Introduction

Successful repair of ruptured flexor tendons, as measured by restoration of digital flexion function, is a great challenge to hand surgeons because the biological cascade of events during healing often causes the tendon proper to adhere indiscriminately to its surrounding tissue.¹ Clinical and experimental observations suggest that formation of adhesions is precipitated by injury to the tendon sheath, surgical manipulation, and immobilization.^2,3 This problem is most challenging with regard to tendon injuries in the "no-man's land" or Zone II. In the past, these were left unrepaired because of the poor prognosis associated with these injuries.⁴ As an alternative to primary repair, the transplantation of a tendon graft allows the surgeon to place the graft junctions outside of the confines of the flexor sheath in zone II, where they can be attached distally in Zone I (where no gliding motion takes place) and proximally in Zone III, to the flexor digitorum profundus tendon. However, even simple surgical manipulation of live flexor tendon grafts can result in cellular necrosis and inflammation, leading to adhesion.⁴ Therefore, devitalized structures such as freeze-dried tendon allografts or tissue-engineered biomaterial scaffolds are potentially attractive alternatives to live autografts in reconstructing the digital flexor mechanism.

Current tissue engineering strategies using synthetic biomaterial scaffolds have yet to yield clinically usable tendon substitutes. The appeal of these engineered scaffolds is that they can potentially be impregnated with growth factors or genes for targeted and timed release at the site of implantation in order to improve healing. However, many of these "manufactured" scaffolds do not match the mechanical strength of native tissue necessary for the expeditious restoration of function, and they do not remodel in response to daily activity; rather, they break down, producing byproducts that induce inflammation and compromise the repair process.⁵ As an alternative, naturally derived materials processed from animal tissue or produced using recombination technology may be better tolerated when implanted. Arguably, the most suitable choice for a naturally derived biomaterial scaffold for tendon tissue engineering would be one that is derived from "allogeneic" tendon tissue. Such scaffolds must meet several functional criteria. As aptly described by Whitlock et al., a naturally derived biomaterial scaffold from tendon tissue must be "amenable to host cell-mediated remodeling", "devoid of cellular material to minimize inflammatory potential", and "distinguished by sufficient biomechanical integrity".⁵

In order to test this concept in a pre-clinical model of tendon adhesion, we recently developed a mouse distal flexor digitorum longus (FDL) tendon-grafting model in which a 3-mm intercalary live autograft or a freeze-dried allograft is implanted.⁶ We demonstrated that both autografts and allografts led to significant reductions in the range of motion of the metatarsophalangeal (MTP) joint at 14 and 28 days, which resolved 42 days after surgery. Interestingly, we also observed that the gene expression of the growth and differentiation factor 5 (Gdf5) was significantly increased in 28-day grafts, thereby implying that this factor plays a role in the remodeling that leads to the functional improvements observed thereafter. We were intrigued by this observation, because GDF-5 deficiency in mice significantly delays the healing of the Achilles tendon (AT)⁷ and adenovirus-mediated Gdf5 gene expression in the treatment of AT injuries in rats leads to increased strength.⁸ On the basis of these findings, and in view of our previous success with recombinant adeno-associated virus (rAAV)-coated cortical bone grafts,^9,10 we hypothesized that the loading of freeze-dried mouse FDL tendon allografts with rAAV expressing Gdf5 gene would improve the functional properties of the reconstructed tendon and abolish the fibrotic adhesions. On the basis of the data from this study, we report that the remarkable hydrophilic capacity of freeze-dried tendon allografts can indeed be exploited for efficient loading of gene delivery vectors such as rAAV-Gdf5 to improve the functional outcome of flexor tendon reconstruction.

Results

Processing freeze-dried FDL tendon allografts as gene-delivery scaffolds

We have previously shown that two freeze-drying cycles produce no adverse effects on the mechanical properties of the tendon grafts.⁶ Pairs of freeze-dried FDL tendon grafts were reconstituted in a buffer containing 5 10⁹ U of rAAV-lacZ, and freeze-dried again. Using polymerase chain reaction (PCR) analysis, the rAAV retention efficiency was determined to be 10% (2.61 10⁸ 1.44 10⁸ genomes per graft, mean value SD; see Supplementary Figure S1). In order to assess the efficacy of cell transduction in vitro, rAAV-lacZ-loaded FDL grafts were individually placed in culture wells containing 293 human embryonic kidney cells (Figure 1a). X-gal staining showed that large numbers of cells were transduced after a 48-hour incubation with the rAAV-lacZ-loaded FDL grafts (Figure 1b), whereas a control culture incubated with unloaded FDL grafts was negative for transduction (not shown). Furthermore, the random and attenuated -galactosidase expression in regions of the wells not proximate to the grafts indicates that the transduction depends on diffusion of the virus after rehydration (Figure 1c). In order to assess the transduction efficacy of rAAV-loaded FDL grafts in vivo, freeze-dried FDL tendon allografts loaded with rAAV-lacZ were implanted in mouse intercalary FDL tendon defects, as previously described.⁶ The mice were killed at 7 or 14 days, and the grafted tissues were removed and fixed, paraffin-embedded, and processed for immunohistochemistry with antibodies specific to -galactosidase. Although the grafts remained mostly acellular, they were surrounded by exuberant hypercellular fibrotic tissue that exhibited intense staining specific for -galactosidase, thereby suggesting that the host cells in the peripheral tissue were transduced by the rAAV-lacZ vector as it slowly diffused out of the implanted graft, with more intense staining on day 7 than on day 14 (Figure 1d and e).

Figure 1.

