Brain Advance Access originally published online on May 6, 2004
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Brain, Vol. 127, No. 6, 1237-1251, 2004
© 2004 Guarantors of Brain
doi: 10.1093/brain/awh132
Cortical selective vulnerability in motor neuron disease: a morphometric study
Departments of 1 Neuropathology, 2 Neurology and 3 Biostatistics and Computing, Institute of Psychiatry and 4 Department of Anatomy and Developmental Biology, University College London, London, UK and 5 Department of Psychiatry, Education and Resource Centre, Royal College of Surgeons in Ireland, Beaumont Road, Dublin 9, Dublin, Ireland
Correspondence to: P. Nigel Leigh, PO 41, Department of Neurology, Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK. E-mail: spgtpnl{at}iop.kcl.ac.uk
 |
Summary |
|---|
 Top Summary IntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Neuroimaging and neuropsychological studies have revealed that the primary motor cortex (PMC) and the extramotor cortical areas are functionally abnormal in motor neuron disease (MND, amyotrophic lateral sclerosis), but the nature of the cortical lesions that underlie these changes is poorly understood. In particular, there have been few attempts to quantify neuronal loss in the PMC and in other cortical areas in MND. We used SMI-32, an antibody against an epitope on non-phosphorylated neurofilament heavy chain, to analyse the size and density of SMI-32-positive cortical pyramidal neurons in layer V of the PMC, the dorsolateral prefrontal cortex (DLPFC) and the supragenual anterior cingulate cortex (ACC) in 13 MND and eight control subjects. There was a statistically significant reduction in the density of SMI-32-immunoreactive (IR) pyramidal neurons within cortical layer V in the PMC, the DLPFC and the ACC in MND subjects compared with controls [t (19) = 2.91, P = 0.009; estimated reduction 25%; 95% CI = 8%, 40%]. In addition, we studied the density and size of interneurons immunoreactive for the calcium-binding proteins calbindin-D28K (CB), parvalbumin (PV) and calretinin (CR) in the same areas (PMC, DLPFC and ACC). Statistically significant differences in the densities of CB-IR neurons were observed within cortical layers V (P = 0.003) and VI (P = 0.001) in MND cases compared with controls. The densities of CR- and PV-IR neurons were not significantly different between MND and control cases, although there were trends towards reductions of CR-IR neuronal density within the same layers and of PV-IR neuronal density within cortical layer VI. Loss of pyramidal neurons and of GABAergic interneurons is more widespread than has been appreciated and is present in areas associated with neuroimaging and cognitive abnormalities in MND. These findings support the notion that MND should be considered a multisystem disorder.
Key Words: motor neuron disease; selective vulnerability; primary motor cortex; dorsolateral prefrontal cortex; anterior cingulate cortex
Abbreviations: ACC = anterior cingulate cortex; CB = calbindin-D28K; CBP = calcium-binding protein; CR = calretinin; DLPFC = dorsolateral prefrontal cortex; DMND = motor neuron disease with frontotemporal dementia; ICI = intracortical inhibition; IR = immunoreactive; LMN = lower motor neuron; MND = motor neuron disease; PMC = primary motor cortex; PMI = post-mortem interval; PV = parvalbumin; SMI-32 = mouse monoclonal antibody against an epitope of non-phosphorylated neurofilament heavy chain
Received October 22, 2003. Revised December 19, 2003. Accepted December 23, 2003.
|
Introduction |
|---|
|
TopSummaryIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Motor neuron disease (MND, amyotrophic lateral sclerosis) involves progressive degeneration of upper (corticospinal) and lower (brainstem and spinal cord) motor neurons. However, several lines of evidence suggest that extramotor cortical regions are also affected in this disease. For example, a PET study revealed a significant reduction in cerebral glucose metabolism in the motor and frontal cortices in MND subjects compared with controls (Ludolph et al., 1992
). In addition, activation studies have revealed significant involvement of the medial prefrontal cortex (Brodmann areas 9 and 10), the anterior cingulate region (Brodmann areas 9 and 32), the parahippocampal gyrus and the anterior thalamic nuclear complex as well as other non-motor areas (Kew et al., 1993a, b). There is now compelling evidence that a substantial proportion of non-demented MND patients have significant cognitive deficits, typically of the frontal lobe type (Gallassi et al., 1985; David and Gillham, 1986; Kew et al., 1993a; Massman et al., 1996; Chari et al., 1996; Abrahams et al., 1995, 1996, 2000; Bak and Hodges, 1999; Strong et al., 1999; Lomen-Hoerth et al., 2003).
Involvement of extramotor areas is also evident in patients in whom MND is associated with dementia of the frontal lobe type (DMND). In the majority of these cases, the lower motor neuron (LMN) pathology is identical to that seen in typical MND without dementia (Wightman et al., 1992). This and other observations suggest that typical MND and DMND form a phenotypic spectrum, with ubiquitin-immunoreactive (IR) inclusions in the most affected neurons as the characteristic feature (Okamoto et al., 1992; Wightman et al., 1992; Jackson et al., 1996; Nakano, 2000; Al-Sarraj et al., 2002). Thus, MND can be regarded as a multisystem disorder with a predilection for motor and frontotemporal regions of the cerebral cortex.
While specific mechanisms underlying the selective degeneration of motor neurons in MND remain unknown, glutamate-mediated excitotoxicity may contribute to neuronal damage (Shaw, 1994; Zeman et al., 1994; Rothstein, 1995, 1996; Leigh and Meldrum, 1996; Shaw and Ince, 1997; Cleveland and Rothstein, 2001). In addition to changes in glutamate uptake by astrocytes (Maragakis and Rothstein, 2001), abnormal GABAergic neurotransmission in MND might contribute to excitotoxicity. Using paired conditioning-test magnetic brain stimulation, significantly reduced intracortical inhibition (ICI) was observed in MND subjects (Ziemann et al., 1997; Caramia et al., 2000; Zanette et al., 2002a, b). PET studies using the benzodiazepine GABAA receptor ligand [11C]flumazenil have shown decreased flumazenil binding in the primary motor and premotor cortices (Brodmann areas 4 and 6), the right dorsomedial prefrontal cortex (Brodmann areas 9 and 10) and the left dorsal prefrontal cortex (Brodmann area 9) in MND patients (Lloyd et al., 2000), suggesting a disturbance of GABAergic neurotransmission.
