The role played by imprinted genes in placental development is gaining wider attention as the impact of imprinting disruptions on placental pathology is explored. Both hyper- and hypoplastic placentation are linked to imprinted gene function (Kanayama et al., 2002
; Murdoch et al., 2006
; reviewed in Kaneko-Ishino et al., 2003
). The most dramatic example of this is diploid, uniparental embryos. Gynogenetic embryos (two maternal/no paternal genomes) and androgenetic embryos (two paternal/no maternal genomes) both exhibit defects in the earliest differentiated tissue in mammals, the extraembryonic trophoblast (Barton et al., 1984
; McGrath and Solter, 1984
; Mann, 2005
). At mid-gestation, gynogenetic embryos possess only a few residual trophoblast giant cells while androgenetic embryos are characterized by a mass of hypertrophic trophoblast. These observations indicate that a suite of genes required for correct placental development is silenced on maternally inherited chromosomes, and must be expressed from the paternally derived chromosomes.
Following the discovery of mammalian genomic imprinting in the mid 1980's, cytogenetic experiments were used to identify imprinted regions of the mouse genome. Balanced Robertsonian translocations are normally viable, except where an imprinted region is involved. Uniparental disomy for specific domains of the mouse genome that resulted in developmental defects or inviability revealed the presence of imprinted genes involved in embryogenesis (Beechey et al., 2005
). One region of interest is proximal Chromosome 2, where an imprinted gene(s) was inferred to reside but for which no gene has yet been identified.
The search for imprinted genes has been aided in recent years by the use of differential screens involving either uniparental embryos (Miyoshi et al., 1998
; Mizuno et al., 2002
; Nikaido et al., 2003
; Ruf et al., 2006
), uniparental cell lines (Piras et al., 2000
), or Robertsonian translocations (Schultz et al., 2006
). All of these approaches are dependent on the use of quantitative amounts of starting material for the generation of probes. Of necessity, therefore, the focus was either on mid-gestation uniparental embryos, which display significant pathology, or on only mildly affected balanced translocation embryos, which are able to survive to birth. Extraembryonic tissues from parthenogenetic embryos are non-existent at mid-gestation, and in one study comparing androgenetic and fertilized embryos, extraembryonic tissues were discarded (Miyoshi et al., 1998
). In the case of uniparental cell lines, only those genes expressed in tissue culture cells and retaining their imprints were open to query. All approaches were unable to address the very early effects of parental-origin on trophoblast development.
Phenotypic differences between uniparental mouse embryos commence shortly after implantation, and affect the trophoblast most profoundly at these early stages (Surani et al., 1986
; Varmuza et al., 1993
). In order to identify imprinted genes contributing to trophoblast development before the onset of severe placental pathology, we performed microarray analysis on androgenetic, gynogenetic and biparental blastocyst cDNA. From this screen we were able to identify the PcG gene Sfmbt2
ex combs on the midleg with f
domains), which maps to the proximal region of Chromosome 2, as a putative imprinted gene. Our experiments demonstrate that Sfmbt2
is paternally expressed in early embryos and later stage extraembryonic tissues, confirming its imprinted status. The recent identification of a third PcG silencing complex in Drosophila (Klymenko et al., 2006
) composed minimally of dSfmbt and Pleiohomeotic (Pho) or Pho-like (Phol), Drosophila ortholgues of the mammalian YY1 transcription factor, prompted us to investigate the possibility that murine SFMBT2 and YY1 may form a similar silencing complex. We found that Yy1
is expressed in quantitatively similar amounts as Sfmbt2
in extraembryonic tissues, and moreover that SFMBT2 and YY1 proteins interact in mammalian tissue culture cells following transfection of expression contructs.
