Smchd1 regulates a subset of autosomal genes subject to monoallelic expression in addition to being critical for X inactivation

Abstract

Background: Smchd1 is an epigenetic modifier essential for X chromosome inactivation: female embryos lacking Smchd1 fail during midgestational development. Male mice are less affected by Smchd1-loss, with some (but not all) surviving to become fertile adults on the FVB/n genetic background. On other genetic backgrounds, all males lacking Smchd1 die perinatally. This suggests that, in addition to being critical for X inactivation, Smchd1 functions to control the expression of essential autosomal genes.

Results: Using genome-wide microarray expression profiling and RNA-seq, we have identified additional genes that fail X inactivation in female Smchd1 mutants and have identified autosomal genes in male mice where the normal expression pattern depends upon Smchd1. A subset of genes in the Snrpn imprinted gene cluster show an epigenetic signature and biallelic expression consistent with loss of imprinting in the absence of Smchd1. In addition, single nucleotide polymorphism analysis of expressed genes in the placenta shows that the Igf2r imprinted gene cluster is also disrupted, with Slc22a3 showing biallelic expression in the absence of Smchd1. In both cases, the disruption was not due to loss of the differential methylation that marks the imprint control region, but affected genes remote from this primary imprint controlling element. The clustered protocadherins (Pcdhα, Pcdhβ, and Pcdhγ) also show altered expression levels, suggesting that their unique pattern of random combinatorial monoallelic expression might also be disrupted.

Conclusions: Smchd1 has a role in the expression of several autosomal gene clusters that are subject to monoallelic expression, rather than being restricted to functioning uniquely in X inactivation. Our findings, combined with the recent report implicating heterozygous mutations of SMCHD1 as a causal factor in the digenically inherited muscular weakness syndrome facioscapulohumeral muscular dystrophy-2, highlight the potential importance of Smchd1 in the etiology of diverse human diseases.

Figure 1

Differential expression in Smchd1^{MommeD1/MommeD1} versus Smchd1^+/+ embryos at E9.5. Heat maps for differentially expressed genes identified in the microarray analysis comparing the expression of (A) female and (B) male embryos. The lists of differentially expressed genes were sorted in order of level of significance with multiple testing correction and those showing a significant difference between the genotypes (adjusted P < 0.05) are displayed on the heat maps. For the female comparison, the top 20 genes are shown, while for the male comparison all genes that were significantly different are shown.

Figure 2

A subset of imprinted genes in the Snrpn cluster shows increased expression in Smchd1 mutants. (A) The Snrpn imprinted gene cluster, where those genes expressed from the paternal allele are shown in blue and those expressed from the maternal allele are in red. CpG islands are represented by circles on the line and where these represent an sDMR (small circles) or imprint control region (ICR) (large circle) the methylation status is indicated (M inside the filled circle for the methylated allele or unfilled circle for the unmethylated allele). Quantification of the expression levels of imprinted genes or transcripts within the Snrpn cluster by qRT-PCR measured using RNA derived from: (B) male and female E9.5 embryos, (C) mouse embryonic fibroblasts (MEF) derived from male E14.5 embryos, and (D) whole brain from adult male mice. The synthesis of first-strand cDNA from E9.5 embryo RNA was primed with oligo dT, but for the analysis of transcript levels in MEFs and brain the cDNA synthesis was primed with a cocktail of the reverse primers used for qRT-PCR. In each case the qRT-PCR signal was normalized relative to that of Rala and plotted relative to the corresponding Smchd1^+/+ sample. The genotype, sex, and number of replicates are indicated in each case. Statistical analysis was performed using the t test. ** P < 0.01 and * P < 0.05 compared with wildtype. Error bars indicate standard error.

