- Email: [email protected]
Searching for the holy grail: What is the structural correlate of cognition?
298
COMMENTARY ON DICKSON ET AL.
42. Scheff, S. W.; Price, D. A. Synapseloss in the temporal lobe in Alzheimer's disease. Ann. Neurol. 33:190-199; 1993. 43. Scheff, S. W.; DeKosky, S. T.; Price, D. A. Quantitativeassessment of cortical synapticdensityin Alzheimer'sdisease. Neurobiol. Aging 11:29-37; 1990. 44. Scheff, S. W.; Sparks, L.; Price, D. A. Quantitative assessment of synapticdensityin the entorhinal cortex in Alzheimer'sdisease. Ann. Neurol. 34:356-361; 1993.
45. Terry, R. D.; Masliah, E.; Salmon, D.; Butters, N.; DeTeresa, R.; Hill, R.; Hansen, L.; Katzman, R. Physicalbasis of cognitivealterations in Alzheimer's disease: Synapse loss is the major correlate of cognitive impairment. Ann. Neurol. 30:572-580; 1991. 46. Tomlinson, B. E.; Blessed, G.; Roth, M. Observationson the brains of nondemented old people. J. Neurol. Sci. 7:331-356; 1968. 47. Wilcock, G. K.; Esiri, M. M. Plaques, tangles and dementia: A quantitative study. J. Neurol. Sci. 56:407-417; 1982.
0197-4580(95)00011-9
COMMENTARY SEARCHING FOR THE HOLY GRAIL: WHAT IS THE STRUCTURAL CORRELATE OF COGNITION? S T E V E N T. D E K O S K Y
Alzheimer's Disease Research Center, University of Pittsburgh, 3600 Forbes Avenue, Suite 400, Pittsburgh, PA 15213
THE data presented by Dickson and colleagues is a valuable addition to the literature about the nature of brain changes, normal and pathological, in normal aging and dementia. The Bronx Aging Project is one of a number of large-scale, long-term projects aimed at identifying relationships between these normal and pathologic age changes and the age-associated cognitive impairments of aging and the neurodegenerative dementias. The study concludes that measures of neuritic plaque density and markers of neurofibrillary tangle counts are "better correlates of dementia" than were assessments of synapse number, marked by homogenate assays of a synaptic protein. The statistical findings of the study support this conclusion; however, these conclusions have limited application to the issue of what tissue marker is "best" correlated with cognition in any specific dementia. In summary, in this study plaque and tangle markers correlated well with level of dementia across all cases, although it is not clear why plaques and tangles, the hallmarks of Alzheimer's disease (AD), were measured in diseases other than AD. The measure of synapse density employed (ELISA of homogenates of the full cortical depth) may not be the accurate way to assess synapses in specific dementias and thus the conclusion that "synaptic markers may not be the best structural correlate of dementia" is at least premature. The background of the search for the structural correlates started with the original studies by Blessed, Tomlinson, and Roth (1) where analyses including controls showed there was a good correlation between plaque counts and cognition. Since then, several studies have shown significant correlations between cognition and plaque density or tangle density (2,8), and others have not (5,7,9,11,12). The idea that one can utilize plaque or tangle counts as a continuous variable to correlate with cognition is dashed by looking at the scatter plots themselves in the report by Dickson et al. The only really continuous variable is the synapse data. The tangle counts appear to be either very high
or very low and do not suggest any sort of continuous correlation when one looks at the graphical display of these data. Most of the high tangle counts had Blessed scores above 30, whereas the mild and moderate cases had near "normal control" counts. Both floor and ceiling effects are problems with both cognitive tests and quantitative assessment of structural markers, and could be a reason why there is not consistent agreement in the literature about correlations between cognition and these neuropathological markers. Looking at biopsy material is especially instructive. Because patients are assessed at milder levels of dementia, their cognitive scores are more reliable, and because the amount of tissue to be examined is necessarily limited, correlations which emerge from such studies carry some strong weight, even though the total amount of tissue to be evaluated is less than one would like. In our study of synapse counts in biopsied patients (3) there was a strong correlation of synapse number with cognition, whereas the correlation of cholinesterase transferase activity in cortex was weakly correlated with cognition, and neuritic plaques did not correlate at all with cognition in this series (5 neuropathological and cognitive correlations). This finding was similar to that of Neary who found that pyramidal cell loss, and not plaques or tangles, had the best correlation with cognition (9). Terry et al. (11) subsequently confirmed the correlation of synapse counts with cognition, using autopsy samples taken from patients who had been cognitively evaluated within a reasonable period prior to death. It is worth noting that our synapse counts were guided by earlier studies we performed looking at the locus within the frontal association cortical lamina of the maximal neurodegeneration in AD (4). That study indicated that the greatest loss of sialogangliosides per neuron, a neurochemical marker for axodendritic expanse, were in lamina 2-3 and 5. We, therefore, did initial ultrastructural synapse counts in biopsy cases (as well as earlier autopsy cases in those lamina) (10). Terry
COMMENTARY ON DICKSON ET AL.
