechanical properties improved with time, but no group differences	83
MT1-MMP-transduced MSCs in fibrin clot	Supraspinatus	• Same methodology as^83 • MSCs in fibrin clot vs. MT1-MMP-transduced MSCs at weeks 2 and 4	• Greater chondrogenesis (Safrinin O staining) in MT1-MMP-transduced MSCs at week 4, as compared with MSC alone. No difference at week 2. • No group difference in collagen alignment (picrosirius red) at either time point • Mechanical properties greater in MT1-MMP MSCs than MSCs alone at week 4 only	88
BMP13-transduced MSCs in fibrin clot	Supraspinatus	• Same methodology as [2] • MSCs in fibrin clot vs. BMP13-transduced MSCs at weeks 2 and 4	• Improved biomechanical properties over time, but no group differences • No group differences in histology (safranin O for fibrocartilage formation and polarized picrosirius red for collagen organization)	84
Scx-transduced MSC in fibrin clot	Supraspinatus	• Same methodology as^83^,^88 • MSCs in fibrin clot vs. Scx-transduced MSCs at weeks 2 and 4	• Ad-Scx MSCs improved mechanical properties at 2 and 4 wk, but absolute values not clearly better than MT1-MMP-mediated improvements [3] • Increased fibrocartilage formation (as determined by Safranin O staining) in Scx-MSC group, but no group differences in collagen organization	87
BMSCs on poly-glycolic scaffold	Infraspinatus	• Japanese White Rabbit • Full-thickness (10x5mm) defect made in infraspinatus tendon • Defect filled with (1) nothing (control), (2) PGA scaffold, or (3) PGA scaffold seeded with 5x10^6 cells	• Left untreated, defect healed with poorly organized fibrovascular scar, with no improvements in mechanical properties from 4 to 16 wk • At 4 wk, both PGA and PGA+MSC groups showed partially degraded scaffold with foreign body reaction and multinuclear cell infiltrate • At week 8, chondrocytes seen in PGA+MSC group only, but at week 16, both groups showed fibrocartilage formation and positive Safranin-O staining • PGA group showed greater Collagen 3 to Collagen 1 staining at weeks 8 and 16; Conversely, PGA+MSC group showed Col1 > Col3 staining at week 16 • No group differences in biomechanical properties at week 4, but by week 16 PGA+MSC > PGA > > > controls (defect only)	93
MSCs in fibrin coating allograft	Anterior cruciate ligament	• NZW rabbit • Bilateral ACL reconstruction with Achilles tendon allograft • Graft coated with (1) fibrin or (2) fibrin + 4 × 10^6 BMSCs	• Fibrin group showed fibrovascular scar in interface at weeks 2, 4, 8 • MSC group showed chondrocyte-like cells in interface at 2 wk; by week 8, differentiated enthesis with 4 zone transition; Copious GAG and Col2 staining • At 8 wk, load to failure higher in MSC than controls. No significant groups differences at earlier time points, or in stress, stiffness, Young’s modulus • Major mode of failure was graft pullout (wk 2) and midsubstance rupture (wk 8); no group difference	82
BMSCs in fibrin sealant coating tendon allograft	Anterior Cruciate ligament	• NZW Rabbits • Bilateral ACL reconstruction with hamstring allograft, coated with (1) fibrin or (2) fibrin + BMSCs	• Fibrin group showed fibrovascular interface at weeks 2, 4, 8 • Fibrin + MSC group showed disorganized chondrocytes at week 2. Increasing fibrocartilage organization at weeks 4 and nearing native enthesis histology at week 8, with graded fibrocartiaginous interface • Load to failure and stiffness did not change over time in fibrin controls; Both increased over time in MSC group, with significant group differences at week 8	80
Synovial MSCs	Medial patella tendon reconstruction	• Sprague-Dawley rats • Medial patella tendon resected and replaced with ipsilateral ½ Achilles tendon autograft; Proximal graft sutured to patella periosteum and distal portion inserted into bone tunnel from tibial plateau to tibial tuberosity • Bone tunnel injected with (1) collagen gel (control) or (2) collagen gel + 1x10^7 synovial MSCs	• At week 1, both control and MSC-group showed fibrovascular scar in interface • By week 4, implanted tendon attached directly to bone, with no observable fibrocartilage in either group • At week 4, MSC-group had higher percentage of oblique (i.e., Sharpey) fibers relative to total interface area, as compared with controls • No biomechanical testing	120
