Introduction
Hydrocephalus which literally means any increase in cerebrospinal fluid (CSF) within the skull has been more precisely defined by the International Hydrocephalus Working Group which describes “an active distension of the ventricular system resulting from inadequate passage of cerebrospinal fluid from its point of production within the cerebral ventricles to its point of absorption into the systemic circulation” [
24]. In infants, its prevalence varies between 1 and 32 per 10.000 births and has been estimated by Munch et al. at 1.1 per 1,000 infants when including cases diagnosed before 1 year of age in the absence of other extrinsic causes and after exclusion of neural tube defects [
14]. Several classifications have been proposed depending on the pathophysiological mechanisms, aetiology, or treatment modalities. In the aetiological classification proposed by Tully and Dobyns [
24], congenital hydrocephalus is considered either as acquired representing about half of the cases and mainly due to haemorrhage, infection or neoplasia, or of developmental nature, also termed “intrinsic hydrocephalus”. This pathological condition is also separated into two subgroups, i.e., communicating i.e., with no point of obstruction or resistance to cerebro-spinal fluid dynamics, or obstructive, knowing that most of the time obstruction is the major cause of hydrocephalus. As regards obstructive hydrocephalus physiopathology, multiple points of obstruction have been recognized, including foramina of Monroe, aqueduct of Sylvius, 4
th ventricle foramina, spinal and cortical subarachnoid spaces, and from birth arachnoid villi and venous hypertension [
17]. Moreover, several points of obstruction may coexist in a same patient. Regarding developmental causes, hydrocephalus is classically divided into syndromic (representing about 75% of the cases and due to chromosomal abnormalities in 30% of them, thus accounting for 6% of all causes of hydrocephalus), and non-syndromic forms. But the distinction between them is most of the time difficult, since additional anomalies may be present in apparently non-syndromic forms. No specific cause is found in more than half of the patients although they present a syndromic form in 11% of the cases, with only 0.6% of their whole infantile series having an identifiable genetic syndrome [
24]. In individuals with apparently isolated hydrocephalus or with no major additional clinical findings,
L1CAM mutations represent the most common genetic form with a prevalence of approximately 1:30,000 and account for about 5–10% of males with non-syndromic congenital hydrocephalus [
19].
L1CAM pathogenic variants are responsible for a wide spectrum of phenotypes, now termed L1, the most severe form being Hydrocephalus with Stenosis of the Aqueduct of Sylvius (HSAS; MIM#307000). More than 200 different deleterious variants spanning over the entire gene have been reported so far [
21,
28,
30]. In HSAS, the stenosis of the aqueduct of Sylvius is a hallmark of the disease together with hydrocephalus, adducted thumbs, pyramidal tract agenesis/hypoplasia, corpus callosum agenesis/hypoplasia and cerebellar anomalies [
1,
27]. More recently, mutations in other genes have been reported to be causative for non-syndromic hydrocephalus, notably in the
MPDZ gene (MIM#615219)
. In 2013, a founder mutation in this gene was identified in two consanguineous Saudi families in whom the foetuses presented ultrasonographically massive bilateral hydrocephalus [
2], but no post-mortem examination could be obtained, so that the underlying mechanism of hydrocephalus remained unknown.
In a previous work, we reviewed the neuropathology of 138 cases genetically tested for X- linked hydrocephalus [
1] that allowed us to classify patients who did not display any pathogenic variant in the
L1CAM gene into four distinct subgroups. Among them, atresia/forking of the aqueduct of Sylvius represented 27% of the cases, associated in some foetuses with rhombencephalosynapsis, fusion of the colliculi, atresia of the 3
rd ventricle, corpus callosum abnormalities consisting of partial/complete total agenesis or hypoplasia, and pyramidal tract hypoplasia or asymmetry. Though atresia of the aqueduct of Sylvius is usually considered as resulting from haemorrhage or infection, an autosomal recessive mode of inheritance was suspected in two consanguineous families, one in whom a child died from severe hydrocephalus and a foetus was interrupted for recurrent hydrocephalus, and another family in whom three foetuses were affected. Since then, recurrent massive hydrocephalus leading to the termination of the pregnancy occurred in an additional consanguineous family. We describe three novel homozygous
MPDZ null mutations in these three families along with the prenatal phenotype and foetal neuropathological lesions, and provide some insights into the potential pathological mechanisms.