Figure 3

The epigenetic signature of deregulated genes in the Snrpn cluster is consistent with biallelic expression. The relative level of DNA methylation was quantified by qPCR of MeDIP recovered DNA in samples from (A) male and female E9.5 embryos, and (B) MEFs isolated from male E14.5 embryos. (C) The relative level of H3K4me2 for each gene was quantified by qPCR of ChIP recovered material in chromatin derived from MEFs isolated from male E14.5 embryos. The qPCR signal was normalized relative to that of Rhox6/9 for MeDIP or relative to input for the H3K4me2 ChIP. Results are plotted relative to the corresponding Smchd1^+/+ sample. The genotype, sex, and number of replicates are indicated in each case. Statistical analysis was performed using the t test. In (A) **P < 0.01 or *P < 0.05 for Smchd1^{MommeD1/MommeD1} compared with Smchd1^+/+. Statistical significance not shown for the comparison of Smchd1^+/+ compared with Smchd1^MommeD1/+. In (B) and (C) ** P < 0.01 compared with wildtype. Error bars indicate standard error. ChIP, chromatin immunoprecipitation.

Figure 4

RNA-FISH showing biallelic expression of deregulated genes in the Snrpn cluster.Snrpn (green) and Ndn (red) RNA-FISH signals on (A)Smchd1^+/+ and (B)Smchd1^{MommeD1/MommeD1} MEFs. Snrpn (green) and Magel2 (red) RNA-FISH signals on (C)Smchd1^+/+ and (D)Smchd1^{MommeD1/MommeD1} MEFs. Images representative of the predominant hybridizing pattern for each genotype and gene combination are shown. The number of RNA-FISH signals per nucleus for (E)Snrpn/Ndn and (F)Snrpn/Magel2 were counted (100 randomly selected nuclei counted in each case) and plotted for each line of MEFs (Smchd1^+/+, Smchd1^{MommeD1/MommeD1}#1, and Smchd1^{MommeD1/MommeD1}#2). The genotypes are indicated in each case.

Figure 5

Loss of imprinting at the Igf2r imprinted gene cluster in placenta in the absence of Smchd1. (A) The Igf2r imprinted gene cluster shows imprinting in the placenta, where genes expressed from the paternal allele are shown in blue and those expressed from the maternal allele are in red. CpG islands are represented by circles on the line and where these represent an ICR (large circle) the methylation status is indicated (M inside the filled circle for the methylated allele or unfilled circle for the unmethylated allele). (B) In Smchd1^+/+ F1 (C57Bl6/J sire × FVB/n dam) placental expression of both Igf2r and Slc22a3 is predominantly from the maternal FVB/n allele with only a minor proportion from the paternal C57Bl6/J allele, but in Smchd1^{MommeD1/MommeD1} placentas, expression of Slc22a3 becomes biallelic while expression of Igf2r remains imprinted with predominant expression of the maternal allele. (C) The same result is seen in the reciprocal cross F1 (FVB/n sire × C57Bl6/J dam) except here the maternal allele is derived from the C57Bl6/J dam. The genotype and number of replicates are indicated in each case. Statistical analysis was performed using the t test. ** P < 0.01 compared with wildtype. Error bars indicate standard error.

Figure 6

The clustered protocadherin genes show altered expression levels in Smchd1 mutants. (A) The clustered protocadherin genes (Pcdhα, Pcdhβ, and Pcdhγ) lie in a cluster on mouse Chr:18 and are expressed in a unique manner. Transcripts from the Pcdhα and Pcdhγ genes show alternative first exons that are monoallelically expressed in a random combinatorial manner and spliced to downstream exons that are biallelically expressed. The Pcdhβ genes exist as individual genes. (B) The expression levels of the protocadherin genes (Pcdhα, Pcdhβ, and Pcdhγ) were quantified by qRT-PCR using RNA derived from whole adult male mouse brains. The synthesis of first-strand cDNA was primed with a cocktail of the reverse primers used for qRT-PCR. In each case, the qRT-PCR signal was normalized relative to that of Rala and plotted relative to the corresponding Smchd1^+/+ sample. Statistical analysis was performed using the t test. ** P < 0.01 and * P < 0.05 compared with wildtype. Error bars indicate standard error.