Holoprosencephaly (HPE; MIM#236100) is the most common brain malformation in humans. It results from incomplete midline division of the prosencephalon between days 18 and 28 of gestation.[1],[2] The estimated prevalence of HPE in liveborn children is less than 1/10,000, but the incidence may be as high as 1/250 in first-trimester conceptuses.[3, 4] HPE is usually associated with facial abnormalities, such as cyclopia, proboscis, or cleft lip/palate in cases of severe HPE. It has been said that “the face predicts the brain” but this rule is not applicable in all cases.[5, 6] Three classical anatomical classes have been described, in increasing order of severity: lobar, semi-lobar and alobar HPE. The full spectrum of HPE also includes middle interhemispheric variants (MIH) or syntelencephaly, septopreoptic HPE and microforms characterised by midline defects (e.g. single maxillary median incisor (SMMI) or hypotelorism) without the brain malformations typical of HPE.[7, 8, 9, 10]

Not only is HPE phenotypically highly variable, its aetiology has also been shown to be very heterogeneous.[11, 12, 13] Indeed, HPE may be caused by chromosome abnormalities, such as trisomy 13, 18 and triploidy in particular, or may be one of the components of a multiple malformation syndrome, such as Smith-Lemli-Opitz or CHARGE syndrome. HPE may also result from exposure to maternal diabetes or hypocholesterolaemia during gestation.[14, 15] Four major genes responsible for HPE have been characterised — SHH, ZIC2, SIX3 and TGIF — and another nine seem to play a lesser role in the occurrence of HPE.[13] The “multiple-hit hypothesis” is now the most widely accepted model of HPE. According to this hypothesis, combinations of mutations in major and/or minor HPE genes lead to the occurrence of HPE and may account for variability in terms of severity.[16] Genetic counselling thus remains very complex for families affected by HPE.[17] There are clearly many interactions between the various signalling pathways involved in forebrain development that remain to be elucidated.

It is therefore difficult to establish phenotype-genotype correlations for this disease, due to the large number of interacting factors. Studies of large series are therefore required. We and others have reported specific findings, including, in particular, the description of particular microforms often associated with SHH, the predominance of severe forms in subjects with SIX3 mutations and a ZIC2-specific phenotype with minor facial features.[6, 9, 18, 19, 20, 21]

We report here clinical and molecular data for the largest ever European series, including 645 probands and their relatives and a high proportion of foetuses; only one larger series has been described, based solely on children with HPE (NIH, Muenke’s team).[21] We aimed to identify new phenotype-genotype correlations in an updated large series of probands with at least one mutation in an HPE-related gene and/or at least one rearrangement identified on array-CGH analysis. Our statistical analyses confirm previous reports and focus, for the first time, on additional features, such as associated brain malformations, neural tube defects and extra-craniofacial defects. These new phenotype-genotype correlations should help to guide molecular analysis strategies and, thus, genetic counselling.

In 1996, a European network for HPE was set up in Rennes, France.[1, 9, 18, 22] Clinical data for about 1500 patients, including 645 HPE probands and 699 relatives, were analysed and DNA was prospectively collected. All the data were input into a specific secure online database.

This study was approved by the institutional review board. All samples and clinical data were collected after informed consent had been obtained.

The main data, including HPE type, facial abnormality category, associated brain malformations, neural tube defects and extracraniofacial malformations, are shown in table 1.

HPE is a very complex disorder in terms of both clinical variability and genetic signalling pathways. We report here the clinical and molecular data for the largest ever European series, encompassing 645 probands and 699 relatives, with a higher proportion of foetuses (51%) than in previous series.[19, 20, 21] The large number of patients included in this series resulted in sufficient power for the detection of statistically significant phenotype-genotype correlations.

In the series as a whole, the sex ratio (F:M) was 1.2, in favour of a slight predominance of female subjects. Female subjects were in the minority among probands with SHH mutations, for which the sex ratio was 0.9, whereas a high predominance of female subjects (sex ratio of 1.8) was observed among those with ZIC2 and SIX3 mutations. The SIX3 sex ratio was consistent with that reported by Lacbawan.[19] This suggests that mutations in ZIC2 or SIX3 may be embryonic-lethal in males.

