Recombinant human bone morphogenetic protein-2 (rhBMP-2) is a potent osteoinductive agent that is approved by the U.S. Food and Drug Administration for anterior lumbar interbody fusion and long-bone fracture repair1. Since its approval, rhBMP-2 has gained popularity as an effective bone-graft substitute2 as it obviates the need for autologous bone-graft harvesting and eliminates associated complications and morbidity3,4. Many surgeons, therefore, began so-called off-label use of the product in all spinal regions5-8, after which new complications associated with rhBMP-2 emerged, including extreme soft-tissue swelling in the cervical spine, heterotopic bone formation, and vertebral body osteolysis in the thoracic and lumbar spine9-12.
Vertebral column trauma is frequently associated with neurologic deficits13. Surgical management of multilevel spinal instability by spinal fusion with rhBMP-2 is becoming more common. However, limited data exist as to the direct effects of rhBMP-2 on the course of posttraumatic spinal cord pathology. Following spinal cord injury, the blood-spinal cord barrier is disrupted regardless of the presence of a frank meningeal tear14. Multilevel spinal trauma often involves high-energy forces, which have been associated with a dural tear rate of up to 74%15.
Previous studies have shown that rhBMP-2 penetrates intrathecally when used for spinal arthrodesis adjacent to a dorsal hemisection spinal cord injury16. Collagen sponges with rhBMP-2 implanted at different time points after spinal cord injury triggered an increase in direct intraparenchymal BMP signaling in all spinal cord cell types, with the extent dependent on the integrity of the blood-spinal cord barrier. However, little is known of the long-term morphologic or functional effects of rhBMP-2 on the spinal cord after injury.
The current study was undertaken to (1) evaluate the acute and chronic effects on the morphology of a spinal cord lesion with the application of rhBMP-2 on the posterolateral rat spine and (2) determine whether there were functional differences in recovery from spinal cord injury after spinal arthrodesis with rhBMP-2.
Animals
A total of fifty-two adult, female Sprague-Dawley rats (weight, 250 to 275 g) were used in this study. All protocols were approved by our Institutional Animal Care and Use Committee.
Randomization Schedule and Treatment Groups
Rats were assigned to either the vehicle control group (twenty-four rats) or the rhBMP-2 group (twenty-four rats). Animals were further subdivided into an acute (one-week) treatment group (eight rats) or a chronic (six-week) treatment group (sixteen rats). An additional group of four rats received recombinant human albumin (rhAlbumin) for one week to control for any nonspecific cross-species immune response elicited by a human protein in a rat. In the experimental treatment group, 43 μg of rhBMP (per side) was implanted onto the posterolateral vertebrae on the absorbable collagen sponges (20 × 15 × 3 mm) (Infuse; Medtronic Spine and Biologics, Minneapolis, Minnesota)16. Control animals received absorbable collagen sponges soaked in an equal volume of sterile water. The control albumin group received 43 μg of rhAlbumin (per side) on absorbable collagen sponges.
Surgical Procedures
Spinal Cord Injury
General anesthesia was induced with intraperitoneal ketamine (80 mg/kg) and xylazine (10 mg/kg). Following aseptic preparation, T8-T10 laminae and transverse processes were exposed through the midline approach, and partial laminectomy of the caudal aspect of the T9 and the cephalad aspect of the T10 laminae was performed followed by the dorsal hemisection of the spinal cord. The spinal cord was incised under the microscope by a single surgeon to a depth of 1.25 mm16. The meninges were left unrepaired. Following hemisection, the wound was covered with sterile, saline solution-soaked gauze, and the animals were left undisturbed for thirty minutes.
Following the wait period, the wound was gently reexplored and two sponges were implanted bilaterally over the transverse processes of T8-T10. The incision was then closed with 6.0 ETHILON suture (Ethicon, Somerville, New Jersey), and the skin was reapproximated with staples. Postoperative pain was managed with use of subcutaneous buprenorphine (0.03 mg/kg), while subcutaneous injection of cefazolin sodium (35 mg/kg) was given once daily for five days as prophylaxis against infection.
