Sixty-four white leghorn chickens, weighing 1.5 to 1.8 kg each, were used as the experimental model because they have a flexor mechanism similar to that of human digits28,29. The Institutional Animal Care and Use Committee of Nantong University Medical School approved the use of the animals in this study. The chickens were anesthetized by means of an intramuscular injection of ketamine (50 mg/kg of body weight). The chicken feet were scrubbed with povidone-iodine, and the operation was performed under tourniquet control. The tendon repairs were carried out with the aid of an operating microscope at six times magnification. The AAV2-bFGF vector system was developed in our laboratories, and the effectiveness of this vector in the promotion of matrix production was tested in an in vitro study27 and was further verified in an in vivo pilot study.
AAV2-bFGF Vector Production
Genetic cloning of the bFGF gene and construction of an AAV2-bFGF gene transfer unit were detailed in a previous publication27. Briefly, the AAV2-bFGF gene transfer unit was constructed with a series of genetic clonings followed by the production of infectious AAV2-bFGF viral particles. A segment containing the bFGF gene preserved in a plasmid was cut through the EcoRI restriction site (sequence, AATTC) and the XhoI restriction site (sequence, CTCGAG) and was cloned to the pBluescript II KS vector (Stratagene, La Jolla, California) at the EcoRI and XhoI restriction sites to lengthen the segment. Subsequently, a 2.7-kb fragment containing the bFGF gene sequences was cut from the pBluescript II KS vector and was ligated to the multiple cloning site of a pAAV-MCS vector (Stratagene) with use of the LigaFast Rapid DNA Ligation System (M8221; Promega, Madison, Wisconsin) to produce pAAV2-bFGF (Fig. 1). The pAAV2-bFGF harbors a nuclear localization signal under the regulation of the cytomegalovirus immediate early promoter30. The promoter and the bFGF gene inserted at the multiple cloning site are flanked by left and right AAV2 inverted terminal repeats (ITRs), the sequences directing viral duplication.
The plasmid AAV2-bFGF obtained from the above preparation then cotransfected a packaging cell line, human embryonic kidney (HEK) 293 cells, with a helper plasmid (pHelper vector) and a pAAV-RC vector. The former carries the subset of adenovirus genes. The latter carries subtype 2 AAV rep and cap genes but lacking the AAV ITRs. After seventy-two hours, the cells were harvested and lysed in Tris-HCl buffer through cycles of freezing and thawing, sucrose precipitation, and CsCl density-gradient ultracentrifugation to isolate AAV2-bFGF. The particle titer of the AAV2-bFGF was determined by means of real-time polymerase chain reaction (Prism 7000 Sequence Detection System; Applied Biosystems, Foster City, California). AAV2-luciferase containing the firefly luciferase gene was produced in a similar way, serving as sham vector, because the luciferase gene does not produce biologically active protein in other animals.
The efficacy of transduction of the chicken tendons by the AAV2-bFGF was tested in a pilot study by injecting the vectors into eight flexor digitorum profundus tendons of four chickens at a dose of 2 × 109 viral particles per tendon and injecting saline solution into the other eight tendons (Fig. 1, E). Four tendons that had been injected with the AAV2-bFGF and four that had been injected with saline solution were harvested at two and four weeks after surgery, respectively. The harvested tendons were fixed in formalin and embedded in paraffin, and sections were stained immunohistochemically with rabbit anti-rat bFGF antibody (AB1459; Chemicon, Temecula, California) at a dilution of 1:4000, observed, and photographed with an image-analysis system (Leica Q550IW; Leica Microsystems, Wetzler, Germany). The control sections underwent identical staining procedures but were stained with either phosphate-buffered saline solution or rabbit serum instead of rabbit anti-rat bFGF antibody.
Experimental Groups and Operation
A zig-zag incision was made in the plantar skin of the long toes of both feet between the proximal interphalangeal and distal interphalangeal joint levels to expose the flexor tendon system. The anatomy of the chicken flexor tendon systems in the long toes resembles that of the human fingers8. In chicken toes, a major annular pulley distal to the proximal interphalangeal joint is the equivalent of the human A2 pulley29. Two flexor tendons exist between the proximal interphalangeal and distal interphalangeal joints. No vincula insert into the tendons under this pulley. These features are similar to those of flexor tendons in the area covered by the A2 pulley (zone IIC) (Fig. 1, D)8,29. Complete laceration of flexor tendons in this area represents the worst-case scenario in terms of the rupture of repaired tendons or the formation of adhesions.
