OUP user menu

Contralateral hemimicrencephaly and clinical–pathological correlations in children with hemimegalencephaly

Noriko Salamon, Marissa Andres, Dennis J. Chute, Snow T. Nguyen, Julia W. Chang, My N. Huynh, P. Sarat Chandra, Veronique M. Andre, Carlos Cepeda, Michael S. Levine, Joao P. Leite, Luciano Neder, Harry V. Vinters, Gary W. Mathern
DOI: http://dx.doi.org/10.1093/brain/awh681 352-365 First published online: 16 November 2005


In paediatric epilepsy surgery patients with hemimegalencephaly (HME; n = 23), this study compared clinical, neuroimaging and pathologic features to discern potential mechanisms for suboptimal post-hemispherectomy developmental outcomes and structural pathogenesis. MRI measured affected and non-affected cerebral hemisphere volumes for HME and non-HME cases, including monozygotic twins where one sibling had HME. Staining against neuronal nuclei (NeuN) determined grey and white matter cell densities and sizes in HME and autopsy cases, including the non-affected side of a HME surgical/autopsy case. By MRI, the affected hemisphere was larger and the non-affected side smaller in HME compared with non-HME children. The affected HME side showed enlarged abnormal deep grey and white matter structures and/or T2-weighted hypointensity in the subcortical white matter in 75% of cases, suggestive of excessive pre-natal neurogenesis and heterotopias. Histopathological examination of the affected HME side revealed immature-appearing neurons in 70%, polymicrogyria (PMG) in 61% and balloon cells in 45% of cases. Compared with autopsy cases, in HME children NeuN cell densities on the affected side were increased in the molecular layer and upper cortex (+244 to +18%), decreased in lower cortical layers (−35%) and increased in the white matter (+139 to +149%). Deep grey matter MRI abnormalities and/or T2-weighted white matter hypointensity correlated with the presence of immature-appearing neurons and PMG on histopathology, decreased NeuN cell densities in lower cortical layers and a positive history of infantile spasms. Post-surgery seizure control was associated with decreased NeuN densities in the molecular layer. In young children with HME and epilepsy, these findings indicate that there are bilateral cerebral hemispheric abnormalities and contralateral hemimicrencephaly is a likely explanation for poorer post-surgery seizure control and cognitive outcomes. In addition, our findings support the hypothesis that HME pathogenesis probably involves somatic mutations that affect each developing cerebral hemisphere differently with more neurons than expected on the HME side.

  • seizures
  • MRI volumetric
  • cortical dysplasia
  • malformations of cortical development
  • unilateral megalencephaly
  • cell cycle
  • neurogenesis
  • corticogenesis
  • tuberous sclerosis complex
  • HME = hemimegalencephaly
  • NeuN = neuronal nuclei
  • PMG = polymicrogyria


Hemimegalencephaly (HME; also termed unilateral megalencephaly) is a relatively rare but clinically impressive malformation of cortical development characterized by marked cerebral asymmetry (Friede, 1989). Originally described by Sims (Sims, 1835), HME can occur in isolation or in association with neurocutaneous syndromes or developmental disorders such as Klippel–Trenaunay syndrome, Ito's hypomelanosis, neurofibromatosis and tuberous sclerosis complex (Vigevano et al., 1996). Non-syndromic HME occurs in multiple ethnic groups without a gender preference and the clinical presentation typically consists of early onset epilepsy, psychomotor retardation and contralateral hemiparesis and hemianopia (Flores-Sarnat, 2002). The aetiology of HME is unknown, although it is presumed to be from abnormalities of neuroglial differentiation and cell migration involving a single hemisphere (De Rosa et al., 1992; Vinters et al., 1992; Arai et al., 1999; Hoffmann et al., 2000).

An approach in determining HME pathogenesis is by evaluating clinical, neuroimaging and histopathological findings for both cerebral hemispheres based on an understanding of normal developmental neurobiology (Rakic, 1988; Uher and Golden, 2000; Volpe, 2000; Miyata et al., 2004; Andres et al., 2005). In mammals, cerebral cortical development begins with progenitor cell proliferation in the periventricular germinal zones, and neural cell divisions and cell cycling are tightly controlled (Marin-Padilla and Marin-Padilla, 1982; Kostovic and Rakic, 1990; Marin-Padilla, 1999; Zecevic and Rakic, 2001; Rakic and Zecevic, 2003). In mice, for example, the number of progenitor cell divisions is limited to 10 or 11, and each successive cell cycle becomes progressively longer (G1-phase) as a larger proportion of progenitor cells differentiate into neurons (Caviness et al., 1995, 2003; Sommer and Rao, 2002). The earliest differentiated cells migrate to the cortical surface forming the preplate, which is partitioned by subsequent generations of neurons into a primordial plexiform layer (eventual molecular layer or Layer 1) and subplate. The pyramidal neurons that form the hexalaminar post-natal neocortex migrate in successive waves to form the cortical grey matter from the inside out (i.e. earliest neurons in lower grey matter). After pyramidal cell migration, cells in the molecular layer and subplate degenerate (probably via apoptosis), and this coincides with secondary gyral folding in the last trimester of human gestation (deAzevedo et al., 2003; Rakic, 2003; Zecevic, 2004).

The larger forebrain and gyral folds of the human cerebral cortex are thought to be from an increase in the number and/or cell cycles of progenitor cells (Rakic, 1995; Haydar et al., 1999; Kuan et al., 2000; Chan et al., 2002). The number of pre-neurogenesis precursors is influenced by mitotic rates, programmed cell death and the proportion of cells that terminally differentiate into neurons at the end of each cell cycle (Rakic, 1988; Hayakawa et al., 1991; Uher and Golden, 2000; Kuzniecky and Barkovich, 2001; Chenn and Walsh, 2002; Caviness et al., 2003). Consequently, small alterations in the number of progenitor cells, especially at later cell cycles, have a dramatic impact on final cerebral size, gyral shapes and cell densities in different layers of the cortex (Chenn and Walsh, 2003; Putz et al., 2005; Tarui et al., 2005).