Despite many neuropathological studies describing loss of giant pyramidal cells of Betz within cortical layer V of the primary motor cortex (PMC) (Brownell et al., 1970; Hammer et al., 1979; Hughes, 1982; Udaka et al., 1986; Kiernan and Hudson, 1991; Murayama et al., 1992; Troost et al., 1992; Bergmann, 1993; Ince et al., 1993; Nihei et al., 1993; Sasaki and Maruyama, 1994), there have been few rigorous studies of layer V pyramidal neurons, including Betz cells, in the PMC of MND subjects and the findings have been inconsistent. In a previous study using two-dimensional methods of neuronal quantification, a significant loss of Betz cells and other pyramidal neurons within layer V was observed in the precentral gyrus in 10 MND cases (Kiernan and Hudson, 1991). Similarly, Nihei et al., 1993 found loss of SMI-32 (mouse monoclonal antibody against an epitope of non-phosphorylated neurofilament heavy chain)-IR pyramidal neurons including Betz cells within cortical layer V in the PMC in four MND cases (two cases with <60% depletion of Betz cells and two cases with 90–100% depletion of Betz cells). However, using a stereological approach, Gredal et al. (2000) could not detect neuronal loss in the neocortex or PMC in eight MND subjects (including one case with MND and dementia). It is possible that methodological differences, such as post-mortem interval (PMI), fixation time, number of cases used, and counting methods, account for these discrepancies.
Some studies have suggested that sporadic and familial MND may not always be associated with loss of large pyramidal neurons or Betz cells (Davison, 1941; Brownell et al., 1970; Troost et al., 1992; Chou, 1995; Ince et al., 1996). Indeed, it is assumed that pyramidal projection neurons other than Betz cells degenerate because Betz cells account for only 2–3% of the total number of pyramidal tract fibres (Lassek, 1940; Nyberg-Hansen and Rinvik, 1963). Studies on the pathology of the corticospinal system in MND without dementia have focused on the PMC and have not systematically examined other extramotor cortical areas which have been shown by neuroimaging and cognitive studies to be involved in MND and where loss of cortical projection neurons and interneurons might be expected (Ellis et al., 2001).
In addition to loss of pyramidal neurons in the PMC, there is evidence that GABAergic interneurons may be involved in MND. Subpopulations of GABAergic inhibitory neurons can be immunohistochemically labelled for the calcium-binding proteins (CBPs) parvalbumin (PV), calbindin-D28K (CB) and calretinin (CR) (Demeulemeester et al., 1988; Celio, 1990; Celio et al., 1990; DeFelipe, 1997). These CBPs are thought to provide intracellular calcium buffering, preventing hyperpolarization of cells, and possibly protecting them from excitotoxic damage (Heizmann and Braun, 1992; Ince et al., 1993). Using PV as a marker for a subpopulation of GABAergic inhibitory neurons, previous studies have demonstrated a loss of PV-IR neurons in the PMC of MND subjects in comparison with control subjects (Nihei et al., 1992, 1993). Also, in two DMND cases, a reduction in the number of CB-IR neurons in the frontal cortex but preservation of PV-IR neurons was indicated (Ferrer et al., 1993).
The purpose of this study was to determine whether MND is associated with a loss of SMI-32-IR pyramidal neurons and GABAergic interneurons in the PMC (Brodmann area 4), but also in two of the extramotor cortical areas shown by functional imaging and cognitive studies to be involved in MND, in particular the dorsolateral prefrontal cortex (DLPFC; Brodmann area 9) and the supragenual anterior cingulate cortex (ACC; Brodmann area 24c) (Gallassi et al., 1985; David and Gillham, 1986; (Kew et al., 1993a; Abrahams et al., 1995, 1996, 2000; Chari et al., 1996; Massman et al., 1996; Bak and Hodges, 1999; Strong et al., 1999; Lomen-Hoerth et al., 2003). Densities and sizes of layer V pyramidal neurons, labelled by SMI-32 in the PMC, the DLPFC and the ACC, in MND and control subjects were measured using quantitative image analysis. We used SMI-32, a widely-used mouse monoclonal antibody raised against a non-phosphorylated form of neurofilament protein, because it preferentially stains both pyramidal neurons (and Betz cells) and motor neurons (Morrison et al., 1987; Campbell and Morrison, 1989; Hof et al., 1990, 1996, 2002; Mesulam and Geula, 1991; Del Río and DeFelipe, 1994; Carriedo et al., 1995, 1996; Baleydier et al., 1997; Macdonald et al., 1997; Urushitani et al., 1998; Vandenberghe et al., 2000a, b, 2001; Comoletti et al., 2001; Saroff et al., 2000; Tsang et al., 2000). In addition, we applied a similar quantitative approach with antibodies against CBPs to investigate selective vulnerability of GABAergic interneurons in the same cortical areas. This study advances our understanding of the selective vulnerability in MND and of the cellular basis of the cognitive and functional imaging abnormalities now recognized as integral to the disease.
|
Material and methods |
|---|
|
TopSummaryIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Tissue samples
Human post-mortem brain tissue samples were obtained from the MRC Brain Bank (Department of Neuropathology, Institute of Psychiatry, King’s College London, London, UK). Informed consent for autopsy and use of tissues for research was given by subjects and next of kin, according to the protocols approved for the Institute of Psychiatry Brain Bank by the Institute of Psychiatry Ethics Committee. Tissue donation was approved by the local research ethics committee. There were 13 cases diagnosed with MND according to El Escorial World Federation of Neurology criteria. The mean age of MND subjects was 60.9 ± 3.5 (SEM) years. There were eight control subjects without previous history of neurological or psychiatric disease with a mean age of 59.5 ± 2.8 (SEM) years. The demographic details of individual cases used in the investigation are shown in Table 1. All the brain specimens had been evaluated both macroscopically and microscopically by neuropathologists (Department of Neuropathology, Institute of Psychiatry, King’s College London, London, UK) for any evidence of coincidental disease. MND subjects with dementia were excluded from this study.