1.1 Microarray Screen
Early defects in uniparental embryos are difficult to study because of the limiting amounts of biological material available for analysis. We relied on a PCR based technology for generating quantitative pools of cDNA from very small samples (Iscove et al., 2002
). Three different pools of cDNA derived from androgenetic, gynogenetic or biparental blastocysts were labeled and applied to Affymetrix MOE430 A and B expression arrays. The data were analyzed with ArrayAssist 2.0 software, using a p<0.01 statistical limit on variation, and selecting genes displaying a minimum 4 fold (ln=2) difference between experimental and baseline (biparental) samples. The statistical parameters allowed the capture of several known imprinted genes. Data were exported to Excel, and sorted according to average absolute difference between androgenetic and gynogenetic samples (). An arbitrary difference of 3.0 (8 fold) was chosen as the threshold. One hundred and twenty seven (127) genes reached or surpassed the threshold difference in expression levels in uniparental embryos. While 41.7% of these genes map to regions of the mouse genome known to contain imprinted genes from the translocation studies (Beechey et al., 2005
), it should be pointed out that there may be imprinted genes that do not affect either viability or morphology because of functional redundancy with non-imprinted genes, and that the imprint map is therefore an under-representation of imprinted regions. In a similar vein, significant numbers of non-imprinted genes are expected to be identified in this kind of screen because of pathological differences between uniparental embryos, although we tried to minimize this effect by using blastocysts. The small number of genes displaying differential expression in our screen likely reflects both the reduced size of the transcriptome at the blastocyst stage, and the relatively similar phenotype of uniparental and biparental blastocysts. The two genes with the greatest difference, Slc38a4
, and Mcts2
, are both imprinted, paternally expressed genes (Beechey et al., 2005
). The next gene with the third largest expression difference is Sfmbt2
, a PcG gene mapping to MMU2A1. For comparison, the expression differential for four other known imprinted genes is shown in . Quantitative expression analysis of Sfmbt2
confirmed that Sfmbt2
mRNA is essentially absent in gynogenotes (), while displaying slightly elevated expression in androgenotes. The control Eef1a1
gene was expressed at levels very similar among the three embryo types, as expected for a non-imprinted gene, confirming successful cDNA amplification ().
Differential Expression of Imprinted Genes in Androgenetic and Gynogenetic Blastocysts
Figure 1 Sfmbt2 expression in biparental and uniparental embryos during preimplantation development. (a) Confirmation of differential expression of Sfmbt2 in androgenetic (A), gynogenetic (G), and biparental reconstituted (F) blastocysts. Data were obtained by (more ...)
1.2 Sfmbt2 Displays Monoallelic Expression
Imprinted genes are expressed from only one parental allele in at least one tissue or at one stage of development. To determine the imprint status of Sfmbt2, allelic expression was assessed in M. mus. domesticus X M. mus. castaneus F1 embryos. Two SNPs mapping within a 340 bp segment of the 3′ UTR generate strain-specific restriction polymorphisms; SnaBI cleaves the M. mus. domesticus allele into 226 and 114 bp fragments, while HinfI digests the M. mus. castaneus allele into 199 and 141 bp fragments. cDNA derived from embryo or extraembryonic tissues from M. mus. domesticus X M. mus. castaneus, or the reciprocal M. mus castaneus X M. mus. domesticus F1 progeny was assessed for allelic expression. Sfmbt2 is expressed preferentially from the paternal allele in blastocysts, embryonic day (e) 7.5 tissues, e11.5 and e14.5 yolk sac and placenta, but is biallelically expressed in e11.5 and e14.5 brain (), and other somatic tissues (not shown). Thus, Sfmbt2 is imprinted in early embryos and extraembryonic tissues, but loses its imprint in somatic tissues after e7.5.
Figure 2 Sfmbt2 is expressed from the paternal allele in early embryos and extraembryonic tissues. (a) Genomic DNA (gDNA), or cDNA derived from Domesticus X Castaneus (DC) F1 e3.5 blastocysys (blast), e7.5 embryonic ectoderm (ee), e7.5 extraembryonic ectoderm (more ...)
1.3 Sfmbt2 Promoter Region Displays Maternal Methylation in Placenta
Imprinted genes are regulated by imprinting control regions (ICRs) which are associated with blocks of tandem repeats and differential methylation (Hutter et al., 2006
; Verona et al., 2003
). We examined the published sequence for Sfmbt2
and found a CpG cluster near the transcriptional start site. Two starts are annotated for Sfmbt2
, suggesting the presence of two promoters. For one of these, the most 5′, and herein referred to as Promoter 1, there are three leader exons that are differentially spliced onto the first common exon (published EST database, and not shown). The second start site (Promoter 2) splices a different leader exon onto the first common exon. The CpG island, which is approximately 6500 bp in length, spans both starts, including all of the leader exons and part of the first major intron. Across the CpG island there was only one SNP differentiating Mus castaneus
and Mus domesticus
, an MseI
site upstream of the Promoter2 start. This necessitated a strategy involving nested PCR of surviving allelic sequences following digestion of F1 genomic DNA with methylation sensitive enzymes (). Adjacent to the start site for Promoter 2 lies a block of imperfect repeats that are highly GC-rich (83%), and that inhibited PCR (); we therefore queried the segment of the CpG island upstream of Promoter 2 that contained the single diagnostic SNP. Analysis of the 1340 bp region 5′ of the repeats revealed that the inactive maternal allele is hypermethylated while the active paternal allele in hypomethylated in e14.5 placentas ().
The promoter region of Sfmbt2 displays maternal methylation.