Murine mesenchymal cell line (C3H10T[1/2] and human BMSC virally-mediated overexpression of Smad8ca and BMP2	Heterotopic (intramuscular)	• Nude mouse • Intramuscular injection of murine C3H10T[1/2] mesenchymal cell line or human BMSC with virus-mediated overexpression of constitutively active Smad8 (Smad8ca) and BMP2	• At 1 mo, C3H10T[1/2] cells with Smad8ca and BMP2 overexpression formed bone ossicle and tendon-like structure with interposing fibrocartilage-like structure • Murine cells infected with Smad8ca alone produced tendon-like tissue • Human BMSCs with lentiviral-mediated overexpression of Smad8ca and BMP2 formed bone ossicle and tendon-like structure with no intervening fibrocartilage • Neither uninfected cells nor wild-type Smad8 expression produced neotissue	121
BMSC	Heterotopic (Hallucis longus tendon in calcaneal bone tunnel)	• NZW rabbit • Hallucis longus tendon inserted into calcaneal bone tunnel • Augmented with (1) fibrin gel (control) or (2) fibrin gel + 1 × 10^7 BMSC	• Control group showed increasing organization of fibrovascular scar at 2, 4, 6 wk • BMSC group showed fibrocartilage formation over 4 wk, but only at 50% of bone-tendon interface • Both groups stained positive for Collagen 1 and 3; Only BMSC group stained positive for collagen 2 • No mechanical testing	81
Periosteum/Periosteum progenitor cells
Autologous periosteum or bone marrow aspirate	Heterotopic (extensor digitorum longus tendon in bone tunnel)	• NZW rabbits • Femoral insertion of extensor digitorum longus resected and inserted in tibial bone tunnel • (1) non-augmented (C) and (2) wrapped with periosteum (P), or (3) bone marrow aspirate (BMA)	• P and BMA groups showed increased interface organization as compared with controls at both weeks 6, 12 • Chondrocyte-like cells seen in BMA group at week 6; Fibrocartilage islands seen in both BMA and P group at week 12; both stained positive for Safranin O • At week 6, load to failure and stiffness P > BMA > C; At week 12, load to failure significantly increased over controls in P alone, with stiffness statistically equivalent across groups	96
Autologous periosteum	Heterotopic (extensor digitorum longus tendon in bone tunnel)	• NZW rabbits • Femoral insertion of extensor digitorum longus resected and inserted in tibial bone tunnel • (1) non-augmented (control) and (2) wrapped with periosteum	• For periosteum-treated group, fibrovascular scar seen at 4 wk; At week 8, maturation of new bone growing into fibrous tissue, with increasing mineralization; At 12 wk, fibrocartilage formation identified; No histology for control groups was shown or described • Load to failure increased over time • Load to failure significantly higher in periosteum-treated group than controls at weeks 8 and 12	97
Autologous periosteum	Infraspinatus	• NZW rabbits • Acute transection of infraspinatus and footprint debridement • Transosseus repair with or without autologous tibial periosteum interposed between tendon and bone	• Both control and periosteum-treated groups showed healing of infraspinatus to humerus • Periosteum-treated group showed fibrovascular scar in interface at week 4, chondrocyte-like cells at week 8, and anatomy similar to native enthesis with tidemark at week 12; No histology for control group was shown • Both groups showed increased load to failure over time; Periosteum-treated group significantly greater than controls at weeks 8 and 12	98
Periosteum progenitor cells with BMP2-tethered hydrogel	Infraspinatus	• NZW white rabbit • Infraspinatus transected and enthesis destroyed • Transosseus repair with interposed hydrogel; 3 groups – (1) native control, (2) PEGDA hydrogel, (3) PEGDA hydrogel + BMP2 + PPCs	• At weeks 4 and 8, both groups demonstrated increasing fibrocartilage formation • Hydrogel containing BMP2 and PPCs demonstrated improved fibrocartilage and bone formation determined by histology, as compared with hydrogel alone • Ultimate load increased over time in both groups, but was higher in PPC groups at both weeks 4 and 8	99