|
Cortical tissue blocks were removed during clinical neuropathological examination. Brains were hemisected, and then either the left or right hemispheres were cut into coronal slices of 2 cm thickness. Three cortical areas, PMC, DLPFC and ACC, were selected for analysis (Fig. 1). Tissue blocks were fixed in 10% formalin and processed into paraffin wax.
|
Histological labelling of tissue sections
Twenty 10 µm thick serial sections of PMC, DLPFC and ACC from each case were cut using a sliding microtome (model SM2400; Leica, Reichert, Austria). From these 20 sections, four series of five sections were sampled for assessment of SMI-32 and for the immunostaining of neurons for the CBPs PV, CR and CB. The remaining one in five serial sections was stained with luxol fast blue and cresyl violet in order to determine cytoarchitecture and cellular morphology. The cytoarchitecture of the PMC, DLPFC and ACC was identified according to well-defined microscopic criteria, as described by Brodal (1992
), Petrides and Pandya (1999) and Vogt et al. (1995), respectively.
Immunohistochemical labelling of tissue sections
Tissue sections were dewaxed in xylene and rehydrated through a series of graded alcohols to water. In order to reduce endogenous peroxidase activity, the sections were pretreated with methanol containing 2.5% H2O2 for 50 min, and then rinsed with deionized water. Sections were microwaved for 30 min in deionized water (for SMI-32) or in 0.01 M sodium citrate buffer (pH 6.0) (for PV, CR and CB) to aid antigen retrieval. Following a brief rinse with Tris-buffered saline (TBS, pH 7.6), sections were preincubated with normal rabbit serum (Dako, Cambridge, UK; 1 : 10), diluted in TBS, 0.5% bovine serum albumin and 0.2% Triton X-100 for 1 h at room temperature. Sections were then incubated with a mouse monoclonal antibody to non-phosphorylated neurofilament protein (SMI-32, ascites fluid, 1 : 2000; Sternberger Monoclonals, MD, USA) and mouse monoclonal antibodies to the CBPs PV (lyophilized ascites, 1 : 10000; SWant, Bellinzona, Switzerland), CR (lyophilized ascites, 1 : 5000; SWant) and CB (lyophilized ascites, 1 : 500–1000, SWant), diluted in TBS and 0.1% Triton X-100 incubated for 24–48 h at 4°C. The sections were washed three times with TBS and incubated with biotinylated secondary antibody (rabbit anti-mouse IgG; Dako; 1 : 200) for 1 h at room temperature. After several washes, the sections were subsequently incubated with avidin–biotin–label complex (Dako), washed in TBS, and finally incubated for 3–10 min with 0.2 mg/ml 3,3'-diaminobenzidine chromogen (Sigma Chemical, Poole, UK) in TBS containing 0.07% H2O2 under microscopic observation to visualize immunoreactive cells. Nuclei were lightly counterstained with Harris’s haematoxylin. Sections were then dehydrated through a graded series of alcohols, cleared in xylene and mounted in DePeX (BDH, Merck Eurolab, UK) for analysis. To minimize variability in immunostaining, three sections from every case and control were processed at the same time for each antigen, and all conditions were maintained constant throughout the procedures. Negative controls were processed without primary antibodies.
Quantitative analysis: neuronal density and size
A computerized image analysis system with a standardized setting was used to analyse immunolabelled tissue sections. Three tissue sections per case were viewed using a 20x objective with a colour video camera mounted onto a Leica DMLB microscope (Leica, Wetzlar, Germany). Using Image Pro-Plus 4.1 software (Media Cybernetics, MD, USA) with a Marzhauser 100 x 100 x, y motorized stage, a series of contiguous images were captured from a demarcated area extending from the pial surface to the grey–white matter border, forming a single tiled image. The dimensions for each frame were 26.560 x 19.7760 µm. Final composite images spanning 3195 µm in the x-axis and 3017–3218 µm in the y-axis, depending on the cortical width, were used for counting individual cases. All immunopositive neurons were manually outlined and neuronal size and density were recorded according to laminar location (except for weakly labelled CB-IR pyramidal neurons in cortical layers II and III). For each individual cortical layer, an areal density value was calculated by dividing the total number of immunopositive cells by the entire area of grey matter per layer and a neuronal size value as the median size of all immunopositive cells within the layer. The thickness of the section was measured by using a z-axis microcator (Heidenhain, London, UK). All cases were assessed blind to case details.
Statistical analysis
SMI-32-IR neuronal density and size data were first log-transformed to normalize the distributions, and then analysed separately using a repeated measures analysis of variance (SPSS program, version 10.0 for Windows) for group and regional comparisons. Overall group differences (main effects of MND and control groups) were tested using t-tests and any dependence of group differences on region (interaction between groups and areas 4, 9 or 24c) was assessed by Wilk’s lambda multivariate test. If the main effect of group was significant, the density or size changes of MND cases relative to controls were estimated and 95% confidence intervals were calculated. A significance level of 0.05 was used for assessing data from SMI-32-IR neurons (1 layer-wise comparison).
Repeated measures analysis of variance was also used for group and regional comparisons of different subpopulations of CBP-IR neuronal densities and sizes. Neuronal size and density were analysed separately for each of the six cortical layers of areas 4 and 9 and the five cortical layers of area 24c. Due to the limited number of cases expressing PV-IR neurons within cortical layer I (n = 1) and CB-IR neurons within cortical layers I (n = 5) and IV (n = 1) in the three regions studied, statistical analysis of neuronal densities and sizes was not appropriate. A value of 0.05 was added to the density values in the three regions (areas 4, 9 and 24c) of a cortical layer (I, II, III, V or VI) and in the two regions (areas 4 and 9) of a cortical layer IV when zero densities were recorded. Density and size data were then log-transformed to normalize the distributions. Overall group differences (main effects of MND and control groups) were tested using t-tests and any dependence of group differences on region (interaction between groups and areas 4, 9, or 24c) was assessed using Wilk’s lambda multivariate test. The Bonferroni correction was used to adjust for multiple layer-wise neuronal density and size tests, yielding a Bonferroni-adjusted significance level of 0.01, 0.0083 and 0.0125 for assessing data from neurons positive for PV (five layer-wise comparisons), CR (six layers) and CB (four layers), respectively. If the main effect of group was significant, the densities or size reductions of MND cases relative to controls were estimated and 95% confidence intervals were calculated. Repeated measures analysis of variance was also used for group and regional comparisons of mean cortical thickness in the two groups.
