Neuronal and Glial Somal Size in the Prefrontal Cortex:  A Postmortem Morphometric Study of Schizophrenia and Huntington Disease FREE

CONSIDERABLE evidence indicates that the prefrontal cortex of schizophrenic patients is subject to a pathological process, but the nature of this process is far from clear. Morphometric studies of prefrontal cortex in schizophrenic brains undertaken to date have generally focused on estimates of neuronal density and number.^1- 3 Although evaluation of neuronal cell size provides a measure of the structural integrity and viability of neuronal populations, only 1 study of neuronal size in the prefrontal cortex (Brodmann area 10) has been undertaken previously, and no size differences were observed between schizophrenic and control brains.³ In regions of the schizophrenic brain outside the prefrontal cortex, decreased neuronal size has been reported for pyramidal neurons in the hippocampus,^4- 5 substantia nigra,⁶ and locus ceruleus.⁷

Quantitative analysis of glial size is also an unexplored issue in schizophrenia; although clearly relevant to theories of pathogenesis, morphometric analyses have thus far focused on glial number or density.^6,8- 12 Whether there is active gliosis in the schizophrenic brain (ie, an increase in the number of glial cells) has become a focal point of research into causal mechanisms for the disease. Although not without exception,⁸ most reports indicate that glial cell number is unchanged in the schizophrenic cortex and limbic nuclei.^6,9- 12 The absence of gliotic processes has been interpreted as support for a developmental cause for the disease.^13- 15 We have previously observed a trend increase in glial density in Brodmann area 9 of schizophrenic cortex in conjunction with increased neuronal density and slight cortical thinning.² The parallel increases in neuronal and glial density were interpreted as due to tighter cell packing as a consequence of reduced interneuronal neuropil, rather than as an indication of increased glial cell number. Despite the fact that glial enlargement is an obligatory component of gliosis, to our knowledge there are no studies to date that have examined glial cell size in schizophrenia.

Comparison of the morphometric abnormalities observed in the schizophrenic cortex with those found in Huntington disease (HD) could provide insight into whether cellular changes in schizophrenia are the result of neuronal loss, because neurodegeneration of large pyramidal cells is a well-established feature of the cortical abnormalities in late-stage HD.^16- 18 Previous analysis of cell density in HD brains revealed a pattern of cytoarchitectonic abnormality distinct from that found in the schizophrenic prefrontal cortex with markedly (50%) increased glial density accompanied by pronounced (28%) cortical thinning.² The morphometric pattern observed in HD brains is consistent with ongoing neurodegenerative processes and reactive gliosis. Indeed, gliosis has been described previously in both the neostriatum and the prefrontal cortex in HD brains.^18- 21 However, it remains to be determined whether changes in cell size occur in association with the neuronal loss and glial proliferation described in the HD prefrontal cortex.

To further explore the cellular abnormalities underlying schizophrenia and to compare the observed abnormalities with those of an established neurodegenerative disorder, neuronal and glial cell bodies were measured in a subset of the same population of schizophrenic, HD, and control brains analyzed in Selemon et al.² Our results indicate a differential vulnerability of neuronal and glial cell types and involvement of different cortical layers or sublayers in the 2 diseases. Further, they provide evidence for a cellular deficiency short of degeneration in neuronal populations in schizophrenia.

Neuronal and glial cell bodies were measured in the same 40-µm Nissl-stained sections used for our previous study of cell density in the schizophrenic and HD cortex.^2,22 From the 51 cases examined previously, 26 brains with the shortest storage time informalin were selected. Cell sizes were analyzed in prefrontal area 9 of the left hemisphere in 9 schizophrenic, 10 neurologically normal controls, and 7 (grade 3 or 4) HD cases (Figure 1; Figure 2, A; and Table 1). Additionally, 7 schizophrenic and 7 control brains in which the occipital pole was available (Table 1) were selected from the original sample for analysis of cell soma size in area 17 of the left hemisphere. Prefrontal area 9 was defined according to cytoarchitectonic criteria described previously,^22- 23 and area 17 was delineated according to Brodmann²⁴ (Figure 1). The software for the 3-dimensional image analyzer used for measuring neuronal and glial cell bodies was developed at Yale University, New Haven, Conn.²⁵