A total of 104 toes from fifty-two chickens that were to undergo surgery and to be evaluated at two, four, eight, and twelve weeks postoperatively were divided into three groups of thirty-eight, thirty-eight, and twenty-eight. The assignment of chickens and their long toes to groups is detailed in Figure 2. The chickens were first randomized into four clusters that corresponded to each of the four postoperative evaluation time-points. Within each cluster of chickens, the long toes of the left and right feet of the chickens were randomly assigned to the three study groups. The flexor digitorum superficialis tendons were exposed through a longitudinal cut of 1.5 cm centered over the pulley. A 1.5-cm segment of the flexor digitorum superficialis tendon was resected to provide exposure to the flexor digitorum profundus tendon. Each flexor digitorum profundus tendon was transected completely with a sharp surgical blade and was subjected to one of three treatments. In Group 1 (the AAV2-bFGF injection group; thirty-eight toes), 0.5 × 109 particles of AAV2 containing the bFGF gene were injected to each of four sites in both stumps of the cut tendon ends before repair, yielding a total injected dose of 2 × 109 in each flexor digitorum profundus tendon; the method of injection of vectors is detailed in the legend of Figure 1. In Group 2 (the AAV2-luciferase injection group; twenty-eight toes), AAV2-luciferase in the same number of viral particles was injected into the tendon stumps in a manner identical to that in Group 1. In Group 3 (the non-injection group; thirty-eight toes), the tendons did not receive any injection prior to the repair. The AAV2-luciferase injection group had fewer samples than the other two groups because we did not include it at the twelve-week-evaluation time-point.
Immediately after the above treatments, each flexor digitorum profundus tendon was repaired with use of the modified Kessler method4 with 5-0 suture (ETHILON; Ethicon, Somerville, New Jersey). A single surgeon repaired all of the tendons; tension and the length of core suture purchase were kept identical in all tendons. Peripheral sutures were not added. The incised skin was closed surgically, but the sheath was not closed. After surgery, the toes were immobilized in semiflexed position with adhesive tape for the first three weeks and were checked regularly. They were released to allow free motion thereafter. Two chickens (one each in Groups 1 and 3) died during the postoperative course and were replaced with an additional two animals.
The flexor digitorum profundus tendons from twenty toes of ten other chickens served as time-zero controls. Ten toes were tested for the strength of the repaired flexor digitorum profundus tendon immediately after surgery; the other ten toes were not injured and were tested with regard to the work of toe flexion to provide baseline data. The surgical approach used for the ten toes tested for the time-zero repair strength was identical to that in the above three groups.
Evaluation
We evaluated the outcomes in the three treatment groups at four postoperative time-points ranging from two to twelve weeks, with an emphasis on variables reflecting both healing status and adhesion formation at early and late healing stages. Two and four weeks corresponded with the early healing period, and eight and twelve weeks corresponded with the late healing period. At two and four weeks, all of the flexor digitorum profundus tendons were first evaluated morphologically with regard to their healing status and adhesion formation and were then tested to determine the load-to-failure strength. At eight and twelve weeks, the adhesions had matured and solid tendon healing was achieved; in all tendons, we first measured the work of toe flexion, then scored the severity of adhesions, and finally tested the tendon strength.
Tendon Morphology and Biomechanical Testing at Two and Four Weeks
Tendon Morphology and Adhesions
The tendon was exposed through a volar longitudinal incision through the entire length of the long toe. The morphological appearance of the tendon, granulation tissue, and adhesions around the tendon was observed under the operating microscope. The diameter of the tendon, the smoothness of the tendon surface, and gapping between the tendon stumps were used to assess the healing status macroscopically. The length and density of the adhesions were evaluated, and the adhesions were scored according to established grading criteria (Table I)29. At two weeks, the adhesions were immature, and a mixture of granulation tissue and primitive adhesions was present around the tendons. At four weeks, the adhesions were more mature. The scores for each toe were a combination of the points derived from the quantitative and qualitative features of the adhesions.