Transduction efficacy of freeze-dried tendon grafts in vitro and in vivo. (a) 3-mm Freeze-dried mouse flexor digitorum longus (FDL) tendon allografts were loaded with 5 10⁹ transducing units of rAAV-lacZ, and incubated on a confluent monolayer of human embryonic kidney 293 cells for 48 hours (arrow). Representative micrographs of X-gal stained cultures show (b) large numbers of LacZ+ cells proximal to the graft, and (c) sparse staining in peripheral fields away from the graft. rAAV-lacZ loaded FDL allografts were also transplanted into FDL tendon defects of mice (n = 4). Representative micrographs of one end of the rAAV-lacZ loaded FDL allografts stained with antibodies against -galactosidase at (d) 7 days after transplantation and (e) 14 days after transplantation. It is important to note the lack of viable cells and absence of any staining in the freeze-dried allografts (asterisks) that are surrounded by hypercellular and intensely stained fibrotic tissue. (c) The specificity of the staining was verified by the absence of non-specific staining in negative controls (f, secondary antibody only). S indicates remnants of the repair suture. rAAV, recombinant adeno-associated virus.

Full figure and legend (202K)

Kinetics and biodistribution of tendon allograft–mediated gene delivery

We next set out to determine the kinetics and biodistribution of the tendon allograft–mediated gene delivery in vivo. FDL tendon allografts (3 mm in length) were freeze-dried, reconstituted in a phosphate-buffered saline solution in a vial containing 5 10⁹ particles of rAAV-Luc, and freeze-dried again pending surgical implantation in FDL tendon defects. In order to assess transduction in vivo over time, the mice were imaged on days 3, 7, 14, and 21 after grafting (n = 4 mice) using a real-time bioluminescent imaging system. As hypothesized, the only detectable bioluminescent imaging signal was localized to the site of tendon grafting, further supporting the efficacy of targeted gene delivery using processed tendon allografts. Furthermore, the transduction was transient; the bioluminescent imaging signal peaked at day 7 but persisted, albeit at declining levels, up to 21 days after implantation (Figure 2b). More sensitive analysis (e.g., PCR) will be needed, however, to determine the reporter gene biodistribution in distant tissues and organs.

Figure 2.

Kinetics and biodistribution of recombinant adeno-associated virus (rAAV)-mediated transduction through the use of processed tendon allografts. (a) Temporal bioluminescence images (BLIs) of a representative mouse grafted with a freeze-dried flexor digitorum longus tendon allograft loaded with rAAV-Luc, recorded over 21 days, show the localized biodistribution of rAAV-Luc transduction (heat map-yellow arrows) at the site of allograft implantation in the hind foot. (b) Kinetics of in vivo rAAV transduction, based on average BLI signal intensity, computed from measurements of total integrated light signal (photons emitted/cm²/s) emitted from a standardized region of interest in a standard 3-minute time interval (mean value SEM; n = 4).

Full figure and legend (52K)

Functional verification of rAAV-Gdf5

The complementary DNA for Gdf5 was PCR-amplified and used for creating a plasmid (pAAV-Gdf5), which was then used in the production of the rAAV by a helper virus–free method, and purified as previously described.¹⁰ For verifying the specificity of the vector, we performed reverse-transcribed PCR on pAAV-Gdf5- transfected human embryonic kidney 293 cells, and demonstrated the predicted 485-base pair PCR product (Figure 3a, top). Western blots on culture supernatants from rAAV-Gdf5- infected human embryonic kidney 293 cells also demonstrated the predicted 13.7-kd GDF-5 protein (Figure 3a, bottom). The effects of rAAV-Gdf5 gene delivery were evaluated in vitro using a standard microwound monolayer assay (Figure 3b). These experiments demonstrated that the infection of NIH 3T3 cells with rAAV-Gdf5 leads to accelerated wound healing when compared with the action of rAAV-lacZ- treated controls (Figure 3c). We further estimated the healing time constant and found significant differences between the healing rate associated with the rAAV-Gdf5-treated wells and that of the controls (P < 0.05; Figure 3d). It is likely that the effect of rAAV-Gdf5 in this experiment was masked by the innate ability of the 3T3 cells to proliferate even under serum-deprived, control conditions. Real-time PCR analysis indicated that the accelerated microwound healing rates were attributable to significant early induction of Cyclin D1 and 1-integrin messenger RNA expression, thereby suggesting a synergistic proliferation and migratory effect of rAAV-Gdf5 (data not shown). In parallel experiments, we treated microwound cultures of 3T3 cells with various concentrations of rmGFDF5 protein, and demonstrated a dose-dependent acceleration of healing with the treatment (Figure 3e). Interestingly, the effects of rAAV-Gdf5 delivery on the microwound healing rate were comparable to the effects of bolus delivery of the GDF-5 protein to these cultures.

Figure 3.

Functional verification of the rAAV-Gdf5 vector. Human embryonic kidney 293 (HEK293) cells were grown in 6-well plates and transfected with: (1) pUC19, (2) pSPORT6-Gdf5, or (3) pAAV-Gdf5, and 48 hours later total RNA was harvested from the cells. The messenger RNA was reverse transcribed and used as the template for polymerase chain reaction (PCR) with Gdf5-specific primers. (4) The pSPORT-Gdf5 plasmid was used as template in the positive control. (a, top) The ethidium bromide–stained agarose gel shows the predicted 485-base pair PCR product. HEK293 cells were grown in 6-well plates and infected with the indicated amount of rAAV-lacZ or rAAV-Gdf5 (5.0 10⁷ particles/ml). After 48 hours in culture, the supernatants were collected and 30 l was used for Western blotting with GDF-5-specific antibodies. Ten nanograms of recombinant murine GDF-5 was used as a positive control. (a, bottom) Autoradiography of the Western blot reveals the predicted 13.7-kd GDF-5 protein. Microwound monolayer assay: (b) 80% confluent 3T3 cells were growth-arrested for 24 hours, and then microwounded by passing a pipette tip across the culture well and treated with 0.5% bovine calf serum (BCS) and 5.0 10⁷ particles/ml of either rAAV-lacZ or rAAV-Gdf5. (c) The average width of the defect was digitally measured over time and the wound width normalized to the time zero width [w (t)/w (0)] versus time was plotted. (d) Healing time constants () for the different treatments were computed and plotted as mean values SEM. Note that higher values indicate slower wound healing rates. (e) In a separate experiment, 3T3 cells grown to 80% confluence were microwounded and treated with 0.5% BCS and incremental doses of rmGDF-5. The data presented are mean values SEM for the healing time constant () for the different doses of the GDF-5 protein treatments. Asterisks indicate significant differences (P < 0.01; n = 6 per treatment) compared to untreated controls. GDF-5, growth and differentiation factor 5.