299
1.2 1 sic
O
O
0.8
O
0.6 0.4-
0.2-
/
°
0 0
0.2
0.4
0.6
0.8
1
O.D. (SY 38) FIG. 1. Comparison of EP10 and SY38 by ELISA. Cortex from both AD and neurologically normal control patients were assessed for both EP l0 and SY38 immunoreactivity under identical conditions. The correlation of the two antigens is r 2 = 0.926, p < 0.0001.
and Masliah's data from confocal microscopy studies of synaptophysin immunochemistry also looked in lamina 2-3 and 5. Thus, the laminar cortical areas maximally affected by Alzheimer's disease were the specific regions evaluated in the studies that turned up strong (and i:a fact remarkably similar) correlations between synapse counts and cognition. The quest for the "Holy Grail" needs to address several specific issues. First, does EP10, the antigenic marker utilized for the quantitation of synapse,,; by Dickson et al., measure something similar to that measured by SY38, the synaptophysin antibody which was used in the studies by Terry and Masliah? The answer to that appears to be yes. Figure 1 shows a correlation of the amount of SY38 antigen compared to EP 10 antigen from identical aliquots of homogenates of cerebral cortex. We utilized the standard ELISA for SY38 and the EPI0 ELISA used the method of Honer et al, which initially reported a decrement in this antigen in AD (6). So it appears that EP10 does reflect the same relative amount of synaptic antigen as is detected by the more traditional SY38 antibody. Second, were the measurements done in a fashion in the Dickson et al, study which were comparable to the studies which found a strong correlation of synapses with cognition? The answer here is no for two reasons. Dementias other than AD were included and homogenates were used, which is not a fair comparison. Homogenates increase the inclusion of cortical lamina which may be much less affected, specifically those which have fewer large neurons (9-11). Homogenate studies would include these less affected lamina. In addition, we (3,10) have shown
that the mean size of residual synapses is enlarged in AD, thus synaptic biochemical markers in homogenates do not represent the same number of individual synapses within any unit volume of AD tissue. Because the synapses are enlarged, homogenates will tend to underestimate the number of synapses in a given tissue volume although there may be fewer of them. This might distort or blunt a correlation with cognition. Third, Dickson et al. assert, on the basis of their data, that "synapses aren't the best measure of cognition." There were no claims made in either our initial studies or in the Terry 1991 study (11) for a correlation of cortical synapse counts with any dementia other than Alzheimer's disease. The regions were selected for analysis because of the topography of Alzheimer pathology. Dickson et al. show that homogenate assays of synaptic protein content do not provide the best correlation with cognition in a mixed group of dementias. However, these data do not appear to alter the conclusion that synapse counts done in the vulnerable place in the cortex and focused in regions known to be affected by the disease, are the best correlate of cognition in Alzheimer's disease. The fact that they measured plaques and tangles (AD-specific pathology) might lead the unwary reader to assume that this finding of better correlation of cognition with plaques or tangles rather than with synapses