A reassessment of a classic neuroprotective combination therapy for spinal cord injured rats: LPS/pregnenolone/indomethacin

Surgical procedures

Female Sprague–Dawley (SD) rats (~200 g; Harlan, Indianapolis, IN) were anesthetized with a ketamine/xylazine cocktail (80/10 mg/kg, i.p.). Guth et al. used 4% chloral hydrate as an anesthetic but OSU veterinarians advised against using this drug because it can cause ileus (Fleischman et al., 1977). The skin overlying the thoracic spinal cord was shaved then swabbed with a sequence of betadine scrub, 80% ethanol then betadine solution. Body temperature was maintained at 36–37 °C using a homeothermic blanket (World Precision Instruments Model ATC1000). The thoracic vertebrae from T₆–T₁₀ were exposed via incision and a laminectomy was performed at T₈. The tips of modified Dumont No. 5 jewelers forceps were placed on either side of the spinal cord between the dura and bone. The forceps were closed for 2 seconds, ensuring that the tips closed completely then were quickly released. Muscle was sutured and the skin incision closed with wound clips. Animals were placed in cages on top of 37 °C warming plates during recovery. Animals received Gentocin (5 mg/kg; s.c.) and 2–5 ml 0.9% saline (s.c.) post-operatively for 6 days. Bladders were expressed manually 2 ×/day until spontaneous voiding was re-established.

Pilot studies were done using n = 12 rats to best replicate the lesions produced by Guth et al. For the formal replication, n = 22 female SD rats were entered into the study protocol. To minimize variability, a single surgeon performed all spinal cord injuries. Prior to surgery, animals were randomly assigned (using GraphPad's QuickCalcs software) to one of two groups: (1) control (vehicle: n = 11) or (2) LPS/Indomethacin/Pregnenolone (LIP: n = 10). One animal died during recovery from surgery in the LIP group.

Preparation and delivery of study compounds

Immediately after SCI, pregnenolone (PREG) was delivered by placing an implantable pellet (SP-111 Innovative Research of America, FL) subcutaneously in a pocket rostral to the laminectomy incision. Although Guth et al. terminated their experiments at 21 days post-injury (dpi), we wanted the option of evaluating rats for longer post-injury survival times. Therefore, to avoid replacing 21 day pellets (50 mg/21 days = 2.38 mg/day) as used by Guth et al. and risk introducing additional surgical variables or changes in circulating and tissue levels of drug, we implanted a 60 day/150 mg pellet (150 mg/60 days = 2.5 mg/day). Animals in the control (vehicle) group received placebo pellets (#SC-111, Innovative Research), which were blank pellets comprised of the same matrix as the PREG pellets. Guth et al. also used time-release pellets prepared by Innovative Research.

Lipopolysaccharide (LPS) from Salmonella enteritidis (Sigma #L6761; 4 mg/ml) was dissolved in sterile pyrogen free water. Guth et al. used an identical species of LPS that was also obtained from Sigma. LPS was delivered via i.p. injection (0.2 mg in 0.1 ml) followed immediately by an i.p. injection of indomethacin (0.16 mg dissolved in 0.09 ml carbonate buffer). The indomethacin used by Dr. Guth was in an intravenous (i.v.) injectable form (Indocin^© IV; indomethacin sodium trihydrate: C₁₉H₁₅ClNNaO₄•3H₂O). We were not able to obtain Indocin^© IV. Instead we used indomethacin from Sigma (# I-7378: C₁₉H₁₆ClNO₄). To solubilize a stable preparation of indomethacin, the drug was prepared using the technique described by Curry et al. (1982). Briefly, 16.5 mg of indomethacin was slurried with 2 ml of 0.9% NaCl, followed by slow addition with vortexing of 0.022 M Na₂CO₃ made in 0.9% NaCl. Solution was vortexed until dissolved and then further diluted with 5 ml of 0.9% NaCl. The final solution was filter sterilized and used within 1.5 hours of preparation. The active amount (molar equivalent) of indomethacin used in this study was identical to that used in Guth et al. Rats in the vehicle (control) group received a placebo pellet (above) along with i.p. injections of 0.1 ml sterile pyrogen free water (control for LPS) or 0.09 ml carbonate buffer (control for indomethacin).

Behavioral analysis

Two individuals who were blind to group assignment performed open-field locomotor analysis at 1, 3, 7, 10, 14, 21 and 28 days post-injury using the Basso–Beattie–Bresnahan locomotor rating scale (BBB). Guth et al. reported modified Tarlov scores only at days 10 and 21 post-injury. To facilitate comparison of locomotor recovery scores reported by Guth et al., BBB scores were extrapolated to three different published modifications of the Tarlov scoring system (Table 1).

