Article activity feed

  1. Evaluation Summary:

    Mechanisms for controlling gene dosage and uniparental gene expression vary widely across the eukaryotic tree, with many such mechanisms still unknown. Montgomery et al. describe an epigenetic mechanism used to modulate paternal chromosome gene dosage during the transient diploid state of the primarily haploid plant, Marchantia polymorpha. This fascinating case of genome-wide genomic imprinting will be of broad interest to evolutionary biologists, epigeneticists, and those focused on understanding the context and mechanisms of gene dosage control.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  2. Reviewer #1 (Public Review):

    Montgomery and colleagues expand our understanding of gene dosage control by examining embryonic expression in a liverwort, the emerging model system Marchantia. Marchantia alternates between haploid and diploid phases, with the diploid, biparental embryo dependent on the haploid maternal parent. It has long been theorized that a form of genomic imprinting might exist in this context, but this has not been examined until now. By performing crosses between polymorphic parents and examining allele-specific gene expression, the authors show that transcription of the embryonic genome is primarily from the maternally-inherited alleles. Additionally, approximately half of the embryonic chromosomes are positive for H3K27me3 by immunofluorescence. By allele-specific CUT&RUN profiling, it is shown that H3K27me3 is biased towards paternally-inherited DNA. It appears that this difference is already present before the maternal and paternal pronuclei fuse. The authors take a genetic approach to determine whether H3K27me3 modifications are of consequence in the embryo. Disruption of E(z)2 and E(z)3 in the maternal parent leads to reduced H3K27me3 enrichment on paternal chromosomes and decreased maternal allele transcriptional bias. Ultimately the embryos are not viable. Taken together, the data support the idea that the maternal genome maintains widespread dominance over the paternal genome.

    1. For most experiments and analyses, embryos are the result of crosses between Cam-2 females and Tak-1 males. Since a reciprocal cross is not possible with these genotypes, the authors examine previously published data from Tak-2 females crossed to Tak-1 males. The analyses show that expression is strikingly biased towards the maternally-inherited DNA in both cases. The issue is that Tak-1 is the male in both sets of experiments. Thus, an alternative explanation is that the effect is specific to Tak-1 - that Tak-1 chromosomes are silenced in combination with either Cam-2 or Tak-2. This would be more akin to a phenomenon like nucleolar dominance - i.e. genotype-dependent rather than parent-of-origin-dependent.

    2. Some aspects of the data analysis need additional rigor. From the methods, it does not appear that any statistics were applied to determine whether genes were significantly biased away from the expected 1:1 maternal:paternal ratio. It is essential to do this - please refer to any mammalian or plant imprinting study.

    3. The authors show that e(z) mutant embryos grow more slowly and that most do not survive. They also show that mutants have reduced maternal allele bias. In terms of linking the phenotype to the gene expression change, it would be important to show that the total expression level of individual genes was altered in the mutants (for example, increased paternal allele expression might be compensated for by decreased maternal allele expression, in which case it would harder to connect the mutant phenotype to PCI). The authors should evaluate how many genes are differentially expressed between wild-type and mutant embryos.

    4. It's satisfying that when ez2 and ez3 are disrupted (Fig 5, Fig 5-fig supp 1D), that the IF for mutant embryos looks like H3K27me3 in vegetative nuclei (Fig 3A). But the paternal bias of H3K27me3 is still quite prevalent (Fig 5-fig supp1F) as is the maternal bias in transcription - the transcriptional ratio is not close to 50:50 (Fig 5B). The authors should comment or speculate on how paternal bias of H3K27me3 persists in this mutant, given their model. Perhaps the remaining H3K27me3 is from paternally supplied E(z). Since the paternal and maternal pronuclei are segregated for quite some time, a paternally supplied factor could also specifically mark one chromosome set (although it is less clear why this would be so from an evolutionary perspective). Generating paternal E(z) mutants would be interesting, but is likely beyond the current scope.