ACL-derived stem cells

Rupture of the ACL occurs most frequently in the ligament midsubstance and fails to spontaneously heal, necessitating ACL reconstruction. The failed healing response is traditionally attributed to the hypovascular nature of the ACL, but recent work suggests that the ACL possesses a source of multipotent stem cells that exist in the vasculature of the septum that divides the two ACL bundles.⁶⁸ From this vasculature, isolation of CD34- and CD146-postitive cells revealed the capacity for multipotent differentiation. In addition, these vasculature-associated stem cells were found in higher concentrations in ruptured ACLs than uninjured ligaments.⁶⁸ In a series of three studies, investigators hypothesized that these CD34⁺ stem cells could augment ACL reconstruction by improving integration of a tendon autograft into the bone tunnels.

ACL reconstructions in animal models were augmented with resected autologous ACL tissue,⁶⁹ CD34⁺ cells alone,⁷⁰ or monolayer sheets of CD34⁺ cells,⁷¹ either injected intra-articularly or sutured to the bone-tunnel portion of the autograft. Across these studies, the cell-treated groups showed enhanced angiogenesis and osteogenesis in the bone tunnel, a smaller tibial tunnel cross-sectional area, and enhanced mechanical properties. Furthermore, these exogenous cells were capable of inducing angiogenic and osteogenic differentiation of host cells, as well as directly differentiating into bone and endothelial cells.^70,71 Taken together, the above studies suggest that the ACL possesses a multipotent cell population that can improve the tendon graft integration following ACL reconstruction. However, the practicality of isolating these cells for exogenous application remains dubious, unless processing can occur intraoperatively or future technologies can augment the contribution of these cells to the intrinsic healing process.

Chondrocytes/cartilage tissue

Rather than applying a multipotent cell source capable of undergoing chondrogenesis, the interposition of mature chondrocytes between tendon and bone has shown promise as a means of restoring the transitional anatomical structure. However, the ability of differentiated chondrocytes to improve the mechanical properties of the healing interface has been limited so far. In two studies using a rabbit⁷² and goat⁷³ model, Wong et al. resected the distal 1/3 of the patella and sutured the patella tendon to the remaining proximal 2/3 bone. Both a chondrocyte pellet⁷² and an autologous cartilage plug derived from condylar hyaline cartilage⁷³ were placed between the tendon and bone, resulting in more fibrocartilage formation without enhancing the mechanical properties, as compared with controls.

In a rat model of Achilles insertion repair, Nourissat et al.⁷⁴ compared MSCs with chondrocytes in terms of their ability to restore enthesis structure and function. Both MSCs and chondrocytes produced fibrocartilage starting at day 15, as demonstrated by collagen type II expression and positive glycosaminoglycan staining. However, only the MSC-treated interfaces recapitulated the columnar structure by day 45. Additionally, both chondrocytes and MSCs restored the ultimate load to uninjured levels by day 45, with no difference between cell groups.⁷⁴ In a related study, cells isolated from the bone-tendon interface of the rat supraspinatus were concentrated in a gelfoam and inserted between the bone and tendon during suture repair.⁷⁵ The interface cell group, as compared with no cells and gelfoam alone, accelerated the healing process as demonstrated by increased cellularity, vascularity, and inflammation at week 3. By week 12, the cell-treated group showed superior collagen organization.⁷⁵

Mesenchymal stem cells (MSCs)

MSCs have emerged as the gold standard for cellular therapies in musculoskeletal diseases, given their ease of expansion and multipotency.⁷⁶ MSCs are capable of differentiating into tenocytes, chondrocytes, and osteocytes,^76-78 thereby potentially aiding in restoration of the native structure of the healing tissues. Beyond differentiating into site-appropriate epithelial lineages, MSCs secrete bioactive molecules that provide a regenerative microenvironment for a variety of injured adult tissues.⁷⁹ For these reasons, several investigators have explored the effect of augmenting enthesis repair with MSCs.