Four cases of late-onset urinary tract infections were successfully managed with antibiotics. Throughout the study, five of sixteen control rats and nine of sixteen animals treated with rhBMP-2 developed autophagia, a known manifestation of neuropathic pain after spinal cord injury in rats17. All but one were successfully managed with acetaminophen and did not need to be killed prematurely.
Anterograde Labeling of the Motor Cortex
To produce fluorescent labeling of the descending axons within the spinal cord, 5% mini-ruby dye (Molecular Probes, Eugene, Oregon) was stereotaxically injected into the motor cortex in sixteen rats (eight per group) at thirty-two days after the lesion was created. Following skull exposure, a 1.0-mm drill-bit connected to a high-speed burr (Midas Rex; Medtronic, Memphis, Tennessee) was used to prepare two troughs parallel to the sagittal suture. The tracer was then injected into the motor cortex at six previously validated coordinates (2 μL per site)18. Rats were killed ten days later at six weeks after the injury.
Functional Testing
Open Field Locomotion
Open field ambulation and gross motor recovery were assessed according to the Basso, Beattie, and Bresnahan (BBB) scoring system19. The animal's functional performance was graded with a 21-point scale (ranging from 0 indicating complete paralysis to 21, normal ambulation) on the basis of the extent of hind limb movement, weight-support, stepping, and coordination. The BBB testing was performed on the first postoperative day and once weekly thereafter until the animals were killed. Two graders, blinded to the treatments, observed the animals and agreed on a functional score during a four-minute testing period, as advocated by Basso et al.19.
Footprint Analysis with Use of the CatWalk System
The digital CatWalk System (Noldus Information Technology, Leesburg, Virginia) was utilized to assess fine motor recovery and in-line ambulation. The system has been previously validated for evaluating motor recovery following dorsal hemisection of the spinal cord20. In short, animals were trained to cross an elevated glass plate with internal fluorescent illumination that reflects footprint outline to a digital camera positioned below. Various ambulation data parameters were measured, including the angle of paw rotation from the midline, stride length, base of hind limb support, and paw print area.
Baseline data were obtained preoperatively, and the tests were repeated once weekly starting at one week after the injury. At each time point, rats were allowed to cross the walkway three times and data from the best two trials (based on walking velocity and uniformity of pace) were pooled and analyzed.
Tissue Collection and Preparation
At the respective end points, all animals were transcardially perfused with phosphate-buffered saline solution and 4% paraformaldehyde. Spinal cord sections extending 5 mm rostral and caudal to the lesion were collected, postfixed, cryoprotected in 30% sucrose, embedded in optimal cutting temperature (OCT) compound, and stored at –80°C. Six 20-μm sagittal sections spaced at 320-μm intervals across the lesion were mounted per slide to provide evenly distributed sections for each antiserum. In animals injected with mini-ruby, spinal cord sections were cut at 5 mm rostral and 12 mm distal to the lesion. For axonal counting, every sixth section was collected starting from the dorsal surface of the spinal cord until the level of gray commissure.
Immunohistochemistry
Prior to staining, spinal cord sections were rehydrated in phosphate-buffered saline solution and blocked for one hour at room temperature in 5% goat serum with 0.03% Triton X-100 (Sigma-Aldrich, St. Louis, Missouri) in phosphate-buffered saline solution. Primary antisera that were used included rabbit anti-glial fibrillary acidic protein (anti-GFAP, 1:500; Dako, Glostrup, Denmark) for reactive astrocytes; mouse anti-ED-1 (1:175; Millipore, Billerica, Massachusetts) for macrophages and microglia; mouse anti-vimentin (1:20; Sigma Aldrich) for fibroblasts; and mouse anti-CS56 (1:200; AbD Serotec, Raleigh, North Carolina), rabbit anti-NG2, and mouse anti-neurocan (1:500; Millipore) for chondroitin sulfate proteoglycans. After three washes in phosphate-buffered saline solution, sections were incubated for ninety minutes with the appropriate fluorescent secondary antibodies. Primary antibody omission control was included with each batch of slides stained.
