Deregulated expression of EZH2 in congenital brainstem disconnection

Congenital brainstem disconnection (CBSD) is an enigmatic embryo-fetal defect presenting as (sub)total absence of a segment between mesencephalon and lower brainstem. Rostro-caudal limits of the defect vary while the basal pons is always involved and the cerebellum is globally hypoplastic. A recent update and review[1] lists 14 cases, including 3 brain autopsy studies[1-3]. Necrosis and glial- or inflammatory reactions were absent. Inferior olivary nuclei were small or absent, pontine nuclei depleted, and the cerebellar dentate nuclei dysplastic. Supra-tentorial parts were normal in size, shape and microscopic structure. Frequent associated findings were vertebral segmentation defects(4/14) and hydronephrosis(3/14). No intra-familial recurrence or consanguinity were recorded. We considered interference with the developing rhombencephalon by an epigenetic mechanism as possible cause of CBSD. We probed the role of PRC2 (Polycomb Repressive Complex 2), member of the polycomb group of chromatin modifying proteins (PcG) with a cell-fate conserving function in organ development[4-7]. PRC2 acts by promotor binding as well as by modification of histone H3 through its catalytic subunit EZH2 (Enhancer of Zeste2) that tri-methylates the free ending tail of H3 at lysine 27 to H3K27me3 (with H3K27me2 as intermediate step). H3K27me3 blocks transcription of developmental genes to consolidate the mature stage of cell lineage on completed development[7]. Advantageous use can be made of anti-H3K27me3 antibody staining of nuclear chromatin on routinely prepared autopsy tissue sections to study EZH2 activity, an application more commonly used in neoplastic studies[8], but not routinely applied in human malformations. The absence of immunoreactivity in a developmental setting is due to immaturity, as in embryonic stem cells, or to an abnormal persistence (or even reversal) to that state with loss of control over tissue specific development. A large number of genes are known to undergo maturational silencing by histone modification. For example, EZH2 stabilizes the identity of individual vertebral body segments as well as brainstem development by silencing HOX genes[9,10]. Another role for EZH2, sustaining normal pontine neuron migration from the embryonic rhombic lip to the mature site of function has been identified by rhombomere specific knockout of Ezh2 in mouse embryos[11]. This finding prompted the present study. 
 
This article is protected by copyright. All rights reserved.

Congenital brainstem disconnection (CBSD) is an enigmatic embryo-fetal defect presenting as (sub)total absence of a segment between mesencephalon and lower brainstem. Rostro-caudal limits of the defect vary while the basal pons is always involved and the cerebellum is globally hypoplastic. A recent update and review [1] lists 14 cases, including three brain autopsy studies [1-3]. Necrosis and glial or inflammatory reactions were absent. Inferior olivary nuclei were small or absent, pontine nuclei depleted and the cerebellar dentate nuclei dysplastic. Supra-tentorial parts were normal in size, shape and microscopic structure. Frequent associated findings were vertebral segmentation defects (4/14) and hydronephrosis (3/ 14). No intra-familial recurrence or consanguinity was recorded.
We considered interference with the developing rhombencephalon by an epigenetic mechanism as possible cause of CBSD. We probed the role of PRC2 (Polycomb Repressive Complex 2), member of the polycomb group of chromatin modifying proteins (PcG) with a cell-fate conserving function in organ development [4][5][6][7]. PRC2 acts by promotor binding as well as by modification of histone H3 through its catalytic subunit EZH2 (Enhancer of Zeste2) that tri-methylates the free ending tail of H3 at lysine 27 to H3K27me3 (with H3K27me2 as intermediate step). H3K27me3 blocks transcription of developmental genes to consolidate the mature stage of cell lineage on completed development [7]. Advantageous use can be made of anti-H3K27me3 antibody staining of nuclear chromatin on routinely prepared autopsy tissue sections to study EZH2 activity, an application more commonly used in neoplastic studies [8], but not routinely applied in human malformations. The absence of immunoreactivity in a developmental setting is due to immaturity, as in embryonic stem cells, or to an abnormal persistence (or even reversal) to that state with loss of control over tissue-specific development. A large number of genes are known to undergo maturational silencing by histone modification. For example, EZH2 stabilizes the identity of individual vertebral body segments as well as brainstem development by silencing HOX genes [9,10]. Another role for EZH2, sustaining normal pontine neuron migration from the embryonic rhombic lip to the mature site of function has been identified by rhombomere-specific knockout of EZH2 in mouse embryos [11]. This finding prompted this study.
Antibody staining to H3K27me3 on routinely prepared autopsy material was applied in this study to probe a possible role for EZH2. In situ hybridization of EZH2 mRNA was used to monitor EZH2 transcription at sites of negative H3K27me3 staining. Antibody staining to histone H3K14ac, which is independent of EZH2 was used as tissue quality control for histone staining.