Ultimate Strength of the Healing Tendons
We harvested the flexor digitorum profundus tendon through its entire length from the distal part of the leg to its insertion into the distal phalanx and used a materials testing machine (model 4411; Instron, Norwood, Massachusetts) to measure the load-to-failure strength of the tendon. The distal phalanx with its attachment of the terminal flexor digitorum profundus tendon was preserved during dissection and was mounted in the lower clamp of the testing machine. The proximal tendon end was mounted in a clamp. The total length of the tendon was 8 cm between the two clamps. The repair site of the flexor digitorum profundus tendon was maintained at the middle of the tested tendon segment. The mounting of the flexor digitorum profundus tendon was secure, and pilot studies confirmed that there was no slipping or tearing of the tested tendon within the clamps at the loading speed used in the test. An overhead crossbar secured to the upper clamp was connected to a force transducer on the testing machine and was advanced at a constant speed of 25 mm/min until the repaired tendon failed completely. This loading speed was intended to imitate a loading rate in the human hand during gentle finger flexion, and this speed has been used in previous studies in the chicken model29,31,32. During the testing, the load of the tendons was continuously measured by the machine with use of a testing software program (Series IX; Instron). The ultimate failure of the repair was indicated by a sharp decline in load displacement as shown on the monitor and abrupt disruption of the repair site. The forces were measured to the nearest 0.1 N.
Biomechanical Testing and Scoring of Adhesions at Eight and Twelve Weeks
At these two evaluation time-points, we first measured the energy required to simulate active toe flexion over a fixed range of angulation (the work of toe flexion), then scored the severity of the adhesions, and finally tested the load-to-failure strength of the repaired tendons.
Work of Toe Flexion
The toes were disarticulated through the knee joint and were mounted to the measurement board attached to the lower clamp of the materials testing machine. The proximal phalanx, the metatarsal, and the bones of the distal part of the leg were fixed with Kirschner wires to the board, but the distal three phalanges were not restricted. The distal distal interphalangeal joint (the most distal interphalangeal joint of the chickens28,29) and distal interphalangeal joint were immobilized with a cast, but the proximal interphalangeal joint was free to move. The proximal end of the flexor digitorum profundus tendon to the long toe was connected to the upper clamp of the testing machine (model 4411; Instron) and was pulled proximally until the proximal interphalangeal joint reached 70° (Fig. 3). The angle of flexion was controlled by means of an attached goniometer. A computer simultaneously recorded the gliding excursion and force applied to the proximal end of the flexor digitorum profundus tendon. The force was plotted against the excursion. The area under the force-excursion curve represents the work of the forces that resist tendon gliding31,33-35. We consistently measured the work of toe flexion during the first run of toe flexion because the energy required to flex the toes differs among runs and the first run was considered to generate the most accurate data33,34.
Adhesion Scores
Just as was done for the tendons that were evaluated at two and four weeks, the repaired tendons were exposed through plantar longitudinal incision. The length and density of the adhesions around the flexor digitorum profundus tendon were evaluated under the operating microscope, and the adhesions were scored according to the grading criteria. The scores in each group of eight or ten toes were used for comparison.
Ultimate Strength of the Tendon
With an approach of sample harvesting that was identical to that used at two and four weeks, the flexor digitorum profundus tendon was obtained, freed from other tissues and adhesions, and tested in the materials testing machine. The mounting method, loading speed, and test software used in this step were the same as those used at two and four weeks.
Statistical Analysis
The data on ultimate strength and the work of flexion of the toes obtained from the mechanical tests were analyzed statistically by means of one-way repeated-measures analysis of variance. When analysis of variance indicated significant differences in these data, the Tukey test was used to determine the differences between groups. Similarly, for the recorded scores of adhesions, the Tukey test was used to determine differences. As the present study made use of data from more than one toe from the same chickens, the analysis had to take into account clustering effects due to the use of toes from the same chickens. In order to correct for the effects of clustering, general estimating equation models were used in all analyses in this study, which allowed us to improve standard error and parameter estimates by controlling for clustering. In all cases, two-tailed statistical analysis was used. The level of significance was set at p < 0.05.
A power analysis was performed to ensure appropriate sample sizes in the experimental arms. We assumed that substantive biological effects of gene therapy of the tendon should at least increase the healing strength by 25% or decrease the energy of digital flexion by 30% of those without treatment. We determined that the minimum sample size would be eight for the test of the strength and nine for the test of the work at each time-point, with the desired statistical power of 0.80, based on another pilot study aiming at determination of an effective dose of AAV2-bFGF delivered to the flexor digitorum profundus tendon in order to bring about substantive biological effects, rather than statistically significant yet still not biologically substantive effects.
The results of the pilot study to determine the increase in the amount of bFGF in the tendon receiving gene therapy demonstrated more immunoreactive bFGF staining in the tendons injected with AAV2-bFGF as compared with saline solution-injection controls at both two and four weeks after surgery.