In each area, cell sizes were analyzed in 3 cortical probes consisting of an uninterrupted series of 3-dimensional counting boxes (70×50×25 µm) that spanned the entire depth of the cortex from the pial surface to the white matter (Figure 1, C). Probes were analyzed in 3 coronal sections representing different rostrocaudal levels of area 9 and area 17, respectively. In each counting box, neuronal and glial cell bodies were traced manually at a magnification of ×2000, using a digitizing pad and a computer mouse. The tracings were made at the focal plane in which cell soma were maximal in size and cell nuclei were clearly visible. Distinctions between neuronal and glial somata were based on morphological criteria described in our previous article.² Soma sizes were estimated by calculating the diameter of every traced cell profile as an equivalent circle (D-circle) with the area measured for that individual neuron or glial cell in its equatorial plane (Figure 1, D). The measured soma size of glial cell was equal to the D-circle of the cell nucleus. In this manner, 300 to 500 neuronal and 400 to 1000 glial cell profiles were measured per brain for a total of 10722 neurons and 19913 glial cells overall.

Four statistical analyses were performed. (1) Multiple regression analysis using cell size as the dependent variable was performed with age, postmortem interval, and time in formalin as independent variables for neuronal and glial sizes in areas 9 and 17. Correlation between cell size and neuronal density, glial density, and cortical thickness also were examined. The correlation between neuronal size and duration of illness was examined for schizophrenic brains. A Bonferroni correction was used for these analyses; the effective α=.006. (2) The mean soma size for neurons and glia measured across all cortical layers represented an average of values from 3 probes. Group means (schizophrenic vs control brains; HD vs control brains) were compared using analysis of variance (ANOVA), followed by the Dunnett test, with α=.05. (3) Mean soma size in individual layers was compared between groups by using single-factor (disease), repeated-measures (cortical layers) ANOVA. Following a significant ANOVA, individual ANOVAs were used for cortical layers to determine differences between the groups by using the Dunnett test. (4) The density (number of cells divided by 0.001 mm³) of different cell size classes was obtained by using the mean (M) and standard deviation (δ) of the control cases to segment cases into small (smaller than M−1δ), medium (M−1δ to M), large (M to M+1δ), and extra large (larger than M+1δ) classes. Differences between groups were analyzed with a repeated-measures (cell-size classes) ANOVA followed by selected contrast analyses with α=.02.

The cytoarchitectonic features of area 9 in normal human brain are described in detail in our previous article.²² Briefly, the presence of large pyramidal neurons in sublayers IIIc and Va and layer VI intermixed with medium and small pyramids is the most prominent feature of prefrontal area 9 (Figure 2, B). Glial architecture is more homogeneous than neuronal architecture. Mean glial size in area 9 was very similar in all cortical layers.

Analysis of variance revealed a significant effect of disease (F[2,127]=4.532, P=.01) on neuronal size. Overall mean neuronal soma size and neuronal sizes in layers III through VI were 4% to 9% smaller in area 9 of schizophrenic brains, reaching significant levels (P=.05) only in layer III (Table 2 and Figure 3, A). Although mean neuronal size was smaller for the schizophrenic group compared with the control group, individual variability in soma sizes was evident. Six of 9 schizophrenic brains used for our study had smaller-than-normal neuronal sizes, while 3 schizophrenic brains fell within the normal range in at least some prefrontal layers (Figure 3, A).

In HD prefrontal cortex, overall mean neuronal size was significantly smaller compared with normal control brains (9%; P=.004). However, significant 10% reductions in soma size were not confined to layer III but also extended to layers V (Vb) and marginally to layer VI of the HD cortex (Table 2; Figure 2, B; and Figure 3, A). No significant differences in mean soma size were found in granular layers II and IV of HD prefrontal cortex. Individual variability was high in HD brains in layers III and V while neurons of layer VI exhibited more consistent decreases in neuronal sizes (Figure 3, A).

Multiple regression analyses of relationships between neuronal somal size and age, postmortem interval, time in formalin, and duration of illness were not significant in any layer or sublayer of area 9 in the 26 brains examined in this study (Figure 4). However, there was a significant correlation between neuronal size and cortical thickness in sublayers IIIb (r²=0.288, P=.003) and IIIc (r²=0.378, P<.001); sublayer Vb (r²=0.328, P=.001); and layer VI (r²=0.345, P=.001). Significant inverse correlations were found between neuronal size and neuronal density in sublayers IIIb and IIIc (r²=0.290, P=.003 and r²=0.538, P<.001, respectively); sublayer Vb (r²=0.287, P=.003); and layer VI (r²=0.320, P=.002); as well as between neuronal size and glial density in layer II (r²=0.254, P=.005); sublayer IIIc (r²=0.278, P=.003); and sublayer Vb (r²=0.308, P=.002). Thus, smaller neuronal size was associated with higher neuronal and glial density and with reduced cortical thickness.