Demographic and other variables (Table 1), which showed large differences between the groups in the sample, were considered as potential confounding variables. These included PMI, fixation time and hemisphere side. All repeated measures analyses were carried out with adjustment for each confounding variable by including the variable as a covariate. To assess the effects of potential confounding variables, P-values and estimated reductions in neuronal density and size were compared between the unadjusted analyses and the adjusted analyses.
Two-dimensional areal densities could provide biased estimates with the size of the estimation bias, depending on cell sizes in relation to section thickness (Abercrombie, 1946). Therefore, if there was any statistical evidence of a group difference in neuronal size after Wilk’s lambda test, neuronal density data were re-analysed using three-dimensional density estimates calculated according to the Abercrombie correction formula (Abercrombie, 1946). For convenience, to be able to compare Abercrombie-corrected densities with uncorrected values, three-dimensional densities were multiplied by a factor to change them into two-dimensional values.
|
Results |
|---|
|
TopSummaryIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Distribution of SMI-32 immunoreactivity
SMI-32 primarily labelled the perikarya and dendrites of a subpopulation of pyramidal neurons predominantly located within cortical layers III and V in all regions (Fig. 2A). Some labelled pyramidal neurons were also found within cortical layer IV, close to the border of layer III in the PMC and the PFC. Staining intensity increased with neuronal size, large pyramidal neurons showing more intense immunoreactivity than smaller ones. Giant pyramidal cells of Betz within layer V of the PMC were also intensely immunoreactive for SMI-32.
|
SMI-32 neuronal density and size
The results of the regional and group comparisons of the median neuronal densities and sizes of SMI-32-IR neurons are shown in Tables 2A and B. There was a statistically significant overall group reduction in density of SMI-32-IR pyramidal neurons within cortical layer V in the PMC, the DLPFC and the ACC in MND cases compared with controls [t (19) = 2.91, P = 0.009; estimated reduction 25%; 95% confidence interval (CI) = 8%, 40%] (Fig. 3, Table 2A).
|
|
Potential confounding variables
Although the independent sample t-tests and the Fisher’s exact tests (Table 1) showed no significant population differences in any of the variables between the two groups, there were considerable group differences in PMI, fixation time and side of hemisphere in the sample. Therefore, these variables were considered as potential confounders and added to the repeated measure analyses. These analyses showed that the P-values of overall group comparisons of the density of the SMI-32-IR neurons were still statistically significant. The estimated density reductions for SMI-32-IR neurons within cortical layer V of the three regions before adjusting for confounders was 25%, compared with 27, 25 and 28% after adjusting for PMI, fixation time and hemisphere side.
Comparisons of the median sizes of SMI-32-IR pyramidal neurons between MND and control cases are also shown in Table 2B. The SMI-32-IR pyramidal neurons within cortical layer V of the MND group appeared larger than those in the control group, and this difference approached significance [t (19) = 2.04, P = 0.056; estimated increase 11%; 95% CI = 0%, 24%] (Table 2B). This became significant when allowance was made for fixation time (P = 0.043; 95% CI = 0%, 21%). The estimated neuronal size increase for SMI-32-IR neurons within layer V of the three regions before adjusting for confounders was 11%, compared with 10, 12 and 11% after adjusting for PMI, fixation time and hemisphere side.
Since we detected neuronal size differences between the two groups, group differences in two-dimensional densities of SMI-32-IR neurons observed in our study might have been affected by cell size differences. However, this interpretation was not affected by application of the Abercrombie formula (Abercrombie, 1946) to the density data [t (19) = 3.27, P = 0.004; estimated reduction 29%; 95% CI = 12%, 43%].
Distribution of calcium-binding protein immunoreactivity
PV-IR neurons were found in all cortical layers except the molecular layer, but predominated in cortical layers III, IV and V. Morphologically, these cells were mostly multipolar, showing many long processes (Fig. 2B). In layer IV, these processes were oriented vertically towards cortical layer III (not shown). CR-IR neurons showed small rounded perikarya, multipolar or bipolar in shape, and were found in all cortical layers, predominantly in layers II and III (Fig. 2C). CB-IR neurons were also present in all cortical layers, but were predominantly found in layers II and III. Morphologically, they were small multipolar or bipolar neurons with ascending dendrites in the molecular layer, small bipolar neurons in cortical layers II and III (Fig. 2D). The CB antibody used in this investigation also weakly immunolabelled pyramidal neurons in cortical layers II and III.
CBP neuronal density
Layer-wise comparisons of the median neuronal densities of each CBP subpopulation between MND and control cases are shown in Tables 3A, 4A and 5A.
|
Calbindin
Overall group comparisons showed that the density of CB-IR neurons in MND cases was significantly decreased within cortical layers V [t (19) = 3.48, P = 0.003; estimated reduction 66%; 95% CI = 35%, 83%] (Fig. 4A) and VI [t (19) = 4.07, P = 0.001; estimated reduction 64%; 95% CI = 39%, 79%] in all regions (Fig. 4B) when overall group differences were assessed. Wilk’s lambda multivariate test for group and regional interactions revealed a trend for reduced density of CB-IR neurons within cortical layers II [F(2,18) = 4.45; P = 0.03] and V [F(2,18) = 5.45; 18; P = 0.014] in MND cases compared with controls (Table 3A).
|
Parvalbumin
The densities of PV-IR neurons in MND cases were not significantly reduced in any layer in comparison with control cases. However, Wilk’s lambda multivariate test for group and regional interactions showed a decreasing trend within cortical layer VI [F(2,18) = 4.32; P = 0.03] (Table 4A). This decreasing trend was observed in the PMC and the DLPFC but not in the ACC within this layer (Table 4A).
|
Calretinin
The layer-wise densities of CR-IR neurons in MND cases were not significantly different from those of control cases. However, the overall group comparisons revealed a trend towards reduced density of CR-IR neurons in MND cases within cortical layers V [t (19) = 2.31; P = 0.03] and VI [t (19) = 2.74; P = 0.01] compared with controls (Table 5A).
|
Neuronal size for CBP
Layer-wise comparisons of the median neuronal size of each CBP subpopulation between MND and control subjects are also shown in Tables 3B, 4B and 5B. There were no significant reductions in the sizes of PV-, CR- or CB-IR neurons of MND subjects in any cortical layers compared with controls (Tables 3B, 4B and 5B).