In a previous study of less severe paediatric epilepsy surgery patients with cortical dysplasia, we found increased upper cortical neuronal packing densities without megalencephaly supporting the hypothesis that non-HME dysplasia is probably due to increased neurogenesis in later progenitor cell cycles and partial failure of post-neurogenesis programmed cell death in the molecular layer and subplate (Andres et al., 2005). We further hypothesized that intractable epilepsy was the consequence of incomplete cortical maturation with preservation of ‘pro-epileptic’ immature neurons and synaptic circuitry (Mathern et al., 2000; Cepeda et al., 2003, 2005a, b; Andre et al., 2004). The purpose of the current study was to apply similar clinical–pathological comparisons to HME cases, a more severe form of cortical dysplasia, to discern if the probable developmental aetiology involved similar or different abnormalities of cerebral development than non-HME dysplasia cases.


Clinical cohort and pre-surgery evaluation

Patients with HME (n = 23) were identified through the University of California, Los Angeles (UCLA) Paediatric Epilepsy Surgery Program database. The cases had surgery between 1990 and 2004, and the clinical protocols have been previously published (Mathern et al., 1999; Jonas et al., 2004, 2005). For this study, HME patients were defined as those surgical cases where most (at least three lobes) or all of one cerebral hemisphere was larger compared with the opposite hemisphere on MRI. Informed consent was obtained to use clinical data for research studies. The standardized pre-surgery evaluation included detailed history and neurological examinations, and interictal and ictal scalp EEG recordings. Neuroimaging studies included the aforementioned MRI and cerebral 2-[18F]fluoro-2-deoxy-d-glucose PET.

MRI and cerebral volume measurements

Quantitative assessments were performed on MRI scans obtained since 2000 using the same inpatient machine and sequence protocols, and were available for 11 HME cases and were compared with 6 non-HME cases. At UCLA, there were very few scans performed in children under 2 years of age without neurological injury or disease on the same MRI machine. Hence, three of the non-HME cases were children without epilepsy, including the monozygotic twin of a HME case (see Fig. 3), and the other three were undergoing epilepsy pre-surgery evaluations for non-HME pathologies (stroke and mild cortical dysplasia) with ages similar to the HME group (Andres et al., 2005). The MRI scans were performed on a General Electric 1.5 tesla Signa scanner (Milwaukee, WI, USA). Sequences included high-resolution coronal T1-weighted spoiled gradient-recalled echo pulse sequences [SPGR; repetition time (TR) 13 ms, echo time (TE) 2.8 ms, inversion time 300 ms, flip angle 25°, field of view (FOV) 24 cm, 1.5 mm coronal thickness slices, 78–124 slices per patient, matrix 256 × 256, number of excitations (NEX) = 1], and coronal T2-weighted images (TR 2000 ms, TE 120 ms, FOV 24 cm, 5 mm thickness slices, matrix 192 × 192, NEX = 2). Volumetric MRI analysis was performed with custom-designed commercial software (Silhouette; CEDARA, Ontario, Canada; www.cedera.com), and coronal T1 SPGR images were used for the analyses. The images were transferred into the Silhouette program and each coronal section was segmented into CSF and brain parenchyma. The volumes from each MRI coronal section were summated and the cerebellum and brainstem regions removed to obtain cerebral hemispheric volumes. For data analysis, each hemisphere was defined as the one affected by the HME (i.e. side of surgery) or the non-affected non-operated side (Andres et al., 2005). For the non-HME cases, the same assignments were applied for the three pre-surgery patients except for the perinatal stroke case where the affected side was excluded, and the affected side was randomly selected for the other three non-HME/non-epilepsy cases.

In addition to the volumetric assessments, the MRI scans of 16 HME cases were qualitatively assessed for neuroimaging abnormalities (11 cases for volumetric plus 5 scans on older MRI machines). The qualitative assessments included (i) the lobes involved with megencephaly; (ii) ventricular enlargement; (iii) a straightened frontal horn of the lateral ventricle; (iv) deep grey matter abnormalities of the caudate nucleus, thalamus, basal ganglia and olfactory tract; (v) enlargement of the cortical grey and white matter at the anterior corpus callosum; and (vi) T2-weighted signal changes (hypo- and hyper-intensity) in the hemispheric white matter.


Brain tissue from surgery for all 23 HME cases was routinely fixed in buffered formalin and multiple serial sections reviewed as previously published (De Rosa et al., 1992; Vinters et al., 1992; O'Kusky et al., 1996). Two neuropathologists (D.J.C. and H.V.V.) reviewed each case and assessed for histopathological signs of cortical dyslamination, polymicrogyria (PMG), dysmorphic cytomegalic neurons and balloon cells, immature-appearing neurons, excessive white matter neurons, glial/neuronal heterotopias and calcifications. PMG was defined as an abnormal arrangement of cell layers with excessive folding of upper layers, and fusion of gyral surfaces (Friede, 1989) (see Fig. 4B and C). Immature-appearing neurons were defined as cells with neuroblast-like features consisting of round to oval nuclear configurations with a thin rim of cytoplasm (Prayson et al., 2002) (see Fig. 4F). One previously reported HME patient died intra-operatively, and homotopic tissue blocks from both hemispheres were available for histopathological review and neuronal nuclei (NeuN) immunocytochemistry (ICC) (Jahan et al., 1997).

Tissue selection for NeuN staining

From 14 HME cases operated since 1998, 1.5–2 cm blocks of neocortex and adjoining white matter involving the crown of a gyrus in an area of severe dysplasia were microsurgically removed at surgery and the blocks immersion fixed for ICC. The remaining brain tissue from the surgical resection was processed for histopathology as described above. Neocortical blocks from 11 autopsy cases of similar ages without known neurological disease were also collected for comparison. Death in the autopsy group was from acute cardiac, septic or traumatic causes, and brain tissue was collected between 3 and 11 h after death (mean ± SD; 6.6 ± 2.3 h).