Discussion
The
MPDZ gene, located on chromosome 9p24-p22, was first involved in severe non-syndromic congenital hydrocephalus by the characterization of a homozygous nonsense variant in exon 6 in two Saudi consanguineous families, c.628C > T; p.(Gln210*), suggesting a founder effect [
2]. In the two index case foetuses, US examination showed severe bilateral symmetrical dilatation of the lateral ventricles with dangling of the choroid plexuses, probable callosal dysgenesis or agenesis and inadequate visualization of the cavum septum pellucidum. The third ventricle could not be visualized and there was a notable thinning of the cerebral cortex. Neuropathological examination was not performed and therefore, the morphological characteristics remained unknown. Genetic testing using a gene panel for non-syndromic congenital hydrocephalus led us to identify a second
MPDZ homozygous pathogenic variant in a foetus born to consanguineous Senegalese parents (Family 1; Fig.
1) who displayed atresia/forking of the aqueduct of Sylvius, suggesting that the specific phenotype of hydrocephalus due to atresia/forking of the aqueduct of Sylvius results from
MPDZ mutations. We therefore decided to screen the
MPDZ gene using this gene panel in the sub-group of 21 foetuses with atresia-forking of the aqueduct of Sylvius belonging to 17 families we previously identified [
1]. This approach allowed the identification of two other
MPDZ homozygous pathogenic variations (Families 2 and 3; Fig.
1). All these four pathogenic variants are null variants: 2 nonsense variants, 1 frameshift deletion and 1 splice variant. It is worth noting that all these pathogenic variants were only detected in foetuses born to consanguineous parents, suggesting that the frequency of heterozygous pathogenic variants in the general population is very low. To assess this prevalence, we examined the ExAC database variants in the
MPDZ gene. Among 2,210
MPDZ variants in the 60,706 ExAC control individuals, 80 are canonical splice site, nonsense or frameshift mutations. These 80 inevitably truncating variants have been found in 155 heterozygous individuals, resulting in an estimation of heterozygous carriers of ~1/624, highlighting the rarity of loss-of-function
MPDZ alleles in the general population. Accordingly, MPDZ constitutes a rare cause of congenital hydrocephalus and should be considered first and foremost in consanguineous families. Nevertheless, the phenotype associated with less severe mutations such as missense or non-canonical splice variants may be different and remains to be described.
In humans, HSAS encompasses approximately 5–15% of congenital hydrocephalus cases with a genetic cause. When excluding X-linked hydrocephalus, the frequency of non-syndromic forms is very low. Empiric recurrence risk rates range from <1 to 4% [
32], indicating the rarity of autosomal recessive congenital hydrocephalus. In HSAS, stenosis of the aqueduct of Sylvius is a hallmark of the disease together with hydrocephalus, adducted thumbs, pyramidal tract agenesis/hypoplasia and corpus callosum agenesis/hypoplasia [
1,
27].
AP1S2 gene pathogenic variants (MIM#300629) which cause hydrocephalus with mental retardation are associated with calcifications and iron deposits in the basal ganglia [
20].
Contrasting with the rarity of non-syndromic hydrocephalus known disease-causing genes in humans, a plethora of hydrocephalus mouse models have been generated. Many congenital hydrocephalus genes/loci have been recognized, allowing the search for possible molecular and cellular pathophysiological mechanisms. About ten years ago, Zhang et al. [
32] proposed a classification into four subgroups. The first consisted in disruption of neural cell membrane proteins. Second, hydrocephalus could be caused by malfunction of ependymal cell cilia and related proteins; third, by malfunction of mesenchymal cells and perturbation of growth factor signalling pathways and fourth by extracellular matrix disruption. From the present work and from several previous studies, another key pathophysiological mechanism should be discussed, consisting in cell membrane junction component alterations, i.e., adherens and gap junctions as well as cell to cell adhesion molecules (especially N-cadherin) which normally join neuroepithelial cells together from embryonic stages, as early as the 4
th post-conception week in humans. These alterations have been shown in spontaneous or engineered mutant animal models to be responsible for development of congenital hydrocephalus by disrupting the VZ with subsequent loss of neuroepithelial cells and later ependyma denudation [
3,
12,
16,
18].