Full figure and legend (38K)

Gdf5-targeted Gene delivery for freeze-dried flexor tendon allografts

In order to investigate whether tendon allografts processed as delivery vehicles for therapeutic genes can reduce adhesions and improve the biomechanical properties of the grafted tendons, we performed experiments with FDL tendon allografts loaded with rAAV-lacZ (controls) or rAAV-Gdf5 (treated) in our murine model. MTP flexion tests (See Supplementary Figure S2) demonstrated that rAAV-Gdf5-loaded allografts were associated with a significantly greater range of joint flexion and a lower gliding coefficient than the lacZ control (P < 0.05; Figure 4) at 14 and 28 days after surgery. The flexion function improved over time between 14 and 28 days after both treatments, but the improvement associated with the rAAV-Gdf5-loaded grafts was still significantly greater than that seen in the lacZ controls. There were also trends of increasing tensile mechanical properties (maximum force and stiffness) over time, but there were no significant differences between the Gdf5-and lacZ-treated grafts in this regard. Tendons from mice killed at 14 days after surgery were removed and fixed, paraffin-embedded, and processed for immunohistochemistry with anti-mouse GDF-5 antibody. The data demonstrate positive anti-GDF-5 staining of host cells (arrows) surrounding the grafts loaded with rAAV-Gdf5, whereas this is absent in the rAAV-lacZ-loaded controls. This finding further validates the efficacy of Gdf5 gene delivery (Figure 5). Next, we histologically examined the implanted allografts at 14 days after surgery (Figure 6). Both Gdf5-treated and lacZ-treated control allografts were surrounded by hypercellular fibrotic tissue at the junction with the host tendon; this could have contributed to impairment of gliding, and consequent reduction in the flexion range of motion (Figure 6c and d). However, there were marked differences in morphology in the middle segment of the grafts. Whereas the rAAV-Gdf5-treated graft was surrounded by organized tissue that resembled neotendon and integrated with the graft (which itself appeared to have been repopulated by cells) (Figure 6f), the rAAV-lacZ control allograft was mostly acellular and was surrounded by disorganized and hypercellular fibrotic tissue (Figure 6e). However, additional assays and immunohistochemistry (for collagen types I and III, for example) are needed in order to confirm these observations.

Figure 4.

rAAV-Gdf5 loading of freeze-dried allografts improves the metatarsophalangeal (MTP) flexion range of motion and the gliding function of reconstructed flexor digitorum longus (FDL) tendons while maintaining their biomechanical properties. Mice had their FDL tendons reconstructed with freeze-dried allografts loaded with rAAV-Gdf5 (treated) or rAAV-lacZ (controls) and killed at 14 and 28 days after surgery (n = 9 per treatment per time point). The operated hind feet were removed and subjected to the MTP flexion test to determine (a) the MTP joint flexion range, and (b) the gliding coefficient. The tendons were then isolated and tested biomechanically to determine (c) their breaking (maximum) tensile force, and (d) their linear tensile stiffness. The data presented are mean values SEM. Asterisks indicate significant differences compared to time-matched controls (P < 0.05). GDF-5, growth and differentiation factor 5; rAAV, recombinant adeno-associated virus.

Full figure and legend (21K)

Figure 5.

rAAV-Gdf5 loading of freeze-dried allografts mediates de novo GDF-5 protein synthesis by the host cells at the periphery of the implanted allograft. Representative immunohistochemical sections of (a) the rAAV-lacZ-loaded and (b) the rAAV-Gdf5-loaded flexor digitorum longus tendon allografts at 14 days after grafting, stained with anti-mouse GDF-5 antibody. It is important to note the matrix-bound GDF-5 (positive staining indicated by arrows), presumably synthesized by the transduced host cells surrounding the rAAV-Gdf5-treated allografts (asterisk), that is absent in the rAAV-lacZ-treated graft. GDF-5, growth and differentiation factor 5; rAAV, recombinant adeno-associated virus.

Full figure and legend (92K)

Figure 6.

rAAV-Gdf5 loading of freeze-dried allografts mediates cellular repopulation of the graft and remodeling of the fibrotic scar tissue. Representative histological sections of (a,c,e) the rAAV-lacZ-loaded and (b,d,f) rAAV-Gdf5-loaded flexor digitorum longus (FDL) tendon allografts at 14 days after grafting, stained with Alcian Blue and Orange G. (a,b) Micrographs at 4 show the implanted grafts with their anatomical relationships to the surrounding tissue. Boxed regions shown in the magnified micrographs (20) depict (c,d) the distal ends of both grafts, and (e,f) the middle segment of the grafts. The tissues represented by numbers are: 1) talus, 2) tarsal bones, 3) metatarsal bone, 4) FDL tendon allograft, and 5) fibrotic/inflammatory tissue. S indicates remnants of suture. (f) Arrows indicate a remodeled tissue that appears to align and integrate with the rAAV-Gdf5-loaded allograft, and that also seems to have been repopulated with host cells compared to e the mostly acellular rAAV-lacZ-loaded allograft. rAAV, recombinant adeno-associated virus.