NeuN ICC processing

NeuN was chosen over traditional Nissl stains because of the specificity of this antibody in identifying differentiated neurons, and the fixation protocols were regimented so that autopsy and surgical tissue were processed in a similar fashion. Surgical and autopsy ICC tissue blocks were immediately immersion fixed in freshly prepared phosphate-buffered 4% paraformaldehyde for 24–48 h, and then cryoprotected overnight in 20% buffered sucrose and stored at −80°C. Cryostat-cut sections (30 μm) were collected and placed in individual 3 ml wells containing 0.05 mol/l Tris–HCl-buffered saline (TBS; pH 7.4). The free-floating sections were processed the same day as follows with 10 min TBS rinses (three changes) between each step. Five minutes in 3% hydrogen peroxide/10% methanol in TBS; 60 min in a blocking solution of 2% normal horse serum in TBS; overnight in primary antisera against NeuN (mouse anti-neuronal nuclei; Chemicon International; Temecula, CA, USA; Catalog # MAB377; 1 : 2000 dilution) diluted in 2% normal blocking serum; 60 min in diluted biotinylated anti-mouse antibody (ABC kit, Vector Laboratories, Burlingame, CA, USA); and 30 min in a solution of excess avidin and biotinylated horseradish peroxidase (ABC Kit, Vector Laboratories). The sections were developed for 7–8 min in 0.5 mg/ml 3,3′-diaminobenzidine tetrahydrochloride and 0.01% hydrogen peroxide. After sufficient colourization, the reaction was halted by washing in several rinses of cold PBS, the sections were mounted on subbed slides, air dried, treated for 35 s in 0.1% osmium tetroxide in 0.1 mol/l phosphate buffer (pH 7.4), dehydrated and coverslipped (Mathern et al., 1995, 1997). For the single autopsy HME case with matched bilateral formalin fixed paraffin blocks of the frontal, parietal and temporal neocortex, 10 μm sections were placed on slides and processed for NeuN ICC as noted above.

NeuN defined cell densities

Nine regions per tissue section were selected in a standardized manner for NeuN cell counts as previously published (Andres et al., 2005). Because of the neocortical dyslamination associated with HME, we selected cell density sample sites based on pre-determined distances from the pial surface or bottom of the cortical ribbon instead of identified neocortical cell layers (see Fig. 5A, D and G). An ocular grid was positioned over the tissue section with the pial surface at the top. For Layer 1, a 10 × 10 box at ×40 magnification (31 μm × 31 μm) was positioned with the superior line parallel to the pial surface, and all NeuN labelled cells within the box were counted except those touching the upper and right borders of the grid. The neocortical grey matter sample sites were labelled Levels 1 (superior) to 6 (inferior), and their location determined by measuring the distance from the bottom of Layer 1 to the junction of the neocortex and white matter and dividing by 6. A 5 × 5 box at ×40 magnification (15.2 μm × 15.2 μm) was positioned at each location, and NeuN-positive cells within the box counted. In the grey matter, the distance between each 5 × 5 box varied by 18.6–21.7 μm from case to case. For the NeuN cells in the superficial white matter, a 3 × 10 box at ×10 magnification (37.2 μm × 124 μm) was positioned just below the neocortical-white matter junction, and cells were counted. At a distance of 24.8 μm below the bottom of that box another 5 × 10 box (62 μm × 124 μm) was positioned, and cell counts performed in the deep white matter. Cell counts were calculated as the number of NeuN cells/10 000 μm2.

It must be emphasized that NeuN-labelled cell densities are estimates of the number of neurons per unit area (i.e. packing density) and not an absolute calculation of the total number of neurons per hemisphere. Experimental techniques used to determine absolute neuronal quantities within a hemisphere or brain region of autopsy cases were not practical in this surgical study because the entire cerebral hemisphere or area of pathology was not available for sampling. Likewise, it is nearly impossible to correct for tissue volume changes that occur from fixation shrinkage, although it can be assumed that it should be the same for HME and autopsy cases with our protocol. However, neuron densities, as used in this study, are reliable relative estimates of packing densities, and statistical differences between groups of patients that are similarly processed and counted can be accurately determined (Mathern et al., 1995, 1997).

Cortical thickness

To assess the average thickness of the neocortex, an image computer was used as previously described (Mathern et al., 1997). The same NeuN stained sections used for neuronal counts were imaged using a video charge-coupled device camera (SPOT RT CCD; v3.2; Diagnostic Instruments, Inc.; Sterling Heights, MI, USA) attached to a Zeiss microscope interfaced with a PC. Once captured, the image was analysed using image system software (Image-Pro Plus, v4.1; Media Cybernetics, Silver Spring, MD, USA). The operator imaged the tissue section at low magnification and outlined for the computer straight portions of the cortical ribbon in a shape as close as possible to a rectangle or trapezoid. Once outlined, the computer measured the perimeter (P) and area (A). The average cortical thickness (CT) was calculated using the quadratic equation: CT = [P − (P2 − 16 × A)1/2]/4. One investigator performed these measures blinded to the pathology classification, and as previously indicated for neuronal densities neocortical thickness measurements should be considered as relative estimates.

NeuN cell size

The same imaging system was used to assess NeuN labelled cell size. Images at ×50 were captured sampling regions from Layer 1, the upper (Levels 2 and 3) and lower (Levels 5 and 6) grey matter and superficial and deep white matter regions. The operator outlined for the computer all individual NeuN labelled cells within the captured image and the computer calculated the average area per cell for that section and region (μm2). Typically, 20 or more NeuN labelled cells were measured for each sample site.

Data analysis

MRI volumes, histopathological findings and NeuN measurements were entered into a database and analysed using a statistical program (StatView 5; SAS Institute, Inc., Cary, NC, USA). Differences between autopsy or non-HME cases and HME patients involving continuous dependent variables were statistically compared using an analysis of covariance (ANCOVA) that included the log of age at surgery or autopsy as co-independent variables. Comparisons using nominal variables were performed using Chi-square tests. Results were considered different at a minimal level of significance of P < 0.05.