Comparison of histograms of somal size distribution for individual cortical layers revealed departures from the normal distribution of different cell size classes in both schizophrenic and HD brains (Figure 5). The curve of the normal distribution was shifted toward smaller sizes in cortical layers III and V of both the schizophrenic and HD brains compared with that of the controls. A marked shift of the normal distribution for layer VI was observed only in HD brains.

To determine more precisely which population of neurons was most susceptible to changes in soma size, the whole population of measured neurons was divided into 4 cell-size classes (small, medium, large, and extra large; see "Materials and Methods" section). In the schizophrenic cortex, the density of small and/or medium neurons was significantly increased by 70% to 140% in cortical layers II, Va, and Vb, and marginally in layers IV and VI (Table 3). In contrast, the population of large and extra large neurons in the schizophrenic cortex did not exhibit significant changes in packing density in any of the cortical layers except sublayer IIIc. In this sublayer, containing the largest pyramidal neurons, the density of extra large neurons was decreased by 40% (P=.02) compared with control brains. Further sublaminar analyses revealed an increase in the density of large neurons in sublayer IIIa, where the size of large neurons was more like that of medium neurons in other sublayers of layer III and V. We interpret this finding to be consistent with the increases in the density of medium neurons in sublayers Va and Vb.

Statistical analysis of the same prefrontal layers of HD cortex revealed a different pattern of change in the density of the 4 cell-size classes. In all cortical layers of the HD cortex except layers II and IV, the density of ex tra large or large neurons was dramatically (50%-80%) decreased in comparison with control brains (Table 3). In addition, in layer VI, the 23% decrease in the density of large neurons was accompanied by a significant 150% increase in the density of small neurons. The above statistical descriptions are consistent with our microscopic observations that pyramidal neurons are smaller in the prefrontal cortex in both schizophrenic and HD brains (Figure 2, B).

Analysis of variance revealed a significant effect of disease on glial size (F[2,153]=3.879, P=.02). Post hoc contrast analyses revealed that mean glial sizes were not statistically different between schizophrenic and control prefrontal cortexes (Table 4). In contrast, glial cell bodies were significantly larger (P=.006) in the same cortical area of HD brains as compared with control brains.

Multiple regression analysis revealed no significant interactions between glial size and age, postmortem interval, time in formalin, neuronal density, glial density, or cortical thickness in area 9 in the 26 brains examined in this study.

Analysis of histograms of glial size distribution followed the paradigm used for analysis of neuronal size distribution. The analysis did not reveal significant differences in glial size in any of the 4 size classes between normal and schizophrenic prefrontal cortexes (Table 4). In contrast, in all cortical layers (I-VI) of the HD brains there was a marked and significant 100% to 150% increase in the density of large and/or extra large glial cells compared with control brains. These statistical findings confirmed our microscopic observations of the presence of enlarged, chromatin-light glial cell nuclei in all cortical layers of the HD brains and only in these cases.

In contrast to the prefrontal cortex, no significant differences in mean neuronal size or size-dependent density were found between schizophrenic and control brains in any cortical layers in area 17 (Figure 3, B). In addition to the 6 main cortical layers, neuronal sizes were compared in the 3 sublayers of layer IV (IVa, IVb, and IVc) in area 17; again, no significant differences in neuronal sizes were found. Glial sizes also did not differ between the 2 groups.

In area 17, there were no significant correlations between neuronal or glial size and age, postmortem interval, or time in formalin. Neuronal size was, however, related to neuronal density in layers III (P=.005) and sublayers IVa (P=.003) and IVb (P=.001; r²=0.854, n=13). Glial size was not correlated with neuronal density, and neither neuronal nor glial size correlated w ith glial density or cortical thickness in area 17.

