Vasculogenic gene therapy: No role for revitalization of structural bone allografts

Abstract

Segmental bone defects are often performed with cryopreserved allografts. They provide immediate stability, but risk non-union, infection and late stress fracture. Improving the rate and extent of bone revitalization may improve results. Angiogenesis from surgically placed arteriovenous(AV) bundles improves bone blood flow and vitality in cryopreserved rat femora, augmented by vasculogenic growth factors. This study tests the same principal in Yucatan mini-pigs with a tibial diaphyseal defect, combining surgical angiogenesis with angiogenic gene therapy within cryopreserved orthotopically-placed allografts.

Tibial diaphyseal defects were reconstructed with cryopreserved allografts and rigid internal fixation in 16 minipigs. Half of the cranial tibial AV bundles placed within the allograft medullary canal were transfected with an adeno-associated virus containing vascular endothelial growth factor(VEGF) and platelet-derived growth factor(PDGF) genes(AAV9.VEGF.PDGF). Bone remodeling, angiogenesis and allograft healing were assessed.

During the postoperative survival period 5 of 8 transfected animals developed cutaneous benign vascular lesions at sites remote from the operated hindlimb, causing excessive bleeding. Within the allograft, both medullary(p=0.013) and cortical(p=0.009) vascular volumes were higher and vessels more mature than non-transfected allografts. Bone turnover(p=0.013), bone mineralization(p=0.018), bone healing(p=0.008) and graft incorporation(p=0.006) were all significantly higher in the gene therapy group.

In a large animal tibial defect model, gene therapy of implanted AV bundles improved revascularization, remodeling and healing of cryopreserved allografts used for limb reconstruction. However, benign vascular lesions causing excessive bleeding developed in 5 out of 8 pigs transfected with AAV containing genes for VEGF and PDGF. This unforeseen complication makes vasculogenic gene therapy unacceptable for clinical use.

Introduction:

Large segmental bone defects are a challenging reconstructive problem. Autologous bone flaps, chiefly fibula and iliac crest, have superior healing and remodeling properties when compared to allogenic bone (1, 2). They generally poorly match most areas of bone loss, however and result in some donor site morbidity (3). Cryopreserved, banked allogenic bone can be matched to virtually any defect but is initially necrotic and only incompletely revitalized over time. Allografts have a failure rate of 25–35% within 3 years due to non-union, infection and stress fracture. The latter is due to microfractures that develop due to repetitive loading. Over time, their inability to heal results in loss of structural integrity (4, 5). An ideal method to reconstruct segmental loss would provide size-and shape-matched bone with adequate endosteal circulation and viable osteocytes. One such solution would be to revitalize necrotic cryopreserved allografts by surgical angiogenesis, placing vascularized tissue within the medullary canal to mimic normal endosteal blood flow (6, 7). Prior studies in small animal models have shown this method to be useful, generating an autogenous circulation and remodelling of the allograft with autogenous osteocytes (8, 9). In a rat femoral model, direct delivery of vascular endothelial growth factor (VEGF) to the allograft via biodegradable microspheres enhanced this effect (10, 11).

This study sought to demonstrate the potential clinical applicability of VEGF and FGF2-augmented surgical angiogenesis in a large animal, large bone defect model as a proof-of-concept demonstration prior to potential clinical application. To do so, we have used gene therapy through an adeno-associated viral vector to provide local production of the same angiogenic growth factors. Others have reported the use of an identical technique to promote bone healing using RANKL, bone morphogenic protein (BMP), or the BMP receptor caALK2 (12–14).

Material and methods:

This study was approved by the Institutional Animal Care and Use Committee at Mayo Clinic, Rochester MN. All animals were housed in their own individual stalls. They were fed twice a day and had continuous access to fresh water.

A total of twenty female Yucatan minipigs (20 kg, 4 months) were used in the study. These included 16 animals used in the experimental groups, and 4 additional pigs who provided eight cryopreserved tibial segments harvested in a single non-survival surgery. The remaining 8 required allografts were obtained from the segmental defect created in some of the experimental animals. Cryopreservation was performed by storage at −80°C for at least one month. Swine leukocyte antigen (SLA) haplotyping was used to mismatch donor and recipient pairs for experimental rigor.

Sixteen Yucatan mini pigs weighing an average of 20 kg and age of 4 months received a cryopreserved segmental tibial bone allograft placed orthotopically to span a 3,5cm tibial defect. These sixteen pigs formed two experimental groups of eight.