Potential confounding variables
The repeated analyses showed that, even after adjusting for the confounders (PMI, fixation time and side of hemisphere), the P-values of overall group comparisons were little affected. The estimated density reduction for CB-IR neurons within layer V of the three regions before adjusting for confounders and after adjusting for PMI, fixation time and hemisphere side was 66, 64, 65 and 67%, respectively. Similarly, the estimated density reduction for CB-IR neurons within layer VI of the three regions before adjusting for confounders and after adjusting for PMI, fixation time and hemisphere side was 64, 63, 61 and 64%, respectively. When the potential confounders were included in the repeated measures analyses of layer-wise neuronal sizes, there was no effect. Therefore, our findings were not affected by any of these potential confounding variables.
Cortical thickness
There was no difference in cortical thickness from the pial surface to the grey–white matter border between MND and control groups (Table 6). Even when potential confounders were included in the analyses, P-values of overall group comparisons of cortical thickness did not reach statistical significance at the 5% level.
|
|
Discussion |
|---|
|
TopSummaryIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
There have been few systematic studies of selective vulnerability of cortical neuronal systems in MND. Recent neuroimaging and cognitive investigations strongly support the notion that the functions of cortical areas other than the PMC are implicated in MND. It is likely that the syndrome of MND comprises a spectrum of cortical involvement ranging from a typical MND (in which there are no detectable cognitive deficits) though subtle degrees of cognitive impairment (typically frontal or frontotemporal in type) to frontotemporal dementia (Gallassi et al., 1985; David and Gillham, 1986
; Kew et al., 1993a; Abrahams et al., 1995, 1996, 2000; Massman et al., 1996; Chari et al., 1996; Bak and Hodges, 1999; Strong et al., 1999; Lomen-Hoerth et al., 2003). In this study we sought to analyse the cellular basis for these clinical features and to clarify the nature of selective vulnerability, since this may be relevant to our understanding of the mechanisms of the cell death.
Pyramidal neurons labelled with SMI-32
We first analysed cortical pyramidal neurons (presumed glutamatergic neurons) using an antibody against an epitope in non-phosphorylated neurofilament heavy chain (SMI-32). The study revealed a significant overall group reduction in the density of SMI-32-IR pyramidal neurons within cortical layer V of the PMC, the DLPFC and the ACC of MND subjects relative to controls. The finding in the PMC confirms the results reported by Nihei et al. (1993). In addition, we have for the first time shown reduced densities of pyramidal neurons in the DLPFC and the ACC. This has important implications for understanding the clinical features, neuroimaging abnormalities and basis of selective vulnerability in MND.
SMI-32 recognizes non-phosphorylated, or partially phosphorylated, neurofilament heavy chain but, because immunoreactivity can be inhibited when phosphorylation is extensive (Sternberger, 1985), it is possible that reduced SMI-32 immunoreactivity in the present study might have been due to abnormal phosphorylation. However, this is unlikely because staining intensity for phosphorylated neurofilament (SMI-31) in the PMC is similar in MND and control cases (Nihei et al., 1993).
We also found that the size of pyramidal neurons in MND brains was significantly greater than in controls. This is somewhat surprising in view of other studies in which the size of pyramidal neurons (and Betz cells) was reduced in MND subjects (Murakami, 1990; Kiernan and Hudson, 1991, 1993; Nihei et al., 1993). However, a small proportion of surviving neurons were reported to be larger (Munoz et al., 1988; Nihei et al., 1993), probably because of neuronal swelling (Barr and Hamilton, 1948; Hammer et al., 1979). We noted that the increase in median neuronal size in the PMC and DLPFC was correlated with decreased neuronal density, suggesting perhaps that smaller pyramidal neurons were more susceptible to neuronal degeneration than larger pyramidal cells, thereby increasing overall neuronal size. However, this correlation was not found in the ACC. Recently, Cabello et al. (2002) attributed an age-dependent increase in the size of melanin-positive neurons in the substantia nigra to hypertrophy of surviving cells in order to compensate for natural cell loss. However, in our study there were no correlations between age and SMI-32-IR neuronal densities in MND and in the DLPFC and ACC (but not the PMC) of controls.
Methodological consideration
The pyramidal neurons in fixed human post-mortem tissues were labelled with SMI-32, a widely used mouse monoclonal antibody raised against a non-phosphorylated form of neurofilament protein (Sternberger, 1985). Loss of SMI-32-IR cortical pyramidal neurons has been reported in patients with Alzheimer’s disease (Morrison et al., 1987; Hof et al., 1990) and Huntington’s disease (Cudkowicz and Kowall, 1990; Macdonald et al., 1997). SMI-32 is known to stain cortical pyramidal neurons, including Betz cells and motor neurons; however, it is not exclusive to these cell types (Morrison et al., 1987; Campbell and Morrison, 1989; Hof et al., 1990, 1996, 2002; Mesulam and Geula, 1991; Del Río and DeFelipe, 1994; Carriedo et al., 1995, 1996; Baleydier et al., 1997; Macdonald et al., 1997; Urushitani et al., 1998; Saroff et al., 2000; Tsang et al., 2000; Vandenberghe et al., 2000a, b; 2001; Comoletti et al., 2001). A small population of SMI-32-IR pyramidal neurons are also immunoreactive for CB (Kondo et al., 1999) and SMI-32-IR pyramidal neurons are usually positive for acetylcholinesterase (Mesulam and Geula, 1991). Nevertheless, most cortical SMI-32-IR neurons are pyramidal neurons. This marker is therefore useful as a means of investigating the selective vulnerability of neuronal populations in MND and other degenerative disorders.