We studied a new patient (case 1) and a previously published patient with CBSD [1] (case 2 in this report) for local involvement of EZH2. We included brain material from the first reported autopsy of pontine tegmental cap dysplasia (PTCD) [12] (case 3) because characteristic aspects of brainstem morphology in PTCD are similar to CBSD [12,13]. Prominent features linking these three patients are summarized in Table 1.
Clinical and neuropathological findings of the three patients are as follows.
Patient 1 is the second male child of nonconsanguineous parents of European ancestry. Birth at 35 weeks was normal following an uncomplicated pregnancy. Birthweight was 2945 g, length 48.5 cm, head circumference 34 cm. Breathing was spontaneous. He had an absent gag reflex and hypertonia of jaw muscles. Radiographs showed multiple segmentation defects of the ribs and spinal column ( Figure 1D). MRI of the brain on day 3 showed a defect separating the tegmentum from the ventral pons ( Figure 1A,B) at the exit level of the trigeminal nerves. The cerebellum was normally shaped but hypoplastic ( Figure 1C). He went home with tube feeding. At 8 months he was found dead at home by the parents. A coroner's autopsy was performed.
With legal and parental consent formalin-fixed specimens of the brain were made available for this study.
General autopsy revealed bilateral hydronephrosis. The fresh brain weight was 745 g. The midbrain was present but the pons was only 1.5 cm in diameter with a nodular shape. The junction between pons and medulla was just 0.5 cm in diameter. Histopathological examination after routine staining demonstrated no abnormalities in any supra-tentorial structures.
Brainstem sections stained with antibodies to MBP and SMI 311 revealed an abnormal course of the cortico-ponto-cerebellar tracts: instead of joining the ventral pons each of these followed a circular route at the surface of the lower mesencephalon to end up in the anterior medullary velum of the cerebellum (Figure 1E-H). The cerebral peduncles were normally myelinated but too small. The decussation of the superior cerebellar peduncles was indistinct. The pontine tegmentum bordering the fourth ventricle lacked a midline depression separating left and right halves. A large smooth edged defect separated the pontine tegmentum from the ventral pons. The middle cerebellar peduncles were absent and the trigeminal roots could be identified at the upper margin of the defect. At the same transverse level, a disc with a smooth margin and diameter of 8 mm represented the ventral pons ( Figure 1I) in which the pyramidal tracts could be identified. The structure dorsal to the pyramidal tracts contained sparse transverse fibres, few neurons and lacked a midline raphe. The medulla oblongata appeared more normal. The inferior cerebellar peduncles were diminished in size. The medial lemnisci were indistinct. The pyramids were present. The sites of the inferior olivary nuclei were empty ( Figure 1J). An abnormal rim was seen surrounding the lower brainstem, immunonegative for axons (SMI311) ( Figure 1I, J), but immunopositive for glial staining (GFAP, not shown). The cerebellar cortex was normal on routine staining, but the cerebellar dentate nucleus lacked its continuous undulated structure, instead appearing as collections or aggregates of neurons resembling islands ( Figure 1K).
The spinal cord was not available for study. The routine neuropathological results are summarized in Table 1 together with those from patients 2 and 3.
We obtained the following results for specific histone staining and EZH2 in situ hybridization (Figure 2). In controls, H3K27me3 immunoreactivity was strong in nuclei of cerebellar cortex (all layers), cerebellar dentate nucleus (not shown) and temporal cerebral cortex (Figure 2A-D). In case 1, H3K27me3 immunoreactivity was undetectable in the cerebellar cortex and dentate nucleus ( Figure 2E,F); expression was reduced in both internal and external granular cell layer and Purkinje cells; H3K27me3 signal was observed in other brain areas including temporal cortex ( Figure 2G). Areas rostral and caudal to the affected brainstem segment showed H3K27me3 immunopositivity in cellular nuclei and immunonegativity in the sparse remaining neurons of the pons (not shown). The nuclear pattern of H3K14ac immunoreactivity was preserved in cerebellum confirming tissue quality for general histone staining ( Figure 2H). Results for H3K27me2 were identical to those for H3K27me3 and therefore not shown. In case 2 H3K27me3 immunoreactivity was reduced in cerebellar cortex (Figure 2I,J) with nuclear staining in a small quantity of internal granule cells and molecular layer cells, while it was well presented in temporal cerebral cortex ( Figure 1K). H3K14ac staining was preserved in case 2 ( Figure 2L). H3K27me3 immunopositivity was also found in mesencephalon and medulla oblongata in areas flanking the affected pontine region. Not enough ventral pontine neurons were available for reliable histone mapping. In case 3, nuclear H3K27me3 was observed in brain in all areas including the cerebellum, in contrast to the CBSD cases. (PTCD; Figure S1A-C). In all three cases paired results using two different batches of anti H3K27me3 antibody were similar. Results for H3K27me2 were identical to those for H3K27me3 in all three cases, and therefore not shown. In situ hybridization of EZH2 in control cerebellum showed strong uniform expression in Purkinje cells ( Figure 2M); neuronal EZH2 expression was also detected in temporal cortex (not shown). In case 1, EZH2 expression was absent in Purkinje cells (Figure 2N); however, normal EZH2 expression was observed in other brain areas (not shown, but similar to case 2). EZH2 expression was also reduced in the cerebellum from case 2 ( Figure 2O); marginally positive expression was found in a few samples, but less than temporal cortex (Figure 2P) or control cerebellum ( Figure 2M).