For all rats, BBB subscores were derived to quantify recovery of toe clearance, paw position, trunk stability and tail use independent of forelimb–hindlimb coordination. Each rat could achieve a maximum subscore of 13 using the following scale: toe clearance → max score = 6 (3/HL)–[none = 0; occasional = 1; frequent = 2; consistent = 3]; paw position → max score = 4 (2/HL scored at initial contact/lift off) [rotated/rotated = 0; parallel/rotated = 1; parallel/parallel = 2]; trunk → max score = 1; [trunk stable = 1; trunk unstable = 0]; tail use → max score = 2; [tail always down = 0; tail up and down = 1; tail always up = 2].

To provide a non-subjective quantitative analysis of hindlimb function, rats also were tested on a gridwalk task at 28 dpi. Rats crossed a grid (1 in. × 1 in. squares, 36 cm × 28 cm area) in both directions for ~6 minutes. Trials were recorded using a Sony HDR-SR11 HD 1080 video camera. An individual rater unaware of treatment evaluated each hindlimb separately from videos displayed at half speed. Successful hindlimb placements were quantified only when the trunk was moving forward during limb advancement and successful placement occurred when the hind paw was placed on a rung with weight support. Guth et al. did not evaluate precision locomotor function on a grid.

Tissue preparation and histology/immunohistochemistry

Animals were anesthetized with ketamine/xylazine and then transcardially perfused with cold 0.1 M PBS pH 7.4, followed by fixation with 4% paraformaldehyde in 0.1 M PBS, pH 7.4. Spinal cords were removed, post-fixed for 2 hours then placed at 4 °C in 0.2 M PBS pH 7.4 overnight and then cryoprotected in 30% sucrose until the tissue sank. A 20 mm length of spinal cord tissue centered at the injury epicenter was embedded in OCT compound (Thermo Shandon, Pittsburgh, PA), frozen on dry ice and stored at −80 °C. Spinal cords were serially sectioned in the coronal plane using a Microm cryostat at 10 µm. Adjacent sets of cross-sections spaced 100 µm apart were collected on Superfrost slides. A series of sections spanning the entire rostro-caudal extent of the lesion were stained using antibodies specific for neurofilament (NF; 1:2000; RT97 clone; Developmental Hybridoma Studies Bank) and were subsequently stained with eriochrome cyanine (Sigma) to label spared or “intact” myelin. Guth et al. analyzed paraffin-embedded spinal cords prepared at 21 dpi. From these samples, longitudinal sections were cut in either the coronal or sagittal plane then were stained with hematoxylin/eosin for cellular detail or were impregnated with Protargol (silver stain) to visualize axons.

To illustrate representative cellular components at the lesion site using additional contemporary methods, a subset of frozen sections from both groups were prepared for confocal microscopic analyses using a triple immunofluorescent labeling technique. Antibodies specific for Iba-1 (microglia; 1:1500, Wako, 019-19741), NF (see above) and GFAP (astrocytes; Aves; 1:500) were applied to sections overnight at 4 °C then for an additional 2 hrs at room temperature. Slides were rinsed with PBS followed by a wash in high salt PBS (23.38 g NaCl₂/1 L of PBS) to reduce non-specific ionic interactions. Primary antibodies were visualized using species-specific directly conjugated secondary antibodies: Iba-1 (AlexaFluor 488 @ 1:500; goat anti-rabbit IgG; Molecular Probes, A11034); NF (AlexaFluor 546 @ 1:1000; goat anti-mouse IgG ; Molecular Probes, A11030); GFAP (AlexaFluor 633 @ 1:500; goat anti-chicken; Molecular Probes, A21103). Sections were incubated with secondary antibodies for 1 hour at room temperature after which sections were rinsed with PBS then coverslipped with Immu-Mount (Thermo, 9990402). All antibodies were diluted in a solution of 4%BSA/0.1%Triton x-100/PBS.

Quantitative morphometric analysis of injured spinal cord sections

Guth et al. prepared sagittal sections of the spinal cord for analysis of lesion volume. Specifically, using individual sections that spanned a width of 360 µm centered on the central canal, Guth et al. quantified changes in lesion pathology. Based on our previous experience and expectation that measurement of lesion volume and cavitation should be independent of plane of section, we cut all specimens in the transverse plane. This allowed unequivocal visualization of cellular changes in gray and white matter while still allowing for quantitative analysis of lesion length and volume.

First, to match the analysis of the region of interest defined in sagittal sections by Guth et al., we placed a rectangular sample box 360 µm wide spanning the dorso-ventral axis of the spinal cord. The sample box was centered on the spinal cord midline of equally spaced cross-sections spanning the rostro-caudal extent of the lesion. The area of lesioned tissue (injured tissue) or cavity was quantified (see Fig. 2). Cavitation was distinguished from histological artifact by microscopically comparing continuity of tissue adjacent to the suspected area of cavitation. Cavities that were not bounded by obvious glial scars or that did not contain inflammatory cells were considered artifact. Using these criteria, the magnitude of cavitation was likely underrepresented in our analyses. However, all tissues were treated identically so changes within and between groups are comparable.