Experiments investigating the utility of MSCs in augmenting bone-tendon healing were first performed in rabbit models.^80,81 Ouyang et al.⁸¹ fixed the hallicus longus tendon in a calcaneal bone tunnel with the interface augmented with bone marrow MSCs (BMSCs) in a fibrin gel, or fibrin gel alone. While the study did not include any mechanical testing, the BMSC group enhanced collagen type II staining, though only 50% of the graft-tunnel interface contained fibrocartilage. Lim et al.⁸⁰ and Soon et al.⁸² reported similar findings, with the MSC group showing chondrocyte-like cells at the interface by 2 wk and a differentiated enthesis with four-zonal transition by 8 wk. Conversely, the fibrin only group contained fibrovascular scar between tendon graft and bone at every time point. The stark contrast in histological appearance between groups translated into to dramatic group differences in mechanical properties.^80,82

These early studies provided strong evidence to support the augmentation of reconstructive surgeries involving tendon grafts fixed in bone tunnels, but the utility of cell therapy in healing the fibrocartilaginous enthesis following rotator cuff repair remained unexplored. Over the course of four studies, Gulotta and colleagues manipulated BMSCs by virus-mediated overexpression of bioactive agents that are known to affect tendon development and healing. In all cases, the rat supraspinatus tendon was transected and repaired with sutures. BMSCs alone⁸³ or BMSCs overexpressing BMP-13,⁸⁴ a growth factor capable of inducing ectopic tendon formation⁸⁵ and improving in vivo tendon healing,⁸⁶ were localized to the wound site in a fibrin clot. At both 2 and 4 wk, there was no difference between groups in mechanical properties, new cartilage formation, or collagen fiber organization. In two related studies, BMSCs were infected with a viral vector designed to overexpress the tenogenic transcription factor Scleraxis⁸⁷ or membrane type 1 matrix metalloproteinase (MT1-MMP, also called MMP-14).⁸⁸ MT1-MMP is a membrane-bound matrix metalloproteinase that plays a role in embryonic development, being upregulated in areas of chondrocyte differentiation as well as in areas of tendon, perichondral bone, ligament, and joint capsule development.^88,89 In comparison to suture repairs augmented with uninfected BMSCs both Ad-Scx MSCs (Adenoviral-transfected MSC with constitutively active Scx) and Ad-MT1-MMP MSCs increased fibrocartilage formation, and increased the mechanical properties (ultimate load, ultimate stress, stiffness) of the healing bone-tendon interface, as compared with controls.

In all the above studies, BMSCs were delivered to the wound site in a fibrin gel carrier. While fibrin clots offer sufficient form and durability to localize stem cells to wound site, they offer little mechanical support to the healing tissue.⁹⁰ Recent work by Butler et al.⁹¹ suggests that matching the mechanical properties of a cell-seeded scaffold to that of the native tissue improves the in vivo function of the regenerating tissue. Therefore, combining BMSCs with biomimetic interfacial scaffold may work synergistically to enhance enthesis regeneration. While multiphasic scaffolds have been fabricated,^59,92 their utility in augmenting enthesis repair has not been explored in vivo. However, Yokoya et al.⁹³ did seed BMSCs on a polyglycolic (PGA) sheet before implanting the cell-seeded scaffold into a full thickness infraspinatus defect that included the insertion site. By week 8, only the BMSC+PGA group showed fibrocartilage formation, while both the BMSC+PGA and PGA sheet alone produced positive glycosaminoglycan staining by week 16. At this later time point, the shoulders augmented with the cell-seeded scaffold possessed superior mechanical properties to those which received the PGA sheet alone, and both experimental groups drastically enhanced healing when compared with untreated defects.⁹³ Unfortunately the delivery of BMSCs alone was not explored as an additional control group.