1.4 Sfmbt2 is Expressed Robustly in Extraembryonic Tissues
Two start and two stop sites are annotated for Sfmbt2
. RT-PCR confirmed that all combinations of starts and stops are used (not shown). The shorter isoform, which is predicted to encode a truncated protein containing only one MBT domain, results from the use of a polyA site within exon 6. Expression of Sfmbt2
in preimplantation embryos is characteristic of a group of genes with a mid-preimplantation gene activation pattern, when major embryonic expression commences between the 8-cell and blastocyst stage (), as assessed from published microarray data (Zeng et al, 2004
; Hamatani et al., Dev Cell, 2004
). Genes with this expression pattern have been purported to have roles in blastocyst formation and the emergence of distinct cell types in the preimplantaion embryo. We examined expression of Sfmbt2
by quantitative (q) RT-PCR in e7.5 and e14.5 tissues. Using primers that distinguish the two alternative starts and stops, we established that Promoter 1 is the predominant transcription start site, and that both termination sites are used equivalently (). Overall, expression of Sfmbt2
is high in extraembryonic tissues, and low in somatic tissues. Sfmbt2
is paralogous with the closely related Sfmbt1
. For tissues examined, expression of the two genes is roughly reciprocal (). In early embryos, Sfmbt2
is the only paralogue detected.
Figure 4 Quantitative RT-PCR Analysis of Sfmbt1, Sfmbt2 and Yy1 Transcripts at embryonic day 7.5 and 14.5. (a) Expression profiles of long and short Sfmbt2 isoforms and of transcripts originating from promoter 1 and 2 in various tissues. (b) Expression profiles (more ...)
Gene silencing during development is mediated by the PcG complexes PRC1 and PRC2, which together promote histone modifications (Sparmann and van Lohuizen, 2006
). Recent evidence in Drosophila
revealed the existence of a third silencing complex called Pleiohomeotic-repressive complex (PHO-RC) (Klymenko et al., 2006
), that is similar to but functionally distinct from PRC1 and PRC2. PHO-RC is composed minimally of dSfmbt and either Pho or Phol, orthologues of the mammalian YingYang1 (YY1) transcription factor. We examined expression of Yy1
in e7.5 and e14.5 tissues and found that it is coincident with Sfmbt2
in early embryos and extraembryonic tissues (), lending support to the idea that mammalian SFMBT2 and YY1 proteins may act in concert, similar to their fly counterparts.
1.5 SFMBT2 and YY1 Interact in Mammalian Cells
To assess the potential for formation of a PHO-RC in mammals, we expressed myc-tagged SFMBT2 long or short isoforms, with a murine YY1 expression vector in HEK293 cells. Co-immunoprecipitation with either anti-myc or anti-YY1 antibodies revealed that both long and short SFMBT2 isoforms interact strongly with YY1 () whereas extracts precipitated using a control GFP antibody failed to show any interaction (data not shown). These experiments indicate that both isoforms of SFMBT2 and YY1 are components of a stable protein complex, possibly PHO-RC. As the shorter isoform is predicted to encode a protein possessing no sterile alpha motif domain, the YY1 interacting domain likely resides in the SFMBT2 NH3 terminus, which harbours one MBT domain. Whether this lone MBT domain is sufficient for SFMBT2-YY1 interaction remains to be tested.
Figure 5 Mouse SFMBT2 interacts with YY1 to form a mammalian PHO-like complex. (a) Western blot analysis of Myc-tagged SFMBT2 isoforms. HEK 293 cells were transfected with the indicated combinations of expression vectors. Immunoprecipitations were performed using (more ...)
We have discovered a new imprinted gene, Sfmbt2
, which is preferentially expressed from the paternal allele in the developing placenta. As a likely member of a recently identified chromatin regulatory complex, PHO-RC, Sfmbt2
is distinguished from other known murine imprinted genes as the first imprinted Polycomb group gene. Only one other imprinted gene is known to encode a chromatin regulatory protein. Human L3MBTL
is expressed from the paternal allele in human blood cells (Li et al., 2004
); however, the murine orthologue of L3mbtl
is not imprinted (Li et al., 2005
maps to proximal mouse chromosome 2, a known imprinted region (Beechey et al., 2005
) for which no imprinted genes have been identified until now. Sfmbt2
is the first imprinted gene within this region to be identified. Previous studies had indicated that six different translocations involving proximal chromosome 2 resulted in lethality when present as a maternal uniparental duplication (Cattanach et al., 2004
). In this same paper, the authors also reported that three more distal duplications of proximal chromosome 2 were not lethal when maternally duplicated, paternally deleted. Instead, maternal disomy resulted in fetal and placental growth reduction while paternal disomy produced placental growth enhancement, pointing to at least one imprinted gene involved in placental development. While the authors interpret their results as evidence that proximal chromosome 2 does not contain any genes required for viability, we offer an alternative interpretation of the presented data, and suggest that the three distal translocation chromosomes may be providing evidence of an epistatic interaction, likely with a the central region of chromosome 2. Our explanation would explain the viability of the three distal translocation chromosomes and the lethality of the other six translocation chromosomes when maternally duplicated, paternally deleted. The most proximal of the lethal translocation chromosomes, T68H, has a breakpoint mapping to 2A3; Sfmbt2
maps to the middle of 2A1, well within the presumed lethal region. Thus, Sfmbt2
is a candidate for the placental growth effect in this epistatic interaction, as well as lethality in the absence of this interaction.