Next, instead of limiting our quantitative analyses to a 360-µm width of spinal cord, we quantified the area of spared myelin (defined by normal appearing eriochrome cyanine staining) and lesion cavitation (defined as areas devoid of myelin, neurofilament or background labeling) throughout the entire cross-section using an unbiased point-counting technique (Cavalieri's method) (Gundersen and Jensen, 1987). Mean area was quantified from 5 evenly spaced sections spanning the rostral–caudal length of the lesion containing obvious lesion cavitation.

Finally, as we have described previously (Kigerl et al., 2007), densitometric thresholding and image analysis was used to quantify spared or growing neurofilament-positive (NF+) axons, both within the region confined by the 360-µm sample box or across the entire cross-section. One specimen was damaged during processing leaving n = 10 animals/group for quantitative morphometric analyses.

Statistics

Repeated measures ANOVA was used to compare BBB or Tarlov scores (time vs. treatment) with t-tests for post-hoc comparisons (Scheff et al., 2002). Accuracy of hindlimb placement on the grid was analyzed using χ² test. Pearson product moment correlation test analyzed the relationship between success on the grid and BBB subscores. Histological analyses were analyzed using 2-way ANOVA or unpaired t-tests.

Pilot studies: calibrating the forceps compression injury and Tarlov/BBB conversions

Experimental spinal cord compression injuries are easily produced using forceps; however the primary trauma varies as a function of compression rate and duration. The magnitude of primary trauma also is expected to vary as a function of the rostro-caudal length of spinal cord that is compressed (determined by the width of the forceps blades) and the magnitude of compression across the ventral-most part of the spinal cord. After reviewing several manuscripts by Guth et al., it became clear that in order to approximate the magnitude of pathology and extent of functional recovery in SCI control rats as described by Guth et al., pilot studies would be needed to calibrate the forceps compression model.

In the manuscript targeted for replication, Guth used a #5 Dumont jewelers forceps to compress the thoracic (T8) spinal cord for 2 seconds. No additional details about the injuries were provided. In other studies Guth et al. described using “thinned and blunted” forceps or jeweler's forceps in which the “tips had been ground to 0.3 mm in width” (Guth et al., 1994b; Zhang and Guth, 1997). Because we did not know which forceps to use for replication, we compared histological and behavioral outcome in rats subjected to SCI with one of two modified pairs of #5 Dumont jeweler's forceps. Forcep tips were ground such that the thickness of the forcep blade in contact with the spinal cord increased from the tip (ventral spinal cord) to the shaft over a range of 0.3–0.5 mm (“thin”) or 0.3–0.8 mm (“thick”). Based on the formula for an area of trapezoid, which is the approximate shape of the forceps in contact with spinal cord, the thick forceps are in contact with ~30% more surface area of spinal cord than the thin forceps. This difference in tip size significantly affected tissue pathology and functional recovery.

Acute (4 days) histological preparations revealed even spinal cord compression across the dorso-ventral axis with both forceps (not shown); however, the thick forceps caused more severe injuries with limited or no functional recovery over the first 5 dpi (average BBB scores = 0 or 1). By 21 dpi, n = 1/8 rats achieved a BBB score of 4.5 (average of both hindlimbs) but the remaining subjects achieved only slight movement of one or two hindlimb joints (BBB scores = 1 or 2). For comparison, these scores correspond to 0–1 on the modified Tarlov scale, which is more severe than the Tarlov scores achieved at 21dpi by control rats in Guth et al. (~2.4 or hindlimb movement but without weight support). Histology at 21 dpi confirmed that the injuries were more severe than those published by Guth et al. (not shown).

“Thin” forceps were then tested on additional rats (n = 4). By 5 dpi, some recovery of function was evident in all rats culminating with weight-supported stepping at 21 dpi. This level of recovery was superior to the control groups described by Guth et al. (Tarlov scores of ~2–3; clear-cut HL movements but without weight-supported stepping) but was much closer than the functional recovery achieved by rats injured with “thick” forceps. Consequently, “thin” forceps were used for the formal replication experiment.

It was also necessary to calibrate techniques for behavioral scoring. All behavioral data from Guth et al. were generated using modifications of the Tarlov scale, citing Wrathall et al. (1985) as the source of these modifications. However, across the different publications from Dr. Guth's lab, inconsistencies were found in the reported methods (see Table 1). Therefore, we applied the more widely used BBB open-field locomotor rating scale then extrapolated those data to different Tarlov equivalents (Table 2 and Fig. 1).