Periosteal progenitor cells/periosteal tissue

The periosteum is a thin layer of connective tissue that lines the outer surface of bone. It is a bilayered membrane with an outer layer of aligned collagen, elastin, and fibroblasts, and an inner (cambrium) layer that consists mostly of progenitor cells that contribute to bone repair.⁹⁴ The periosteum has received growing interest as an autologous biomaterial for musculoskeletal tissue regeneration as it contains progenitor cells and noncollagenous proteins (including growth factors) embedded in a native extracellular matrix.⁹⁵ The periosteum may have particular relevance to augmentation of bone-tendon healing, as the outer, fibrous layer resembles the soft connective tissue of T/L, while the cambrium layer may provide progenitor cells for chondrogenesis and osteogenesis,⁹⁵ thereby restoring the four zones of native entheses.

Karaoglu et al.⁹⁶ compared the effects of periosteum tissue (P) against bone marrow aspirate (BMA). Relative to untreated controls, both P- and BMA-treated groups showed increased interface organization by week 6. Chondrocyte-like cells were present at 6 wk only in the BMA group, but both experimental conditions produced fibrocartilage islands by week 12. At week 6, both load-to-failure and stiffness were significantly increased in P, as compared with BMA, which was also superior to controls.⁹⁶ Similarly, Chen and colleagues demonstrated that periosteum^97,98 or hydrogels containing periosteal progenitor cells with supplemented BMP-2⁹⁹ could drastically improve both the histological and biomechanical properties of the healing bone-tendon interface. When examining periosteum-augmented repairs at 12 wk, the interface anatomy resembled that of the native enthesis, with a clearly visible tidemark.⁹⁸ Since the effects of periosteum and periosteum progenitor cells with BMP-2 supplementation were investigating in independent studies, it is not possible to compare their relative efficacy.

The same can be said when comparing cell sources (e.g., MSCs vs. periosteum). As highlighted above, multiple cell sources can be used to improve fibrocartilage formation and mechanical properties of the healing bone-tendon interface. But, with the great diversity of animal models used and dependent variables measured, it is impossible to identify an ideal cell source at this time. Future work should directly compare multiple cell types within the same injury model, similar to that performed by Gulotta and colleagues.^83,84,87,88 Furthermore, combining cells with growth factors⁹⁹ and scaffolds⁹³ may offer synergistic effects and should be further explored.

Cell therapy in bone-t/l interface regeneration—clinical studies

Despite the growing in vivo evidence demonstrating the benefit of cellular therapies in bone-tendon healing, the translation of these results into clinical care has been slow. To date, there have been no randomized, prospective trials investigating cell-augmented enthesis repair. However, three cases series have been performed, the results of which encourage future research.

Chen et al.¹⁰⁰ performed 372 single-bundle ACL reconstructions, all of which were augmented with periosteum harvested from the anterior tibial cortex. The median Lysholm knee scores increased from 56 points preoperatively to 95 point postoperatively. Following reconstruction, 85% of patients were able to return to moderate to strenuous activity, and 93% of patients had normal or near-normal ratings according to the International Knee Documentation Committee assessment.¹⁰⁰ In a similar study of 20 patients, a periosteal flap lateral to the rotator cuff footprint was created and subsequently interposed between the tendon and humeral head before securing with suture anchors.¹⁰¹ At a mean follow-up of 14.4 mo, the average Constant score had increased from 51.7 points to 80.0 points. But, four patients (20%) displayed a tendon re-tear on postoperative MRI. Conversely, all patients (14/14) demonstrated tendon integrity at 12-mo follow-up in a different study in which a rotator cuff repair was augmented with autologous bone marrow mononuclear cells.¹⁰² In this clinical study, the UCLA score increased from 12 ± 3.0 before surgery to 31 ± 3.2 after surgery.