To our knowledge, this is the first report of a significant reduction in neuronal size in the schizophrenic prefrontal cortex, size-dependent changes in neuronal density, and quantitative analysis of glial size. In the schizophrenic prefrontal cortex, a downward shift in neuronal size was observed in all layers with a significant reduction in layer III. Despite an overall 17% elevation in neuronal density in area 9,² only the density of small and medium neurons was increased by 70% to 140%; neuron density was not increased for large neurons in schizophrenic brains. If large cells had degenerated, the 70% to 140% increases in laminar density of smaller size classes would have been comparable to those observed for the entire neuronal population (11%-24%)²; instead they were orders of magnitude larger. Therefore, the reduction in neuronal size can be attributed to a downward shift, specifically in the size of large and extra large neurons, such that they inflated the smaller size classes. The observed reduction in mean neuronal size in the context of dramatic increases in the density of smaller neurons suggests that subtle cellular changes, rather than neuronal loss, may be present in the schizophrenic prefrontal cortex. In the present study, the reduction in neuronal sizes averaged across all layers was not significant in the schizophrenic prefrontal cortex, replicating previous negative findings with respect to mean neuronal size.^3,26

Our findings of smaller neurons, revealed only by layer-by-layer analysis, are concordant with earlier qualitative ultrastructural observations on prefrontal cortical neurons in schizophrenic brain biopsy specimens and are indicative of previously unsuspected size-dependent vulnerabilities with respect to cortical architecture and circuitry. In a unique study,²⁷ abnormalities were limited to intracellular alterations in organelle structure, eg, prominent and irregular Golgi apparatus, the presence of small dense granules in the cytoplasm, and synapses without synaptic vesicles. We interpret the smaller neurons observed in our study as consistent with our previous evidence of diminution of neuropil in schizophrenic cortex, as cells with reduced dendritic arbors may have reduced metabolic needs and therefore require less somal cytoplasmic volume.

Differences in mean neuronal size between schizophrenic and control brains, though present in other layers, were significant only for layer III. Furthermore, layer III was the only compartment in the schizophrenic brains in which the increase in density of smaller neurons was accompanied by a significant decrease in the density of extra large cells. These observations suggest that the large pyramidal neurons of layer III (IIIc) may undergo more severe pathological changes in the schizophrenic cortex than large neurons in other layers. It is unclear, however, whether the prominent pathological characteristics of pyramidal cells in layer III of area 9, which are among the largest neurons in the prefrontal cortex,²² represent a selective vulnerability of these cells alone or a vulnerability of large neurons in general. The finding that reductions in the largest prefrontal neurons of layer Va were nearly significant (P=.07) would be compatible with a size-dependent process. However, whether general or selective, the large cell vulnerability observed in the present study was not observed in the visual cortex, and thus appears specific to prefrontal cortex.

In patients with Alzheimer disease, large pyramidal neurons have been found to be susceptible to degenerative changes as decreases in the number of large SMI32-immunoreactive neurons have been found in layers III and V of prefrontal area 9.²⁸ As layer III pyramidal cells are one of the major sources of corticocortical connections,^29- 32 our results and evidence for reduction of spine density on layer III pyramidal neurons^33- 34 suggest a prominent role of corticocortical circuitry in schizophrenia.

In contrast to findings in the anterior cingulate cortex and prefrontal area 10, where a reduction in the density of small neurons has been reported in layers II through VI,³⁵ evidence for loss of small neurons was not observed in any cortical layer in our study. Indeed, the absence of significant decreases in neuronal size in layers II and IV suggests that small nonpyramidal cells are less impaired than large pyramidal cells. Whether the apparent discrepancies in findings represent regional variation or reflect differences in methodological approaches is unclear.

Significant abnormalities in neuronal size were not found in any cortical layers in area 17 in the schizophrenic brains despite the 10% increase in neuronal density observed previously.² These findings suggest that the morphologic changes in the primary visual area are less marked that those in the prefrontal region. Differences in the cytoarchitectonic composition of these 2 areas may account for the lack of change in somal sizes in area 17. Primary sensory cortical areas, including area 17, have a much less prominent layer III and smaller pyramidal cells than associational cortical regions like area 9. Although it could be argued that size changes in somal diameter may be most easily detected in layer III of area 9 because the pyramids are very large, similar decreases in somal size were not detected in the very large nonpyramidal cells in layer IV of area 17, lending further weight to the hypothesis that pyramidal cells, perhaps prefrontal pyramidal cells, are selectively impaired in the disease. Corticocortical synapses, particularly afferent input from other cortical areas, also may not be as numerous in area 17 as in the associational cortical regions.