Formal replication experiment

Guth et al. quantified spontaneous recovery of motor function every day for 21 dpi using a modified Tarlov scale. Only average scores at days 10 and 21 were reported (see Tables 1 and 2 in Guth et al., 1994a). We scored recovery of motor function from 1 to 28 dpi using the individual criteria defined in 3 different modifications of the original Tarlov scale (see Table 1) and the BBB locomotor rating scale.

In all rats, forceps compression injury caused significant locomotor impairment with some spontaneous functional recovery evident over the course of 28 days; however, there were no differences between vehicle and LIP-treated animals (p > 0.05) (Fig. 1A). Similar patterns of recovery were noted when BBB scores were converted to modified Tarlov scores (Table 2; Fig. 1C and D). Signs of a treatment effect were evident using BBB subscore analyses. BBB subscores showed that most rats in the control group failed to achieve a milestone of frequent stepping and had greater deficits in HL use during stepping. This resulted in a median subscore of 0 at 10 (n = 9), 14 (n = 9) and 21 dpi (n = 7) (Fig. 1B). In contrast, 50% of LIP-treated rats stepped frequently or consistently by 14 dpi, increasing to 70% of the group at 21dpi with a median subscore of 2 (Fig. 1B).

Although group differences in BBB subscores were underwhelming in terms of total score, they were predictive of improved hindlimb function on a grid-walking task (Fig. 1D, E). Indeed, a significant treatment effect was observed when analyzing paw placement accuracy during grid walking. The lowest performers in both groups accurately placed their HLs on 18–36% of their steps. Most of these rats were control animals. Rats with intermediate HL placement accuracy achieved accurate placement with 41–50% of their steps. An equal number of control and LIP-treated rats were classified as “intermediate” performers. High performers were 54–91% accurate during grid walking and were mostly from the LIP-treated group. Using a χ² analysis, a significant effect of treatment was detected with more LIP-treated rats being able to accurately place > 50% of the time (p< 0.01; LIP vs. vehicle) (Fig. 1E). Grid performance correlated with BBB subscores (r² = 0.72, p < 0.001) indicating that rats with greater accuracy on the grid had improved recovery of toe clearance and paw position (Fig. 1E).

Histological analysis

Guth et al. evaluated lesion area and spared ventral white matter (VWM) using paraffin-embedded tissue sections cut in the sagittal plane. To ensure that observations were made at the same level in all samples, those analyses were completed only on sections in which portions of the central canal were visible in proximity to the lesion (mid-sagittal section). In total, a width of 360 µm of spinal cord centered on the central canal was analyzed across the entire dorsoventral axis of the spinal cord for each animal at 21 dpi. To ensure that the same regions of spinal cord were analyzed in cross-sections, we quantified areas of lesion (damaged tissue) and cavity (devoid of tissue) within a rectangular box of width 360 µm centered on the spinal cord midline (Fig. 2, inset). Using this approach, no differences were detected in lesion area, cavity area or lesion length; however, cavity size and lesion length were more consistent and tended to be reduced in spinal cords from LIP-treated rats (Fig. 2A–C).

We also analyzed the mean area of spared white matter or lesion cavity throughout the entire cross section of up to 5 tissue sections spanning ~1.2 cm of spinal cord. This length of tissue corresponded with the area in which lesion cavities were grossly obvious in both groups. Using stereological point counting techniques, significantly more spared white matter was detected in LIP-treated spinal cords as compared to vehicle controls (Fig. 3A). Although lesion cavities were not different between groups, there was a trend for smaller cavities in spinal cords of LIP-treated rats (Fig. 3B; p = 0.06 vs. vehicle).

During blinded analyses of these sections, marked differences in axonal sparing and/or sprouting were noted between sections. Quantitative analyses confirmed that axonal density (quantified within a rectangular box of width 360 µm centered on the central canal) was significantly greater in LIP-treated rats, both in regions of spared white matter and within the center of the lesion (Fig. 3C, D). Using Protargol (silver stain) to visualize axons, Guth et al. described “remarkable injury-attenuating effects” of LIP in midsagittal plane at 21 dpi with a “modest number of nerve fibers characteristic of regenerating axons coursing within the lesion”. In the present study, triple immunofluorescent labeling and confocal microscopy was used to visualize these cellular responses. In sections that were taken from specimens that were representative of mean neurofilament-labeling in the lesion center of vehicle or LIP-treated rats, a modest but consistent increase in axonal sprouting was co-localized with a robust macrophage response, especially in sections from LIP-treated subjects (Fig. 4). Similar patterns were noted in vehicle-treated rats but fewer axon profiles were detected and the morphology of the macrophage reaction was distinct; i.e., the cells were smaller and exhibited a more ramified morphology. GFAP-immunoreactivity also appeared to be decreased in sections from LIP-treated rats (Fig. 4). Detailed quantitative analyses were not completed since Guth's observations also were qualitative.