Reply to Pembrey et al: ‘ZNF277 microdeletions, specific language impairment and the meiotic mismatch methylation (3M) hypothesis’

In a recent paper,¹ we described a homozygous exonic microdeletion in ZNF277 in a girl with specific language impairment (SLI). This microdeletion was also identified in the heterozygous form in eight families of the SLI Consortium (SLIC) cohort and four families with ASD cases from the IMGSAC Cohort. We observed an increased allelic frequency of ZNF277 microdeletions in SLI probands (1.1%) compared with both ASD family members (0.3%) and unrelated controls (0.4%), suggesting that these microdeletions might be a risk factor for SLI. However, as the ZNF277 microdeletions showed incomplete segregation with the SLI phenotype, as they were also identified in unaffected family members and, in some cases, they were not inherited by the affected children (reverse discordance), we hypothesised that these CNVs might contribute to the SLI susceptibility in a complex manner, acting as a risk factor with a reduced penetrance.

Pembrey and colleagues have suggested that an epigenetic mechanism may account for the reverse discordance observed for the maternal ZNF277 microdeletions. According to their hypothesis, called meiotic mismatch methylation (3 M),² during the maternal meiosis I, a deletion on one chromosome might lead to abnormal methylation of the normal allele on the homologous chromosome, which would result in the silencing of the non-deleted allele. This hypothesis implies that all the children with a mother who carries a ZNF277 microdeletion would be at risk, irrespective of the copy number of their inherited allele. In contrast, children in families where the father carries a ZNF277 microdeletion would only be at risk if they directly inherit the event. Thus, one would expect to see an increased frequency of SLI in families in which the mother carries a ZNF277 microdeletion.

In our discovery pedigree, we did observe reverse discordance: the proband inherited a deleted copy of ZNF277 from both parents, whereas the affected sister and the brother, who had mild language impairment, did not carry the microdeletion. In the additional SLI families identified to carry a ZNF277 microdeletion, we identified three maternal and three paternal ZNF277 microdeletions. In both family types, we observe a similar level of affected children (four out of nine children are affected in maternal lines, while four out of eight children are affected in paternal lines). Moreover, in the maternal lines, we do observe unaffected children supporting the incomplete penetrance of this allele. Therefore, although the sample size is very small, in the SLI families we did not observe a clear maternal pattern that would support the 3 M hypothesis. In our study, however, the autism families did display aspects of maternal inheritance; in all four identified ASD families (of 252 screened), the ZNF277 microdeletions were always carried by the mothers and a high number of offspring were affected (eight of nine offspring have ASD or PDD; Figure 1).

Figure 1

The figure shows the four ASD families carrying the ZNF277 microdeletions. Black filling indicates a diagnosis of autism, grey filling indicates a diagnosis of pervasive developmental disorder (PDD). Lymphoblastoid cell lines from underlined children were available for the analysis.

It is known that DNA methylation profiles are a combination of inherited epigenetic marks and de novo modifications that can be spatio-temporally heterogeneous and dynamically regulated throughout development (as reviewed by Zhou³). The 3M hypothesis assumes a constitutive epigenetic effect on the expression of the risk gene, established at the very early stage of the life span (maternal meiosis). DNA methylation of ZNF277 has not been directly examined in any of the SLI or ASD families included in our original article. However, as part of our previous study, we did examine ZNF277 transcript levels in lymphoblastoid cell lines derived from the four ASD families (Figure 1). When we analysed all the available individuals carrying the ZNF277 microdeletion, we observed a reduced expression compared with that of the housekeeping gene GUSB, which was chosen to have comparable expression levels to ZNF277 (as shown in Ceroni et al¹). In contrast, when we analysed all the children of mothers carrying the microdeletion, the ZNF277 gene did not display altered expression (normalised average expression level of 1.01 across six samples), nor did we observe reduced expression when we restricted the analysis only to the children who did not inherit the ZNF277 maternal deletion (normalised average expression level of 1.07 across four samples with lymphoblastoid samples available). Together, these data suggest that there is no constitutive silencing of the maternal copy of ZNF277 by methylation. We recognise, however, that the number of available individuals is too small to conclusively rule out the possibility of an epigenetic regulation. Larger sample sizes and more extensive investigations, involving methylation sequencing in male and female lines, would be required to fully inform the 3 M hypothesis.