In the HD prefrontal cortex, mean neuronal size was decreased relative to normal controls in layers III (IIIb, IIIc), Vb, and VI, and in addition, the density of extra large or large neurons was decreased by 50% to 80% in all cortical layers except layers II and IV. Because these decreases were not accompanied by comparable increases in the density of smaller neurons, the reductions cannot be easily attributed to a mere shift in cell size like that observed in the schizophrenic cortex. It is likely that large cells degenerate in HD. Even in layer VI, which exhibited a 150% increase in the density of small neurons, greater cell packing of the small neuron population can be mostly explained by the dramatic thinning of infragranular layers² rather than by the addition of "new" small neurons that have shrunk from the large neuron population. These findings of reduced density of large neurons are in accord with previous observations of loss of large pyramidal neurons in layers III, V, and VI of area 10¹⁸ and in layers V and VI of areas 8 and 9¹⁷ in HD brains. Reductions in somal size and density of large neurons in layers III, V, and VI of the HD prefrontal cortex suggests that corticocortical, corticostriatal, and corticothalamic projection neurons all are compromised in the advanced stages of HD.

In our study, mean glial size and the density of large and extra large glial cells was greatly increased in HD brains relative to normal brains. In addition, increased glial density was found in the same HD brains in our previous study.² These findings concur with reports of increased numbers of oligodendrocytes and density of astrocytes in the prefrontal cortex for all HD grades.¹⁸ Enlarged glial size and increased glial density strongly indicate that neurodegenerative processes and gliotic reactions are present in the HD prefrontal cortex. Whereas glial proliferation in the cortex and reactive gliosis in the striatum have been reported before,^19- 21 glial enlargement has not been well-documented in prior studies of HD.

In contrast to HD brains, in schizophrenic brains only a trend increase in mean glial size was observed in area 9 and in occipital area 17; mean glial size was virtually identical in schizophrenic and control brains. Prior to this study, glial cell size has not been examined despite the fact that glial enlargement is an obligatory component of gliosis. Glial density was estimated previously in the schizophrenic prefrontal cortex and was found to be indistinguishable from normal control brains,^2- 3 although in the latter study there was a trend elevation in the schizophrenic brains with storage time less than 5 years. Thus, the observation of normal glial somal sizes in conjunction with modest and variable increases in glial density suggest that glial cell reactions are not a cardinal feature of the cortical pathophysiology of schizophrenia. The lack of reactive gliosis also provides further evidence that cell loss does not occur in prefrontal and occipital cortexes of the schizophrenic brain.

Although the absence of gliotic reactions in the schizophrenic cortex has been interpreted as evidence for neuronal loss during early development,^13- 15 our findings suggest that degenerative changes in all layers of cortex, with the possible exception of sublayer IIIc, stop short of neuronal loss and therefore would probably not trigger gliosis even in an adult brain. It is unclear whether the smaller neuronal volumes are due to a developmental failure or to cytoplasmic atrophy occurring early in the course of the disease. While it is not entirely possible to rule out cell loss for the select population of very large cells in layer IIIc as cell size changes in this sublayer are similar in schizophrenic and HD brains, absence of gliosis in the schizophrenic brain indicates that if neuronal loss occurred, it was very limited in scope and occurred either developmentally or at the onset of illness, because there is no evidence of ongoing neurodegeneration in the schizophrenic cortex.

Accepted for publication September 22, 1997.

This work was supported by grant 44866 from the National Institute of Mental Health (Center for the Neuroscience of Mental Disorders), Rockville, Md.

Presented at the International Congress on Schizophrenia Research, Hot Springs, Va, April 10, 1995.

The authors thank F. M. Benes, MD; J. E. Kleinman, MD, PhD; M. M. Herman, MD; and I. Delalle, MD, for the donation of postmortem brain tissue. We also acknowledge F. M. Benes, MD; J. E. Kleinman, MD, PhD; I. Delalle, MD; and E. Radonic, MD, for psychiatric diagnoses of these cases. We thank J. Coburn, J. Enwright, J. Paquette, and J. Musco for excellent technical assistance, and M. Richmond and S. Pittman for help in analyzing data. Finally, we are grateful to I. Paul, PhD, and Michael Andrew, PhD, for valuable comments on statistical analyses.

Reprints: Grazyna Rajkowska, PhD, Department of Psychiatry and Human Behavior, University of Mississippi Medical Center, 2500 N State St, Jackson, MS 39216 (e-mail: grazyna@fiona.umsmed.edu).