Tendon Rupture Rate
The postoperative rupture rate was determined when the animals were killed at each evaluation time-point, and all ruptured tendons were found to have failed at the site of the repair. Overall, postoperative rupture of the repaired tendon was noted in one tendon in the AAV2-bFGF group, three tendons in the AAV2-luciferase group, and three tendons in the non-injection control group. Ruptures of the tendons in each group as recorded at four time-points are detailed in Table II. The rupture rate was lower for AAV2-bFGF group (2.6%) than for the AAV2-luciferase group (10.7%) or the non-injection control group (7.9%).
Ultimate Strength
At the end of two weeks, the ultimate strength (given as the mean and the standard deviation) of the repaired flexor tendons in the AAV2-bFGF group (10.9 ± 2.0 N) was significantly greater than that of the tendons in the AAV2-luciferase group (7.8 ± 0.7 N) (p < 0.01) and that of the tendons in the non-injection control group (6.3 ± 2.6 N) (p < 0.001). No significant difference in ultimate strength was found between the tendons in the AAV2-luciferase group and those in the non-injection control group. Similarly, at four weeks, the strength of tendons in the AAV2-bFGF group (8.9 ± 1.9 N) was significantly greater than that of tendons in the AAV2-luciferase group (6.1 ± 1.0 N) (p < 0.01) and those in the non-injection control group (5.7 ± 1.1 N) (p < 0.001) (Fig. 4).
At the later healing period (eight weeks), the strength of the repaired flexor tendons in the AAV2-bFGF group (84.8 ± 22.0 N) was increased compared with the value recorded during the early period, and it continued to be significantly greater than that of tendons in the non-injection control group (56.7 ± 17.6 N) (p < 0.05). The strength of the tendons in the AAV2-luciferase group (69.6 ± 20.8 N) did not differ significantly from that of the tendons in the non-injection control group.
At twelve weeks, the repaired tendons had healed very solidly. Tensioning of the repaired tendons by the testing machine failed to disrupt the repairs. Instead, with high forces, the tendons in these groups were disrupted at the insertion site of the flexor digitorum profundus to the most distal phalanx or at the tendon-clamp junction; no failure at a surgical repair site was recorded during the test.
Extent of Peritendinous Adhesions
At two and four weeks, the tendons mostly had predominately loose or moderately dense adhesions, and the adhesions extended over a short length. No tendon had dense or extensive adhesions, except in one case (Fig. 5). The nature of the adhesions differed between two and four weeks. At two weeks the tendons were covered with a mixture of granulation tissue and filmy, soft, adhesion tissue fibers, but at four weeks the tendons were surrounded by more mature adhesions. At two weeks, the grading of the adhesions (along with the granulation tissue) demonstrated significantly greater scores in the AAV2-bFGF group than in the AAV2-luciferase group or the non-injection control group (p < 0.05). At both four and eight weeks, the adhesion scores for the tendons in the AAV2-bFGF group did not differ significantly from those for the tendons in the non-injection control group. However, at twelve weeks, the adhesion scores for the tendons in the AAV2-bFGF group were significantly less than those for the tendons in the non-injection control group (p < 0.05) (Fig. 6). This finding indicates that AAV2-bFGF may decrease adhesions later in the healing period by strengthening the intrinsic biological healing process. The adhesions generally were less severe at twelve weeks than at eight weeks in both the AAV-bFGF group and the non-injection control group. Interestingly, between eight and twelve weeks, the tendons in the AAV2-bFGF group had a greater decrease in the severity of adhesions than did those in the non-injection control group.
Work of Toe Flexion
The work of toe flexion was not significantly different among the three groups at eight weeks. However, at twelve weeks, the work of flexion of the toes in the AAV2-bFGF group was significantly less (by approximately 40%) than that of the toes in the non-injection control group (p < 0.05) (Fig. 7).
The ideal method to improve the outcome of treatment of the injured digital flexor tendon, whether mechanical or biological, should substantially increase the healing strength of the tendon but should not simultaneously increase adhesion formation. With use of a primary tendon repair model, we delivered AAV2-bFGF to the cut tendon ends and demonstrated that the healing strength of the repaired tendon increased over the critical tendon-healing period compared with that in the group that had not received the bFGF gene. The ultimate strength of the tendon after the gene therapy was 150% to 170% that of the tendon of non-injection (simple suture) controls in the early healing period (two to four weeks postoperatively). Because this period is critical for healing of the flexor tendon, and more aggressive motion regimes are applied starting from this period, a substantial increase in the strength would enable the tendon to resist increased tensile loads during finger motion.