Extraembryonic trophoblast is particularly sensitive to perturbations in genomic imprinting. It is the first differentiated tissue to arise in mammalian embryos, and has been demonstrated to be adversely affected in various pathological states arising from dysregulation of imprinting. Mutations in several imprinted genes affect the trophoblast exclusivley (Coan et al., 2005
; Kaneko-Ishino et al., 2003
). Defects in the trans-acting machinery that regulates imprinting, such as DNA methyltransferase deficiency (Lei et al., 1996
; Bourc'his et al., 2001
; Hata et al., 2002
; Kaneda et al., 2004
; Arima et al., 2006
) or mutations causing familial hydatidiform moles (Judson et al., 2002
; Murdoch et al., 2006
) result in abnormal placentation. Finally, interspecific hybrids (Vrana et al., 1998
; Vrana et al., 2000
), somatic cell nuclear transfer embryos (Dindot et al., 2004
; Inoue et al., 2002
) and uniparental embryos (Surani et al., 1986
; Varmuza et al., 1993
; Mann 2005
) all suffer from dysmorphic trophoblast, in part or in whole because of abnormal expression of imprinted genes.
Of the approximately 55 protein-encoding imprinted genes that have been identified, roughly 20% can be classified as having a “placental only” imprinting (Ferguson-Smith, et al., 2006
; Wagshcal and Feil, 2006), that is while expressed in tissues of both embryonal and placental origin, expression is monoallelic in placenta but biallelic in embryonic cell lineages. Our analysis of Sfmbt2
imprinted expression has revealed that this gene fits the classification of a placental only imprinted gene. Interestingly, all of the genes exhibit paternal allele-specific expression, including Sfmbt2
(Wagschal and Feil, 2006
; herein). Functions prescribed to placental only imprinted genes are placental development, growth and function (Wagschal and Feil, 2006
; Reik and Lewis, 2005
). It is postulated that the imprinted expression patterns of this class of genes is mechanistically similar to imprinted X inactivation, and that imprinting of these genes may have arisen in ancestral placental mammals (Wagschal and Feil, 2006
; Reik and Lewis, 2005
The PcG silencing complexes PRC1 and PRC2 have been implicated in mammals with the maintenance of stem cell populations and the development of cancer (Sparmann and van Lohuizen, 2006
). Recently, a new repressive complex called PHO-RC was identified in Drosophila
that is composed minimally of dSfmbt and PHO or PHOL, the fly orthologues of the transcription factor YY1 (Klymenko et al., 2006
). The role played by PHO-RC in development is a new avenue of research. It will be particularly interesting to determine whether Sfmbt2
regulates TS cell maintenance.
We observed that SFMBT2 and YY1 interact strongly in tissue culture cells transfected with expression constructs, suggesting that they do so in vivo. Support for this finding comes from expression studies indicating that the Sfmbt2
genes are co-expressed in early embryos and extraembryonic tissues (Donohoe et al., 1999
; Frankenberg et al., 2007
; this study). Furthermore, the phenotype of Yy1
null embryos is very similar to that of gynogenetic embryos, with lethality occurring just after implantation (Donohoe et al., 1999
null blastocysts form outgrowths bearing a striking resemblance to gynogenetic blastocyst outgrowths, with reduced proliferation of both inner cell mass and trophoblast cells (Donohoe et al., 2007
). A coherent working hypothesis posits that SFMBT2 and YY1 form the core of a mammalian PHO-RC silencing complex that is required in early mouse embryos to maintain the TS cell population. This hypothesis is currently under investigation.
The discovery that the PcG gene Sfmbt2 is paternally, and robustly, expressed in early embryos opens up an exciting new avenue of research in trophoblast biology and placentation. Investigation of the downstream targets of SFMBT2 regulation across a range of mammalian species will provide insight into trophoblast, a specialized cell type that is characterized in mice and humans by its ability to invade other tissues. In addition, the study of a new gene silencing complex in mammals will add to our growing understanding of epigenetic regulation of both normal and pathological processes.