Quantitative Analysis
Immunohistochemistry
Immunofluorescent analysis of spinal cord tissues was performed on the samples from at least four animals randomly selected from each group. Six spinal cord sections per animal were digitally photographed with use of an Olympus BX61 microscope (Olympus, Tokyo, Japan) with a charge-coupled device camera. Depending on the signal intensity of a particular antibody, images were obtained at a magnification of either 2× (single image) or 4× (two images: left and right of the lesion). Relative immunofluorescent intensity of each antibody was quantified within a 2-mm circumference surrounding the lesion center with use of iVision-Mac Software (Bio-Vision Technologies, Exton, Pennsylvania). Immunofluorescence intensity was calculated according to the threshold method21. In brief, pixel intensity threshold for positively stained tissue was manually selected for an antibody and remained constant throughout the analysis. This allowed for a comparative analysis of the average fluorescence intensity within tissues between the control and experimental groups.
Anterograde Axonal Labeling Assessment
Mini-ruby-labeled axons were counted in two sections per animal. Axonal counting was performed at 0.25-mm intervals (magnification, 40×) that extended 5 mm rostral and 10 mm caudal from the lesion center, which was identified as an area devoid of GFAP immunoreactivity but filled with DAPI (4,6-diamidino-2-phenylindole-dihydrochloride)-positive cell nuclei. The total number of labeled axons rostral and caudal to the lesion as well as the average axon lengths were calculated as previously described18,22.
Microcomputed Tomography
At six weeks after the injury, all spinal columns were radiographically evaluated en bloc with use of a microcomputed tomography system (SkyScan 1172; SkyScan, Kontich, Belgium). Scans were performed at an 11-μm resolution, and images were reconstructed in the sagittal, coronal, and axial planes for analysis. The extent of bone formation into the spinal canal and neural foramina was qualitatively evaluated.
Data and Statistical Analysis
Numerical data are presented as the mean and the standard error of the mean. Intergroup differences between control and rhBMP-2-treated animals were compared with use of a Student t test at each follow-up point. At one week after the injury, intergroup comparison of the inflammatory response between the three groups were evaluated with use of a one-way analysis of variance (ANOVA) followed by the Tukey honestly significant difference test as post hoc comparison. All statistical computations were performed with use of the SPSS 16.0 software (SPSS, Chicago, Illinois), and a difference at p < 0.05 was considered significant.
Source of Funding
This study was funded by a grant from the translational research program of the Blast Spinal Cord Injury Program, U.S. Department of Defense.
Effects of rhBMP-2 on the Composition of the Lesion Scar at One Week After Injury
Significant changes in the morphology of the spinal cord with hemisection were observed in rhBMP-2-treated rats (Fig. 1). In this group, ED-1 staining (macrophages and microglia) around the lesion was increased compared with either vehicle controls (84% increase; 95% confidence interval [CI], 4% to 165%) or rats receiving rhAlbumin (81% increase; 95% CI, 2% to 161%) (ANOVA; F = 5.56, p = 0.027) (Fig. 1, A). In contrast, ED-1 immunolabeling was similar between the vehicle and albumin controls, indicating that a nonspecific human protein itself did not elicit postinjury inflammation.
As astrocytes and fibroblasts express BMP receptors, we examined the spinal cords with lesions for markers of reactive astrocytes (GFAP and vimentin) and invading fibroblasts (vimentin only)23-26. Both GFAP and vimentin immunolabeling was significantly increased in the rhBMP-2 group compared with the control group (181% increase; 95% CI, 70% to 290%; t test, p = 0.002) (Fig. 2). Additionally, rats treated with rhBMP-2 demonstrated a profound intraparenchymal fibroblast invasion, as evidenced by a 157% increase (95% CI, 33% to 282%; t test, p = 0.021) in vimentin-positive GFAP-negative immunostained area.
