Locally Delivered Ascorbic Acid and β-Glycerophosphate Augment Local Bone Graft in a Murine Model of 2-Level Posterior Spinal Fusion

Collagen Sponge Preparation

Collagen sponges were used as a delivery vehicle for ascorbic acid. Commercially available hemostatic collagen sponges (Gelfoam, Pfizer, New York, New York) were sterilely cut using a disposable No. 11 blade scalpel into 1 mm³ blocks, measured by a sterile ruler, and performed under magnification. One hour prior to surgery, sponges were prepared with either primary media alone (Dulbecco modified Eagle medium:F12/fetal bovine serum [DMEM:F12/FBS]) or with ascorbic acid (50 μg/mL) and β-glycerophosphate (3 mM). β-Glycerophosphate provides a high local concentration of phosphate ions to assist with mineralization.

Allograft Harvest and Preparation

All animal protocols were approved by our Institutional Animal Care and Use Committee (IACUC). This study adheres to the ARRIVE guidelines. To obtain consistent volumes of fresh morselized bone graft representing both cortical and cancellous bone, 12 donor C57BL/6 mice were randomly selected from a batch of 36 healthy male mice. The remaining mice were randomly allocated to 3 test groups. Bilateral femurs were harvested, with a single femur allocated to each of 8 mice in 3 test groups. Soft tissue was removed, and the remaining bone was morselized using sterilized household nail clippers (KAI, Seki, Japan).

Surgical Procedure

A total of 24 healthy male C57BL/6 mice (6 weeks old) underwent an uninstrumented posterior L3–L5 lumbar fusion. After isoflurane anesthesia, the lumbar region of the dorsal skin was shaved and prepped with povidone-iodine. Buprenorphine was injected subcutaneously for perioperative pain control. A 15-mm midline incision centered over L3–L5, referencing the iliac crests as L5–L6, was made with a No. 11 blade scalpel using surgical microscopy. The paraspinal muscles covering L3–L5 were gently dissected off the lamina using the scalpel. The periosteum overlying the spinous processes and lamina were vigorously scraped by the scalpel for 20 seconds, and 1 mm³ allograft obtained from a donor mouse was placed directly on the vertebrae and was held in place by surrounding paraspinal muscles and overlying fascia upon closure of the fascial layer. The control group (control, n = 8) received a total of 1 mm³ morselized bone only. A second control group to define the effect of the collagen sponge/primary media (sponge, n = 8) received 1 mm³ morselized bone and a 1 mm³ collagen sponge prepared with primary media alone. The third group (ascorbic acid, n = 8) received 1 mm³ morselized bone and a 1 mm³ collagen sponge prepared with ascorbic acid and β-glycerophosphate. Collagen sponges were placed superficial to the allograft. The fascial layer was closed, holding the collagen sponges in place, and skin was closed overlying it. Bupivacaine was injected into the wound during closing as a splash block. Postoperatively, animals were given meloxicam wafers, required by our IACUC, and allowed activity without restriction. Animals were monitored daily by research and veterinary medical unit staff for humane end points, which include unrelieved pain or distress, moribund state, inability to obtain water or feed, weight loss, and decreased appetite.

Micro–Computed Tomography Scans

At 3 months postoperatively, mice were killed, and the lumbar spine was harvested en bloc with retention of the surrounding musculature. Dissected specimens were placed in 70% ethanol and scanned using a SCANCO Medical μCT 50 Specimen Scanner (Zurich, Switzerland) at a slice thickness of 10 μm.

Bone Volume and Density Quantification

Data were quantified by a single blinded analyst to reduce variability. Reliably distinguishing between host native bone comprising the lamina and spinous processes from newly formed bone proves difficult without a comparison scan (Figure 1). This is complicated by morphologic differences between individual mice. Although in vivo whole mouse micro–computed tomography (micro-CT) scans are helpful for gross assessment of spinal fusion, imaging studies with sufficient resolution to accurately evaluate bone volume, density, and bridging bone in the murine lumbar spine require ex vivo specimen micro-CT scans that would represent lethal radiation exposures for a mouse. Therefore, we quantified bone volume using the absolute volume of bone per lumbar level and used a correction factor derived from the unoperated portion of the same level. Namely, we compared the amount of dorsal bone volume relative to the unoperated ventral bone volume, additionally allowing us to account for mouse-to-mouse variability in bone morphology. Because the dorsal bone encapsulates both newly formed bone and native lamina and spinous processes, this would decrease the measured effect of our experimental interventions. This results in a more statistically rigorous approach that makes it more difficult to achieve statistical significance.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Axial micro-CT images (ascorbic acid group). The border between native lamina and newly formed bone is poorly defined, requiring the referencing of the anterior vertebral body to assess bone volume.