Clinical characteristics

Twenty-three HME cases were identified for this clinical–pathological study, and, with the exception of one case of tuberous sclerosis complex, none of the other patients had neurocutaneous or other developmental syndromes associated with HME. In other words, in this HME cohort all but one patient were non-syndromic clinical presentations. All 23 HME cases underwent hemispherectomy, which included 6 anatomical, 5 functional and 12 modified functional procedures (Cook et al., 2004). The HME patients constituted 17% of hemispherectomy cases (n = 138), 9% of extratemporal resections (hemispherectomy + multilobar + non-temporal lobar; n = 255) and 7% of all paediatric epilepsy surgery procedures at our institution (above + temporal lobe; n = 316). There were 12 females and 12 right-sided cases (52% respectively). Mean age (years ± SD) at seizure onset was 0.1 ± 0.15 (range 0–0.5), age at surgery was 1.5 ± 1.2 (range 0.2–4.1) and seizure duration before surgery was 1.4 ± 1.2 (range 0.2–4). A history of infantile spasms was noted in 15 cases (65%), and post-surgery seizure control at last follow-up (mean 4.3 years; range 1–10) was noted in 15 out of 22 (68%) cases. Perioperative complications were noted in four (17%) HME cases, and included one intraoperative death (Jahan et al., 1997), a second case that survived cardiac arrest in the operating room from excessive blood loss, a permanent third nerve palsy and a post-surgery cranial infection successfully treated with antibiotics (Di Rocco and Iannelli, 2000). Four (17%) HME children underwent re-operations for recurrent seizures and possible incomplete disconnections with successful seizure control in one case after repeat surgery.

Cerebral hemisphere volumes and neuronal packing densities are known to dramatically change during the early post-natal period as the brain rapidly grows (Andres et al., 2005). To statistically control for post-natal brain growth we performed an analysis of covariance (ANCOVA), and the results (F/P-values) are shown for MRI-assessed cerebral hemisphere volumes and NeuN-defined cell densities and sizes (Table 1). The ANCOVA incorporated the pathology groups (MRI, HME versus non-HME; NeuN, HME versus autopsy) and log of age at surgery as the independent variables. Statistically significant interactions would indicate different changes with age between pathology groups. No significant interactions were identified meaning that we found either age-related change for all categories and/or differences between HME and non-HME (or autopsy) comparison groups. The remainder of the Results section will sequentially discuss clinical–pathological findings for HME cases.

View this table:
Table 1

ANCOVA statistical results (F/P-values) comparing autopsy (NeuN) or non-HME (MRI) cases with HME patients

MRI/cortical measurePathology category*Log of ageInteractionFigures
MRI assessment
    Total cerebral volumes1.33/0.27364.3/<0.00012.70/0.1261–3
    Affected hemisphere5.75/0.03530.6/0.00012.84/0.1171–3
    Non-affected hemisphere5.65/0.03950.7/<0.00011.13/0.3091–3
    Aff-non-aff hemisphere7.97/0.0151.62/0.2271.0/0.3361–3
    Cortical thickness1.47/0.2410.034/0.8570.382/0.5494–6
Neuronal densities
    Layer 1 NeuN5.61/0.0400.156/0.6950.001/0.9794–6
    Level 1 NeuN1.06/0.3172.62/0.1240.489/0.4944–6
    Level 2 NeuN0.009/0.9242.08/0.1680.382/0.5454–6
    Level 3 NeuN0.492/0.4932.34/0.1451.39/0.2544–6
    Level 4 NeuN7.72/0.0133.67/0.0720.066/0.8014–6
    Level 5 NeuN1.03/0.3253.10/0.0961.40/0.2534–6
    Level 6 NeuN5.39/0.0331.24/0.2812.37/0.1424–6
    Superficial WM NeuN1.23/0.2813.94/0.0630.576/0.458
    Deep WM NeuN1.68/0.2113.62/0.0730.235/0.634
Neuronal size
    Layer 1 NeuN cell size2.38/0.1410.559/0.4650.124/0.729
    Upper GM NeuN cell size0.658/0.4291.71/0.2080.820/0.378
    Lower GM NeuN cell size0.877/0.3610.109/0.7451.68/0.211
    Super. WM NeuN cell size1.89/0.1850.593/0.4510.131/0.722
    Deep WM NeuN cell size0.529/0.4770.205/0.6560.027/0.872
  • Significant values indicated in bold type.

  • * For MRI assessments: HME (n = 11) versus non-HME (n = 6). For all others categories the comparison is HME (n = 14) versus autopsy controls (n = 11).

MRI and cerebral hemisphere volumes

Pre-surgical studies for quantitative analysis were available on 11 (48%) HME cases using the same in-patient MRI machine and scanning sequences. These 11 MRI scans were compared with 6 scans from infants without epilepsy (n = 3) or children undergoing pre-surgery epilepsy evaluation with different pathologies (n = 3; one stroke and two with subtle non-HME cortical dysplasia). The mean age (years ± SD) at MRI for all 17 cases was 0.70 ± 0.5 (range 0.1–1.87), and there were no statistically significant differences for age between the HME and non-HME groups (P = 0.58). An additional five MRI scans (total of 16) were available for qualitative assessments.

Qualitatively, all 16 HME cases showed enlargement of the affected compared with non-affected cerebral hemisphere with apparent thickening of the grey matter, and poor grey/white matter differentiation (Fig. 1). HME involved all four lobes of the brain in nine cases (56%). In the remaining seven cases there was no preference for which lobes of the brain were enlarged with the frontal, temporal and occipital lobes involved in five patients, and the parietal lobe in six cases (21 out of 28 lobes; 75%). Enlargement of the ipsilateral lateral ventricle (Fig. 1D and I) and a straightened appearance of the ipsilateral frontal horn of the lateral ventricle was noted in 11 out of 16 HME cases (69%; Fig. 1C arrow). In six HME cases (37%), deeper grey matter structures, such as the caudate nucleus, thalamus, basal ganglia and olfactory tract, were larger than expected, often displacing neighbouring white matter structures or the ventricular system (Fig. 1D arrow, E and F arrowhead). Likewise, in eight HME cases (50%) there was enlarged abnormal grey and white matter structures at the anterior corpus callosum medial to the ventricle that displaced the lateral frontal horn (Fig. 1A and C arrow, G and I). Thus, a total of 12 out of 16 (75%) HME patients showed enlargement of deep grey and white matter structural abnormalities on MRI. Seven HME cases (44%) demonstrated a thinner corpus callosum than would be expected for age (data not shown). Two HME cases showed normal white matter signal symmetric to the contralateral side, three cases had hypointense T2-weighted signal in the affected white matter (Fig. 1D), four cases demonstrated hyperintense T2-weighted foci and seven cases showed both hypo- and hyper-intense T2-weighted white matter abnormalities in the central or subcortical regions on the affected side (Fig. 1). Ten cases showed a band-like appearance of T2-weighted hypointense signal in the periventricular white matter suggestive for heterotopia (Fig. 1B arrow and D). No statistically significant correlations were found comparing the presence and absence of the above qualitative MRI abnormalities with each other (Chi-square; P > 0.09).