    5. From the genetic results, one can conclude that E(z)2 and E(z)3 are essential for viability and fecundity. But it is not yet clear, as claimed on line 339, that PCI is essential for viability and fecundity, as E(z)2 and E(z)3 may also have roles beyond or in addition to PCI. I suggest dividing this sentence into what one can conclude from the genetics, and what this suggests about the possible importance of PCI.

    6. Finally, paternal chromosome inactivation is perhaps too strong of a phrase to describe this very interesting phenomenon. There are thousands of genes for which expression is biased toward the maternal allele, but detectable paternal allele transcript is present in the embryos. It is important to get the name right now, because it may influence the field for a long time. For example, we now know that many genes on the "inactive X" are not inactive at all. But this phrase - X chromosome inactivation - continues to be the framing for much of the field, even though extensive caveats must be applied.

  3. Reviewer #2 (Public Review):

    The authors show that a strong maternal bias attributable to silencing and H3K27me3 enrichment of the paternal genome is a feature of bryophyte embryonic development. Paternal H3K27me3 enrichment is observed both by chromatin profiling and cytologically in the paternal chromosomes of the pronucleus and depends upon embryonic PRC2. The data that support the authors' conclusions are compelling, and I have only a few suggestions for improvement.

    1. At 3 days after fertilization both H3 and H3K27me3 are present, but since histones replaced protamines, the critical events of histone deposition and Polycomb marking might have occurred simultaneously or more likely successively. Can the authors distinguish these possibilities?

    2. Comparisons are made to X-chromosome inactivation in mammals and male X upregulation in Drosophila as silencing phenomena that are not conserved in evolution, but there are others that might be more relevant, such as paternal genome elimination in mealybugs, which is a spermatocyte-specific event that follows whole genome silencing during embryogenesis. Another is Meiotic Sex Chromosome Inactivation, a conserved phenomenon in animals that targets unpaired chromosomes.

    This is very nice work in a fascinating area.

  4. Reviewer #3 (Public Review):

    The manuscript by Montgomery et al. describes an interesting phenomenon in the Marchantia plant, where the entire paternal genome is silenced by Polycomb-mediated repression in the diploid embryo. This species spends most of its life cycle as haploid, yet has a diploid stage during embryogenesis. By analyzing the transcriptome of embryos derived from genetically distinct strains of parents, the authors show that transcription is heavily maternally biased in such embryos (at least for genes, for which distinct SNPs could be used), suggesting that paternal genes are widely silenced. The authors further demonstrate via CUT&RUN that enrichment of the Polycomb histone mark H3K27Me3 is biased towards paternal alleles. Interestingly, the authors also find that nuclei of embryos of this early stage display DAPI-bright, condensed foci that are specifically enriched for the H3K27Me3 signal, suggesting that these represent paternal chromosomes. Further immunofluorescence characterization revealed a strong H3K27Me3 signal specifically in the male pronucleus at the 3 daf stage, before the genomes fuse, presenting a possible time point and mechanism for paternal-specific deposition of Polycomb marks.

    Importantly, the authors show that when embryonic-specific E(z) homologues are knocked out, the large nuclear foci disperse. CUT&RUN and RNA-seq analysis on the mutant embryos further demonstrate the H3K27Me3 mark is widely reduced (although not abolished), especially on paternal alleles, and that expression becomes less maternally biased and more biallelic overall. The spores produced from such mutants are inviable. Together, these results support the authors' model that PRC2-mediated chromatin modification silences the paternal genome, revealing genomic imprinting on a global scale as a necessary part of the embryonic development of this species.

    Overall, this is an interesting and newly described (at least to my knowledge) phenomenon of genome-wide and post-fertilization genomic imprinting, which expands our knowledge of gene dosage mechanisms. Although the involvement of Polycomb is not particularly surprising, such global paternal silencing itself seems to be. Therefore, this represents an additional system, where mechanisms of gene dosage, Polycomb repression, nuclear organization, and genomic imprinting can be investigated in a unique context. The data appears to be of high quality and generally supports the conclusions made by the authors.