JOURNAL OF NEUROTRAUMA Volume 9, Number 2, 1992 Mary Ann Lieben, Inc., Publishers
Anatomic and Behavioral Outcome After Spinal Cord Injury Produced by a Displacement Controlled Impact Device MICHAEL S. BEATTIE
My
charge is to Discuss HiSTOLOGic end points. I want to do that in the context of correlations with behavioral recovery and relate our recent experience using the Ohio State injury device discussed earlier by Dr. Stokes. These studies were done with Donald Behrmann in our laboratory. The lesion we are all dealing with in impact models extends not only across the transverse aspect of the cord but also rostrally and caudally. We have made three-dimensional computer reconstructions of a number of these lesions and find that they have characteristic shapes, depending on the severity of impact and the spinal level of the injury. The very long rostral and caudal extensions through the base of the dorsal columns are of interest. They may represent an ongoing process of degeneration. Dr. Wrathall's group has some data suggesting that may be true. This reconstruction process yields a lot of information, including the volume of the lesion, which may be important in evaluating the effects of treatments on sparing of gray matter and white matter. However, it is extremely time-consuming, requiring serial sectioning of paraffin or plastic blocks and evaluation of a large number of tissue sections. A number of detailed analyses may be useful for evaluating histologie outcome, including counts of myelinated axons and evaluation of spared tracts or cord reorganization using sophisticated tracing techniques. However, the question before us is how to evaluate efficiently the relationship among impact severity, behavioral end points, and histology. A very useful technique is to embed cords in plastic, as for electron microscopy, and cut 1 p,m semithin sections for staining with toluidine blue. This reveals all the cellular elements and provides a beautiful picture of the pathology, but even this relatively straightforward method requires considerable effort as well as the use of an ultramicrotome. In the studies I report, we have used the simple technique of scoring cross-sectional lesion extent at the injury site. Evaluations are made on paraffin-embedded, formaldehyde-fixed cords, sectioned and stained for myelin (luxol fast blue) and cell bodies (cresyl violet). This method has lower resolution compared to 1 p,m toluidine blue-stained sections, but it allows assessment of the pattern and extent of necrosis, cell loss, and demyelination. Criteria for spared tissue are established: in practice, the observer (blinded to the treatment condition) simply draws around the area of damage using a camera lucida device attached to the microscope. Interrater reliability is surprisingly high for trained observers. The numerical measure is the percent of the total cross-sectional area of the section that is judged to be normal, i.e., the percentage of spared tissue. The revised injury device was used to produce groups of rats with different displacement injuries: 0.80, 0.95, and 1.10 mm. These displacements were chosen from pilot studies indicating that they produced a range of injuries within the desired outcome of a severe initial paraplegia followed by partial recovery. Measured displacements were within 0.01 mm of those intended. Measured resultant peak force for each group was 264.2 + 3.6, 289.2 + 10.5, and 302.2 + 8.5 kdynes, respectively. Histologie outcomes, using the simple cross-sectional percentage of tissue spared, were as follows: 0.8 mm displacement group, 41.5 + 2.3%; 0.95 group, 20.9 + 0.9%; 1.1 group, 14.6 + 2.9%. Actual delivered displacement correlated with histologie outcome (r 0.91, p < 0.00002). Peak force also correlated with outcome, although not quite as reliably, at 42 days postinjury. =
Department of Anatomy, Ohio State University, Columbus, Ohio. 157
BEATTIE
Behavioral outcomes were strongly correlated (p < 0.0001) with histologie outcome for the three simplest behavioral measures, i.e., open field walking score, inclined plane test, and grid walking. In addition to the three displacement groups, a laminectomy control (n 3) and complete surgical transection group (n 4) were included. No deficits were seen for the laminectomy controls over the course of 42 postoperative testing days. Complete transections resulted in profound deficits that showed no recovery over 42 days, with the exception that open field scores of 1 (rather than 0) occasionally were seen after several weeks. This reflects spontaneous reflex activity in the spared caudal segments. All three displacement groups showed substantial deficits on the first day postinjury, which recovered to varying degrees over 42 days. The 0.8 and 1.1 displacement groups were