Screening for mutations in the major genes SHH, ZIC2, SIX3 and TGIF and the minor gene GLI2 in this large series gave an estimated global mutation rate of 25.8%, with specific mutation rates of 10.4% (SHH), 8.2% (ZIC2), 5.1% (SIX3), 1.7% (TGIF) and 0.4% (GLI2). These results are consistent with our previous findings for smaller series.[1, 9, 18, 34] Point mutations in SHH, SIX3 and TGIF were highly heritable, with heritabilities of 73%, 88% and 100%, respectively, whereas most ZIC2 mutations were de novo (70% of cases).[6, 20] This may be due to a higher penetrance and the high frequency of severe HPE in individuals with ZIC2 mutations, resulting in the identification of a larger number of sporadic cases. By contrast to what was reported by Lacbawan et al., no predominance of maternal inheritance was observed for SIX3 mutations in our series, whereas such a pattern of inheritance was observed for SHH mutations.[19]

Overall, more severe types of HPE (atelencephaly, alobar or semilobar HPE) were reported in individuals with SIX3 (74%) and ZIC2 (75%) mutations than in individuals with SHH (62%) and TGIF (57%) mutations. Moreover, atelencephaly/aprosencephaly was associated exclusively with SIX3 mutations. Pasquier et al. reported this particular association and other teams have reported a high proportion of severe cases in subjects with SIX3 mutations, including ophthalmological defects in particular.[19, 20, 33] Thus, alobar HPE tends to be associated with SIX3, semilobar HPE with ZIC2 and microforms with SHH, consistent with the findings of Solomon.[21] Furthermore, our findings confirmed the specific association of MIH or syntelencephaly with ZIC2, as the mutations identified in four of the 11 cases of MIH in the series concerned the ZIC2 gene.[6]

This study highlights the considerable intra- and interfamilial phenotypic variability in SHH probands.[35] We report here, for the first time, the high prevalence of brain microforms (74%) in probands with SHH mutations presenting, essentially, pyriform sinus and choanal stenosis (category 3). Associations with midline defects, coloboma and sacral agenesis (systematically linked to a large 7q36 deletion) were also observed. By contrast, facial morphology in subjects with ZIC2 mutations was generally characterised by moderate facial dysmorphia or even a normal face (category 4), associated with alobar or semilobar HPE. This feature seems to be specific to ZIC2 mutations and has also been reported by other teams.[6, 20] Solomon et al. mentioned a particular facial phenotype found in some of the cases in our series, with a high proportion of abnormal noses (9 cases) usually short, sometimes flat or large, and ear dysplasia (4 cases). Furthermore, statistical analyses revealed a lack of correlation between the severity of HPE brain malformation and facial features in probands with ZIC2 mutations, but not in those with SHH, SIX3 and TGIF mutations.

We also considered other associated brain malformations, focusing in particular on RES, which was associated with HPE in six of the 645 HPE probands. This extremely rare malformation is the histological equivalent of HPE for the cerebellum. The signalling pathways involved in this disease has yet to be identified, but we suggest that there may be interactions with HPE signalling pathways, as a ZIC2 mutation has been reported in two siblings with partial RES and HPE.[36, 37] SHH and SIX3 mutations have recently been associated with schizencephaly, but only three cases of schizencephaly were reported in our series, with no mutation identified in the probands concerned.[31]

We also identified two other previously unknown associations with ZIC2 mutations: rachischisis and neuronal migration abnormalities. Neural tube defects are frequently reported in HPE, but rachischisis seems to be more specifically associated with ZIC2 mutation, as such mutations were found in the only two cases of rachischisis in our series. In addition, neuronal migration abnormalities have already been described, but without phenotype-genotype correlations.[2, 38, 39] We focus here on neuronal migration abnormalities, such as gyral pattern abnormalities and heterotopia, which were almost exclusively restricted to probands with ZIC2 mutations (8 cases). Hahn et al. found subcortical heterotopic grey matter or cortical dysplasia in two thirds of MIH cases.[38] In our series, neuronal migration abnormalities were observed in two of 11 cases of MIH, but no ZIC2 mutation was identified in these cases.