Full figure and legend (188K)

Discussion

Tendon, ligament, and joint capsular injuries represent 45% of the almost 33 million musculoskeletal injuries that occur each year in the United States, and hand injuries account for 5–10% of annual emergency department visits nationwide.¹¹ While flexor tendon injuries might represent only a small fraction of these numbers, adhesion formation associated with tendon surgery in general is a much more widespread problem.¹² This problem is not limited to primary repair, but has also been reported in response to tendon reconstruction using autografts or allografts¹³ as an inevitable byproduct of the biological cascade of events in tendon healing. It is well established that tendon healing consists of three phases: inflammatory, proliferative, and remodeling. The inflammatory phase involves the recruitment of fibroblasts and macrophages to the site of injury, and phagocytosis of the necrotic tissue. The second phase involves the proliferation of fibroblasts and the formation of a repair scar formed from immature and disorganized collagen matrix. Remodeling then follows in the final phase as the immature collagen fibers in the scar tissue become organized and align with the tendon fibers. While the etiology of adhesion formation has been clinically linked to this last phase,¹ it is hypothesized that adhesions result from post-operative inflammation that extends to surrounding tissue.¹⁴

Clinical and experimental strategies to abrogate adhesion formation can be categorized as either mechanical or biochemical intervention protocols. Mechanical strategies include post-operative passive motion and rehabilitation protocols early after surgery^3,15; optimized surgical techniques that involve minimally traumatic manipulation of the tendon, graft, and surrounding tissues¹⁶; and the use of antiadhesion surface coating of the graft as a physical barrier against adhesion formation.^12,17 Biological and biochemical intervention strategies primarily rely upon growth factor delivery to accelerate the rate of tendon healing and remodeling.^18,19 While a number of growth factors^{19,20,21,22,23,24} could potentially improve the repair of tendons, their effects on tendon adhesion have been left largely unexplored. A rational design for a growth-factor delivery therapy should arguably be based on the natural history of gene expression of growth factors during the different phases of tendon repair, a thorough understanding of the molecular action of these factors, and a sustained delivery mechanism to maximize the therapeutic effects of these factors.

In our mouse FDL tendon–grafting model, we previously observed that that Gdf5 messenger RNA levels were significantly elevated in 28-day autografts.⁶ The increase in Gdf5 levels was concomitant with the observed increases in Vegfa expression, in agreement with findings in other reports.²⁵ Interestingly, the timing of the peak in Gdf5 and Vegfa gene expression was coincidental with the marked improvements in MTP joint flexion observed thereafter.⁶ On the basis of this observation, we hypothesized that GDF-5 plays an important role in the remodeling phase, and that exogenous delivery of paracrine GDF-5 signals will accelerate remodeling and lead to functional improvements in joint flexion. GDF-5 belongs to the category of "cartilage-derived morphogenetic proteins-1" [brachypodism or bone morphogenetic protein-14 (BMP-14)]. The transforming growth factor- subfamily of proteins which, in addition to GDF-5, includes GDF-6 and GDF-7 (cartilage-derived morphogenetic proteins-2 and -3 or BMP-13 and -12), has been shown to be important for skeletal development in general and for tendon formation and repair in particular.^26,27 The effects of GDF-5 on cell recruitment, migration/adhesion, differentiation, proliferation, and angiogenesis in vitro and in vivo,⁷ as well as its effects on the ultrastructure of the collagen fibrils and the biomechanical properties of normal and repaired tendon tissue have been reported.^7,28 However, the exact mechanism of GDF-5 action in tendon repair has yet to be explored. It has been reported that GDF-5 binds with activin receptor–like kinase 3 and/or activin receptor–like kinase-6 (also termed "BMP type IA" and "BMP type IB" receptors, respectively),^29,30 and this presumably activates the Smad signaling pathway. A recent study demonstrated that constitutive activation and nuclear translocation of Smad8 led to upregulation of the Scelraxis transcription factor and promoted tendon formation in C3H10T1/2 cells (a murine multipotent cell line) which, when implanted on a collagen sponge into a rat AT gap tenotomy model, led to formation of tendon-like tissue.³¹ The hypothesis that GDF-5 might be the paracrine signal that leads to Smad8 activation remains an intriguing possibility despite some preliminary observations that contradict this.³¹

Next we were faced with the decision regarding the selection of the GDF-5 delivery mechanism. The therapeutic window of bolus or topical delivery is not long, because the signaling is almost instantly initiated and short-lived.³² Alternatively, local transfer of genes that express the relevant healing factors may mediate sustained expression of these factors. The efficacy of various viral vector systems (including retrovirus and adenovirus) in mediating targeted and transient gene transfer in tendon repair has been demonstrated in vitro and in vivo.^8,33 In a recent study, direct injection of rAAV vectors expressing green fluorescent protein (AdGFP) or BMP-13 (AdBMP-13 or AdGDF-6) into rabbit flexor tendons demonstrated a transient dose-dependent transgene expression up to 12 days in vivo.³⁴ These reports are consistent with our data that demonstrated that rAAV loading of freeze-dried FDL allograft mediates targeted and transient gene expression by host cells at the implant site, with the expression peaking at 7 days but persisting up to 21 days. Notwithstanding the known safety concerns (that are abated by low-dose vector delivery), rAAV-based gene therapy can potentially be a therapeutic option for musculoskeletal tissue (including tendon) reconstruction, in view of the localized and transient expression achieved.

Finally, in the light of our previous findings which suggested that freeze-dried tendon allografts appear to be tolerated well in the host mouse and provide biomechanical scaffolding functions equivalent to those afforded by live autografts,⁶ we examined whether tendon allografts can serve as a delivery scaffolds for therapeutic factors, in order to mediate adhesion-free reconstruction of flexor tendon gap defects. Our data indeed show that, despite the modest retention efficiency (10%), freeze-dried FDL allografts loaded with rAAV-Gdf5 did transduce local expression of the GDF-5 protein at 14 days, and that this was associated with a significantly improved range of flexion as compared to that achieved by rAAV-lacZ controls. While previous studies reported the presence of small foci of bone and fibrocartilage within ectopic tendon/ligament tissue in response to Ad-GDF-6 (or BMP-13) injections in athymic rats,³⁵ we did not observe such untoward effects in our model.