statistically distinguishable from each other for each of the simple tests (open field, inclined plane, grid walk) at 42 days. The 0.95 group had scores between these two groups. This study, with rather small numbers, provided us with the appropriate range of injury severities with which to begin testing compounds for potential therapeutic value. The range of behavioral outcomes between the 0.8 and 1.1 mm displacement groups also served as the basis for estimating required numbers for later studies using a power analysis. The first drug trial using the new version of the impactor tested an opiate agonist (U-50488) vs an antagonist (nalmefene) and a putative nonopioid TRH analog (YM-14673). The drug doses were chosen from previous studies showing protective effects of these compounds in the rat. All three compounds yielded enhanced open field performance on combined days 28 and 42, compared with injured vehicle controls. The TRH analog also improved inclined plane performance and increased the number of subjects able to be tested on grid walking and footprint tasks. These animals had displacement injuries in the 1.1 mm range, that is, the most severe injury. Histologie analysis showed that animals receiving YM-14673 had significant fiber sparing at the lesion epicenter when compared with controls (22.4 ± 2.6% for YM-14673 vs 14.2 ± 2.6% for controls). Thus, the compound that at these doses and this injury severity, produced the most profound behavioral effect was the most effective at tissue protection. Methylprednisolone (doses of 30, 15, and 15 mg/kg given immediately, 2 h, and 6 h after injury) and tirilazad mesylate at three dose regimens (3, 1.5, 1.5 mg/kg; 10, 5, 5 mg/kg; 30, 15, 15 mg/kg; same postoperative intervals as methylprednisolone) were compared to controls in animals receiving a smaller trauma dose (0.95 mm displacement). No differences were seen. However, a subsequent trial using several doses of methylprednisolone (MPSS) and a higher trauma dose of 1.1 mm yielded MPSS effects similar in magnitude to those seen with YM-14673. Doses of MPSS were given immediately, then 2, 4, and 6 h after injury. Two dose regimens (30, 30, 30, 30 mg/kg and 60, 60, 60, 60 mg/kg; n 16 per group) had small, but highly reliable effects on open field and inclined plane outcomes. These behavioral improvements were accompanied by a significant sparing of tissue in the MPSS groups. Two points can be made: first, given the magnitude of the effects, we may not have seen significant outcome differences without low intragroup variability provided by the model, and second, trauma dose as well as drug dose may substantially determine whether protective effects are manifest. Our data suggest as a preliminary hypothesis that the neuroprotective effects of the agents we employed are more easily detectable when the injury is more severe. We conclude that with adequate control over the displacement impact injury to the rat cord, it is possible to produce groups with low between-animal variability in histologie outcome as measured using a routine histopathologic analysis. Further, these histologie outcomes correlate highly with three simple behavioral measures of outcome, and in some cases, drug-related behavioral improvements are reflected by histologie evidence of increased sparing of fibers. The consistent outcomes in our model and the establishment of an appropriate range of outcomes in behavioral tests have allowed us to estimate the numbers of animals needed to detect significant neuroprotective effects of treatments (using power analysis). These numbers are in the range of 5-15 animals per group, which makes multiple trauma and drug dose studies with histologie outcome measures feasible. =
=
=
REFERENCES BEATTIE, M.S., and BRESNAHAN, J.C. (1989). Longitudinal assessment of locomotor recovery in rats after feedback-controlled spinal cord impact lesions, in: Proceedings of the Conference on Criteria for Assessment of Recovery and Function: Behavioral Methods. M. Brown and M. Goldberger (eds). American Paralysis Association: Springfield, NJ, pp. 16-25. 158
END POINT MEASURES: MORPHOLOGIC BLIGHT, A.R. (1983). Cellular morphology of chronic spinal cord injury in the cat: Analysis of myelinated axons by line
sampling. Neuroscience 10, 521-543. BRESNAHAN, J.C. (1978). An electron microscopic analysis of axonal alterations following blunt contusion of the spinal cord of the rhesus monkey (Macaca mulatto). J. Neurol. Sei. 37, 59-82. BRESNAHAN, J.C, BEATTIE, M.S., STOKES, B.T., and CONWAY, K.M. (1990). Three-dimensional computerassisted analysis of graded contusion lesions in the spinal cord of the rat. J. Neurotrauma 8, 91-101. STOKES, B.T., and REIER, P.J. (1992). Fetal grafts alter chronic behavioral adult rat spinal cord. Exp. Neurol. (In press.)