applied to AD specifically. The data in this paper do not address that issue because the AD cases were not separated out for analysis. The synaptic protein assays were done from regions selected for involvement in Alzheimer's disease. The cases examined, however, included dementia of several different etiologies. Why would one expect a synapse change in the cortex in multi-infarction dementia or Lewy body dementia to correlate strongly with cognition if samples were taken from association regions selectively vulnerable to Alzheimer's disease? In vascular dementia the pathologically affected regions are variable and in the latter they may well not be the same as those in Alzheimer's disease or involve the same neuronal loss in that region. The wealth of cognitive studies done in this research study does enable the correlation of more specific cognitive functions (e.g., visuospatial vs. verbal performance) with specific regions of brain, regardless of dementia etiology. The inclusion of other types of dementia besides AD in the analysis, however, even with the removal of controls from the equation, is comparing apples with oranges in respect to synapses. One would not necessarily expect, for example, parietal synapses in the right temporal lobe to be impaired if in a case of vascular dementia there was no specific ischemic structural change in that region. Similarly, if a subcortical denervating lesion was the reason for parietal dysfunction (disconnection syndrome). A similar rationale applies to other neurodegenerative dementias. To correlate plaque and tangle counts, the defining pathologic hallmarks of Alzheimer's disease, with cognition in dementias other than AD does not have an obvious rationale. If one only uses the AD cases for such correlations, it may be that correlations will be lost because the N is much smaller. Alternatively the correlations might be strengthened. In either case, the methodology for quantitative synapse assessment must be carefully considered, and the rationale for looking at neuritic plaques and neurofibrillary tangles in diseases other than AD should be carefully explained and justified.
REFERENCES
I. Blessed, G.; Tomlinson, B.; Roth, M. The association between quantitative measures of dementia and of senilechanges in cerebral grey matter of elderlysubjects. Br. J. Psychiatry 114:797-811; 1968.
2. Bowen, D. M.; Neary, D. Cerebral biopsy in the study of Alzheimer's disease. Bull. Clin. Neurosci. 50:44-8; 1985. 3. DeKosky, S.; Scheff, S. Synapse loss in frontal cortex biopsies in
300
4. 5.
6. 7.
COMMENTARY ON DICKSON ET AL. Alzheimer's disease: Correlation with cognitiveseverity. Ann. Neurol. 27:4577-464; 1990. DeKosky, S. T.; Bass, N. H. Aging, seniledementia, and the intralaminar microchemistryof cerebral cortex. Neurol. 32:1227-1233; 1982. DeKosky, S. T.; Harbaugh, R. E.; Schmitt, F. A. et al. Cortical biopsy in Alzheimer's Disease: Diagnostic accuracy and neurochemical, neuropathological and cognitive correlations. Ann. Neurol. 32:625-632; 1992. Honer, W. G.; Dickson, D. W.; Gleeson, J.; Davies, P. Regional synapticpathologyin Alzheimer'sdisease. Neurobiol. Aging 13:375382; 1992. Mann, D. M.; Marcyniuk, B.; Yates, P. O.; Neary, D.; Snowden, J. S. The progression of the pathologic changes of Alzheimer's disease in frontal and temporal neocortexexaminedboth at biopsy and at autopsy. Neuropathol. Appl. Neurobiol. 14:177-95; 1988.