Interestingly, while we observed beneficial functional effects of rAAV-Gdf5 on the grafted tendon, we did not observe any significant effects on the biomechanical properties. Earlier reports have suggested that GDF-5-deficient mice displayed a delay in the accrual of biomechanical strength during the initial healing of the AT as compared to control mice.⁷ On the other hand, Dines et al. (2007), working with a rat model, reported that, at 3 weeks, lacerated ATs that had been repaired with sutures coated with rhGDF-5 showed a greater rate of healing than the repaired tendons in the controls.³⁶ In both studies, the mechanical properties of the controls, GDF-5 deficient, and GDF-5 augmented tendon repairs were equivalent at later time points. These results suggest that the effects of GDF-5 treatment might be temporally sensitive and dependent on the healing phase. However, our results are not different from those of similar gene therapy–based tendon repair studies. Rickert et al. (2005) reported efficacious adenovirus-mediated transfer of Gdf5 gene by using injections of AdGDF-5 at the site of lacerations in the rat AT, but this was not associated with any significant improvements in mechanical properties.⁸ This was in contrast to the previous report by Lou and co-workers (2001) that AdGDF-7 injections at the site of lacerations in chicken flexor digitorum profundus tendon results in delayed but significant onefold improvements in mechanical properties at 6 weeks after treatment.³⁷ It is therefore possible that other isoforms of the growth and differentiation factor, such as GDF-7, have more potent effects on the mechanical properties of tendon tissue that is undergoing repair. Other possibilities cannot be excluded. The lack of improvement in mechanical properties in the rAAV-Gdf5-treated allografts compared to the controls, as observed by us, may be related to: (i) the dosage used for the treatment (number of rAAV particles transferred), which might have to be optimized in future studies; (ii) the efficiency of rAAV-mediated gene transfer; (iii) the absence of interactions with in vivo forces in our model; and (iv) the observation that the transfected host cells resided in the external callus, resulting in remodeling of the fibrotic callus tissue, a reduction in adhesions, and improved gliding function, but not necessarily any remodeling of the graft tissue proper.

On the basis of these findings, we propose a simplified alternative paradigm in tissue engineering, using freeze-dried allograft tissue to deliver cues to the host cells in situ to reprogram the repair response. Freeze-dried tendon allografts can provide these delivery functions with a number of desirable characteristics that may be unavailable with synthetic and naturally derived biomaterials. Freeze-dried tendon allografts are biochemically unaltered, because the lyophilization is purely a physical process that leads to dehydration of the tissue. These allografts potentially have an indefinite shelf-life and will likely have less regulatory hurdles to clear en route to clinical applications because they can still be classified as "allografts". Furthermore, freeze-dried tendon allografts have biomechanical properties equivalent to fresh or fresh-frozen tendon tissue. Despite being devoid of live cells (which actually confers on them the advantage of not eliciting an immune response leading to graft rejection), they can be readily remodeled and populated by host cells when implanted in vivo. Most important, freeze-dried tendon allografts have remarkable native hydrophilic properties that permit efficient reconstitution of the tissue in a physiologic solution containing therapeutic molecules. This concept could be applied to other tendon and ligament models including the anterior cruciate ligament, the AT, and the supraspinatus "rotator cuff" tendon, and could involve loading gene delivery vectors or recombinant or tissue-derived growth factors (see Supplementary Figure S3). Future developments could also focus on differential processing (multiple genes and proteins) of composite allograft tissue (bone–tendon–bone) so as to address clinically challenging problems such as soft tissue insertion into bone. Furthermore, such technology can potentially be translatable to other musculoskeletal soft tissue models, including articular cartilage and meniscus tissues.

Materials and Methods

rAAV preparation. rAAV-lacZ and rAAV-Luc stock solutions were purchased, and the single stranded rAAV-Gdf5 vector (serotype 2), which was custom cloned from an existing plasmid (pAAV-Gdf5) containing a cytomegalovirus promoter and the Gdf5 complementary DNA, was purified and titered at the Gene Therapy Center of the University of North Carolina, Chapel Hill, North Carolina, USA.

Processing of tendon allografts. FDL tendon allografts were aseptically isolated, placed in sterile vials, frozen at –80 °C, and freeze-dried. In order to load the tendon grafts with rAAV vectors, pairs of tendons were soaked in a vial containing 50 l of phosphate-buffered saline solution containing 5 10⁹ U of rAAV (lacZ, Luc, or Gdf5). After allowing the dehydrated grafts to take up the solution for 1 hour, the grafts were snap-frozen and then freeze-dried and stored awaiting experimental use.

Real-time quantitative PCR assessment of rAAV retention in the allografts. rAAV-lacZ-loaded FDL tendon grafts were digested in a buffer solution of proteinase K (10 g/ml Applied Biosystems, Foster City, CA) at 50 °C for 1 hour, and then at 95 °C for 20 minutes to deactivate the enzyme. Samples from a serial dilution of digested virus at standard concentrations of 10¹⁰, 10⁹, 10⁸, 10⁷, 10⁶, 10⁵, and 10⁴ U were used for creating a standard curve. Duplicate samples (2 l) of each standard dilution, along with samples from tendon digests, were added to real-time PCR Master Mix (SYBR Green PCR Master Mix; Applied Biosystems, Foster City, CA) and allowed to react in a Rotor-Gene 2000 Real-Time DNA detection system (Corbett Research, Sydney, Australia) for 40 cycles on a program of 94 °C for 20 seconds, 61.6 °C for 30 seconds, and 72 °C for 30 seconds. The mean cycle threshold (C_t) values were used for calculating the rAAV content and retention efficiency in the tendon samples, on the basis of the standard curve.