Department
outcome
after contusion
damage to the
Address reprint requests to: Michael S. Beanie, Ph.D. of CelllNeurobiology and Anatomy Ohio State University Graves Hall 333 West 10th Avenue Columbus, OH 43210
DISCUSSION Dr. Faden: In earlier work, you showed a correlation coefficient of around 0.9 between the open field test 8 vs and the area score. Yet when you talked here about the "power analysis," you were talking about n 15. If the two measures correlate that well, shouldn't the numbers of animals needed for each test be n closer? Dr. Beattie: You have to look at every rat that has been run that has been included in the power analysis in order to answer that question. This takes into account variability from trial to trial, so the correlation may not match the power analysis. Dr. Tator: What were the assumptions you made for the power analysis? You wanted to show a difference of what? Did you actually do that? Dr. Beattie: The numbers that we used are the variabilities for each outcome measure, for instance, inclined plane. We pooled all the inclined plane data that had been gathered for all groups and used that variance to say, "The expected standard error of the mean is X across all of these different groups that received different treatments." So, if the treatment contributed a certain amount to the variability, we need this number of animals (n) to show that treatment effect. The more variability there is within each measure across the whole group, the more animals that are needed to reveal the variability due to treatment. This allows us to see relative ranking of the outcome measures that will be good predictors of small drug effects with a few animals. Dr. Tator: Did you use the last value obtained, or did you use the cumulative response by analysis of variance over the period of study? Dr. Beattie: This was done for the last day only, as if you had two groups that were drawn from the population that had X variability. How many animals would you need to distinguish between two groups? If you run the analysis across day s, you really increase your power, and you will see changes that occur early and late. Dr. Wrathall: I went through a series of power calculations with our data for this meeting, and I had to decide how to do it. I decided to use the percentage difference from the different tests rather than using an actual, directly measured score. Is that the way your analysis was done? Did you assume, say, a 10% difference and then pair all the relative facts? Dr. Beattie: The expected differences for the power analysis were initially taken from the behavioral study where three displacements were used. The differences between the most and least severe impact groups were divided by 3; so about 33% of this difference was taken as the expected difference for each measure. As it turned out, that difference was similar to the drug effects we observed in the pharmacologie trials. =
=
159
BEATTIE Dr. Wrathall: We have seen, and you show, that apparently tissue damage correlates with injury. Yet, I know that Dr. Faden and others found a number of effects that were functionally important for particular drugs but did not correlate with the amount of spared tissue. Now you present evidence that even drugs that previously may not have shown a correlation to tissue sparing are doing so. Dr. Beattie: It is certainly possible that a functional effect would be missed by a measure like ours. For example, if a functional measure were related to the sparing of the oligodendrocytes that myelinate certain tracts, our simple measure may not be sensitive enough. We may be able to see the effects that we do see because of the reproducibility of the injury device. Dr. Wrathall: That supports Dr. Young's measure of the preservation of tissue early after injury. Dr. Beattie: I think it is important to describe histologically the status of the cord, the status of the lesion, at any behavioral end point. It is important when you look at the three-dimensional organization and the possibility that the lesion is evolving over time. Dr. Faden: With the Yamanouchi compound, we saw a trend in terms of residual volume, but with relatively small numbers of animals in each group, the residual volume by itself was not statistically significant, even in the presence of rather striking functional differences. For relatively small numbers, there may be superior ways (e.g., immunocytochemistry for transmitters) to tease out that predictiveness. Dr. Walker: The most functionally relevant analysis may be to use several repeated measures, over time, compared to a one-time measure of area. Dr. Beattie: In the more severe injuries, where we do see drug effects, the more complex behavioral tests are not as useful, principally because those animals are just too impaired. For example, animals that are at a level of 3 really cannot perform grid walking at all. The same for footprint analysis. You get sloppy data, not because the measures themselves are sloppy but because the animals are at the edge of being able to walk down the runway. Dr. Wrathall: We had always seen a very tight correlation between chronic behavior and chronic tissue preservation measured by relatively simple measures. Dr. Anderson: Many of the early studies that Dr. Eugene Means and I did were with methylprednisolone and more recently with lazaroid and vitamin E. We do cavity/cord ratios and then a semiquantitive measure of myelinated axons. These two measures correlate closely (0.6 to 0.9) with behavior, and, in our hands, drugs that do not give improved histology, do not give behavioral improvement.