Studies on DNA obtained from blood samples in the two patients with brainstem disconnection and their parents excluded a mutation in the EZH2 gene except for known reported polymorphisms. For case 1, whole genome analysis of DNA isolated from fibroblasts of the patient was performed. We evaluated all introns and noncoding parts as well as the surrounding 530 kb region of the EZH2 gene. No obvious pathogenic mutations or partial deletions were detected. Experimental knockout of the Ezh2 gene from rhombomeric regions in embryonic mice [11] impedes the migration of pontine precerebellar neurons causing heterotopia. The stalled migration is not intrinsic, that is, not due to changed properties of the migrating neurons, but due to loss of migratory guidance in which Ezh2 has an 'orchestrating' role [11]. This takes effect in two ways: (i) by limiting the Sonic hedgehog pathway, which controls the distribution of the chemo-attractant and guidance factor Netrin-1, to the floor plate region. (ii) By restricting the migrating neurons to different HOX (that is, homeobox related) regions. Rescue innervation of the ectopic neurons through their proper afferents (from the cerebral cortex), a precondition for their survival, was also demonstrated in the murine knockout [11]. In CBSD, no collections of ectopic pontine neurons were found. Instead ventral pontine nuclei were almost absent in the present as well as in previously reported CBSD cases. The most likely explanation holds that in CBSD, inappropriate guidance of migrating pontine neurons is compounded by failed rescue innervation, resulting in cell death by apoptosis. This explanation is strengthened by a detail in case 1, not previously encountered in CBSD: the ectopic cortico-pontine fibre tracts, misdirected to the anterior medullary velum ( Figure 1H). No collection of ectopic neurons was found there. The heterotopic cortico-pontine tracts signify that pontine nuclei were not normally innervated by afferents from the cerebral cortex at any time from the moment of their origin. Incidentally, the abnormal tracts signify axon misrouting, a process which partly relies on a gradient of the chemo-attractant Netrin-1 for the directed outgrowth of neurites, which is similar to the process by which migrating young neurons are guided at an earlier stage [16]. No experimental data exist on the impact of the loss of EZH2 on the migration of future inferior olivary neurons. These neurons also originate in the rhombic lip, and are similarly guided by the presence of Netrin-1 [17] in the floor plate; therefore, likely to be affected by loss of migratory control in the case of lost EZH2 function. The loss of pontine and inferior olivary neurons precluded their use to test the presence of H3K27me3. Reliable histone typing in brainstem nuclei was impeded by altered anatomy. By good fortune, the cerebellum, which belongs to the embryonic rhombencephalon [18] was sufficiently preserved, presenting a homogenous cell-rich structure that could be used for histone testing with consistent results. In a recent study in embryonic mice, targeted knockout of EZH2 from the future cerebellar region [19] resulted in global hypoplasia. The loss of EZH2 affected cerebellar precursor cells from both the rhombic lip and the ventricular zone, respectively, resulting in loss of internal granule cells, an altered ratio between GABAergic neurons and loss of Purkinje cells. Histone staining results of the cerebellum for H3K27me3 were remarkably similar to the present cases of CBSD. Experimental loss of EZH2 also results in delayed HOX gene silencing, causing rhombomeric and vertebral segmentation defects [9,10]. The latter effect bears morphological similarity to vertebral segmentation defects as observed in patient 1 as well as a significant portion of published CBSD cases [1]. Table 2 summarizes the multiple roles for EZH2 in the development of the nervous system and vertebral column as presently known.
In conclusion, this study of two cases of CBSD shows the involvement of EZH2 by the diminished presence of its enzyme product histone H3K27me3, throughout the cerebellum, as well as by the absence of EZH2 transcription in the same regions. Comparison to the results of experimental knockout of EZH2 by selective targeting of rhombomeres [11] well explains part of its unusual neuropathology, for example, the vanished pontine and inferior olivary nuclei. Also, cerebellar hypoplasia in CBSD compares well with experimental knockout of EZH2 from the future cerebellar region [19]. The mechanism by which this local gene dysregulation in early human embryogenesis is initiated in the apparent absence of systemic genomic mutations remains to be addressed. A case of PTCD, studied because of important morphologic similarities to CBSD was unaffected by histone changes.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1. H3K27me3 immunoreactivity in case 3 (PTCD; pontine tegmental cap dysplasia). Figure S2. H3K27me3 immunoreactivity in control cerebellum.