|Home | About | Journals | Submit | Contact Us | Français|
To determine the involvement of the midbrain and hindbrain (MHB) in the groups of classic (cLIS), variant (vLIS), and cobblestone complex (CBSC) lissencephalies and to determine whether a correlation exists between the cerebral malformation and the MHB abnormalities.
MRI scans of 111 patients (aged 1 day to 32 years; mean 5 years 4 months) were retrospectively reviewed. After reviewing the brain involvement on MRI, the cases were reclassified according to known mutation (LIS1, DCX, ARX, VLDLR, RELN, MEB, WWS) or mutation phenotype (LIS1-P, DCX-P, RELN-P, ARX-P, VLDLR-P) determined on the basis of characteristic MRI features. Abnormalities in the MHB were then recorded. For each structure, a score was assigned, ranging from 0 (normal) to 3 (severely abnormal). The differences between defined groups and the correlation between the extent of brain agyria/pachygyria and MHB involvement were assessed using Kruskal–Wallis and χ2 McNemar tests.
There was a significant difference in MHB appearance among the three major groups of cLIS, vLIS, and CBSC. The overall score showed a severity gradient of MHB involvement: cLIS (0 or 1), vLIS (7), and CBSC (11 or 12). The extent of cerebral lissencephaly was significantly correlated with the severity of MHB abnormalities (p = 0.0029).
Our study focused on posterior fossa anomalies, which are an integral part of cobblestone complex lissencephalies but previously have not been well categorized for other lissencephalies. According to our results and the review of the literature, we propose a new classification of human lissencephalies.
The term lissencephaly (smooth outer brain surface) refers to a paucity of gyral and sulcal development.1 It encompasses a spectrum of gyral malformations ranging from complete agyria to regional pachygyria and includes subcortical band heterotopia. Lissencephaly has been traditionally classified in two distinct groups: classic (cLIS, formerly called lissencephaly type 1) and cobblestone complex (CBSC, formerly called lissencephaly type 2) based on both brain imaging and pathology. To date, five genes have been identified as causing or contributing to human cLIS: LIS1, 14-3-3 in Miller–Dieker syndrome (MDS), DCX, RELN, and ARX.1–6 A new classification has recently been proposed defining four groups of cLIS differing from each other by genetics and neuropathologic features of the cerebral cortex, cerebellum, and brainstem.3 1) LIS1 mutations and MDS show a four-layered cortex predominantly posterior with white matter heterotopia in the cerebellum, a normal three-layered cerebellar cortex, and normal to small pons. 2) DCX mutations show a four-layered cortex similar to LIS1 but predominantly anterior; cerebellar and pontine abnormalities are similar to those in LIS1 mutations. 3) ARX mutations have a three-layered cortex; the cerebellum is normal, but the pons is small due to hypoplastic corticospinal tracts. 4) Lissencephaly with a two-layered cortex, diffuse involvement of the brain surface with no gradient; the cerebellum is extremely hypoplastic with pontine and medullary disorganization. No RELN-like mutations were studied. The authors divided these lissencephalies into two groups: LIS1 and DCX were classified as cLIS, whereas ARX and two-layered lissencephalies were classified as variant lissencephaly (vLIS).3
Numerous genes have been identified in association with the CBSC (also called dystroglycanopathies).7–12 Most affected patients have symptoms of congenital muscular dystrophies (CMDs) with CNS involvement. Among these disorders, three phenotypes are well defined: Fukuyama congenital muscular dystrophy (FCMD), Walker–Warburg syndrome (WWS), and muscle–eye–brain disease (MEB).1,7–11 These phenotypes have been associated with mutations of multiple genes involved in O-glycosylation of α-dystroglycan, including FCMD, FKRP, POMT1, POMT2, LARGE, and POMGnT1.12
The aim of our study was to determine the involvement of the midbrain and hindbrain (MHB) in these disorders, and to determine whether a correlation exists between the cerebral malformation and the MHB characteristics as determined by MRI.