8. Martin, E. M.; Wilson, R. S.; Penn, R. D.; Fox, J. H.; Clasen, R. A.; Savoy, S. M. Cortical biopsy results in Alzheimer's disease: Correlation with cognitive deficits. Neurol. 37:1201-4; 1987. 9. Neary, D.; Snowden, J. S.; Mann, D. M. et al. Alzheimer'sdisease: A correlative study. J. Neurol. Neurosurg. Psychiat. 49:229-237; 1986. 10. Scheff, S. W.; DeKosky, S. T.; Price, D. A. Quantitativeassessment of cortical synapticdensityin Alzheimer'sdisease. Neurobiol. Aging 11:29-37; 1990. 11. Terry, R. D.; Masliah, E.; Salmon, D. P. et al. Physicalbasis of cognitive alterations in Alzheimer's disease: Synapse loss is the major correlate of cognitiveimpairment. Ann. Neurol. 30:572-580; 1991. 12. Wilcock, G. K.; Esiri, M. M.; Bowen, D. M.; Smith, C. C. T. Alzheimer's disease: Correlation of cortical choline acetyltransferase activity with the severity of dementia and histological abnormalities. J. Neurol. Sci. 57:407-417; 1982.
0197-4580(95)00012-7
COMMENTARY HOW TIGHT IS THE C O N N E C T I O N BETWEEN S Y N A P S E LOSS A N D CLINICAL DEMENTIA? N E I L W. K O W A L L
GRECC (182B), Bedford VAMC, Bedford MA 01730 DENNIS Dickson and colleagues report that the classical markers of Alzheimer neuropathology, i.e., senile plaques (SP) and neurofibrillary tangles (NFT), are better correlates of clinical dementia than synapse loss. This finding is in sharp contrast to the work of Robert Terry and colleagues who showed that the best correlate of dementia in Alzheimer's disease (AD) is synapse loss (19). Neuronal loss is a less powerful correlate whereas NFT and SP weakly or inconsistently correlate with dementia. It is possible that the discrepancy between these two reports can largely be explained by differences in methodology and sample selection. One notable problem, as Dickson et al. notes, is the heterogeneity of their clinical population. Of the 45 patients prospectively studied, only 13 had pure AD. Surely one cannot expect that the clinical severity of ischemic dementia, for example, will correlate with NFT counts. The results, therefore, may be difficult to interpret. Indeed, to make a valid comparison with the results of Terry et al., the patient population and methodologies must be comparable. The staining procedures, antibodies and quantitative methods used by Dickson et al. are significantly different than those used by Terry et al. It is notable that the average cortical synaptophysin immunoreactivity of Dickson et al.'s demented patients was not significantly different from their nondemented patients. They do not report if this is also true, however, in their AD patient subset. Cortical synapse loss has been documented in Huntington's disease (HD) (26), Parkinson's disease with dementia (26), prion diseases (8), and even in AIDS encephalopathy (24). It would
not be surprising if synapse loss were associated with all dementing illnesses as an expected consequence of neuronal depletion. If a pathologic process preferentially affects axons or synapses, the loss of specialized outgrowths may outpace neuronal loss and more closely reflect the extent of damage, but even if true, does this knowledge elucidate the process of cell death or its ultimate cause? The concept of pathoclisis or selective vulnerability expounded by the Vogts (22) has been a guiding principle of modern neuropathology. Specific subsets of neurons are differentially vulnerable to particular insults, for example, in ischemia, epilepsy, trauma, and neurodegenerative diseases. One of the best examples is Huntington's disease (HD) where the pattern of differential vulnerability is identical to that produced experimentally by NMDA-type glutamate excitotoxins (4) and mitochondrial electron transport inhibitors (3). The observed pattern of neuronal damage in postmortem human brain has led to a better understanding of possible disease-related mechanisms and therapeutic strategies (2). The presence of synapse loss in the striatum (11) and cerebral cortex (26) in HD adds little to our conceptual understanding of pathogenesis. The more important question might be whether a specific class or type of synapse is lost. Indeed, Beach and McGeer reported depletion of cholinergic fibers in the visual cortex of AD brain when synaptophysin density is unchanged (1). Is synaptophysin, the most popular synaptic protein identified as an indicator of synapses in general, truly equivalent to raw synapse count? Is it uniformly distributed? Can it be downregulated without synapse loss? These
COMMENTARY ON DICKSON ET AL.