Microwound experiments. The microwound assay was performed as previously described by Hocking and Chang.^38,39 Briefly, mouse embryonic fibroblast (NIH3T3; American Type Culture Collection, Manassas, VA) cells were plated and allowed to grow to 80% confluence. The cells were serum-deprived for 24 hours prior to creating wounds. Using a 100-l pipette tip, wounds were created by scratching the pipette tip across the monolayer, resulting in wounds initially measuring 1.00 mm (0.20). The cells were then cultured with 0.5% bovine calf serum (American Type Culture Collection, Manassas, VA) and 5.0 10⁷ particle U/ml of either rAAV-lacZ or rAAV-Gdf5. Digital photos of the microwound were taken at 0, 2, 4, 8, 12, 24, 36, and 48 hours. Using a custom Matlab program, the average width of each wound was measured at each time and normalized against the initial wound width (w (t)/w (0)). The data were fitted to the equation w (t)/w (0) = A/(B exp(t /) + 1) wherein represents the "healing time constant" such that wounds that heal faster have a lower healing time constant.

Bioluminescent imaging. In order to demonstrate the efficacy of processed tendon allograft-mediated gene delivery, freeze-dried allografts loaded with rAAV-Luc were implanted to reconstruct mouse FDL tendons as described later in this report. Host cells transduced by this virus express the firefly luciferase gene. At each time point following implantation, we injected each mouse with the substrate D-luciferin potassium salt (Xenogen, Cranbury, NJ) which, when cleaved by the transduced luciferase enzyme, emits light that can be captured using a special camera system and software (IVIS 100 Bioluminescent Imaging System, Xenogen, Cranbury, NJ) and the bioluminescence intensity gradients can be represented by a heat map intensity (purple, least intense; red, most intense) computed from measurements of the total integrated light signal (photons emitted/cm²/s) emitted from a standardized region of interest in a standard time interval (3-minute exposure).

Mouse FDL tendon-grafting surgeries. Animal studies were conducted in compliance with principles and procedures approved by the University of Rochester Committee for Animal Resources. Surgeries were performed using an aseptic technique under a 2 microdissection magnifying lens. Briefly, a longitudinal plantar incision was made on the left hind foot. The distal FDL tendon of the C57Bl/6 mouse (Jackson Laboratories, Bar Harbor, ME) was isolated and transected on the plantar surface of the metatarsal bones. A 3-mm freeze-dried tendon allograft, that had been previously removed from a C57Bl/6 mouse and loaded with rAAV, was reconstituted in phosphate-buffered saline and sutured between the ends of the host tendon, using an 8-0 nylon suture in a horizontal mattress suture pattern (similar to a modified Kessler technique). The proximal tendon insertion into the flexor muscle was severed to eliminate early gliding, so as to protect the repair during the early phases of healing and to induce adhesion formation. The skin was closed with 4-0 silk suture.

MTP joint flexion test. Immediately after each mouse was killed, the lower hind limbs were disarticulated from the knee and the proximal FDL tendon along the tibia was released just proximal to the tarsal tunnel without disrupting the skin at the ankle or foot. The proximal end of the tendon was secured between two square pieces of tape using a thin layer of cyanoacrylate, as previously described.³⁶ The lower hind limb was fixed in a custom apparatus, with the tibia rigidly gripped to prevent rotation. In order to standardize the neutral position, the toes of the mouse were passively extended by the examiner and allowed to return to an unloaded position before a digital image was taken medially to determine the neutral position (zero load) of the MTP joint. The FDL tendon was incrementally loaded in the same anatomical direction as the flexor muscle line of force, using dead weights that were statically suspended from a hook and line passing through the proximal FDL tendon/tape composite. The dead weights were suspended for 30 seconds before the digital pictures were taken, so as to avoid "creep" effects. With each increment of load, a digital image was taken to quantify the MTP flexion angle relative to the neutral position. The MTP joint flexion angles were measured from the digital images by two independent observers (P.B. and T.D.) using ImageJ software (http://rsb.info.nih.gov/ij/), and plotted against the applied loads. The flexion data were fitted to a single-phase exponential association equation of the form: MTP flexion angle = [1 – exp(–m/)]; where m is the applied load (Prism GraphPad 3.0; GraphPad Software, San Diego, CA). The curve fit was constrained to the maximum flexion angle () for normal tendons that was previously determined to be 75° for the maximum applied load. The constant (gliding coefficient) that governs the rate of rise of the flexion curve with loading was determined by non-linear regression as a measure of the resistance to MTP flexion on account of impaired gliding.

Biomechanical test. Following the MTP flexion test, the proximal end of the FDL tendon was released at the tarsal tunnel, and dissected medially along the bone. Once the tendon was free from the tunnel, the calcaneus was removed, freeing the proximal end of the tendon for direct gripping in the mechanical test, as described elsewhere.⁴⁰ The distal bones of the foot were directly gripped in custom grips without disrupting the graft or the branching tendon insertion into the phalanges. The specimen was placed in sterile gauze soaked with saline to maintain adequate tissue hydration until tested. The FDL tendon was then mounted on the Instron 8841 DynaMight axial servohydraulic testing system (Instron, Norwood, MA) and tested using published protocols.³⁶ The tendon was loaded in tension in displacement control at a rate of 30 mm/minute until failure. Force–displacement data were automatically logged and plotted, and the maximum tensile force and stiffness were determined.

Histology and immunohistochemistry. The grafted limbs were removed by disarticulating the tibia from the knee joint. With the tibia perpendicular to the foot, the FDL tendon was kept in tension by passing a pin through the flexor muscles and the tibia. The tissues were then prepared for histology and analyzed using routine techniques. Briefly, the removed limbs were fixed in 10% neutral-buffered formalin and decalcified in 10% EDTA at 4 °C for 21 days. The decalcified tissues were dehydrated in a gradient of alcohols and then embedded en bloc in paraffin. Serial 3-m sagittal sections through the FDL tendon plane were prepared and stained with Alcian Blue and Orange G. For immunohistochemistry, the rAAV-loaded tendon sections were stained with primary antibodies against -galactosidase (PAb # GTX26646, GeneTex, San Antonio, TX) or against the murine GDF-5 (AF853; R&D Systems, Minneapolis, MN). The tissue sections were then treated with appropriate biotin-conjugated secondary antibodies, before being developed with streptavidin-conjugated AEC chromogen (Zymed Laboratories, San Francisco, CA).