MRI scans or portions of MRI scans of 120 patients were retrospectively reviewed. The scans were acquired from the private teaching collection of the senior author (acquired over 22 years), from the teaching file of the radiology department at our institution (searching using lissencephaly, agyria, pachygyria, band heterotopia, cobblestone malformation, and congenital muscular dystrophy as key words), and from MRIs reviewed for a study of the genetics of epilepsy. MRIs were only included if good quality images of the cerebrum (to assess the type and severity of the cortical malformation) and posterior fossa were available as determined by a consensus of the authors. Most studies included T1-weighted and T2-weighted images with scans performed in at least two planes. Four cases were excluded because only CT images were available and, therefore, assessment of the posterior fossa content was suboptimal. Five others were excluded because the posterior fossa structures were not assessed on magnetic resonance (MR) images available. The remaining 111 MRI studies were analyzed. Patients were aged 1 day to 32 years (mean 5 years 4 months; median 2 years) at the time of MRI, including 54 men, 41 women, and 16 with unavailable sex information. The MRIs included sagittal and axial T1-weighted, axial, and coronal T2-weighted images with slices of 3- to 5-mm thickness. Cases were classified as “lissencephalies” in the database according to the diagnosis based on brain MRI features or known genetics. Clinical information was not often available; when available, it was usually brief, reporting mental retardation and seizures and, sometimes, family history. After reviewing the brain involvement, the cases were reclassified according to known mutation (LIS1, DCX, ARX, VLDLR, RELN, MEB, WWS) or mutation phenotype (LIS1-P, DCX-P, RELN-P, ARX-P, VLDLR-P) according to characteristic MR features1; many of the patients were imaged before testing for specific mutations was available and were not able to be found; therefore, the number with a confirmed genetic diagnosis is low. Phenotypes were determined as follows:
The VLDLR mutation was proposed (VDLR-P) for patients having the so-called dysequilibrium syndrome.13,14 Patients with cLIS (LIS1, LIS1-P, DCX, DCX-P) were designated as cLIS, and those with ARX, RELN, RELN-P, and ND were designated as vLIS. All cases with defined mutations of O-glycosylation genes (FKRP, POMGnT1) and those without defined mutation (CMD-P) but features of CBSC on brain MRI were classified as CBSC. Cases not fitting the previously defined groups were considered to be a new group. We did not include CMD with Merosin deficiency in this study, because it does not manifest brain features of CBSC and is not of the group of disorders previously called type 2 lissencephaly.15
After this initial classification, the posterior fossa in each case was carefully reviewed with separate attention to the midbrain, the pons, the medulla, the cerebellar vermis, and the cerebellar hemispheres. Morphologic features were then recorded according to a systematic visual analysis as follows:
The degree of involvement was further characterized as mild or severe, based on consensus by the authors.
For statistics, a score was assigned for each main feature: normal = 0; mildly hypoplastic (including inferior vermis hypoplasia) = 1; severely hypoplastic (including vermian hypoplasia) = 2; dysplastic (including cerebellar and vermian cysts as well as pontomedullary “kink”) = 3. To assess the differences between defined groups of patients based on MHB features and to evaluate the correlation between the brain agyria/pachygyria and MHB involvement, nonparametric statistical tests were used, Kruskal–Wallis test for multiple group comparison and χ2 McNemar test (Statistica version 7.1, StatSoft France, 2005, www.statsoft.fr). A p value less than 0.05 was considered significant.
Table e-1 (on the Neurology® Web site at www.neurology.org) shows all patients included in the study listed by number, sex, and age, and classified according to lissencephaly groups and genotype/phenotype. For each patient, corresponding forebrain, midbrain, and hindbrain involvement are reported.
Table 1 summarizes the results of the statistical tests. Only significant p values are reported to show group differences. The discriminating MR feature makes the difference between groups for each MHB structure. Significant differences were found between cLIS and vLIS, and between cLIS and CBSC. Moreover, according to the predefined score assignment, we found a severity gradient of MHB involvement.
Normal mid-hindbrain was found in 85% of cLIS patients, including all LIS1, 60% of LIS1-P, 50% of DCX, and 58% of DCX-P (figure 1, A and B). Moderate mid-hindbrain involvement corresponded to isolated inferior vermis hypoplasia (IVH), small or hypoplastic vermis, cerebellar hemispheres, midbrain, pons, or medulla. Among the cLIS patients with IVH, 75% were either DCX or DCX-P (in equal proportions) (figure 1, C and D). Severe mid-hindbrain involvement included severe hypoplasia or dysplasia of the vermis and cerebellar hemispheres, cysts within the vermis and cerebellar hemispheres, midbrain hypoplasia or dysplasia, severe hypoplasia of the pons and medulla, pontomedullary kink, and pontine midline cleft. These features were remarkably predominant among the vLIS (figure 1, E–G) and CBSC groups, differentiating them significantly from cLIS groups (table 1). Interestingly, all patients with WWS phenotype had a ventral midline pontine cleft associated with severe hindbrain hypoplasia (figure 2).