Consistent with the results regarding the postoperative strength of the tendon as reported by other groups36,37, with surgical suture as the sole treatment, the strength tended not to increase over the initial weeks in the present study. This finding was attributed to softening of the tendon ends after injury and to a consequent decrease in the holding power in the area of the tendon grasped by the suture36. The phenomenon of the detrimental decrease or no gain in strength has long been known, but no methods have successfully overcome this limitation. In the present study, delivery of the bFGF genes was associated with increased strength during initial healing. As tendon healing progressed, superiority of the AAV2-bFGF treatment was shown again at eight weeks, although the strength of the tendon in all three groups increased at this late healing stage. At eight weeks, the strength of the tendon after AAV2-bFGF therapy was 150% that of the tendon treated with simple suture.
We evaluated the results at four time-points, starting two weeks after surgery and continuing until twelve weeks after surgery, with an emphasis on different aspects of the natural healing process of the tendon. These time-points fall into the initial postoperative period of early healing responses (the first four weeks), followed by the period of adhesion maturation (four to eight weeks) and remodeling of the tendon and peritendinous adhesions (after eight weeks). In previous investigations involving a chicken tendon-injury model, the adhesions become mature at about six weeks29,34,35. We measured the work of toe flexion at the end of the eighth and twelfth weeks, when the adhesions had matured and undergone remodeling under the stress of digital motion.
An intriguing finding is that the energy required to flex the toes was actually decreased in the AAV2-bFGF group as compared with the non-injection control group at twelve weeks. This observation reflects a possible advantage of AAV2-bFGF treatment, specifically, that besides strengthening tendon healing at earlier periods, the treatment also may restore better gliding function to the tendon later in healing following remodeling of the tendon and the surrounding adhesion tissue. This beneficial effect was correlated with the significant decrease in adhesion scores in the AAV2-bFGF group as compared with the controls. Although the underlying mechanism is hard to define and has yet to be explored, we assume that augmentation of healing capacity through the delivery of the bFGF gene has beneficial effects on the biological composition and mechanical properties of the adhesions or that adhesions are more likely to be disrupted during motion during the remodeling period of tendon healing. We noted a decline in both the work of toe flexion and the amount of adhesions from eight to twelve weeks, with the tendons in the AAV2-bFGF group showing greater declines. However, we should acknowledge that the adhesion scores were prone to bias because of their semiquantitative nature. We also noted that the adhesion scores were higher in the AAV2-bFGF group than in the non-injection control group at two weeks, which could have been due to AAV2-induced inflammation, because our previous study revealed that AAV2 vector elicits obvious inflammation at the tendon surface at two weeks after surgery26.
Healing strength is of critical importance to the repaired tendon with respect to the prevention of repair rupture. We measured the maximum strength of each repair at all assessment time-points. The work of digital flexion serves to represent the total resistance to tendon motion during simulated active digital flexion, and we measured the work at a late period, after the tendons had healed solidly and the adhesions had become mature. During the test of the work of toe flexion, we immobilized the distal interphalangeal joint and set a maximum range of proximal interphalangeal joint motion during the simulation of active toe motion. This method was a modification of the originally described test method of the work of flexion31,33-35 and was used in a series of biomechanical tests involving the chicken toe model29,31. The modification was intended to standardize the motion range of the proximal interphalangeal joint up to the range likely to be achieved by all of the tested toes. We did not choose to test the force transmitted through the flexor digitorum profundus tendon, a parameter used in a number of recent investigations38,39. Recording of the force of the flexor digitorum profundus tendon can provide direct measures of other forces acting on the tendon by adhesions or peritendinous structures (such as anular pulleys)38,39, but measurement of the work of flexion gives a general picture of how adhesion formation affects the total effort required to flex the toes, particularly when a treatment substantially alters simulated digital flexion. In the present study, the measurement revealed significant and mechanically substantial differences in the energy required to flex the digits at twelve weeks.