Group I: angiogenesis alone via an AV-bundle (AVB group)

Group II: angiogenesis through AV-bundle transfected with an adeno-associated virus for production of 2 growth factors (GF): (AVB + GF).

A recombinant adeno-associated virus 9 (AAV9) expressing vascular endothelial growth factor (VEGF) and platelet derived growth factor (PDGF) was constructed and produced by the Penn Vector Core at the University of Pennsylvania, using chicken beta actin (CB7) as a promotor, and woodchuck hepatitis post-transcriptional regulatory element (WPRE) and Chimeric Intron (CI) to increase gene expression. In the AAV cis plasmid, a rabbit beta-globin polyadenylation sequence was placed after WPRE to ensure transcription termination and mRNA stability. On mRNAs, the poly(A) tail protects the mRNA molecule from enzymatic degradation in the cytoplasm and aids in transcription termination, export of the mRNA from the nucleus, and translation. The completed vector is therefore identified as AAV9.CB7.CI.VEGF.PDGF.WPRE.rBG, with a titer of 5.88×10E13 genome copies/ml (GC/ml), obtained using the droplet digital PCR (ddPCR)-method (15).

Under sterile conditions a 10 cm incision was made lateral to the anterior ridge of the tibia. The anterior and lateral compartment muscles were reflected from the bone to visualize the cranial tibial artery and vein lying on the interosseous membrane. These vessels were divided at the ankle and mobilized proximally to permit their placement within a tibial allograft. In the VEGF + PDGF group the cranial tibial artery was clamped proximally and cannulated distally with a 24-gauge I.V. catheter (Jelco, Dublin, Ohio). We delivered 1×10E13 particles of AAV.VEGF.PDGF into the lumen, suspended in 0.6 ml of PBS, 35 mM NACL and 0.001% Pluronic F68. After injection, the clamp remained in place for an additional 30 minutes to allow transfection of the arterial endothelial cells. The cannula was then removed, and the distal bundle ligated. The proximal clamp was removed, allowing proximal ingress and egress of blood.

A 3.5 cm segmental defect was next created in the tibial diaphysis immediately distal to the tibial tubercle. It was reconstructed with a cryopreserved allogeneic tibia of matched size and shape, stabilized with dual 9-hole 2.7 mm locking compression plates (Synthes, Monument, CO). A burr was used to create an opening at the proximal and distal allograft coaptation sites, through which the AV-bundle was positioned within the allograft in both groups. The fascia and skin were closed in layers using 2–0 Vicryl (Ethicon, Somerville NJ) and 3–0 Monocryl cutaneous sutures (Ethicon, Somerville NJ). A compressive bandage was applied. A 20-week survival time was planned to study bone healing, angiogenesis and remodeling of the bone. Immediately after surgery the pigs were full weight bearing. Excede 5 mg/kg IM every 5 days (Pfizer, NY, NY) and Baytril 10mg/kg IM daily (Bayer Healthcare, Shawnee Mission KS) were given for antibiotic prophylaxis for 2 weeks postoperatively. Buprenorphine SR 0.18mg/kg (10mg/ml ZooPharm, Windsor, CO) was given prior to surgery and repeated after 72 hours. Carprofen 4mg/kg IM (Rymadyl, Zoetis Inc, Kalamazoo, MI) was given in the first 5 days after surgery and continued when necessary. Pigs were housed in individual pens and monitored daily.

After 20 weeks’ survival the pigs were anesthetized with Tiletamine HCL + Zolazepam HCL 5 mg/kg IM (Telazol 100 mg/ml, Zoetis inc, Kalamazoo, MI.), Xylazine 2 mg/kg IM (Xylamed 100 mg/ml, Bimeda-MTC, Cambridge ON, Canada) and euthanized with Pentobarbital Sodium 390mg/ml (Vortech Dearborn MI, 0.22 ml/kg IV). The femoral arteries of both hind legs were cannulated, and the limbs flushed with 150 ml of heparinized saline, followed by infusion of 20ml of Microfil (Flow Tech, Inc. Carver, MA). Both tibias were removed, and the plates and screws removed.

Evaluation Methods

Graft incorporation (union):

To evaluate incorporation of the allograft, micro-CT (Siemens Medical Solutions, Knoxville, TN, USA) images were made of the entire tibia at sacrifice at 20 weeks. The incorporation was scored according to a modified scale based on Taira et al. (16). The images were scored by 2 independent observers.