42. Scheff, S. W.; Price, D. A. Synapseloss in the temporal lobe in Alzheimer's disease. Ann. Neurol. 33:190-199; 1993. 43. Scheff, S. W.; DeKosky, S. T.; Price, D. A. Quantitativeassessment of cortical synapticdensityin Alzheimer'sdisease. Neurobiol. Aging 11:29-37; 1990. 44. Scheff, S. W.; Sparks, L.; Price, D. A. Quantitative assessment of synapticdensityin the entorhinal cortex in Alzheimer'sdisease. Ann. Neurol. 34:356-361; 1993.
45. Terry, R. D.; Masliah, E.; Salmon, D.; Butters, N.; DeTeresa, R.; Hill, R.; Hansen, L.; Katzman, R. Physicalbasis of cognitivealterations in Alzheimer's disease: Synapse loss is the major correlate of cognitive impairment. Ann. Neurol. 30:572-580; 1991. 46. Tomlinson, B. E.; Blessed, G.; Roth, M. Observationson the brains of nondemented old people. J. Neurol. Sci. 7:331-356; 1968. 47. Wilcock, G. K.; Esiri, M. M. Plaques, tangles and dementia: A quantitative study. J. Neurol. Sci. 56:407-417; 1982.
0197-4580(95)00011-9
COMMENTARY SEARCHING FOR THE HOLY GRAIL: WHAT IS THE STRUCTURAL CORRELATE OF COGNITION? S T E V E N T. D E K O S K Y
Alzheimer's Disease Research Center, University of Pittsburgh, 3600 Forbes Avenue, Suite 400, Pittsburgh, PA 15213
THE data presented by Dickson and colleagues is a valuable addition to the literature about the nature of brain changes, normal and pathological, in normal aging and dementia. The Bronx Aging Project is one of a number of large-scale, long-term projects aimed at identifying relationships between these normal and pathologic age changes and the age-associated cognitive impairments of aging and the neurodegenerative dementias. The study concludes that measures of neuritic plaque density and markers of neurofibrillary tangle counts are "better correlates of dementia" than were assessments of synapse number, marked by homogenate assays of a synaptic protein. The statistical findings of the study support this conclusion; however, these conclusions have limited application to the issue of what tissue marker is "best" correlated with cognition in any specific dementia. In summary, in this study plaque and tangle markers correlated well with level of dementia across all cases, although it is not clear why plaques and tangles, the hallmarks of Alzheimer's disease (AD), were measured in diseases other than AD. The measure of synapse density employed (ELISA of homogenates of the full cortical depth) may not be the accurate way to assess synapses in specific dementias and thus the conclusion that "synaptic markers may not be the best structural correlate of dementia" is at least premature. The background of the search for the structural correlates started with the original studies by Blessed, Tomlinson, and Roth (1) where analyses including controls showed there was a good correlation between plaque counts and cognition. Since then, several studies have shown significant correlations between cognition and plaque density or tangle density (2,8), and others have not (5,7,9,11,12). The idea that one can utilize plaque or tangle counts as a continuous variable to correlate with cognition is dashed by looking at the scatter plots themselves in the report by Dickson et al. The only really continuous variable is the synapse data. The tangle counts appear to be either very high
or very low and do not suggest any sort of continuous correlation when one looks at the graphical display of these data. Most of the high tangle counts had Blessed scores above 30, whereas the mild and moderate cases had near "normal control" counts. Both floor and ceiling effects are problems with both cognitive tests and quantitative assessment of structural markers, and could be a reason why there is not consistent agreement in the literature about correlations between cognition and these neuropathological markers. Looking at biopsy material is especially instructive. Because patients are assessed at milder levels of dementia, their cognitive scores are more reliable, and because the amount of tissue to be examined is necessarily limited, correlations which emerge from such studies carry some strong weight, even though the total amount of tissue to be evaluated is less than one would like. In our study of synapse counts in biopsied patients (3) there was a strong correlation of synapse number with cognition, whereas the correlation of cholinesterase transferase activity in cortex was weakly correlated with cognition, and neuritic plaques did not correlate at all with cognition in this series (5 neuropathological and cognitive correlations). This finding was similar to that of Neary who found that pyramidal cell loss, and not plaques or tangles, had the best correlation with cognition (9). Terry et al. (11) subsequently confirmed the correlation of synapse counts with cognition, using autopsy samples taken from patients who had been cognitively evaluated within a reasonable period prior to death. It is worth noting that our synapse counts were guided by earlier studies we performed looking at the locus within the frontal association cortical lamina of the maximal neurodegeneration in AD (4). That study indicated that the greatest loss of sialogangliosides per neuron, a neurochemical marker for axodendritic expanse, were in lamina 2-3 and 5. We, therefore, did initial ultrastructural synapse counts in biopsy cases (as well as earlier autopsy cases in those lamina) (10). Terry
COMMENTARY ON DICKSON ET AL.