160
Anatomic and Behavioral Outcome After Spinal Cord Injury Produced by a Displacement Controlled Impact Device MICHAEL S. BEATTIE
My
charge is to Discuss HiSTOLOGic end points. I want to do that in the context of correlations with behavioral recovery and relate our recent experience using the Ohio State injury device discussed earlier by Dr. Stokes. These studies were done with Donald Behrmann in our laboratory. The lesion we are all dealing with in impact models extends not only across the transverse aspect of the cord but also rostrally and caudally. We have made three-dimensional computer reconstructions of a number of these lesions and find that they have characteristic shapes, depending on the severity of impact and the spinal level of the injury. The very long rostral and caudal extensions through the base of the dorsal columns are of interest. They may represent an ongoing process of degeneration. Dr. Wrathall's group has some data suggesting that may be true. This reconstruction process yields a lot of information, including the volume of the lesion, which may be important in evaluating the effects of treatments on sparing of gray matter and white matter. However, it is extremely time-consuming, requiring serial sectioning of paraffin or plastic blocks and evaluation of a large number of tissue sections. A number of detailed analyses may be useful for evaluating histologie outcome, including counts of myelinated axons and evaluation of spared tracts or cord reorganization using sophisticated tracing techniques. However, the question before us is how to evaluate efficiently the relationship among impact severity, behavioral end points, and histology. A very useful technique is to embed cords in plastic, as for electron microscopy, and cut 1 p,m semithin sections for staining with toluidine blue. This reveals all the cellular elements and provides a beautiful picture of the pathology, but even this relatively straightforward method requires considerable effort as well as the use of an ultramicrotome. In the studies I report, we have used the simple technique of scoring cross-sectional lesion extent at the injury site. Evaluations are made on paraffin-embedded, formaldehyde-fixed cords, sectioned and stained for myelin (luxol fast blue) and cell bodies (cresyl violet). This method has lower resolution compared to 1 p,m toluidine blue-stained sections, but it allows assessment of the pattern and extent of necrosis, cell loss, and demyelination. Criteria for spared tissue are established: in practice, the observer (blinded to the treatment condition) simply draws around the area of damage using a camera lucida device attached to the microscope. Interrater reliability is surprisingly high for trained observers. The numerical measure is the percent of the total cross-sectional area of the section that is judged to be normal, i.e., the percentage of spared tissue. The revised injury device was used to produce groups of rats with different displacement injuries: 0.80, 0.95, and 1.10 mm. These displacements were chosen from pilot studies indicating that they produced a range of injuries within the desired outcome of a severe initial paraplegia followed by partial recovery. Measured displacements were within 0.01 mm of those intended. Measured resultant peak force for each group was 264.2 + 3.6, 289.2 + 10.5, and 302.2 + 8.5 kdynes, respectively. Histologie outcomes, using the simple cross-sectional percentage of tissue spared, were as follows: 0.8 mm displacement group, 41.5 + 2.3%; 0.95 group, 20.9 + 0.9%; 1.1 group, 14.6 + 2.9%. Actual delivered displacement correlated with histologie outcome (r 0.91, p < 0.00002). Peak force also correlated with outcome, although not quite as reliably, at 42 days postinjury. =
Department of Anatomy, Ohio State University, Columbus, Ohio. 157
BEATTIE
Behavioral outcomes were strongly correlated (p < 0.0001) with histologie outcome for the three simplest behavioral measures, i.e., open field walking score, inclined plane test, and grid walking. In addition to the three displacement groups, a laminectomy control (n 3) and complete surgical transection group (n 4) were included. No deficits were seen for the laminectomy controls over the course of 42 postoperative testing days. Complete transections resulted in profound deficits that showed no recovery over 42 days, with the exception that open field scores of 1 (rather than 0) occasionally were seen after several weeks. This reflects spontaneous reflex activity in the spared caudal segments. All three displacement groups showed substantial deficits on the first day postinjury, which recovered to varying degrees over 42 days. The 0.8 and 1.1 displacement groups were statistically distinguishable from each