Although these three clinical studies suggest the safety and feasibility of augmenting bone-tendon healing with autologous cell sources, no further conclusions can be drawn in the absence of control groups and larger patient enrollment. Moving forward, prospective, randomized trials will be required in order to establish the superiority of cell-augmented surgical repairs over the current standard of care. Furthermore, patient populations with a comparable severity of injury should be compared, as this parameter can greatly affect functional outcomes.¹⁰³ Standardized functional scores should also be included to enable valid comparisons between studies. To the same end, the surgical technique can also affect functional and radiological findings,⁴² and so should be controlled as much as possible without jeopardizing patient care.

Translating Cell Therapy into Clinical Care

Regulatory, economic, and practical hurdles

As highlighted above, there is compelling evidence to support the application of cell therapies in augmenting current surgical techniques to repair bone-T/L injuries. However, only three case series involving autologous periosteum or mononuclear stromal cells have been performed in a patient population. In broader terms, less than 25% of promising biomedical discoveries yield a published randomized clinical trial, and less than 10% are established in clinical practice within 20 y.¹⁰⁴ The reasons for the stagnation of medical technologies in the pre-clinical phase are numerous, yet basic scientists and clinicians alike undoubtedly wish to see promising therapeutics make it into clinical trials. Moran et al.¹⁰⁵ have proposed several recommendations to improve bench-to-bedside translation, including changes at the level of the trainee up through governmental funding (e.g., NIH) and regulatory (i.e., FDA) institutions. Others have taken a more humble, but still ambitious approach to accelerate the implementation of regenerative medicine strategies into clinical care. “Facilitated endogenous repair,”¹⁰⁶ also referred to as “in situ guided tissue regeneration,”¹⁰⁷ is an approach to tissue engineering that avoids ex vivo culture of autologous cells and the use of manufactured scaffolds, thereby minimizing costs and the number of invasive clinical procedures. Part of this approach would involve the use of autologous cells and extracellular matrices procured, processed, and re-implanted intraoperatively. Two of the three clinical studies highlighted above could be classified under this concept.^100,101 In addition to obviating the costs associated with establishing good manufacturing practices (GMP)-certified labs for ex vivo cell expansion and/or manipulation, many of these point-of-care approaches bypass the arduous process of receiving FDA approval. On the other hand, facilitated endogenous repair also includes the delivery of viral vectors, which would rightfully be under close regulatory scrutiny. Regardless, the philosophy of this regenerative medicine approach stands in contrast to a tissue engineering strategy involving ex vivo whole organ/tissue fabrication with subsequent in vivo implantation. By allowing the body to serve as the bioreactor, cellular therapies for musculoskeletal disease may be implemented earlier and in a larger patient population.

Limitations of current animal models

Limitations of applying discoveries in animal models to the treatment of human disease are inherent in every field of biomedical research, including orthopedic sports medicine. Even within a specific pathological condition, different animal models may be most appropriate depending upon which disease parameter is being studied. In the case of rotator cuff repair, the rat is ideal for exploring biological approaches to bone-tendon healing, given the many similarities in shoulder anatomy shared with humans, as well as its relatively inexpensive cost.¹⁰⁸ By comparison, the canine shoulder allows investigators to study clinically relevant rehabilitation modalities, including slinging, hobbling, casting, swimming, jumping, and treadmill running.¹⁰⁸ Furthermore, each animal model possesses differing intrinsic healing capabilities, limiting the extension of experimental results to particular injuries that do not heal spontaneously in humans.¹⁰⁹

Another seldom acknowledged limitation of animal models of acute T/L ruptures is that spontaneous tears in humans are almost always preceded by chronic degenerative changes.³⁴ Conversely, acute rupture and repair models involve transection of a previously healthy tendon, often in young adult animals. Degenerative changes, even with the tendon intact, are associated with upregulation of pro-inflammatory and catabolic mediators in the synovial fluid.¹¹⁰ Therefore, it may be argued that the intra-articular milieu of spontaneously ruptured T/L in humans differs greatly from that created in animal models. Fortunately, validated animal models of tendinopathy do exist.¹¹¹ But to our knowledge, no one has yet studied the healing of an acute T/L rupture following a prodromal phase of degenerative changes.