299
1.2 1 sic
O
O
0.8
O
0.6 0.4-
0.2-
/
°
0 0
0.2
0.4
0.6
0.8
1
O.D. (SY 38) FIG. 1. Comparison of EP10 and SY38 by ELISA. Cortex from both AD and neurologically normal control patients were assessed for both EP l0 and SY38 immunoreactivity under identical conditions. The correlation of the two antigens is r 2 = 0.926, p < 0.0001.
and Masliah's data from confocal microscopy studies of synaptophysin immunochemistry also looked in lamina 2-3 and 5. Thus, the laminar cortical areas maximally affected by Alzheimer's disease were the specific regions evaluated in the studies that turned up strong (and i:a fact remarkably similar) correlations between synapse counts and cognition. The quest for the "Holy Grail" needs to address several specific issues. First, does EP10, the antigenic marker utilized for the quantitation of synapse,,; by Dickson et al., measure something similar to that measured by SY38, the synaptophysin antibody which was used in the studies by Terry and Masliah? The answer to that appears to be yes. Figure 1 shows a correlation of the amount of SY38 antigen compared to EP 10 antigen from identical aliquots of homogenates of cerebral cortex. We utilized the standard ELISA for SY38 and the EPI0 ELISA used the method of Honer et al, which initially reported a decrement in this antigen in AD (6). So it appears that EP10 does reflect the same relative amount of synaptic antigen as is detected by the more traditional SY38 antibody. Second, were the measurements done in a fashion in the Dickson et al, study which were comparable to the studies which found a strong correlation of synapses with cognition? The answer here is no for two reasons. Dementias other than AD were included and homogenates were used, which is not a fair comparison. Homogenates increase the inclusion of cortical lamina which may be much less affected, specifically those which have fewer large neurons (9-11). Homogenate studies would include these less affected lamina. In addition, we (3,10) have shown
that the mean size of residual synapses is enlarged in AD, thus synaptic biochemical markers in homogenates do not represent the same number of individual synapses within any unit volume of AD tissue. Because the synapses are enlarged, homogenates will tend to underestimate the number of synapses in a given tissue volume although there may be fewer of them. This might distort or blunt a correlation with cognition. Third, Dickson et al. assert, on the basis of their data, that "synapses aren't the best measure of cognition." There were no claims made in either our initial studies or in the Terry 1991 study (11) for a correlation of cortical synapse counts with any dementia other than Alzheimer's disease. The regions were selected for analysis because of the topography of Alzheimer pathology. Dickson et al. show that homogenate assays of synaptic protein content do not provide the best correlation with cognition in a mixed group of dementias. However, these data do not appear to alter the conclusion that synapse counts done in the vulnerable place in the cortex and focused in regions known to be affected by the disease, are the best correlate of cognition in Alzheimer's disease. The fact that they measured plaques and tangles (AD-specific pathology) might lead the unwary reader to assume that this finding of better correlation of cognition with plaques or tangles rather than with synapses