Fig. 1

Representative MRI of HME cases with intractable epilepsy. Right/left orientation of all panels is shown in A. (A and B) An 8-month old presented in status epilepticus from right HME. Note the diffuse increase in cortical grey matter thickness and hemispheric size on the right relative to the non-affected left hemisphere. Arrow in B indicates periventricular signal change consistent with white matter heterotopia. (C) A 0.33 year old with seizures since birth. In addition to the increased size of the right hemisphere compared with the left notice the thickened cortex and white matter near the corpus callosum (arrow), enlarged and upward straightened tilt of the right frontal horn of the lateral ventricle. (DF) A 5-month old with seizures since birth and left HME. The deep central grey matter is increased in size including the caudate nucleus (arrow) with corresponding increased T2 signal changes, an open sylvian fissure (arrowhead) and white matter changes in the temporal lobe. The left olfactory tract is also enlarged (arrowhead). This patient was seizure free at last follow-up 1-year post-hemispherectomy. (G and H) A 9-month old with seizures since birth and right HME. The enlarged frontal lobe with thickened grey matter is associated with loss of the underlying white matter and enlargement of the lateral ventricle. Again, note the upward tilt of the right frontal horn. (I) A 1-year old with a diagnosis of tuberous sclerosis complex and seizures since birth. There is left HME with increased grey matter thickness and white matter changes.

For the quantitative MRI analysis, results of the ANCOVA found three MRI-assessed cerebral volume measures that logarithmically increased with age and three factors that were different between HME and non-HME patient groups (Fig. 2 and Table 1). Total cerebral volumes along with affected and non-affected hemisphere volumes logarithmically increased with age (Fig. 2B; P = 0.0001). The averaged volumes were increased in the affected cerebral hemispheres (+18%) and decreased in the non-affected hemispheres (−13%) of HME cases compared with non-HME patients (Fig. 2A, left; P < 0.039). Consequently, when the volumes of the affected were subtracted from the non-affected cerebral hemispheres (side-to-side comparison) there were greater differences in HME (33%) compared with non-HME cases (7%; Fig. 2A, right; P = 0.015). Seven of 11 HME cases had affected minus non-affected cerebral hemisphere volume differences that exceeded +2 SDs of the mean for the non-HME group (>63 cc; range 75–293). Cerebral hemispheric asymmetry with contralateral hemimicrencephaly is further illustrated in a presentation of monozygotic twins where the child with HME had a larger affected and smaller non-affected cerebral hemisphere compared with the normal sibling (Fig. 3). In this HME cohort, affected, non-affected and affected minus non-affected MRI-assessed cerebral hemisphere volumes did not correlate with gender (t-tests; P > 0.19), side resected (P > 0.36), history of infantile spasms (P > 0.33) and seizure control post-surgery (P > 0.58). However, T2-weighted hypointensity in the white matter by MRI correlated with a positive history of infantile spasms (Chi-square; P = 0.032).

Fig. 2

Bar graphs (A; mean ± SEM) showing differences in hemispheric MRI-assessed cerebral volumes and scatter plot B indicating volume changes with age. The P-values from the ANCOVA (Table 2) are indicated for each plot. (A) Controlling for age (ANCOVA), total cerebral volumes for both hemispheres were not statistically different between HME and non-HME cases (P = 0.271). However, for HME cases, the affected cerebral hemisphere was larger (P = 0.035), and the non-affected side was smaller (P = 0.039) compared with non-HME cases. The differences in affected minus non-affected cerebral hemisphere volumes were greater in HME cases compared with non-HME cases (P = 0.015). (B) For both HME and non-HME cases, total cerebral volumes logarithmically increased from birth to age 2 years (P < 0.0001). Similar data showed a logarithmic increase for affected and non-affected cerebral hemispheres (Table 1; data not shown).

Fig. 3

Axial T2-weighted MRI and volumetric findings from monozygotic twins with one sibling presenting with HME and seizures. Twin A developed seizures within days of birth, and both infants underwent MRI within 2-weeks of each other at age 4 months. (A and B) The twin with seizures showed enlargement of the left compared with the right cerebral hemisphere (A) while the twin without seizures had a normal scan without hemisphere asymmetry. (C and D) Calculated volumes for each cerebral hemisphere and twin are shown (note the difference in right/left orientation compared with panels A and B). Total cerebral volume for the twin with HME is 539 and 568 cm3 for the normal twin. However, there were differences in hemisphere volumes with the ‘normal’ side of the HME case being smaller than either hemisphere of the normal twin.


HME was often diagnosed in the freshly cut surgical specimen by the obvious blurring of the cortex–white matter junction, and variably sized gyri (Fig. 4A). Tissue sections from all 23 HME surgical specimens were reviewed, and every case showed severe cortical dysplasia with cortical dyslamination, excessive cells in the molecular layer, the presence of dysmorphic neurons and excessive heterotopic neurons in the white matter (Mischel et al., 1995). More variables were other histopathological features of severe cortical dysplasia on the affected HME side (Fig. 4 and Table 2). Balloon cells were detected in 45% and cytomegalic neurons in 96% of HME cases (Fig. 4D, E and Table 2). By comparison, PMG was observed in 61% and immature-appearing neurons in 70% of cases (Figs. 4B, C, F and Table 2). In this HME cohort, PMG, balloon cells, immature-appearing neurons and glial/neuronal heterotopias did not correlate with MRI-assessed total, affected or non-affected cerebral hemisphere volumes (t-test; P > 0.11), gender (Chi-square; P > 0.25), side resected (P > 0.37), age at surgery (P > 0.20), age at seizure onset (P > 0.51), seizure duration (P > 0.22) or post-surgery seizure control (P > 0.15). However, PMG on histopathology was associated with enlarged deep grey matter MRI abnormalities (Fig. 1D–F; Chi-square; P = 0.035). In addition, the presence of immature-appearing neurons in the surgical specimen correlated with a positive history of infantile spasms (P = 0.019), and T2-weighted white matter hypointensity on MRI scans (P = 0.047; Fig. 4F). By comparison, balloon cells in the surgical specimen did not correlate with neuroimaging abnormalities or clinical variables (P > 0.14).