We found an association between agyria/pachygyria of the entire brain (with or without gradient) and severe MHB involvement (p = 0.0029).
Agenesis of corpus callosum (ACC) (1.27% of all lissencephalies) was always associated with severe involvement of the vermis; it was found in vLIS with no defined genotype, as expected in all ARX mutations, and in two patients with complete agyria and undulating band heterotopia (figure 3, A and B). In addition, periventricular nodular heterotopia was seen in patients with normal to mild MHB involvement and subependymal linear heterotopia (SELH) in two patients with near total absence of the vermis and hypoplastic pons (figure 3, C and D).
This study used a database including many patients with lissencephalies to perform a systematic review of MHB structures in a large series including many types of lissencephalies. More than a simple, exploratory approach to the distribution of cerebellar and brainstem abnormalities in lissencephalies, this study tested whether a relationship existed between the type and extent of supratentorial involvement and the involvement of the mid-hindbrain, and whether, as hypothesized, the distribution among different defined groups could help to predict the mutation by associating the pattern of cerebral involvement with mid-hindbrain involvement.
The patterns of cerebral involvement have been well documented in human lissencephalies and have helped to classify them and to correlate them with histology and genetics.1–10 Consistent differences were found in the gyral patterns among cLISs, with the malformation more severe posteriorly in individuals with LIS1 mutations and more severe anteriorly in individuals with DCX mutations.4 These observations were later correlated with neuropathologic findings as mentioned in the introduction.3 The pattern of cortical brain involvement in cLIS is typically different from what is seen in “cobblestone malformations” associated with defects in O-glycosylation of α-dystroglycan.7–11,16 The agyria/pachygyria complex seen in cLIS and vLIS is mainly the result of incomplete migration of neurons,2–6 whereas CBSC is primarily a result of overmigration of neurons, many of which pass through gaps in the glial limiting membrane.7–11 Involvement of the cerebellum and the pons has been described in cLISs that were classified according to mutation analysis. Some authors found hypoplasia of the cerebellar vermis to be more common in patients with DCX mutations, whereas both DCX and LIS1 sometimes had normal cerebellum and brainstem.4 The new pathologic classification recently proposed included cerebellum and brainstem abnormalities associated with cLIS (LIS1 and DCX) and vLIS (ARX and two-layered cortex).3 They found the most severe involvement of the hindbrain (hypoplasia) in vLIS group and, most notably, in two-layered cortex. In our series, the mid-hindbrain in cLIS ranged from normal to small (more frequently among DCX and DCX-P patients) in agreement with previous studies. Our observations were similar, revealing an association between the extent of brain involvement and severity of MHB abnormalities (p = 0.0029). Marked to severe MHB hypoplasia was most often found in patients with pachygyria involving the entire cerebrum, relatively thin cortex (5 mm or less), and no cell-sparse zone (vLIS-ND, likely Forman’s two-layered cortex lissencephaly). As discussed by previously cited authors,3 this association could be related to extensive impairment of cortical fibers projecting to the pons as well as hypoplasia of mid-hindbrain nuclei.
Of interest, we found a high rate of isolated inferior vermis hypoplasia associated with DCX. This sign is probably to be considered as the smallest degree of MHB involvement and would give an interesting clue for differential diagnosis if the finding is substantiated on subsequent studies. MHB was severely hypoplastic in ARX patients (with dysmorphic features of the vermis) and in patients with RELN mutations. Different phenotypes of lissencephaly with cerebellar hypoplasia (LCH) are associated with LIS1, DCX, and RELN genes.6,17
All patients initially classified as VLDLR and VLDLR-P had MHB features similar to patients with RELN mutation. Consequently, the two groups had the same overall score for MHB involvement (table 2). Severe hypoplasia of the vermis was found in all RELN (two cases) and in five of eight VLDLR and VLDLR-P patients. These were grouped together as vLIS. The VLDLR mutation has been described as responsible for the so-called dysequilibrium syndrome and can therefore be presumed in any patient with characteristic clinical and brain MRI features.18
CBSC was formerly called type 2 lissencephaly.12 It is the main CNS feature of CMDs with O-glycosylation defects of α-dystroglycan, even though it is difficult to differentiate the genetic causes on the basis of brain MRI because of overlapping phenotypes.19–23 Cerebral, cerebellar, and brainstem abnormalities have been well documented in the major phenotypes: FCMD, WWS, and MEB. WWS displays severe cerebral cortical dysplasia, hypomyelination, dysplastic tectum, pontomedullary kink, severe hypoplasia of the vermis, and dysplastic cerebellar hemispheres.1,10 A ventral midline cleft of the ventral pons is often associated with MEB and WWS and could also be considered among their typical MR features. The midline cleft is probably due to reduced number of crossing pontine axons, possibly resulting from impaired tangential migration of pontine nuclei, as demonstrated in mice with mutation of Large.24,25
We found that callosal agenesis is not uncommon among patients with lissencephaly. ACC is among the characteristic features of ARX-associated lissencephaly.26,27 The frequent association of ACC with vLIS-ND patients in our series suggests considering this feature as a supplementary criteria to differentiate cLIS from vLIS. It is important to note, however, that ACC may also be found in other types of lissencephalies.28 Interestingly, the cerebellar vermis was dysplastic or severely hypoplastic in 13 of 14 patients with ACC; it will be interesting to see whether this association holds up in other series. A recent study of callosal agenesis associated with DCX found that the doublecortin protein might be involved in the early stages of corpus callosum formation29; however, none of the DCX or DCX-P patients in this study had ACC.