Vessel volume:

Next, a 2 cm segment was removed from the center of the allograft and a corresponding segment of the normal contralateral tibia, fixed in 10% formalin for 48 hours and decalcified for seven weeks. The decalcification solution was changed weekly. After 7 weeks the bones were decalcified and imaged using a micro-CT. Scanning parameters were 60 KeV, 500 μA, 360 projections with 53.9 μm thickness slides and a high magnification. The vascular volume occupied by Microfil was measured using the Bone Microarchitecture Analysis (BMA) module in the Analyze 12.0 software (Mayo Clinic, Rochester, MN) in both cortical and medullary bone.

Quantitative Histomorphometry:

New bone formation was determined by administration of two fluorescent labels, Calcein (Sigma, St Louis, MO 20mg/kg) and oxytetracycline Hydrochloride (Vetrimycin 100mg, Boise, ID. 20mg/kg). These fluorescent labels were administered under sedation at 14 days and 4 days respectively prior to sacrifice. Tiletamine HCL + Zolazepam HCL 5 mg/kg IM (Telazol 100 mg/ml, Zoetis inc, Kalamazoo, MI.) and Xylazine 2 mg/kg IM (Xylamed 100 mg/ml, Bimeda-MTC, Cambridge ON, Canada) provided the necessary sedation. In both experimental groups, the proximal and distal allograft junction sites were sectioned for quantitative histomorphometry. A matched segment from the untreated contralateral side was taken for normalization. The bone sections were embedded in methyl methacrylate and 15µm slides cut using a Diamond band saw (Exakt systems). Unstained slides were analyzed using light microscopy at a magnification of 100x (Olympus BX51) with bone image analysis software (Osteomeasure; Osteometrics, Atlanta, GA). We quantified the proportion of bone actively mineralizing using the Mineralizing bone Surface to total Bone Surface (MS/BS) ratio. The Mineral Apposition Rate (MAR), which represents the average rate at which new bone mineral is being deposited and the Bone. It was calculated from the distance between the two fluorescent labels (calcein and oxytetracycline) divided by the 10 days separating the administration of the labels. Formation Rate to Bone Volume ratio (BFR/BV) were also determined. The bone turnover rate was calculated from the MAR multiplied by the ratio of bone surface to bone volume (BFR/BV (μm³/μm³/year). All histomorphometric parameters were determined for both the inner (endosteal) and outer (periosteal) cortical layer, averaging values obtained by measurement of 9 fields on each surface.

The micro-CT images were used to measure the ratio of bone volume to total volume (BV/TV) and bone surface to bone volume (BS/BV), providing additional quantified values of bone formation within the allograft.

Statistical analysis:

Analysis between groups was performed with absolute graft values and with normalized values (graft values as a percentage of the normal contralateral values). The statistical computer program JMP was used to perform the calculations (SAS, Cary, NC). The Mann-Whitney u test to detect differences between groups (conventional grafts versus revascularized grafts) and the paired Wilcoxon signed rank test to detect differences within each group (graft versus contralateral femora). Significance was set at p<0.05.

Results:

Vascular tumors

Surgery was successfully performed on all 16 pigs without intraoperative or immediate postoperative complications. After 2 weeks all pigs were fully weightbearing on their operated leg. Unexpectedly, however, during the 10^th week postoperatively daily inspection demonstrated the development of cutaneous vascular tumors in 5 transfected animals. No such tumors were seen in the no-vector control group. A detailed analysis of the vascular lesions has been recently published (17). To summarize, all were cutaneous or subcutaneous in location, forming blood-filled bullae that, once ruptured, resulted in hemorrhage. Lesions were found variably in all limbs, as well as the trunk and head [Figure 1]. Multiple surgeries were performed on three out of the five affected pigs to remove the lesions, and one pig was euthanized at 12 weeks due to excessive hemorrhage. Thus 7 pigs in the AVB + GF group, and 8 in the AVB group remained for study at 20 weeks. In all animals, double plating proved adequate to maintain limb function despite immediate unprotected weight-bearing.

Figure 1.

Vascular tumors in 5 of 8 pigs with cranial tibial arteries transfected with VEGF- and PDGF-producing adeno-associated virus vectors.