View this table:
Table 2

Frequency of histopathological features on the affected side of HME cases with intractable epilepsy (n = 23)

Pathologic featureNone (%)Mild (%)Severe (%)
Cortical dyslamination04357
Balloon cells652213
Cytomegalic neurons43957
Immature-appearing cells30700
Excessive white matter neurons04852
Glial/neuronal heterotopias39610
Fig. 4

Histopathological features of HME. (A) Freshly cut surface from a HME surgical specimen showing thickened grey matter with indistinct grey–white matter border (upper), and subcortical heterotopia (lower). (B and C) NeuN stained region of PMG. The box in B is shown at higher magnification (C). The density of NeuN stained cells was visibly increased in the superficial cortical layers compared with the deeper layers. (D) H and E stained section through the cortex shows cortical dyslamination with numerous dysmorphic cells including enlarged cytomegalic neurons and balloon cells. (E) Magnified views of the grey matter showing dysmorphic cytomegalic neurons and balloon cells. Balloon cells were most often found in either upper cortical layers or white matter, and cytomegalic neurons were most frequently observed in the lower grey and upper white matter regions. (F) A cluster of immature-appearing cells in the middle of the grey matter. These neuroblast-like cells are characterized by their round to oval nuclear configurations with a thin rim of cytoplasm in which there are typically no dysmorphic features. Immature-appearing cells were observed in the white and grey matter, and most frequently in the perisylvian tissue blocks. Calibration bars as indicated. Panels E and F of equal magnification.

NeuN densities, cortical thickness and cell size

Tissue blocks prospectively collected at surgery were available from 14 (61%) HME cases, and were compared with 11 autopsy cases (controls). The mean age (years ± SD) of all 25 cases was 2.3 ± 2.4, and there were no statistically significant differences between the HME and autopsy groups (t-test; P = 0.23). Visually, NeuN packing densities in the upper cortical layers were variable in HME cases, some patients showing more and others fewer cells per unit area than aged comparable autopsy cases (Fig. 5A–C). Middle cortical layers were not generally different in most HME compared with autopsy cases (Fig. 5D–F), and in the lower cortex there were fewer neurons per unit area in most HME cases (Fig. 5G–I).

Fig. 5

NeuN-stained sections illustrating differences in cell densities for layer 1 and upper cortex (levels 1 and 2; AC), lower cortex (levels 3–5; DF) and white matter (GI) from an autopsy case (A, D and G), and two HME cases (B, C, E, F, H and I) of similar ages. Boxes in left column indicate approximate locations for NeuN cell counts. Compared with the autopsy case, the first HME patient (middle column) shows increased NeuN densities in the upper cortex and white matter with similar densities in the lower cortex. The increased upper cortex densities were associated with infolding of the neuronal layer near the pial surface (panel B arrow). In contrast, the second HME case (right column) shows lower NeuN densities throughout the cortex with minimal distinction between the grey and white matter. All micrographs are at identical magnification.

ANCOVA found no statistically significant NeuN density or size measurements that changed within the limited age range of this cohort, but three variables were found that were statistically different between the HME and autopsy cases (Table 1, Figs 5 and 6). On the affected side, NeuN cell counts showed increased densities in Layer 1 (molecular layer; +244%; P = 0.04), and decreased densities for lower cortical Levels 4 (−35%; P = 0.013) and 6 (−35%; P = 0.033; Fig. 6A). Numerically, NeuN densities were increased in upper cortical Levels 2 (+18%) and 3 (+56%) and decreased in cortical Levels 1 (−9%) and 5 (−11%), but the differences did not reach statistical significance (P > 0.32). Similarly, NeuN densities in the superficial (+139%) and deep white matter (+149%) were numerically increased in HME compared with autopsy cases (Tables 1 and 3). Cortical thickness was increased in HME compared with autopsy cases (+27%), but the differences were not statistically significant (Fig. 6A; P = 0.24). While again not statistically significant, NeuN cell size measurements were increased for HME compared with autopsy cases in Layer 1 (+42%), lower grey matter (+11%), superficial white matter (+23%) and deep white matter (+20%), but not upper grey matter (−1%; Table 3).

View this table:
Table 3

Clinical, MRI and histopathological correlations with NeuN cell density and size measurements in HME patients

Cells/10 000 μm2 ± SDHMEAutopsyP-value
HME versus autopsy controls
    Superficial white matter117 ± 9749 ± 280.28
    Deep white matter92 ± 6837 ± 260.21
Cell size (μm2)
    Layer I108 ± 6176 ± 100.14
    Upper grey matter152 ± 46153 ± 760.43
    Lower grey matter176 ± 76158 ± 610.36
    Superficial white matter199 ± 81162 ± 490.18
    Deep white matter194 ± 111162 ± 490.48
Cells/10 000 μm2 ± SDInfantile spasmsNo infantile spasms
History of infantile spasms
    Level 4748 ± 3841654 ± 3680.0091
    Level 5719 ± 4231717 ± 9860.0221
    Level 6613 ± 3891249 ± 3530.049
Cells/10 000 μm2 ± SDPositive MRINegative MRI
Enlarged deep grey/white matter by MRI
    Level 4270 ± 2171088 ± 4610.021
    Level 6187 ± 59895 ± 3550.0038
Cells/10 000 μm2 ± SDPMGNo PMG
    Level 6389 ± 300887 ± 4210.037
Cells/10 000 μm2 ± SDSeizure freeNot seizure free
Post-surgery seizure control
    Layer 1 (molecular layer)368 ± 1801136 ± 5740.0026
Fig. 6

Bar graphs showing mean (±SEM) NeuN cell densities and cortical thickness for autopsy (control) and HME cases using paraformaldehyde fixed 30 μm cryostat sections (upper), and affected and non-affected side of a single HME case that died during surgery using 10 μm formalin fixed blocks (lower). The P-values are indicated. (A) For the paraformaldehyde fixed 30 μm cryostat sections, statistically significant differences were found for Layer 1, Level 4 and Level 6 (P < 0.04). (B) For the single case that was formalin fixed, NeuN densities from three homotopic neocortical regions per side were averaged. Numerically, there were increased densities for Level 1 and cortical height in the affected hemisphere compared with the non-affected ‘normal’ side. Because of the different fixation techniques and slice thickness, NeuN density measurement cannot be directly compared between the upper and lower graphs.