Specifically, postprocedural micro-CT images encapsulating L3–L5 were imported into the DICOM viewing software OsiriX MD (Pixmeo SARL, Geneva, Switzerland). These ranges were further narrowed by only including transverse sections with continuous vertebral lamina, pedicles, and spinous processes (Figure 2). A line was drawn in a transverse section of a vertebra from the dorsal-most point of the vertebral foramina to the dorsal-most point of the vertebral body (Figure 3). This line was then bisected by a perpendicular line using the Perpendicular Lines tool (Figure 3). We defined any bone dorsal to this perpendicular line as “dorsal bone,” and any bone ventral to this line as “ventral bone.” This quadrant approach ensures reproducibility between specimens.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Analysis was performed on micro-CT images where there were continuous vertebral lamina, pedicles, and spinous processes, ensuring consistent anterior vertebral body sampling between levels. (Top) Images with noncontinuous vertebral lamina, pedicle, or spinous process. (Bottom) Images used for reference calculation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

(Top) A line is drawn connecting the dorsal-most point of the vertebral foramen to the dorsal-most point of the vertebral body. (Middle) A bisecting perpendicular line is drawn, separating the dorsal from the ventral portion. (Bottom) Example dorsal regions of interest are traced.

To process the large number of micro-CT images, between 60 and 160 images per level, the OsiriX region of interest (ROI)–generating algorithm was used. This algorithm predicts desired ROIs for every image based on a series of interspersed user-generated ROIs in a given range of images. We first partitioned the images to display 1 per 20 images using the subselection tool and manually traced ROIs for the dorsal portion of the vertebrae. When the cortical bone was noncontinuous, the ROI was traced so that a line of least distance was drawn between the break (Figure 4). OsiriX uses these manual ROIs as a guide and generates ROIs for every image, which were checked by the analyst and edited to improve accuracy. OsiriX can subsequently output the collective volume and density for the dorsal portion of each vertebral level. This process was repeated for the ventral portion of the vertebrae and for each of L3–L5. A metric was created that takes the ratio of the dorsal to ventral bone volume, with the ventral bone volume serving as a level-specific correction factor. The resulting unitless ratios, corrected bone volume (cBV), were calculated for each vertebra and used to compare bone volume between mice.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

A line of least distance bridges noncontinuous cortical bone, drawn over a micro–computed tomography image (ascorbic acid group). The circular region of interest in the middle image identifies the noncontinuous cortical bone.

OsiriX additionally provides pixel density histograms of pixels encapsulated within the ROIs, and we observed a bimodal distribution of lower-density and higher-density pixels representing soft tissue and mineralized bone, respectively. Micro-CT images are calibrated to hydroxyapatite (mgHA/ccm) rather than Hounsfield units (HU); we considered pixel densities below 2000 mgHA/ccm to be soft tissue and densities above 3500 mgHA/ccm to be mineralized bone. Pixels with densities between 2000 and 3500 mgHA/ccm could not be reliably distinguished as soft tissue or mineralized bone. To compare bone mineral densities between specimens, for each lumbar level, the total number of mineralized bone pixels was divided by the sum of the number of soft tissue and mineralized bone pixels to yield a mineralized bone fraction. We investigated the mineralized bone fraction for the dorsal portion of each individual L3–L5 vertebra for each specimen. This fraction was used to compare bone mineral density between mice. This approach was used in lieu of dividing mineralized bone pixels by total ROI pixels to improve effect size by removing pixels of uncertain tissue type, with densities between 2000 and 3500 mgHA/ccm.

Fusion Assessment

Full-resolution sagittal imaging sections of the lumbar spine were reviewed. Fusion was considered present or absent and was assessed at each level independently. A motion segment was considered fused if bridging bone consisting of at least 50% of the lamina of 2 adjacent segments and a continuous cortical shell from one vertebra to the adjacent vertebra were present. Three independent individuals, blinded to the test groups, assessed the specimens. Manual palpation was used in addition.

Statistical Methods

We previously determined that a sample size of 5 animals per test group was sufficient to detect significant differences in spinal fusion for rats between control and repurposed osteoarthritic chondrocytes.²¹ We selected a sample size of 8 animals per test group for additional statistical power.

Statistical calculations were performed in GraphPad Prism 8 (GraphPad, San Diego, California) by a single blinded analyst. Significance was set at P < .05. There was a single outlier in the mineralized bone fraction data set for the sponge group identified by the interquartile range that was not omitted; these data pass the Shapiro-Wilk test of normality (P = .0875). There were no additional outliers. Two-tailed Fisher exact tests were used for categoric data analysis (fusion), and Student t-tests were used for analysis of continuous variables (bone volume and mineralized bone fraction). Siegel-Castellan fixed-marginal multirater κ was used to assess interrater agreement for manual palpation.