Graft incorporation (union):

The CT scans of the 15 remaining pigs were evaluated at 20 weeks post-surgery [Figure 2]. The CT scans were evaluated by a scoring system (16) (Table 1). Three bone grafts healed both proximally and distally in the AVB + GF group. No complete unions were seen in the AVB group. Three others in each group healed one of the two graft junctions. The median healing score for the AVB group was 9 (range-1,17) compared to 17 (range 13,21) for the AVB + GF group (p=0.006). Periosteal bridging, callus remodeling and union all seemed significantly higher in the AVB + GF group (Table 2).

Figure 2.

Micro-CT images of the treated tibias in the AVB and AVB + GF group at sacrifice. Union is significantly better in the AVB + GF group.

Table 1.

radiographic evaluation of osseous healing and transplant incorporation modified after Taira et al.

Periostial bridging
none	0
Minimal (<25%)	1
Mediate (25–50%)	2
Moderate (50–75%)	3
Complete (75–100%)	4
Callus remodeling
none	0
partial	2
complete	4
Union
Total line	0
Partial line	2
union	4
Graft appearance
resorption	−1
No change	0
Periosteal reaction from the graft	1
Total	25

Table 2.

TGraft incorporation measured according to Taira et al. compared between the two groups. Group I (AV-bundle) group II (AV bundle with growth factor).

	Group I (AVB) (n=8)	Group II (AVBGF) (n=7)	p-value I-II
Periosteal bridging	2[0,4]	3[0,4]	0.020
Callus remodeling	2[0,2]	2[2,2]	0.012
union	0[0,4]	3[0,4]	0.008
Graft appearance	1[−1,1]	1[1,1]	0.202

Vessel volume:

The intramedullary AV-bundles were patent in all 16 pigs at sacrifice, including the pig sacrificed at 12 weeks. Both intramedullary (p=0.021) and cortical (p=0.006) vascular volumes were significantly higher in the AVB + GF group compared to the AVB group. Both groups had significantly higher medullary and cortical vascular volume when compared to the contralateral side (Table 3). The vessels in the AVB + GF group looked visibly more mature and organized shown by the thick defined vascular layer, and less leaky when compared to the AVB [Figure 3].

Table 3.

Vascular Volume, contralateral side (cl), AVB (I), AVB + GF (II)

	Contra- lateral (cl) (n=15)	Group I(AVB) (n=8)	Group II (AVBGF) (n=7)	p-value cl-I	P-value cl-II	p-value I-II
Total Vessel Volume/Total bone volume	0.90 [0.32,2.0]	1.9 [0.92,3.1]	3.7 [2.9,5.6]	0.000	0.016	0.002
Medullary Vessel Volume/volume medullary bone (%)	1.0 [0.81,2.3]	2.1 [1.5,6.4]	6.1[4.9,11.1]	0.008	0.016	0.021
Cortical Vessel Volume/volume medullary bone (%)	0.91 [0.37,1.9]	1.5 [0.63,2.9]	2.9 [2.0,5.4]	0.039	0.016	0.006

Figure 3.

Typical examples of micro-CT angiographic images for the AV- bundle group, the AV-bundle + growth factor group and the untreated contralateral side. The AVB + GF shows similar images to the untreated contralateral side. The vasculature of the AVB + GF group looks more patent and organized when compared to the AVB group

Quantitative Histomorphometry:

Histomorphometry was performed on the inner cortical surface. The inner cortical surface showed significant differences for all 4 measured parameters when comparing the AVB + GF group to the AVB group MS/BS (p=0.018), MAR (p=0.018), BFR/BS (p=0.013), BFR/BV (p=0.043). The AVB + GF group showed no significant differences (p= 0.109) for BFR/BS when compared to the contralateral side in contrary to the AVB group (p=0.008). When comparing the graft to the contralateral side, the AVB + GF group MAR was significantly higher (p=0.031), unlike the AVB group (p=0.313) (Table 4) [Figure 4].

Table 4.

Quantitive histomorphometry Inner cortical surface

Quantitive histomorphometry Inner cortical surface	Contra- lateral (cl) (n=15)	Group I (control) (n=8)	Group II (VEGF) (n=7)	p-value cl-I	P-value cl-II	p-value I-II
MS/BS %	114[60,157]	62[43,95]	71[63,90]	0.008	0.047	0.018
MAR (μm/day)	2[0.9,2.5]	2[1.4,2.2]	2.2[2.1,3.2]	0.313	0.031	0.018
BFR/BS (μm 3 /μm 2 /year	804[489,1394]	403[229,705]	557[508,849]	0.008	0.109	0.013
BFR/BV (μm 3 /μm 3 /year	28[4,151]	181[123,450]	340[235,554]	0.008	0.016	0.043

Figure 4.