other for each of the simple tests (open field, inclined plane, grid walk) at 42 days. The 0.95 group had scores between these two groups. This study, with rather small numbers, provided us with the appropriate range of injury severities with which to begin testing compounds for potential therapeutic value. The range of behavioral outcomes between the 0.8 and 1.1 mm displacement groups also served as the basis for estimating required numbers for later studies using a power analysis. The first drug trial using the new version of the impactor tested an opiate agonist (U-50488) vs an antagonist (nalmefene) and a putative nonopioid TRH analog (YM-14673). The drug doses were chosen from previous studies showing protective effects of these compounds in the rat. All three compounds yielded enhanced open field performance on combined days 28 and 42, compared with injured vehicle controls. The TRH analog also improved inclined plane performance and increased the number of subjects able to be tested on grid walking and footprint tasks. These animals had displacement injuries in the 1.1 mm range, that is, the most severe injury. Histologie analysis showed that animals receiving YM-14673 had significant fiber sparing at the lesion epicenter when compared with controls (22.4 ± 2.6% for YM-14673 vs 14.2 ± 2.6% for controls). Thus, the compound that at these doses and this injury severity, produced the most profound behavioral effect was the most effective at tissue protection. Methylprednisolone (doses of 30, 15, and 15 mg/kg given immediately, 2 h, and 6 h after injury) and tirilazad mesylate at three dose regimens (3, 1.5, 1.5 mg/kg; 10, 5, 5 mg/kg; 30, 15, 15 mg/kg; same postoperative intervals as methylprednisolone) were compared to controls in animals receiving a smaller trauma dose (0.95 mm displacement). No differences were seen. However, a subsequent trial using several doses of methylprednisolone (MPSS) and a higher trauma dose of 1.1 mm yielded MPSS effects similar in magnitude to those seen with YM-14673. Doses of MPSS were given immediately, then 2, 4, and 6 h after injury. Two dose regimens (30, 30, 30, 30 mg/kg and 60, 60, 60, 60 mg/kg; n 16 per group) had small, but highly reliable effects on open field and inclined plane outcomes. These behavioral improvements were accompanied by a significant sparing of tissue in the MPSS groups. Two points can be made: first, given the magnitude of the effects, we may not have seen significant outcome differences without low intragroup variability provided by the model, and second, trauma dose as well as drug dose may substantially determine whether protective effects are manifest. Our data suggest as a preliminary hypothesis that the neuroprotective effects of the agents we employed are more easily detectable when the injury is more severe. We conclude that with adequate control over the displacement impact injury to the rat cord, it is possible to produce groups with low between-animal variability in histologie outcome as measured using a routine histopathologic analysis. Further, these histologie outcomes correlate highly with three simple behavioral measures of outcome, and in some cases, drug-related behavioral improvements are reflected by histologie evidence of increased sparing of fibers. The consistent outcomes in our model and the establishment of an appropriate range of outcomes in behavioral tests have allowed us to estimate the numbers of animals needed to detect significant neuroprotective effects of treatments (using power analysis). These numbers are in the range of 5-15 animals per group, which makes multiple trauma and drug dose studies with histologie outcome measures feasible. =
=
=
REFERENCES BEATTIE, M.S., and BRESNAHAN, J.C. (1989). Longitudinal assessment of locomotor recovery in rats after feedback-controlled spinal cord impact lesions, in: Proceedings of the Conference on Criteria for Assessment of Recovery and Function: Behavioral Methods. M. Brown and M. Goldberger (eds). American Paralysis Association: Springfield, NJ, pp. 16-25. 158
END POINT MEASURES: MORPHOLOGIC BLIGHT, A.R. (1983). Cellular morphology of chronic spinal cord injury in the cat: Analysis of myelinated axons by line
sampling. Neuroscience 10, 521-543. BRESNAHAN, J.C. (1978). An electron microscopic analysis of axonal alterations following blunt contusion of the spinal cord of the rhesus monkey (Macaca mulatto). J. Neurol. Sei. 37, 59-82. BRESNAHAN, J.C, BEATTIE, M.S., STOKES, B.T., and CONWAY, K.M. (1990). Three-dimensional computerassisted analysis of graded contusion lesions in the spinal cord of the rat. J. Neurotrauma 8, 91-101. STOKES, B.T., and REIER, P.J. (1992). Fetal grafts alter chronic behavioral adult rat spinal cord. Exp. Neurol. (In press.)