Past efforts to enhance tendon healing have mostly been focused on the local application of therapeutics40-48, such as corticosteroids, hyaluronic acid, 5-fluorouracil, transforming growth factor beta-1 (TGF-ß1)-neutralizing antibodies, and mannose-6-phosphate, to preclude adhesion formation. These approaches were effective for diminishing the adhesions, but some led to a reduction in tendon healing strength as well. With regard to the use of bFGF to achieve tendon healing in vivo, two groups reported different effects. Chan et al.15 injected bFGF into cut rat patellar tendons and detected no effects on the ultimate strength of the tendon seven and fourteen days after surgery despite increases in cell proliferation and type-III collagen production. Conversely, Hamada et al.49 found that bFGF-coated nylon suture increased the healing strength of rabbit flexor tendons at three weeks postoperatively but found no effects on the strength at two other time-points (one and six weeks postoperatively). Direct injection of bFGF might have been ineffective because of a limited half-life, and the amount of bFGF released from coated suture might have been insufficient to produce dramatic increases in healing strength. We did not find any report showing an increase in tendon healing strength with a simultaneous decrease in adhesions, although this is obviously the ultimate goal of treatment. The present study demonstrated a molecular approach that significantly increases the healing strength at the critical healing period while at least not increasing adhesion formation.
A number of vector systems have been used to deliver growth factor genes to tenocytes, including nonviral (liposome)26 and viral (adenoviral18-21 and adeno-associated [AAV]27) vectors. Over the past several years, AAV vectors have emerged as an important and promising vector system to deliver genes to cells or tissues22-26. AAV vectors are free of viral coding sequences22,23; hence, transduced cells will not synthesize any viral proteins. AAV vectors cause minimal tissue reactions (or none at all), in contrast to adenoviral vectors, which cause obvious tissue reactions. These considerations had prompted us to compare the efficiency and effectiveness in in vitro cultured tenocytes and tissue reactions of plasmid, adenoviral, and AAV vectors in healing flexor tendons in animals26. We recorded stronger adverse tissue reactions to liposome and adenoviral vectors, and the AAV2 elicited epitenon reactions similar to those seen in healing tendons but did not result in histologically observable tissue reactions in the endotenon area26. This evidence, coupled with the nonpathogenic nature of the AAVs in humans, contributed to our selection of AAV vectors for gene delivery to the injured flexor tendon.
In the present study, at the dosage that we tested, bFGF gene delivery brought about the tendon strength of 150% to 170% of that of the non-injection controls. The dosage was determined on the basis of the transduction rate of tenocytes by the AAV2 vector, the density of tenocytes in the tendon, and results of a pilot study involving animals. However, it has not been determined whether an increase in the amount of AAV2-bFGF used would bring about an even greater increase in strength or whether treatment with a smaller amount of vector would still generate biologically and mechanically substantial effects. Further studies on the effects of AAV2-bFGF at different delivery doses are needed to answer these questions. Additionally, the present study focused on biomechanical outcomes and macroscopic morphology. Our experience with tendon experimentation has identified these two aspects as being the most important outcome measures in the assessment of a treatment aimed at substantial augmentation of healing strength. Nevertheless, histological elucidation of the healing process is necessary, particularly with respect to the activities of growth factors, matrix components, and the transgene using molecular techniques. We could not include histological analysis in the current investigation, and work in this area is needed in the future.
Finally, despite the evidence of reliability in terms of the safety of AAV vectors and the fact that AAV does not randomly integrate into the host genome and inserts only at a specific site of chromosome 19 without known interference to the function of the host genome, one may wonder whether a nonintegrating vector or nonviral vector system could function as well. The potential differences in the effects of growth factor genes delivered by viral or nonviral vectors are unknown. In a previous in vitro study, we found that AAV2 vectors transduced cultured tenocytes more effectively than viral plasmid vectors50. In addition, we verified that serotype 2 of AAV has the higher transduction efficiency to tenocytes than other serotypes of AAV50. The current study was limited to testing AAV2-bFGF vector in an in vivo model. In future studies, the effects of bFGF gene therapy with AAV2 vectors will be compared with the effect of therapy with other viral or nonviral vectors. In the past decade, there have been three preclinical studies involving the use of AAV2 vectors registered in the National Institutes of Health. The information from these studies may eventually shed light on the safety issue in humans. However, the encouraging outcomes of our treatment of injured digital flexor tendons with AAV2-bFGF gene therapy in this animal model can serve as a basis on which an optimization of dosage of gene delivery and vector systems can be built. 
Note: The authors thank Joseph W. Hogan, ScD, Center for Statistical Science, Brown University, and Alan Brookhart, PhD, Harvard Medical School, for advice on data analysis, and Chuan Hao Chen, MD, for immunohistochemical staining.