Combining regenerative medicine with biophysical modalities

All musculoskeletal tissues are responsive to mechanical loading. In fact, the health of the tissue depends upon the homeostasis of anabolic and catabolic mediators, the balance of which is controlled by the stresses exerted on the system.¹¹² Furthermore, the timing and magnitude of mechanical loads placed on surgically repaired structures is carefully monitored by surgeons, while extensive research has been conducted in an effort to optimize rehabilitation protocols aimed at restoring function quickly and completely. In spite of this standard of care, mechanical loading of surgically repaired tissues in animal models is almost never controlled, except when considered as a dependent variable. Of the primary research articles reviewed above, all allowed resumption of normal cage activity immediately following surgery. While most (especially small) animal models tolerate early loading exceedingly well with low rates of re-tear, such a practice is greatly discordant with the current standard of care in almost all orthopedic surgical procedures.

Given the importance of mechanical loading in both T/L interface development and healing, future research examining the relationship between regenerative medicine approaches and physical therapeutics is greatly needed.¹¹³ More specifically, there is reason to believe that providing the appropriate mechanotransducive signals to the exogenously delivered cells is needed to optimize the regenerative effect. For example, improved mechanical properties were found in a healing tendon following an injection of platelet-rich plasma (PRP), an autologous source of growth factors of known importance in wound healing.¹¹⁴ However, the beneficial effect of PRP augmentation was only noted when mechanical loading was introduced following the acute inflammatory phase of the healing cascade (3 d). On the other hand, long-duration immobilization negated any effect of PRP.

Lastly, cellular therapies may be synergistically combined with non-invasive biophysical modalities with known benefit in promoting bone-tendon healing. Low-intensity pulsed ultrasound,¹¹⁵ pulsed electromagnetic fields,¹¹⁶ and extracorporeal shockwave therapy,¹¹⁷ have been shown to independently improve histological and mechanical properties of the surgically repaired bone-tendon interface. While the mechanisms by which such novel therapies exert these effects remain largely unknown, electrical cues are known to regulate cellular orientation, division, and differentiation, as well as the rate of in vivo wound healing.^118,119 Thus electrical stimulation may serve as a novel tool for directing the regeneration of the bone-tendon interface, although the utility of this modality is unexplored. In the meantime, there is sufficient pre-clinical data to support the implementation of autologous MSC or chondrogenic cell sources as a means of enhancing enthesis healing, even if the optimal combination of tissue engineering elements is yet to be determined.

Abbreviations:
ACL	=	anterior cruciate ligament
BMP	=	bone morphogenetic protein
BMSC	=	bone marrow-derived mesenchymal stem cell
BPTB	=	bone-patella tendon-bone
BtxA	=	botulinum toxin A
CF	=	calcified fibrocartilage
d.p.c.	=	days post coitum
GFP	=	green fluorescent protein
Ihh	=	Indian hedgehog
MCL	=	medial collateral ligament
MSC	=	mesenchymal stem cell
MT1-MMP	=	membrane type 1 matrix metalloproteinase
p.c.	=	post coitum
PTHrPR	=	Parathyroid hormone-related peptide receptor
RT-PCR	=	real-time polymerase chain reaction
Scx	=	Scleraxis
SnoN	=	SKI-like oncogene
T/L	=	Tendons and Ligaments
TGF-β	=	transforming growth factor beta
Tgif	=	TGF-interacting factor
TNMD	=	Tenomodulin
UF	=	uncalcified fibrocartilage

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

This work was supported in part by Commonwealth of Pennsylvania Department of Health (SAP4100050913), NIH (T32-EB001026 and 1R01 AR062947), and the U.S. Department of Defense (W81XWH-11-2-0143).

10.4161/org.27404