Following spinal cord injury, the cells within the fibroglial scar are known to produce a number of extracellular matrix molecules including various chondroitin sulfate proteoglycans (CSPGs), which are highly inhibitory to axonal regeneration27. Our analysis revealed that the total CSPG immunoreactivity (CS56) was doubled in the area surrounding the lesion in rhBMP-2-treated animals versus controls (95% CI, 7% to 204% increase; t test, p = 0.04) (Fig. 1, B). Additional staining for two specific CSPG core proteins (NG2 and neurocan) revealed a trend for increased NG2 immunoreactivity within the glial scar of rhBMP-2-treated rats (80% labeling increase; t test, p = 0.17) (Fig. 1, C). In contrast, neurocan staining was similar for both groups (t test, p = 0.699; data not shown).
Chronic Effects of rhBMP-2 on the Composition of the Lesion Scar at Six Weeks After Injury
At six weeks after the injury, arthrodesis with rhBMP-2 elicited a similar effect on the morphology of the scar (Figs. 3 and 4). Intraparenchymal macrophage and/or microglia immunoreactivity remained elevated in rhBMP-2-treated rats (t test, p = 0.157) (Fig. 3, A). Similarly, GFAP immunolabeling was 51% greater (95% CI, 11% to 91%; t test, p = 0.021) in the rhBMP-2 group compared with control animals and qualitatively appeared to cover a larger area than at one week after the injury (Fig. 4). However, the extent of fibrous scarring was more compact at six weeks than at one week after the injury, without significant differences in vimentin immunoreactivity between the groups (t test, p = 0.297). Overall, CS56 immunolabeling (CSPGs) remained elevated in the rhBMP-2 group, although the differences were not significant (t test, p = 0.111) (Fig. 3, B). However, NG2 immunoreactivity was greater in rhBMP-2-treated rats, highlighting its chronic effects on deposition of some CSPGs (95% CI, 8% to 115% increase; t test, p = 0.031) (Fig. 3, C).
Microcomputed tomography evaluation of the six-week specimens revealed a 100% fusion rate in the rhBMP-2 group. Additional reconstructions indicated no cases of bone encroachment into the spinal canal with all fusion masses contained dorsal to its circumference (Fig. 5). Therefore, no morphologic or functional deficits observed in these animals could be attributed to the mechanical compression of neural elements.
Anterograde Axonal Tracing
In animals injected with the anterograde axonal tracer, immunofluorescent dye was localized to the descending axons of the dorsal column, corresponding to the corticospinal tract in the rat28 (Fig. 6, A). Immediately proximal to the lesion, dense areas of axonal sprouting were observed within the region (909 ± 198 axons for control group and 1740 ± 724 axons for rhBMP-2 group; Figs. 6, B and C). In contrast, despite the low overall counts for both groups, we observed distal to the injury a stronger trend for spontaneous regeneration in the control group versus the rhBMP-2 treatment group (total axon counts were 270 ± 45 for the control group and 156 ± 25 for the rhBMP-2 group; p = 0.07) (Fig. 6, D). With respect to the average axonal length, there were no appreciable intergroup differences.
Behavioral Testing of Locomotor Function
Open Field Locomotion
On the first postoperative day, rats in all three groups (rhBMP-2, rhAlbumin, and vehicle control) received identical BBB scores (10.8 ± 0.75, 10.4 ± 0.63, 10.8 ± 0.98, respectively), indicating lesion consistency across treatments (Fig. 7). In contrast, by one week after the injury, the rhBMP-2-treated rats had significantly lower BBB scores than the control animals (t test, p < 0.05). Rats implanted with rhAlbumin performed comparably to the vehicle control group. However, at later time points, BBB scores were similar in the rhBMP-2 and control groups, with only slight differences recorded at three and four weeks after the injury (p > 0.05).
Footprint Gait Analysis with Use of the CatWalk System
Postinjury fine-motor deficits were evaluated with use of the CatWalk System. All data were normalized to the baseline preoperative values for each animal and presented as the percent change. Testing was not performed on the first postoperative day as rats were unable to walk. By one week after injury, rhBMP-2-treated rats showed a 318% increase in paw external rotation compared with a 47% increase in the control group (mean of the difference, 271%; 95% CI, 49% to 493%) (p = 0.019) (Fig. 8, A). Fine-motor control in the rhBMP-2-treated animals fluctuated considerably over the next five weeks, whereas control rats appeared to maintain their initial deficit level. By six weeks, the differences in paw external rotation between the two groups were again significant (a 473% increase in the rhBMP-2 group vs. a 66% increase in the control group; mean of the difference, 407%; 95% CI, 5% to 808%) (t test, p = 0.048).