Tight junctions (TJs) connect adjacent cells so tightly that they constitute a paracellular barrier that prevents the diffusion of solutes, lipids and proteins across the epithelial cell sheets [
26]. Therefore, any disruption of TJs and particularly abnormal cell-cell adhesion by PDZ proteins including MPDZ (also named MUPP1 which is strongly expressed in choroid plexuses) most likely alters the distribution of TJs that leads to uncontrolled secretion of CSF and hydrocephalus. Besides, PDZ proteins are modular proteins that act as adaptors by selective interactions of their PDZ domains to other protein modules [
4]. They are localized to specialized submembranous sites including synaptic, tight, gap and neuromuscular junctions. MUPP1 which is concentrated at TJs contains 13 PDZ domains, and constitutes a scaffold for several other tight junction components such as claudins, occludin (Ocln), JAMs, angiomotin (Amot), Amot-like 1 and 2 [
9,
23,
25]. The neuropathology in humans has been described only in a minority of mutations in other tight junction component genes and differs from the phenotype observed in case of
MPDZ mutations. Recessive mutations in the
OCLN gene (MIM#251290) also named pseudo-TORCH syndrome, cause microcephaly with band–like calcifications, simplified gyral pattern and polymicrogyria. OCLN is mainly expressed in the endothelium, pericytes and surrounding astrocytes [
15], and the defective protein leads to destructive lesions. Among the claudin family, only
CLAUDIN1 mutations have been described (MIM#603718) and cause ichthyosis, vacuolated leukocytes and alopecia but without brain lesions. Besides, human brain lesions remain unknown in case of
JAM1 and
JAM4 mutations (MIM#610638). JAM1 also interacts with another PDZ-containing domain, Afadin which is localized at the ependymal zonulae adherens and TJs of the third ventricle and of the aqueduct of Sylvius and whose genetic deletion induces hydrocephalus with disappearance of ependyma in the mouse midbrain and obliteration of the third ventricle [
31]. But again, no human pathology has been reported until now. Although it has been demonstrated that Amot colocalizes with occludin in mice, no human pathology has been associated with mutations in the
AMOT gene so far. Several individuals from a single large family with autosomal recessive
JAM3 mutations [
13] presented all bilateral cataracts and some of them had hepatomegaly and thrombocytopenia. MRI displayed multifocal intra-parenchymal haemorrhages in the white matter and basal ganglia with secondary ventricular dilatation owing to major JAM3 expression in the vascular endothelium.
MUPP1 expression has been documented using immunocytochemistry in adult mouse brain, with highest expression on the apical surface of the choroid plexuses, strong expression in the hippocampus, amygdala and pyriform cortex, in Layer II of most neocortical areas as well as in all layers of the cerebellar cortex and in several brainstem nuclei [
9,
22], but its expression in human brain is unknown. In our control case, MUPP1 was strongly expressed in the aqueduct ependyma, conversely to the patients in whom a total lack of expression was observed. The ependyma derives from radial glial cells (RGC) which are highly polarized cells joined by zonulae adherens, gap and tight junctions and constitutes a barrier between parenchyma and ventricles from embryonic stage. Consequently, early tight junction alterations lead to abnormal organization of ependyma with rosette formation and narrowed ventricular lumens [
10]. In our foetuses, they were observed at different locations with aberrantly located cells in the surrounding parenchyma through loss of cohesion. These cells had the characteristic immunophenotype of RGC-intermediate progenitors (SOX2 positivity; PAX6 negativity) and of non-differentiated ependyma (nestin positivity; vimentin, GFAP and CD56 negativity). At last, we observed that the SCO was hypoplastic, probably due to early denudation of the specialized ependyma of the aqueduct leading to SCO ependyma malfunction, since SCO secretions support the integrity of ependymal cells in this region and prevent the closure of the cerebral aqueduct [
5,
11,
29].
Acknowledegments
The authors wish to thank the families for their participation to this study, Céline Lesueur and Nathalie Drouot for their technical help.