In one surgical/autopsy HME case, three paired formalin-fixed tissue blocks from the affected HME and non-affected apparently ‘normal’ side were available for NeuN density and cell size measurements (Fig. 6B) (Jahan et al., 1997). Analysis found that there were numerically decreased neuronal densities in the upper cortex of the HME compared with the apparently ‘normal’ side (Level 1, −31%; Level 2, −16% and Level 3, −29%), and an increase in cortical thickness on the HME side (+38%). In addition, NeuN-assessed cell sizes were numerically increased on the HME compared with ‘normal’ side for Layer 1 (+26%), upper grey matter (+79%) and lower grey matter (+87%; data not shown).

Several NeuN measurements correlated with MRI, histopathological features or clinical variables (Table 3). In this HME cohort, a history of infantile spasms was associated with a statistically significant decrease in NeuN cell densities for cortical Level 4 (−55%), Level 5 (−58%) and Level 6 (−51%), but not Layer 1 (P = 0.95), Levels 1–3 (P > 0.076), or superficial or deep white matter regions (P > 0.70). Likewise, deep MRI grey matter abnormalities (Fig. 1D–F) correlated with decreased NeuN cell densities in cortical Level 4 (−75%), and Level 6 (−79%). Similarly, PMG at histopathology was associated with lower NeuN densities in Level 6 (−56%). Finally, patients with post-hemispherectomy seizure control were associated with decreased Layer 1 molecular layer NeuN densities on the HME side (−68%). No statistically significant associations were found comparing cortical thickness and NeuN cell sizes with histopathological, MRI and clinical variables.


This study of children with HME undergoing hemispherectomy for medically refractory epilepsy identified several findings unappreciated in previous clinical, neuroimaging and pathological appraisals. By volumetric MRI, the affected cerebral hemisphere was larger (ipsilateral HME) and the non-affected side smaller (contralateral hemimicrencephaly) than non-HME cases (Figs 13 and Table 1). The HME hemisphere demonstrated deep grey and white matter MRI abnormalities and/or T2-weighted hypointensity in the subcortical white matter in 75% of cases, suggestive for excessive pre-natal neurogenesis and heterotopias (Fig. 1). Balloon cells, considered in some classification systems to be essential histopathological characteristics of HME (Barkovich et al., 2001; Flores-Sarnat et al., 2003; Palmini et al., 2004) were identified in 45% of surgical specimens. By comparison, immature-appearing neurons were identified in 70% and PMG in 61% of paediatric HME cases (Table 2 and Fig. 4). Compared with autopsy cases, in HME children NeuN densities on the affected side were significantly increased in the molecular layer (+244%), decreased in the lower cortex (−35%) and numerically increased in upper cortical layers (+56 to +18%) and superficial and deep white matter (+139 to +149%; Figs 5 and 6; Table 1). The decrease in lower cortical NeuN densities (average −20%) was similar to the increase in cortical thickness (+27%) on the affected HME side. Deep grey matter MRI abnormalities and/or T2-weighted white matter hypointensity positively correlated with the presence of immature-appearing neurons and PMG on histopathology, decreased lower cortical NeuN cell densities and a history of infantile spasms. In HME cases, seizure control post-surgery was associated with decreased NeuN densities in the molecular layer on the affected side (Table 3). These findings indicate the existence of bi-hemispheric abnormalities in HME cases that may explain the suboptimal developmental outcomes observed in these children post-hemispherectomy. Furthermore, based on an understanding of normal developmental neurobiology, our neuroimaging and neuropathological findings support the hypothesis that HME pathogenesis probably involves somatic mosaic mutations that affect cortical development differently for each side.

Previous clinical studies have noted worse post-surgery seizure control along with poorer cognitive and language outcomes in HME cases compared with other children undergoing epilepsy neurosurgery (Vigevano et al., 1996; Battaglia et al., 1999; Carreno et al., 2001; Curtiss et al., 2001; Maehara et al., 2002; Jonas et al., 2004). The reasons for the suboptimal surgical outcomes in HME cases are unknown, but our data support the concept that one reason may be because the contralateral apparently ‘normal’ cerebral hemisphere shows hemimicrencephaly. Consistent with this notion, in our cohort the HME cases with the greatest pre-surgery side-to-side cerebral asymmetries by MRI (>100 cc) were post-surgery developmentally doing the worst, especially for language, despite being seizure free. In our previous quantitative MRI study of less severe paediatric cortical dysplasia patients we reported minimal increases (+3%) in cerebral hemisphere volumes comparing the affected with the non-affected side (Andres et al., 2005). This contrasts with the 33% side-to-side asymmetry in HME cases in our current report. Hence, our findings would agree with prior proposals using qualitative MRI assessments that poorer developmental outcomes in HME children are associated with the severity of cerebral asymmetry (Barkovich and Chuang, 1990; Sener, 1995).

Our neuroimaging and histopathology findings also suggest that previous hypotheses involving HME pathogenesis should be revised to consider developmental mechanisms that affect both hemispheres differently (Barkovich and Chuang, 1990; Barkovich et al., 2001; Flores-Sarnat et al., 2003; Palmini et al., 2004). Previous classification systems, based on qualitative neuroimaging or histopathological findings on a smaller number of patients have speculated that HME pathogenesis is from a very early cortical developmental abnormality on one side that alters neuronal migration and/or neuroglial differentiation to explain the presence of dysmorphic balloon cells on histopathology. If these hypotheses are correct then in HME one would expect (i) a normal sized contralateral cerebral hemisphere because the aetiological process affects one side; (ii) decreased trans-cortical neuronal densities and thickness because of reduced neurogenesis from abnormal neuroblast formation and/or incomplete neuronal migration; and (iii) balloon cells to be present in all HME cases.