Examples of histomorphometry fluorescence microscopy slides with 20X magnification. The green label is calcein and the dark blue label is the tetracycline. The brown staining is vascular contrast inside the vessels. There is significantly more fluorescence label seen in the AVB + GF group compared to the AVB group.

Discussion:

Large segmental bone defects are often reconstructed with cryopreserved bone allografts due to insufficient size and shape of available, expendable bone autografts. The necrotic characteristic of the allograft not infrequently results in non-union or late stress fracture of the allograft. Results may reasonably be improved if the allograft can be ‘revitalized’ by revascularization and subsequent healing and remodeling (18). Willems et al. (10, 11) have previously shown that VEGF enhances revascularization and bone remodeling of frozen allografts in a small animal model. Our objective was to see if results would be similar in a large animal with anatomy and size more similar to man. We chose to transfect the implanted AV bundle with a bi-cistronic AAV vector rather than direct growth factor delivery due to the large size of the porcine tibial allograft and correspondingly large amounts growth factor required to expect a biologic effect.

The development of multiple cutaneous vascular tumors at sites remote from the operated hindlimb was not anticipated. We had selected a recombinant AAV-9 virus as our vector for transfection, based upon a literature review and a pilot study (19). The adeno-associated virus has been described to be a safe vector without known pathogenesis nor integration into the genome (20–22) The adeno-associated viral serotype 9 is an acceptable gene therapy vector for delivery of genes in porcine peripheral arteries, and that number of viral particles we selected demonstrated to be sufficient for successful gene delivery (19).

Although benign vascular lesions had been reported by others using VEGF AAV vectors, the combination of VEGF and PDGF was previously reported to result in more mature neoangiogenic vessels, and to eliminate the risk of vascular tumors in a murine model (23,24). Our experience, unfortunately, does not support this finding. Instead, angioma-like tumors developed in 5 of 8 treated pigs, causing bleeding sufficient to require surgical excision in several, and euthanasia in another. In previous studies in our laboratory, we have used a variety of methods for VEGF delivery, including biodegradable microspheres, local delivery by osmotic pump and gene therapy with a different vector (a replication-deficient adenovirus) (11, 25–27). None of these methods generated vascular lesions. In this case, a high AAV titer and sustained expression of VEGF and PDGF were clearly responsible for this complication.

Within the reconstructed hindlimbs, micro-CT angiography demonstrated a more organized and mature vasculature in the AVB + GF group, as well as more endosteal and cortical neoangiogenesis. We found surgical angiogenesis combined with vascular growth factor production to desirably increase bone turnover and bone mineralization rates on histomorphometry, and to improve bone healing scores at 20 weeks post-surgery. The vasculature in the AVB group was disorganized and capillary-like, while those within the AVB + GF group allografts similar to intramedullary vessels found in the untreated contralateral side. These and reported histomorphometric results are similar to those reported in hindlimb reconstructions of rat femoral defects by Willems et al. (10). This study used surgical angiogenesis augmented by VEGF to treat orthotopically placed femoral allografts. Growth factor was delivered by biodegradable microspheres. In the future, surgical angiogenesis using growth factor augmentation (by a means other than AAV transfection) may prove of value in reconstruction of segmental bone loss using banked allogeneic structural bone. Similar methods might prove of value in fracture nonunions or osteonecrosis as well, although such speculation is beyond the scope of this study.

Study results are limited by a single 20-week survival period, and a relatively small number of animals in each group. Twenty weeks is sufficient for angiogenesis and substantial bone formation, but too short to expect complete healing or remodeling of a large segmental tibial allograft. Both survival period and numbers of experimental animals used are the result of practical limitations of the large animal model.

In conclusion this study proved to be a cautionary tale of an unexpected serious complication of angiogenic gene therapy. The AAV.VEGF.PDGF vector was used to transfect arterial endothelial cells in an implanted AV bundle. The expectation of accelerated angiogenesis from the bundle when implanted into the allograft medullary canal was confirmed, with evidence of improved rate and extent of subsequent bone healing and remodeling. Unfortunately, the formation of cutaneous vascular tumors in 5 of 8 treated animals caused significant morbidity in 4 and mortality in another. These findings make any consideration of angiogenic gene therapy for clinical use inadvisable without further study. Clinical application of surgical angiogenesis with current understanding should be limited to neoangiogenesis from untreated vessels alone, or augmented by direct delivery rather than local production of vasculogenic growth factors.