applied to AD specifically. The data in this paper do not address that issue because the AD cases were not separated out for analysis. The synaptic protein assays were done from regions selected for involvement in Alzheimer's disease. The cases examined, however, included dementia of several different etiologies. Why would one expect a synapse change in the cortex in multi-infarction dementia or Lewy body dementia to correlate strongly with cognition if samples were taken from association regions selectively vulnerable to Alzheimer's disease? In vascular dementia the pathologically affected regions are variable and in the latter they may well not be the same as those in Alzheimer's disease or involve the same neuronal loss in that region. The wealth of cognitive studies done in this research study does enable the correlation of more specific cognitive functions (e.g., visuospatial vs. verbal performance) with specific regions of brain, regardless of dementia etiology. The inclusion of other types of dementia besides AD in the analysis, however, even with the removal of controls from the equation, is comparing apples with oranges in respect to synapses. One would not necessarily expect, for example, parietal synapses in the right temporal lobe to be impaired if in a case of vascular dementia there was no specific ischemic structural change in that region. Similarly, if a subcortical denervating lesion was the reason for parietal dysfunction (disconnection syndrome). A similar rationale applies to other neurodegenerative dementias. To correlate plaque and tangle counts, the defining pathologic hallmarks of Alzheimer's disease, with cognition in dementias other than AD does not have an obvious rationale. If one only uses the AD cases for such correlations, it may be that correlations will be lost because the N is much smaller. Alternatively the correlations might be strengthened. In either case, the methodology for quantitative synapse assessment must be carefully considered, and the rationale for looking at neuritic plaques and neurofibrillary tangles in diseases other than AD should be carefully explained and justified.
REFERENCES
I. Blessed, G.; Tomlinson, B.; Roth, M. The association between quantitative measures of dementia and of senilechanges in cerebral grey matter of elderlysubjects. Br. J. Psychiatry 114:797-811; 1968.
2. Bowen, D. M.; Neary, D. Cerebral biopsy in the study of Alzheimer's disease. Bull. Clin. Neurosci. 50:44-8; 1985. 3. DeKosky, S.; Scheff, S. Synapse loss in frontal cortex biopsies in
300
4. 5.
6. 7.
COMMENTARY ON DICKSON ET AL. Alzheimer's disease: Correlation with cognitiveseverity. Ann. Neurol. 27:4577-464; 1990. DeKosky, S. T.; Bass, N. H. Aging, seniledementia, and the intralaminar microchemistryof cerebral cortex. Neurol. 32:1227-1233; 1982. DeKosky, S. T.; Harbaugh, R. E.; Schmitt, F. A. et al. Cortical biopsy in Alzheimer's Disease: Diagnostic accuracy and neurochemical, neuropathological and cognitive correlations. Ann. Neurol. 32:625-632; 1992. Honer, W. G.; Dickson, D. W.; Gleeson, J.; Davies, P. Regional synapticpathologyin Alzheimer'sdisease. Neurobiol. Aging 13:375382; 1992. Mann, D. M.; Marcyniuk, B.; Yates, P. O.; Neary, D.; Snowden, J. S. The progression of the pathologic changes of Alzheimer's disease in frontal and temporal neocortexexaminedboth at biopsy and at autopsy. Neuropathol. Appl. Neurobiol. 14:177-95; 1988.