We report here, for the first time, the frequency and variety of extracraniofacial malformations in a large series, together with the corresponding molecular data. Molecular defects were more frequent in ZIC2 than in the other three main genes. We found a higher frequency of extracraniofacial defects than reported by Solomon, probably because his study focused on children, rather than foetuses.[6] Renal/urinary defects were significantly associated with SHH or ZIC2 mutations and no such defects were found in probands with SIX3 or TGIF mutations. SIX3 was the only gene for which no heart defect was reported.

The data for TGIF and, above all, for GLI2 revealed no significant phenotype-genotype correlations, but alobar HPE with severe facial features (category 1) seemed to be more frequent in probands with TGIF mutations, although phenotypic variability was high. We suggest that phenotype-genotype correlations for TGIF and GLI2 should be analysed in larger series. Another disease spectrum, characterised by anterior pituitary insufficiency and polydactyly, seems to be more specifically linked to GLI2 mutations[23, 40] (Dubourg, unpublished data).

Several signalling pathways are known to be involved in HPE, and the various mechanisms underlying HPE may account for the observed phenotypic variability. Indeed, SIX3 is involved in eye development, consistent with the eye defects observed in probands with SIX3 mutations, and in ventral forebrain development, through regulation of the SHH signalling pathway. Furthermore, ZIC2 significantly differs from the other HPE-related genes, as shown in phenotype-genotype correlations, and its product also acts by different mechanisms. Indeed, ZIC2 is known to act both early stage, in upstream SHH signalling during mid-gastrulation, and at later stages, in the development of the dorsal telencephalon, potentially accounting for the specific occurrence of MIH and neural tube defects in ZIC2 probands.

Array-CGH studies in this series confirmed the high frequency of rearrangements reported in a preliminary study.[26] Indeed, rearrangements were identified in 22% of the 260 patients studied (14% de novo and 8% inherited). The proportion of foetuses and the severity of HPE phenotype (extracraniofacial malformations included) were not significantly higher in probands with rearrangements than in the series as a whole, as might have been expected. The higher frequency of category 2 and 4 probands may reflect more common features, such as cleft lip or palate (category 2) or mild non-specific features (category 4). This population seemed to be heterogeneous and an understanding of the molecular basis of this heterogeneity should facilitate the definition of more homogeneous subgroups. Indeed, the high frequency of rearrangements was consistent with there being a large number of loci corresponding to new candidate genes. The recurrent regions included the MACROD2 and PATCHD3 loci, in particular. Furthermore, recurrent deletion 6qter led to the identification of DLL1 as a new HPE-associated gene and of NOTCH as a new signalling pathway involved in HPE.[41]

It has already been suggested that HPE involves a multiple-hit process.[26, 42] Animal models have provided evidence of digenism, by implicating either the same or two different signalling pathways. Genetic background also played an important role in determining the severity of the phenotype.[43, 44, 45] In humans, low penetrance or variable levels of expression in HPE multiplex families and reports of patients with two different mutations support this hypothesis.[16, 24, 46, 47] Our series included up to 15 cases of double mutation: eight cases of double rearrangements and seven cases of rearrangement associated with a mutation in an HPE-related gene (6 rearrangements associated with a mutation in SHH, ZIC2 or SIX3 and one ZIC2 deletion associated with a GLI2 mutation). This findings provide solid support for the multiple-hit hypothesis in HPE.

In conclusion, even if the identification of an HPE gene mutation cannot provide a precise prognosis, the new phenotype-genotype correlations identified here should facilitate the definition of a better molecular analysis strategy. We propose a new algorithm based on these correlations (Fig. 3). Evidence of double mutations in HPE probands provides support for the HPE multiple-hit hypothesis underlying the complexity of genetic counselling and research on this disorder. The known clinical variability of HPE also suggests an overlap between environmental and genetic mechanisms. A combination of clinical and functional studies should help to improve our understanding of the complex interactions between the various signalling pathways involved in HPE and brain development.