While decreased lower cortical NeuN packing densities in our study could be envisioned as a sign of abnormal early cortical neuron migration consistent with prior proposals, when all the neuroimaging and neuropathological data are collectively reviewed an alternative ontogenetic hypotheses should be considered. The decrease in lower cortical NeuN packing densities (average Level 4–6; −20%) was associated with an increase in MRI cerebral volumes (+18%) and cortical thickness (+27%). By comparison, on the affected side NeuN cell densities were numerically increased in the upper cortex (average Level 1–3; +22%). Thus, compensating for changes in cortical thickness and cerebral hemisphere volume, total neuronal numbers in the lower cortex are probably similar to autopsy cases and increased in the upper cortex of the HME hemisphere. Furthermore, in the majority of HME cases we found enlarged deep grey matter MRI abnormalities in regions normally associated with pre-natal neurogenesis suggesting excessive neuronal production (Haydar et al., 1999; Rakic and Zecevic, 2000; Chan et al., 2002), and T2-weighted hypointensity indicative of excess heterotopic neurons in the subcortical white matter. Taken together, these neuroimaging and histopathological findings support the view that there appears to be more neurons than expected throughout the ipsilateral hemisphere of HME patients. Assuming that neurons in the upper cortex are the latest born, the pattern of histopathological abnormalities is most consistent with the concept that neurons were overproduced late in corticoneurogenesis, and there is partial failure of post-neurogenesis programmed cell death in the remnants of the molecular layer and subplate. These conclusions are similar to our previous findings in less severe non-HME forms of cortical dysplasia, and support the concept that paediatric cortical dysplasia associated with intractable epilepsy, including HME, may have comparable ontogenetic mechanisms (Andres et al., 2005). In contrast, balloon cells, thought to be products of abnormal neuroglial differentiation, were identified in less than half of our HME cases (Table 2), and their presence in the surgical specimen did not correlate with clinical, MRI or neuropathological variables. Such findings support the notion that abnormal mechanisms that produce balloon cells are not essential for HME pathogenesis (Mischel et al., 1995; Cepeda et al., 2005a).

Of equal relevance was our observation that immature-appearing neurons were frequent histopathological findings in HME patients, and correlated with T2-weighted MRI white matter hypointensity and a history of infantile spasms. Such clinical–pathological correlations suggest that mechanisms that over produce immature neurons in the later stages of cortical development maybe more representative of the ontogenetic process leading to HME and seizures (Caviness et al., 2003; Tarui et al., 2005). In other words, molecular mechanisms that affect pre-neurogenesis progenitor cell cycling may be cellular pathways to investigate as possible ‘causes’ of non-syndromic paediatric HME and cortical dysplasia associated with epilepsy (Cotter et al., 1999a, b). Such conclusions support our previous hypothesis that epileptogenesis in cortical dysplasia and HME is probably due to incomplete cortical maturation with excessive immature-appearing neurons and under developed synaptic circuitry (Cepeda et al., 2005a, b). Finally, whether genetic and/or environmental in origin the aetiology of HME seems to begin after zygote separation as illustrated in our study of monozygotic twins (Fig. 3).

It is important to consider the potential limitations of our human study when interpreting the results. For example, our findings are based on neuroimaging and histopathology obtained from mostly non-syndromic HME patients with epilepsy who were surgical candidates. Our results and interpretations may or may not be applicable to syndromic HME cases, non-surgical HME patients with bi-hemispheric seizures and the few HME cases without intractable epilepsy. Likewise, our MRI studies did not include a very large sample of non-seizure normal controls because such a group was nearly impossible to recruit. However, based on our prior MRI study of children with milder forms of cortical dysplasia and non-dysplasia pathologies, non-affected cerebral hemisphere volumes were only minimally different, and thus non-HME cases are a reasonable comparison group for the HME cohort (Andres et al., 2005). Furthermore, our cohort consisted of very young HME children (less than age 4 years). We do not know if there could be progressive pathological abnormalities with longer seizure histories (Mathern et al., 2002). In addition, we measured NeuN cell densities from a limited number of tissue blocks per patient and did not systematically sample the entire brain (Cook et al., 2004). Despite the expected increased variability due to the limited number of sample sites, we still found statistically significant differences between HME and autopsy patient groups supporting the view that our methods were sufficiently sensitive for this study. Moreover, we counted NeuN positive cells which should accurately identify differentiated neurons and avoid potential confusion with glia or other non-neuronal cell types. However, we cannot discern whether the cells were excitatory or inhibitory. Finally, our hypothesis concerning HME pathogenesis is based on the presumption that the developmental neurobiology of cortical development and migration are functional in these morphologically abnormal brains.


Examining paediatric epilepsy surgery patients with HME, this study found changes in cerebral hemisphere volumes by MRI and NeuN cell densities that correlated with clinical and histopathological variables that in turn provide clues to HME pathogenesis and an explanation for poorer cognitive and seizure control outcomes post-hemispherectomy. Based on our results, we propose that HME pathogenesis is the consequence of over proliferation of progenitor cells in later cell cycles. We also propose that HME pathogenesis involves mechanisms that interfere with late corticoneurogenesis with partial failure of post-neurogenesis apoptosis in the molecular layer and subplate (Andres et al., 2005). Finally, we propose that poorer post-surgery seizure control and cognitive outcomes are due to contralateral hemimicrencephaly in most HME patients. It should be noted that these proposed concepts do not exclude other possible explanations, such as a defect in neuronal migration in later cortical development, and the mechanisms of inducing contralateral hemimicrencephaly with ipsilateral HME will require additional studies. However, our data provide a conceptual framework for proposing hypothesis-driven studies directed toward molecular mechanisms that affect progenitor cell cycling, migration and pre- and post-neurogenesis to explain HME and seizures.


This study was supported by the National Institutes of Health grants R01 NS38992 and P05 NS02808 to G.W.M. and FAPESP (CInAPCe-Project-05/56447-7) to J.P.L.


View Abstract