Additional differences in motor function were observed in a comparison of changes in the base of hind limb support (Fig. 8, B). At one week, rhBMP-2-treated rats showed significantly greater support than control rats (29% increase; 95% CI, 12% to 46%) (t test, p = 0.003), but these differences diminished at later time points. No other comparisons between the two groups were significant (p > 0.05).
The current comprehensive study was undertaken to evaluate the acute and longer-term morphologic and functional changes triggered by a posterolateral spinal arthrodesis with rhBMP-2 implanted in the immediate vicinity of a penetrating spinal cord lesion. A dorsal spinal cord hemisection model of spinal cord injury was chosen for this project as it simulates the worst-case scenario of the laceration of the dura and spinal cord that exposes the spinal parenchyma to exogenous molecules, including rhBMP-2. Although it is rare in clinical practice, this type of spinal cord injury is associated with high-energy spinal trauma15.
At one week, we observed a pronounced increase in the macrophage and microglial staining in animals receiving rhBMP-2 compared with control rats (Fig. 1, A). Invading macrophages and activated microglia are known mediators of inflammation after spinal cord injury29, which in turn is detrimental to recovery from spinal cord injury. Indeed, rhBMP-2-treated rats showed significantly poorer walking performance compared with control rats at one week, despite comparable BBB scores observed on the first postoperative day. Thus, treatment with rhBMP-2 led to a transient decrease in motor function that correlated with an increased presence of inflammatory cells around the lesion.
Spinal cords of animals treated with rhBMP-2 revealed increased staining for GFAP, a marker of reactive astrocytes, at one and six weeks after injury. After spinal cord injury, reactive astrocytes play a critical role in the formation of a glial scar and production of inhibitory CSPGs that impede axonal regeneration30. The use of rhBMP-2 directly alters astrocyte hypertrophy and GFAP expression31 and can also upregulate specific CSPG core proteins in astrocyte culture32,33. The inhibitory properties of CSPGs have been studied extensively in the preclinical models of spinal cord injury treatment34. Currently, one of the main directions in spinal cord repair research focuses on prevention of the postinjury CSPG production and digestion of the existing proteoglycans found within the glial scar34,35. Additionally, a recent study by Buss et al. described the morphologic nature of the human glial scar and documented increased levels of CSPGs within and around the lesion core36. Obtained from spinal cord samples of patients who had died at various time points after a spinal cord injury, these data bridge the gap between postinjury events observed in animal studies and clinical changes in spinal cord morphology.
In the current study, we observed increased expression of total CSPGs (CS56 staining) in rhBMP-2-treated rats at one week. The core CSPG protein, NG2, is expressed by oligodendrocyte precursor cells and also fibroblasts around the lesion37. Thus, the increased NG2 staining that was greater in rhBMP-2-treated rats at one and six weeks likely reflects increases in both infiltrating fibroblasts and endogenous proliferating or migrating oligodendrocyte precursor cells around the lesion.
Analysis of the six-week samples showed similar rhBMP-2-driven inflammatory changes and scar formation (Figs. 4 and 5). These findings are important since the initial studies with use of the absorbable collagen sponge carrier showed that only 50% of rhBMP-2 remains on the absorbable collagen sponge at ten days following implantation and all protein is metabolized or excreted by week 4 after surgery38. Therefore, morphologic changes within the spinal cord observed at six weeks persist long after the protein is cleared from the surgical site. Our data suggest that long-term changes in spinal cord morphology result from the transient use of rhBMP-2 around the injury site.