8. Martin, E. M.; Wilson, R. S.; Penn, R. D.; Fox, J. H.; Clasen, R. A.; Savoy, S. M. Cortical biopsy results in Alzheimer's disease: Correlation with cognitive deficits. Neurol. 37:1201-4; 1987. 9. Neary, D.; Snowden, J. S.; Mann, D. M. et al. Alzheimer'sdisease: A correlative study. J. Neurol. Neurosurg. Psychiat. 49:229-237; 1986. 10. Scheff, S. W.; DeKosky, S. T.; Price, D. A. Quantitativeassessment of cortical synapticdensityin Alzheimer'sdisease. Neurobiol. Aging 11:29-37; 1990. 11. Terry, R. D.; Masliah, E.; Salmon, D. P. et al. Physicalbasis of cognitive alterations in Alzheimer's disease: Synapse loss is the major correlate of cognitiveimpairment. Ann. Neurol. 30:572-580; 1991. 12. Wilcock, G. K.; Esiri, M. M.; Bowen, D. M.; Smith, C. C. T. Alzheimer's disease: Correlation of cortical choline acetyltransferase activity with the severity of dementia and histological abnormalities. J. Neurol. Sci. 57:407-417; 1982.
0197-4580(95)00012-7
COMMENTARY HOW TIGHT IS THE C O N N E C T I O N BETWEEN S Y N A P S E LOSS A N D CLINICAL DEMENTIA? N E I L W. K O W A L L
GRECC (182B), Bedford VAMC, Bedford MA 01730 DENNIS Dickson and colleagues report that the classical markers of Alzheimer neuropathology, i.e., senile plaques (SP) and neurofibrillary tangles (NFT), are better correlates of clinical dementia than synapse loss. This finding is in sharp contrast to the work of Robert Terry and colleagues who showed that the best correlate of dementia in Alzheimer's disease (AD) is synapse loss (19). Neuronal loss is a less powerful correlate whereas NFT and SP weakly or inconsistently correlate with dementia. It is possible that the discrepancy between these two reports can largely be explained by differences in methodology and sample selection. One notable problem, as Dickson et al. notes, is the heterogeneity of their clinical population. Of the 45 patients prospectively studied, only 13 had pure AD. Surely one cannot expect that the clinical severity of ischemic dementia, for example, will correlate with NFT counts. The results, therefore, may be difficult to interpret. Indeed, to make a valid comparison with the results of Terry et al., the patient population and methodologies must be comparable. The staining procedures, antibodies and quantitative methods used by Dickson et al. are significantly different than those used by Terry et al. It is notable that the average cortical synaptophysin immunoreactivity of Dickson et al.'s demented patients was not significantly different from their nondemented patients. They do not report if this is also true, however, in their AD patient subset. Cortical synapse loss has been documented in Huntington's disease (HD) (26), Parkinson's disease with dementia (26), prion diseases (8), and even in AIDS encephalopathy (24). It would
not be surprising if synapse loss were associated with all dementing illnesses as an expected consequence of neuronal depletion. If a pathologic process preferentially affects axons or synapses, the loss of specialized outgrowths may outpace neuronal loss and more closely reflect the extent of damage, but even if true, does this knowledge elucidate the process of cell death or its ultimate cause? The concept of pathoclisis or selective vulnerability expounded by the Vogts (22) has been a guiding principle of modern neuropathology. Specific subsets of neurons are differentially vulnerable to particular insults, for example, in ischemia, epilepsy, trauma, and neurodegenerative diseases. One of the best examples is Huntington's disease (HD) where the pattern of differential vulnerability is identical to that produced experimentally by NMDA-type glutamate excitotoxins (4) and mitochondrial electron transport inhibitors (3). The observed pattern of neuronal damage in postmortem human brain has led to a better understanding of possible disease-related mechanisms and therapeutic strategies (2). The presence of synapse loss in the striatum (11) and cerebral cortex (26) in HD adds little to our conceptual understanding of pathogenesis. The more important question might be whether a specific class or type of synapse is lost. Indeed, Beach and McGeer reported depletion of cholinergic fibers in the visual cortex of AD brain when synaptophysin density is unchanged (1). Is synaptophysin, the most popular synaptic protein identified as an indicator of synapses in general, truly equivalent to raw synapse count? Is it uniformly distributed? Can it be downregulated without synapse loss? These