In our study, the sizes and densities of SMI-32-IR neurons were estimated by two-dimensional image analysis, a technique which tends to over-count large cells, thus creating sampling bias, compared with traditional three-dimensional stereological methods. This error was corrected for in the present study by applying an Abercrombie correction, which adjusts two-dimensional cell densities by incorporating a measure of cell size and section thickness (Abercrombie, 1946). While this correction makes some incorrect assumptions with regard to cell shape and orientation when applied to a two-dimensional study (Hedreen, 1998), it was necessary to correct for potential biases in our two-dimensional SMI-32-IR neuronal density estimates. This was because the tissue sections we used were thin and neuronal size differed between groups after adjusting for fixation time. As a result of this correction, the significance level of our density findings increased. Thus, uncorrected neuronal density underestimated group differences (Table 2A). Three-dimensional stereological methods allow direct measurement of cell density in a small, precisely defined volume of tissue regardless of differences in cell size and shape and section thickness. A stereological approach is therefore generally regarded as the most accurate way of estimating cell density. There are, however, advantages in using two-dimensional methods over three-dimensional stereological methods. First, two-dimensional counts of cells are usually carried out on thinner sections. Antibody penetration is therefore more reliable compared with three-dimensional cell counting, which uses thicker sections. Secondly, three-dimensional methods generally use relatively smaller counting frames than two-dimensional methods. Subsequently, two-dimensional studies are more sensitive in detecting differences in the distribution of cells. As a result, two-dimensional counting methods, like our own, which use larger sampling frames, may provide better estimates of cell density, following Abercrombie correction (Benes and Lange, 2001). It is worth noting that our findings in relation to neuronal size are not biased by two-dimensional counting methods.
Several potential confounding factors (age, sex, fixation time, PMI and hemisphere side) could influence neuronal density and/or immunoreactivity. It is important to determine which variables represent confounders and to adjust the analyses for it. In our study, PMI, fixation time and hemisphere side were found to be potential confounders of neuronal density and size. Our analysis was therefore adjusted to take these variables into account. In our data, age and sex could not have been confounding the density differences since the mean age and gender distributions were very similar between MND and control groups (Table 1).
GABAergic interneurons labelled with CBPs
We also analysed different subpopulations of GABAergic interneurons in the cortex using antibodies against CBPs (CB, PV and CR). We found overall reductions in the density of CB-IR neurons within cortical layers V and VI of MND cases in all regions investigated, relative to controls. There was also a reduction in the density of CB-IR neurons within cortical layer II of MND cases in all regions, but this reduction did not reach statistical significance. Although the densities of PV- and CR-IR neurons were not significantly different between groups, there were trends towards reduced density of PV-IR neurons within cortical layer VI and of CR-IR neurons within cortical layers V and VI.
These findings provide strong evidence supporting the notion that a defect of GABAergic neurotransmission occurs in MND subjects and that this pattern of selective vulnerability is common to motor and extramotor cortical areas. This has significant implications for our understanding of the selective vulnerability of cortical systems in MND.
Calbindin-immunoreactive neurons
In contrast to our observations, previous studies have not detected loss of CB-IR neurons within the paracentral gyri (motor and sensory cortex) in MND subjects (Nihei et al., 1992; Ince et al., 1993). The differences in the findings of these workers may be due to technical differences in counting cells. Here, we used a computerized image analysis system to analyse areal density of CBP-IR neurons in each cortical layer. This program provides more accurate estimates of cell density, involving automated alignment of contiguous microscopic fields as one large image, and provides rapid analysis of very large microscopic regions (Cotter et al., 2002).
While the earlier studies did not investigate CB-IR neurons in the DLPFC, a dramatic reduction in CB-IR neurons within the frontal cortex has been observed in patients with DMND (Ferrer et al., 1993) and in dementia of the frontal type associated with hereditary spastic paraparesis (Ferrer et al., 1995). Our observations are in keeping with the notion that MND and DMND represent a spectrum of neuronal degeneration characterized by ubiquitin-IR inclusions (Jackson et al., 1996; Ince et al., 1998; Al-Sarraj et al., 2002).
It has been suggested (Greene et al., 2001) that the density of CB-IR neurons may increase with age. We therefore analysed separately correlations between age and CB-IR neuronal densities in the two groups by means of Spearman’s non-parametric analysis, and found no correlation in either MND or control cases.
Parvalbumin- and calretinin-immunoreactive neurons
Nihei et al. (1992, 1993) demonstrated a significant reduction in the overall mean PV-IR neuronal density in the PMC but not in the sensory cortex of MND subjects, and Ince et al. (1993) noted preservation of PV-IR neurons within the paracentral gyri (motor and sensory cortex) of MND subjects. When the total cortical density of PV-IR neurons within the three regions (PMC, DLPFC and ACC) of MND cases was compared with that in controls, there was no significant reduction in the density of PV-IR neurons in any cortical regions we studied. We did, however, find a trend towards decreased density of PV-IR neurons within cortical layer VI of the PMC and the DLPFC, relative to the ACC. Thus, the changes we observed are suggestive of region- and laminar-specific reductions in PV-IR neuronal density, with the ACC remaining well preserved.
We also found a trend towards decreased density of CR-IR neurons within cortical layers V and VI in all cortical regions examined. However, using a computer-assisted image analysis system, Hof et al. (1994) reported that parvalbumin and calretinin appeared to be relatively resistant to the degenerative process in the amyotrophic lateral sclerosis/parkinsonism–dementia complex of Guam cases and sporadic MND. Our findings are thus in keeping with the conclusion that PV-IR and CR-IR neurons are relatively well preserved in MND, in contrast to CB-IR neurons.
For all CBP neuronal subpopulations, the density reductions in MND were mainly detected in cortical layers V and VI, suggesting that damage to these neuronal populations in MND is most pronounced in the deeper cortical layers.
Technical and statistical aspects
Two-dimensional image analysis was also used to estimate neuronal density and size in CBP-IR subpopulations. In this study we showed that there were no significant group differences in layer-wise median neuronal sizes in any of the CBP subpopulations between MND and control cases (Tables 2–4). Therefore, the Abercrombie correction is not likely to have a significant impact on the interpretation of our data. We believe that differences in the two-dimensional estimates of the CB- and PV-IR neuronal densities observed between the groups were unlikely to be significantly biased by neuronal size.