Detailed microcomputed tomographic analysis of the spines also produced no evidence of bone formation in the spinal canal in any animals. Thus, all morphologic and functional changes observed at six weeks appear to result from the direct effect of rhBMP-2 on spinal cord tissue rather than its mechanical compression.
Meningeal fibroblasts invading the lesion form the core of the fibroglial scar34. In turn, they secrete several extracellular matrix molecules including class-3 semaphorin and NG2 that further contribute to the nonpermissive qualities of the lesion core37. Our data show that rhBMP-2 contributed to a marked increase in fibroblast invasion and scar formation (Figs. 1, 2, 4, and 5).
Anterograde labeling of the descending corticospinal tract fibers was performed to evaluate the effect of rhBMP-2 on axonal regeneration. Dorsal hemisection spinal cord injury completely transects the corticospinal tract; therefore, labeled axons distal to the injury zone result from regeneration through the lesion. We observed that corticospinal tract axonal sprouting in rhBMP-2-treated rats was moderately increased rostral to the lesion; however, axonal numbers distal to the injury were significantly lower than in controls. Previous reports have suggested that BMPs can either enhance or repress axonal growth dependent on the injury model and the treatment paradigm39,40. In our hands, rhBMP-2 reduced the amount of regenerating axons, while increasing sprouting of the cut axons.
Increased axonal sprouting could result in improper synapse formation in the dorsal horn leading to postinjury onset of neuropathic pain. In the rat, afferent sensory axons synapse on spinal interneurons in laminae 1 and 2 of the spinal cord dorsal horn, whereas descending corticospinal tract fibers terminate on the respective lower motor neurons in laminae 3 to 628. Improper signaling could facilitate progression of chronic pain syndrome, a frequent complication after spinal cord injury41-43. One manifestation of neuropathic pain in rats after spinal cord injury is the development of autophagia. We observed this condition in 56% of rhBMP-2-treated rats versus only 35% of the control animals. Whether rhBMP-2 elicited this response through increased sprouting or other mechanisms remains to be determined; however, macrophages and reactive glia are known modulators of neuropathic pain in rats44,45. Thus, rhBMP-2 could have an indirect effect on postinjury allodynia through upregulated macrophage and astrocyte activity. Clinically, the onset of postoperative radiculitis has been reported in patients undergoing transforaminal lumbar interbody fusion with rhBMP-246. Therefore, additional studies are necessary to fully elucidate the exact role and mechanism of rhBMP-2 on the sensory and motor neurons of the spinal cord, as well as its modulation of pain.
Functionally, we observed significant differences in the degree of hind paw external rotation between the control and rhBMP-2-treated animals at six weeks after injury. Unlike in humans, a corticospinal tract lesion in a rat does not preclude animals from spontaneously regaining hind limb function, including locomotion28. Instead, corticospinal tract lesions are associated with a loss of fine motor skills, including paw position control47. Therefore, rhBMP-2 had a significant detrimental effect on motor skills as evidenced by an increase in hind paw external rotation in that group. BBB testing, although validated for assessing functional recovery after a contusive spinal cord injury, was not a sensitive enough tool to delineate functional changes in this study.
In conclusion, our results demonstrate that spinal arthrodesis with rhBMP-2 performed in the vicinity of a penetrating spinal cord injury with a nonreparable dural laceration has detrimental effects on spinal cord histological structure and fine motor control in the rat model. This establishes a possibility of detrimental effect in a similar clinical scenario in humans but does not prove it as humans and rats have different spinal cords and different responses to rhBMP-2. Additional studies with use of a nonpenetrating contusion spinal cord injury model are necessary, however, to evaluate a more realistic clinical scenario and perhaps identify a safe dose or carrier combination that does not have the detrimental effects identified in rats in our study. It is plausible that, without direct meningeal rupture, the untoward effects of rhBMP-2 on the spinal cord will be subdued as the significance of several immunohistochemistry marker comparisons performed in the current project were just below the 0.05 threshold for Type I error. Despite these limitations, this study serves a critical role in alerting scientists and clinicians alike to the potential negative effects that the exogenous rhBMP-2 can exert on the spinal cord if given direct access to the cells comprising it.