Statistical analysis. Data analysis including analysis of variance with Bonferroni post-hoc multiple comparisons ( = 0.05) and the non-linear regression analyses to estimate the gliding coefficient from the MTP flexion data were performed using Prism GraphPad 4.0 statistical software.

References

REFERENCES

1. Lilly, SI and Messer, TM (2006). Complications after treatment of flexor tendon injuries. J Am Acad Orthop Surg 14: 387–396. | PubMed |
2. Gelberman, RH and Manske, PR (1985). Factors influencing flexor tendon adhesions. Hand Clin 1: 35–42. | PubMed | ChemPort |
3. Silva, MJ, Boyer, MI and Gelberman, RH (2002). Recent progress in flexor tendon healing. J Orthop Sci 7: 508–514. | Article | PubMed |
4. Leddy, JP (1988). Flexor tendons—acute injuries. In: Green, DP (ed.). Operative Hand Surgery vol. 3. Churchill Livingstone: New York. pp. 1935–1968.
5. Whitlock, PW, Smith, TL, Poehling, GG, Shilt, JS and Van Dyke, M (2007). A naturally derived, cytocompatible, and architecturally optimized scaffold for tendon and ligament regeneration. Biomaterials 28: 4321–4329. | Article | PubMed | ChemPort |
6. Hasslund, S, Jacobson, JA, Dadali, T, Basile, P, Ulrich-Vinther, M, Soballe, K et al. (2007). Adhesions in a murine flexor tendon graft model: autograft versus allograft reconstruction. J Orthop Res (in the press).
7. Chhabra, A, Tsou, D, Clark, RT, Gaschen, V, Hunziker, EB and Mikic, B (2003). GDF-5 deficiency in mice delays Achilles tendon healing. J Orthop Res 21: 826–835. | Article | PubMed | ChemPort |
8. Rickert, M, Wang, H, Wieloch, P, Lorenz, H, Steck, E, Sabo, D et al. (2005). Adenovirus-mediated gene transfer of growth and differentiation factor-5 into tenocytes and the healing rat Achilles tendon. Connect Tissue Res 46: 175–183. | Article | PubMed | ChemPort |
9. Ito, H, Koefoed, M, Tiyapatanaputi, P, Gromov, K, Goater, JJ, Carmouche, J et al. (2005). Remodeling of cortical bone allografts mediated by adherent rAAV-RANKL and VEGF gene therapy. Nat Med 11: 291–297. | Article | PubMed | ISI | ChemPort |
10. Koefoed, M, Ito, H, Gromov, K, Reynolds, DG, Awad, HA, Rubery, PT et al. (2005). Biological effects of rAAV-caAlk2 coating on structural allograft healing. Mol Ther 12: 212–218. | Article | PubMed | ISI | ChemPort |
11. Praemer, A, Furner, S and Rice, D (1999). Musculoskeletal Conditions in the United States. American Academy of Orthopaedic Surgeons: Rosemont, IL.
12. Meislin, RJ, Wiseman, DM, Alexander, H, Cunningham, T, Linsky, C, Carlstedt, C et al. (1990). A biomechanical study of tendon adhesion reduction using a biodegradable barrier in a rabbit model. J Appl Biomater 1: 13–19. | Article | PubMed | ChemPort |
13. Zhao, C, Sun, YL, Amadio, PC, Tanaka, T, Ettema, AM and An, KN (2006). Surface treatment of flexor tendon autografts with carbodiimide-derivatized hyaluronic acid. An in vivo canine model. J Bone Joint Surg Am 88: 2181–2191. | Article | PubMed |
14. Cooper, MA, Laverty, PH and Soiderer, EE (2005). Bilateral flexor tendon contracture following onychectomy in 2 cats. Can Vet J 46: 244–246. | PubMed |
15. Boyer, MI, Goldfarb, CA and Gelberman, RH (2005). Recent progress in flexor tendon healing. The modulation of tendon healing with rehabilitation variables. J Hand Ther 18: 80–85. | PubMed |
16. Hatano, I, Suga, T, Diao, E, Peimer, CA and Howard, C (2000). Adhesions from flexor tendon surgery: an animal study comparing surgical techniques. J Hand Surg [Am] 25: 252–259. | Article |
17. Mentzel, M, Hoss, H, Keppler, P, Ebinger, T, Kinzl, L, Wachter, NJ et al. (2000). The effectiveness of ADCON-T/N, a new anti-adhesion barrier gel, in fresh divisions of the flexor tendons in Zone II. J Hand Surg [Br] 25: 590–592. | Article |
18. Towler, DA and Gelberman, RH (2006). The alchemy of tendon repair: a primer for the (S)mad scientist. J Clin Invest 116: 863–866. | Article | PubMed | ChemPort |
19. Hsu, C and Chang, J (2004). Clinical implications of growth factors in flexor tendon wound healing. J Hand Surg [Am] 29: 551–563. | Article |
20. Thomopoulos, S, Harwood, FL, Silva, MJ, Amiel, D and Gelberman, RH (2005). Effect of several growth factors on canine flexor tendon fibroblast proliferation and collagen synthesis in vitro. J Hand Surg [Am] 30: 441–447. | Article |
21. Gelberman, RH, Thomopoulos, S, Sakiyama-Elbert, SE, Das, R and Silva, MJ (2007). The early effects of sustained platelet-derived growth factor administration on the functional and structural properties of repaired intrasynovial flexor tendons: an in vivo biomechanic study at 3 weeks in canines. J Hand Surg [Am] 32: 373–379. | Article |
22. Wang, XT, Liu, PY and Tang, JB (2005). Tendon healing in vitro: modification of tenocytes with exogenous vascular endothelial growth factor gene increases expression of transforming growth factor beta but minimally affects expression of collagen genes. J Hand Surg [Am] 30: 222–229. | Article |