Our sample size could also influence the interpretation of this study. Because of the limited numbers of cases available for this study and the conservative Bonferroni correction for multiple layer-wise testing, the study may have lacked power to detect neuronal loss. Consequently, although we have not found statistically significant differences between groups within some layers of the GABAergic subpopulations, we cannot be sure that densities are unchanged. It is possible that neuronal losses would become statistically significant if the sample size were increased.
Glutamate and GABA interactions in MND
The present study showed that SMI-32-IR pyramidal neurons and GABAergic interneurons degenerate in MND. The similar patterns of selective vulnerability in the PMC and in extramotor areas of MND brains add to the growing body of evidence for extramotor involvement in MND. Loss of these extramotor projection neurons would substantially reduce the connectivity of corticospinal and corticocortical pathways. In addition, widespread loss of cortical pyramidal neurons might indirectly contribute to excitotoxicity. For example, in cats, lesions of the sensorimotor cortex reduced glutamate uptake in its projection areas, such as the ventrolateral thalamic nucleus, the red nucleus, the pontine nuclei, the caudate nucleus and the cervical and lumbar spinal cord (Young et al., 1981). Similarly, lesions of the corticostriatal pathway down-regulated the astroglial specific glutamate transporter (GLT-1) in adult rats (Ginsberg et al., 1995).
The inhibitory activity of GABAergic interneurons is implicated in modulating the overall level of neuronal activity of the cerebral cortex and, in particular, the activity of the glutamatergic pyramidal cells. Obviously, post-mortem studies identify changes present at the end-stage of the disease. Thus, it is not clear whether damage to GABAergic and glutamatergic neurons occur in tandem, or whether damage to GABAergic neurons is an early event leading to increased glutamatergic transmission and excitotoxic damage. Recent electrophysiological studies using transcranial magnetic stimulation suggest that ICI is reduced in the early stages of MND compared with the later stages (Zanette et al., 2002a). Also, cortical thresholds, an index of hyperexcitability, were reported to be lowest early in the course of MND and to increase as the disease progressed (Eisen, 2001). Furthermore, decreased ICI can be reversed by agents that potentiate GABAergic transmission, such as diazepam and gabapentin (Caramia et al., 2000). Riluzole, a glutamate antagonist, has been reported to partially reverse ICI (Stefan et al., 2001) or to have no significant effect on ICI (Sommer et al., 1999; Caramia et al., 2000). Overall, these findings support the concept that an imbalance between GABAergic and glutamatergic neurotransmission in MND might result in increased cortical excitability.
There are several clinical implications of our findings. They may help to explain the neuronal basis of the abnormalities that have found in MND by a variety of neuroimaging techniques (Leigh et al., 2002). In essence, these have shown impaired function in prefrontal areas, including the DLPFC (Brodmann areas 9 and 10) and ACC (Brodmann areas 24 and 32). Furthermore, these findings provide a possible anatomical basis for the cognitive changes that can be detected in a substantial proportion of MND patients (Gallassi et al., 1985; David and Gillham, 1986; (Kew et al., 1993a, b; Abrahams et al., 1995, 1996, 2000; Chari et al., 1996; Massman et al., 1996; Bak and Hodges, 1999; Strong et al., 1999; Lomen-Hoerth et al., 2003). In general, these findings support the notion of a continuum of cerebral involvement from typical MND without dementia to MND with frontotemporal dementia (Kew and Leigh, 1992; Okamoto et al., 1992; Wightman et al., 1992; Jackson et al., 1996; Al-Sarraj et al., 2002; Lomen-Hoerth et al., 2003; Nakano, 2000).
In summary, our studies show that pyramidal neurons and GABAergic interneurons in the PMC and in at least two other cortical areas degenerate in MND. The notion that the degeneration process in the cerebral cortex in MND is restricted to corticospinal tract motor neurons cannot be sustained. Loss of cortical projection neurons for areas such as DLPFC and ACC may explain some of the clinical and neuroimaging changes recorded in MND.
|
Acknowledgements |
|---|
We are grateful to the Motor Neurone Disease Association for funding this work and to the MRC Brain Bank for providing tissue samples. We also thank the people affected by MND who have enabled tissue to be donated for use in this study.
|
References |
|---|
|
TopSummaryIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Abercrombie M. Estimation of nuclear population from microtomic sections. Anat Rev 1946; 94: 239.[CrossRef]
Abrahams S, Leigh PN, Kew JJ, Goldstein LH, Lloyd CM, Brooks DJ. A positron emission tomography study of frontal lobe function (verbal fluency) in amyotrophic lateral sclerosis. J Neurol Sci 1995; 129 Suppl: 44–6.[ISI][Medline]
Abrahams S, Goldstein LH, Kew JJ, Brooks DJ, Lloyd CM, Frith CD, et al. Frontal lobe dysfunction in amyotrophic lateral sclerosis. A PET study. Brain 1996; 119: 2105–20.
Abrahams S, Leigh PN, Harvey A, Vythelingum GN, Grise D, Goldstein LH. Verbal fluency and executive dysfunction in amyotrophic lateral sclerosis (ALS). Neuropsychologia 2000; 38: 734–47.[CrossRef][ISI][Medline]
Al-Sarraj S, Maekawa S, Kibble M, Everall I, Leigh PN. Ubiquitin-only intraneuronal inclusion in the substantia nigra is a characteristic feature of motor neuron disease with dementia. Neuropathol Appl Neurobiol 2002; 28: 120–8.[CrossRef][ISI][Medline]
Bak TH, Hodges JR. Cognition, language and behaviour in motor neurone disease: evidence of frontotemporal dysfunction. [Review]. Dement Geriatr Cogn Disord 1999; 10 Suppl 1: 29–32.
Baleydier C, Achache P, Froment JC. Neurofilament architecture of superior and mesial premotor cortex in the human brain. Neuroreport 1997; 8: 1691–6.[ISI][Medline]
Barr ML, Hamilton JD. A quantitative study of certain morphological changes in spinal motor neurons during axon reaction. J Comp Neurol 1948; 89: 93–121.[CrossRef]
Benes FM, Lange N. Two-dimensional versus three-dimensional cell counting: a practical perspective. Trends Neurosci 2001; 24: 11–7.[CrossRef][ISI][Medline]
Bergmann M. Motor neuron disease/amyotrophic lateral sclerosis–lessons from ubiquitin. Pathol Res Pract 1993; 189: 902–12.[ISI][Medline]
Brodal P. Cerebral cortex. In: Brodal P. The central nervous system: structure and function. New York: Oxford University Press; 1992. p. 398–424.