Department
outcome
after contusion
damage to the
Address reprint requests to: Michael S. Beanie, Ph.D. of CelllNeurobiology and Anatomy Ohio State University Graves Hall 333 West 10th Avenue Columbus, OH 43210
DISCUSSION Dr. Faden: In earlier work, you showed a correlation coefficient of around 0.9 between the open field test 8 vs and the area score. Yet when you talked here about the "power analysis," you were talking about n 15. If the two measures correlate that well, shouldn't the numbers of animals needed for each test be n closer? Dr. Beattie: You have to look at every rat that has been run that has been included in the power analysis in order to answer that question. This takes into account variability from trial to trial, so the correlation may not match the power analysis. Dr. Tator: What were the assumptions you made for the power analysis? You wanted to show a difference of what? Did you actually do that? Dr. Beattie: The numbers that we used are the variabilities for each outcome measure, for instance, inclined plane. We pooled all the inclined plane data that had been gathered for all groups and used that variance to say, "The expected standard error of the mean is X across all of these different groups that received different treatments." So, if the treatment contributed a certain amount to the variability, we need this number of animals (n) to show that treatment effect. The more variability there is within each measure across the whole group, the more animals that are needed to reveal the variability due to treatment. This allows us to see relative ranking of the outcome measures that will be good predictors of small drug effects with a few animals. Dr. Tator: Did you use the last value obtained, or did you use the cumulative response by analysis of variance over the period of study? Dr. Beattie: This was done for the last day only, as if you had two groups that were drawn from the population that had X variability. How many animals would you need to distinguish between two groups? If you run the analysis across day s, you really increase your power, and you will see changes that occur early and late. Dr. Wrathall: I went through a series of power calculations with our data for this meeting, and I had to decide how to do it. I decided to use the percentage difference from the different tests rather than using an actual, directly measured score. Is that the way your analysis was done? Did you assume, say, a 10% difference and then pair all the relative facts? Dr. Beattie: The expected differences for the power analysis were initially taken from the behavioral study where three displacements were used. The differences between the most and least severe impact groups were divided by 3; so about 33% of this difference was taken as the expected difference for each measure. As it turned out, that difference was similar to the drug effects we observed in the pharmacologie trials. =
=
159
BEATTIE Dr. Wrathall: We have seen, and you show, that apparently tissue damage correlates with injury. Yet, I know that Dr. Faden and others found a number of effects that were functionally important for particular drugs but did not correlate with the amount of spared tissue. Now you present evidence that even drugs that previously may not have shown a correlation to tissue sparing are doing so. Dr. Beattie: It is certainly possible that a functional effect would be missed by a measure like ours. For example, if a functional measure were related to the sparing of the oligodendrocytes that myelinate certain tracts, our simple measure may not be sensitive enough. We may be able to see the effects that we do see because of the reproducibility of the injury device. Dr. Wrathall: That supports Dr. Young's measure of the preservation of tissue early after injury. Dr. Beattie: I think it is important to describe histologically the status of the cord, the status of the lesion, at any behavioral end point. It is important when you look at the three-dimensional organization and the possibility that the lesion is evolving over time. Dr. Faden: With the Yamanouchi compound, we saw a trend in terms of residual volume, but with relatively small numbers of animals in each group, the residual volume by itself was not statistically significant, even in the presence of rather striking functional differences. For relatively small numbers, there may be superior ways (e.g., immunocytochemistry for transmitters) to tease out that predictiveness. Dr. Walker: The most functionally relevant analysis may be to use several repeated measures, over time, compared to a one-time measure of area. Dr. Beattie: In the more severe injuries, where we do see drug effects, the more complex behavioral tests are not as useful, principally because those animals are just too impaired. For example, animals that are at a level of 3 really cannot perform grid walking at all. The same for footprint analysis. You get sloppy data, not because the measures themselves are sloppy but because the animals are at the edge of being able to walk down the runway. Dr. Wrathall: We had always seen a very tight correlation between chronic behavior and chronic tissue preservation measured by relatively simple measures. Dr. Anderson: Many of the early studies that Dr. Eugene Means and I did were with methylprednisolone and more recently with lazaroid and vitamin E. We do cavity/cord ratios and then a semiquantitive measure of myelinated axons. These two measures correlate closely (0.6 to 0.9) with behavior, and, in our hands, drugs that do not give improved histology, do not give behavioral improvement.
160