23. Abrahamsson, SO and Lohmander, S (1996). Differential effects of insulin-like growth factor-I on matrix and DNA synthesis in various regions and types of rabbit tendons. J Orthop Res 14: 370–376. | Article | PubMed | ChemPort |
24. Kashiwagi, K, Mochizuki, Y, Yasunaga, Y, Ishida, O, Deie, M and Ochi, M (2004). Effects of transforming growth factor-beta 1 on the early stages of healing of the Achilles tendon in a rat model. Scand J Plast Reconstr Surg Hand Surg 38: 193–197. | Article | PubMed |
25. Zeng, Q, Li, X, Beck, G, Balian, G and Shen, FH (2007). Growth and differentiation factor-5 (GDF-5) stimulates osteogenic differentiation and increases vascular endothelial growth factor (VEGF) levels in fat-derived stromal cells in vitro. Bone 40: 374–381. | Article | PubMed | ChemPort |
26. Mikic, B (2004). Multiple effects of GDF-5 deficiency on skeletal tissues: implications for therapeutic bioengineering. Ann Biomed Eng 32: 466–476. | Article | PubMed |
27. Wolfman, NM, Hattersley, G, Cox, K, Celeste, AJ, Nelson, R, Yamaji, N et al. (1997). Ectopic induction of tendon and ligament in rats by growth and differentiation factors 5, 6, and 7, members of the TGF-beta gene family. J Clin Invest 100: 321–330. | PubMed | ISI | ChemPort |
28. Clark, RT, Johnson, TL, Schalet, BJ, Davis, L, Gaschen, V, Hunziker, EB et al. (2001). GDF-5 deficiency in mice leads to disruption of tail tendon form and function. Connect Tissue Res 42: 175–186. | PubMed | ChemPort |
29. Schreuder, H, Liesum, A, Pohl, J, Kruse, M and Koyama, M (2005). Crystal structure of recombinant human growth and differentiation factor 5: evidence for interaction of the type I and type II receptor-binding sites. Biochem Biophys Res Commun 329: 1076–1086. | Article | PubMed | ChemPort |
30. Aoki, H, Fujii, M, Imamura, T, Yagi, K, Takehara, K, Kato, M et al. (2001). Synergistic effects of different bone morphogenetic protein type I receptors on alkaline phosphatase induction. J Cell Sci 114: 1483–1489. | PubMed | ISI | ChemPort |
31. Hoffmann, A, Pelled, G, Turgeman, G, Eberle, P, Zilberman, Y, Shinar, H et al. (2006). Neotendon formation induced by manipulation of the Smad8 signalling pathway in mesenchymal stem cells. J Clin Invest 116: 940–952. | Article | PubMed | ISI | ChemPort |
32. Eming, SA, Krieg, T and Davidson, JM (2004). Gene transfer in tissue repair: status, challenges and future directions. Expert Opin Biol Ther 4: 1373–1386. | Article | PubMed | ISI | ChemPort |
33. Gerich, TG, Kang, R, Fu, FH, Robbins, PD and Evans, CH (1997). Gene transfer to the patellar tendon. Knee Surg Sports Traumatol Arthrosc 5: 118–123. | Article | PubMed | ChemPort |
34. Mehta, V, Kang, Q, Luo, J, He, TC, Haydon, RC and Mass, DP (2005). Characterization of adenovirus-mediated gene transfer in rabbit flexor tendons. J Hand Surg [Am] 30: 136–141. | Article |
35. Helm, GA, Li, JZ, Alden, TD, Hudson, SB, Beres, EJ, Cunningham, M et al. (2001). A light and electron microscopic study of ectopic tendon and ligament formation induced by bone morphogenetic protein-13 adenoviral gene therapy. J Neurosurg 95: 298–307. | PubMed | ChemPort |
36. Dines, JS, Weber, L, Razzano, P, Prajapati, R, Timmer, M, Bowman, S et al. (2007). The effect of growth differentiation factor-5-coated sutures on tendon repair in a rat model. J Shoulder Elbow Surg 16: S215–S221. | Article | PubMed |
37. Lou, J, Tu, Y, Burns, M, Silva, MJ and Manske, P (2001). BMP-12 gene transfer augmentation of lacerated tendon repair. J Orthop Res 19: 1199–1202. | Article | PubMed | ChemPort |
38. Hocking, DC and Chang, CH (2003). Fibronectin matrix polymerization regulates small airway epithelial cell migration. Am J Physiol Lung Cell Mol Physiol 285: L169–L179. | PubMed | ChemPort |
39. Maffulli, N, Ewen, SW, Waterston, SW, Reaper, J and Barrass, V (2000). Tenocytes from ruptured and tendinopathic achilles tendons produce greater quantities of type III collagen than tenocytes from normal Achilles tendons. An in vitro model of human tendon healing. Am J Sports Med 28: 499–505. | PubMed | ChemPort |
40. Mikic, B, Schalet, BJ, Clark, RT, Gaschen, V, Hunziker, EB (2001). GDF-5 deficiency in mice alters the ultrastructure, mechanical properties and composition of the Achilles tendon. J Orthop Res 19: 365–371. | Article | PubMed | ChemPort |

Acknowledgments

P.B. and T.D. share first authorship as they contributed equally to the work. The authors thank Brendan Boyce for valuable advice and stimulating discussions, Denise Hocking for advice on the in vitro microwound assay, Gwynne Bragdon for advice on surgeries, Krista Scorsone for technical assistance with histology, David Reynolds for technical assistance with biomechanical testing, and Tony Chen for help with data analysis (all from the University of Rochester Medical Center). This work was supported in part by research grants from the National Institutes of Health (PHS AR054041, AR051469, DE017096), Whitaker Foundation, the Danish Medical Research Council, the Musculoskeletal Transplant Foundation, the Orthopaedic Research Education Foundation (OREF), and DePuy J&J.