This study is limited by the small number of cases with established genetic diagnoses. An attempt was made to increase this number by assigning a presumed genetic classification based on established cerebral MRI features in different lissencephaly groups and correlation of imaging features with known pathologic features such as cell-sparse zones, associated callosal anomalies, and cortical thickness. This is obviously not scientifically rigorous and may be considered an artificial classification. However, we observed relative homogeneity of MHB involvement both within groups with presumed classification (LIS1-P) and between groups with known and presumed mutations of the same gene (DCX, RELN, and VLDLR). This homogeneity and our statistical results emboldened us to group together those with presumed and proved mutations for purposes of correlation. Results of these groupings are shown in the two first columns of table 2. For each structure analyzed, a gradient of involvement was found to be associated with each presumed genotype (mutation phenotype) (table e-2). We suggest that these presumed groupings might be useful to determine which tests for specific gene mutations should be performed in such patients.
Based on the data in this project and that in the literature, we propose a classification similar to that recently proposed,3 separating cLIS (LIS1 and DCX) from vLIS (RELN and ARX) (table 2). We included VLDLR and some ND cases with markedly atypical MR features, namely ND1, ND2, and ND3, within the vLIS group. ND3 refers to a case with microcephaly, paracentral pachygyria, and ACC resembling Baraitser–Winter syndrome, but other phenotypes may be associated with this syndrome.30 ND2 refers to two cases with undulating band heterotopia (a configuration that must be extremely rare, because we found no previous reports) and ACC. ND1 refers to two cases with complete agyria, SELH, ACC, and severe MHB involvement. Further investigations or more cases are needed to determine whether the ND groups should be classified with vLIS or in still another category.
As noted in the previous paragraph, it was difficult to determine which entities to include in the vLIS category and how to classify them. More than 80% of vLIS patients did not have characteristic imaging findings (ND). The protein products of VLDLR and RELN act in the same pathway but are coded by different genes that have different patterns of inheritance, so they were included as separate malformations. Based on a previously cited work, we included two-layered lissencephaly as part of the vLIS category.3
Another weakness of this study is a lack of clinical data that could help clinicians to better identify patients with these malformations. It would also have helped to better classify those patients who were ultimately assigned to the ND category. Despite this obvious deficiency, important information has emerged from this study for both radiologists and other physicians interested in brain development. This study expands the categories of lissencephalies and adds to the knowledge of associated malformations. This information will, hopefully, lead to further studies of the molecular disorders involved in the many brain anomalies associated with agyria and pachygyria and, ultimately, a better understanding of the mechanisms by which the brain is formed.
P.J.-T. conducted the statistical analysis.
The authors thank Drs. Chistopher Walsh, William Dobyns, Daniela Pilz, Sean Bryant, and Naci Koçer for the MRI scans that they kindly contributed to this work.
Address correspondence and reprint requests to Dr. Patrice Jissendi-Tchofo, Radiology Department, Erasme Hospital, Route de Lennik 808, B-1070 Brussels, Belgium jissendi/at/gmail.com
Supplemental data at www.neurology.org
Editorial, page 394
e-Pub ahead of print on November 19, 2008, at www.neurology.org.
Supported by R37 NS035129 from the National Institute of Neurological Disorders and Stroke, NIH, Bethesda, MD.
Disclosure: The authors report no disclosures.
Received March 7, 2008. Accepted in final form July 16, 2008.