Brownell B, Oppenheimer DR, Hughes JT. The central nervous system in motor neurone disease. J Neurol Neurosurg Psychiatry 1970; 33: 338–57.[ISI][Medline]
Cabello CR, Thune JJ, Pakkenberg H, Pakkenberg B. Ageing of substantia nigra in humans: cell loss may be compensated by hypertrophy. Neuropathol Appl Neurobiol 2002; 28: 283–91.
Campbell MJ, Morrison JH. Monoclonal antibody to neurofilament protein (SMI-32) labels a subpopulation of pyramidal neurons in the human and monkey neocortex. J Comp Neurol 1989; 282: 191–205.[CrossRef][ISI][Medline]
Caramia MD, Palmieri MG, Desiato MT, Iani C, Scalise A, Telera S, et al. Pharmacologic reversal of cortical hyperexcitability in patients with ALS. Neurology 2000; 54: 58–64.
Carriedo SG, Yin HZ, Lamberta R, Weiss JH. In vitro kainate injury to large, SMI-32(+) spinal neurons is Ca2+ dependent. Neuroreport 1995; 6: 945–8.[ISI][Medline]
Carriedo SG, Yin HZ, Weiss JH. Motor neurons are selectively vulnerable to AMPA/kainate receptor-mediated injury in vitro. J Neurosci 1996; 16: 4069–79.
18 Celio MR. Calbindin D-28k and parvalbumin in the rat nervous system. Neuroscience 1990; 35: 375–475.[CrossRef][ISI][Medline]
Celio MR, Baier W, Scharer L, Gregersen HJ, de Viragh PA, Norman AW. Monoclonal antibodies directed against the calcium binding protein Calbindin D-28k. Cell Calcium 1990; 11: 599–602.[CrossRef][ISI][Medline]
Chari G, Shaw PJ, Sahgal A. Nonverbal visual attention, but not recognition memory of learning, processes are impaired in motor neurone disease. Neuropsychologia 1996; 34: 377–85.[CrossRef][ISI][Medline]
Chou SM. Pathology of motor system disorder. In: Leigh PN, Swash M, editors. Motor neuron disease. London: Springer-Verlag; 1995. p. 53–92.
Cleveland DW, Rothstein JD. From Charcot to Lou Gehrig: deciphering selective motor neuron death in ALS. [Review]. Nat Rev Neurosci 2001; 2: 806–19.[CrossRef][ISI][Medline]
Comoletti D, Muzio V, Capobianco A, Ravizza T, Mennini T. Nitric oxide produced by non-motoneuron cells enhances rat embryonic motoneuron sensitivity to excitotoxins: comparison in mixed neuron/glia or purified cultures. J Neurol Sci 2001; 192: 61–9.[CrossRef][ISI][Medline]
Cotter D, Landau S, Beasley C, Stevenson R, Chana G, MacMillan L, et al. The density and spatial distribution of GABAergic neurons, labelled using calcium binding proteins, in the anterior cingulate cortex in major depressive disorder, bipolar disorder, and schizophrenia. Biol Psychiatry 2002; 51: 377–86.[CrossRef][ISI][Medline]
Cudkowicz M, Kowall NW. Degeneration of pyramidal projection neurons in Huntington’s disease cortex. Ann Neurol 1990; 27: 200–4.[CrossRef][ISI][Medline]
David AS, Gillham RA. Neuropsychological study of motor neuron disease. Psychosomatics 1986; 27: 441–5.
Davison C. Amyotrophic lateral sclerosis: origin and extent of the upper motor neuron lesion. Arch Neurol Psychiatry 1941; 46: 1039–56.
DeFelipe J. Types of neurons, synaptic connections and chemical characteristics of cells immunoreactive for calbindin-D28K, parvalbumin and calretinin in the neocortex. [Review]. J Chem Neuroanat 1997; 14: 1–19.[CrossRef][ISI][Medline]
Del Río MR, DeFelipe J. A study of SMI 32-stained pyramidal cells, parvalbumin-immunoreactive chandelier cells, and presumptive thalamocortical axons in the human temporal neocortex. J Comp Neurol 1994; 342: 389–408.[CrossRef][ISI][Medline]
Demeulemeester H, Vandesande F, Orban GA, Brandon C, Vanderhaeghen JJ. Heterogeneity of GABAergic cells in cat visual cortex. J Neurosci 1988; 8: 988–1000.[Abstract]
Eisen A. Clinical electrophysiology of the upper and lower motor neuron in amyotrophic lateral sclerosis. [Review]. Semin Neurol 2001; 21: 141–54.[CrossRef][ISI][Medline]
Ellis CM, Suckling J, Amaro E Jr, Bullmore ET, Simmons A, Williams SC, et al. Volumetric analysis reveals corticospinal tract degeneration and extramotor involvement in ALS. Neurology 2001; 57: 1571–8.
Ferrer I, Tunon T, Serrano MT, Casas R, Alcantara S, Zujar MJ, et al. Calbindin D-28k and parvalbumin immunoreactivity in the frontal cortex in patients with frontal lobe dementia of non-Alzheimer type associated with amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry 1993; 56: 257–61.[Abstract]
Ferrer I, Olive M, Rivera R, Pou A, Narberhaus B, Ugarte A. Hereditary spastic paraparesis with dementia, amyotrophy and peripheral neuropathy. A neuropathological study. Neuropathol Appl Neurobiol 1995; 21: 255–61.[ISI][Medline]
Gallassi R, Montagna P, Ciardulli C, Lorusso S, Mussuto V, Stracciari A. Cognitive impairment in motor neuron disease. Acta Neurol Scand 1985; 71: 480–4.[ISI][Medline]
Ginsberg SD, Martin LJ, Rothstein JD. Regional deafferentation down-regulates subtypes of glutamate transporter proteins. J Neurochem 1995; 65: 2800–3.