RNA Pol IV induces antagonistic parent-of-origin effects on Arabidopsis endosperm

  1. Prasad R. V. Satyaki
  2. Mary Gehring

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Gene expression in endosperm—a seed tissue that mediates transfer of maternal resources to offspring—is under complex epigenetic control. We show here that plant-specific RNA polymerase IV (Pol IV) mediates parental control of endosperm gene expression. Pol IV is required for the production of small interfering RNAs that typically direct DNA methylation. We compared small RNAs (sRNAs), DNA methylation, and mRNAs in Arabidopsis thaliana endosperm from heterozygotes produced by reciprocally crossing wild-type (WT) plants to Pol IV mutants. We find that maternally and paternally acting Pol IV induce distinct effects on endosperm. Loss of maternal or paternal Pol IV impacts sRNAs and DNA methylation at different genomic sites. Strikingly, maternally and paternally acting Pol IV have antagonistic impacts on gene expression at some loci, divergently promoting or repressing endosperm gene expression. Antagonistic parent-of-origin effects have only rarely been described and are consistent with a gene regulatory system evolving under parental conflict.

Article activity feed

  1. Version published to 10.1371/journal.pbio.3001602
  2. Review Commons

    Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Response to reviewers

    Reviewer #1

    I believe that this is a very sound and authoritative study. The analysis of all data seems appropriate and robust, and many connections between the data (and subsets of data) and their possible interpretations have been considered. In fact, in the massive Results section, some interpretations are supported by cited references (this is not meant as a critique). However, I wonder about the length of the Results section, and the balance between it and the relatively short Discussion section. It is difficult for me to nail down any part of Results that might be shortened, as I could not find clear redundancies. I also think that the level of speculation is absolutely warranted, and I did not find excessive claims being made to this or that end. Rather, I suggest to broaden the perspective somewhat (in their Discussion; see below under Significance), which might allow people with a less mechanistic perspective to grasp the potential relevance of this work for non-model plant systems studied mostly by evolutionary geneticists.

    Response: We thank the reviewer for their kind remarks. We have spent a very large amount of time trying to streamline the results section and we are not sure if it would be possible to shorten it any further without removing critical details.

    We appreciate the reviewer’s comment to add more detail to the discussion to make it more appealing to evolutionary geneticists and we have added the following lines to the discussion section: “The WISO or “weak inbreeder/strong outcrosser” model (Brandvain & Haig, 2005) emerges from the dynamics of parental conflict and parent-of-origin effects. Under this model, a parent from populations with higher levels of outcrossing is exposed to higher levels of conflict and can thus dominate the programming of maternal resource allocation in a cross with an individual from a population with lower levels of outcrossing. Such a phenomenon has been observed in numerous clades including Dalechampia, Arabidopsis, Capsella and Leavenworthia (Brandvain & Haig, 2018; İltaş et al., 2021; Lafon-Placette et al., 2018; Raunsgard et al., 2018). Intriguingly, loss of function phenotypes in the RdDM pathway are more severe in recently outcrossing species than in *A.thaliana *(Grover et al., 2018; Wang et al., 2020) and suggests that RNA Pol IV functions are more elaborate and important in these species. This raises the possibility that the role for RNA Pol IV and RdDM in parental conflict that we describe in A.thaliana here is likely heightened in and mediates the elevated level of parental conflict in species that are currently or have been recently outcrossing.”

    One aspect that might warrant more scrutiny is the mapping of sRNA reads to the reference genome. I found the short section of this (M&M section, page 20, lines 23-25) to be too brief. It is not clear to me which of ShortStack's v3 weighting scheme the authors used, which is relevant for multi-mapping reads (see NR Johnson et al. 2016, G3). In addition, it is not mentioned whether zero mismatches were allowed. Perhaps this is described in more detail in Erdmann et al. (2017), but even if so, it deserves to be clarified here.

    Response: Small RNA reads were aligned after allowing two mismatches. This was indicated in the bowtie command (‘*bowtie -v 2’ *where v 2 indicates two mis-matches). We have added text to expand on the meaning of the commands.

    We have also expanded the commands used for ShortStack. We used the “Placement guided by uniquely mapping reads (-u)” option to divide the multi-mapping reads.

    The manuscript is well-written and concise, despite the length of the Results section. The verbal clarity and absence of typos or grammatical issues is superb. I did find some of the Figures to be somewhat "un-intuitive", in the sense that it takes acute concentration for an outsider (of sorts) to gather and interpret the underlying data. This is probably due to the many cross-comparisons of differences between two genotypes on one axis and those of a different pair of genotypes on the other axis. I am not sure how this issue can be ameliorated (nor whether this is really necessary); however, from a technical point of view, all Figures and Suppl. Figures are really well-done.

    Response: We thank the reviewer for their kind remarks. We have strived to make the figures easier to understand but we are aware that the figures do require a lot of concentration. We haven’t found an easy way to fix this. We thank the reviewer for patiently going through the figures.

    The list of references seems adequate in terms of citing relevant (both older and very recent) publications. However, almost all cited papers concern Arabidopsis or other model species; I suggest to consider adding a few relevant studies on non-Brassicaceae (whether considered model taxa or not), in conjunction with my suggestion (in Significance) to potentially broaden the scope by searching for natural phenomena that also involve parent-of-origin effects on endosperm/seed development. Curiously, many of the references are "incomplete" in the sense of stopping with the journal's name, then stating the doi, i.e. they lack volume numbers and page/article numbers. This should be harmonized throughout.

    Response: We have added references to non-Brassicaceae species and have also fixed the references.

    Reviewer #2:

    This manuscript provides evidence that a loss of either the maternal or paternal copy of NRPD1 have different, and sometime opposite, effects on the accumulation of small RNAs and on expression of a subset of genes, with a loss of the maternal copy having more substantial effects. The manuscript is well written, and the conclusions, as far as they go, are justified by the data, which are effectively presented. The overall effect is subtle but informative and according to the authors support a parental conflict model for imprinting. The experiments failed to find a smoking gun in the form of a mechanism to explain how or why the maternal and paternal alleles have different effects and the explanation for a lack of clear phenotypic differences was reasonable, but untested. I would have like to see it tested by looking in a plant species that is outcrossing and highly polymorphic. However, I do appreciate that the observation that the male and female alleles can have distinct effect when mutant is an important clue. My specific comments below may reflect confusion on my part, rather than real issues. If that is that I hope that confusion can aid in clarifying what are in places quite subtle points.

    Response: We thank the reviewer for their comments. We agree that it would be potentially informative to do similar experiments in an outcrossing species but that this is beyond the scope of this manuscript. Additionally, loss of NRPD1 or other components of the RdDM pathway has dramatic effects on gametogenesis in some examined outcrossing species(Grover et al., 2018; Wang et al., 2020), which could prevent the detection of subtle parent-of-origin effects on seed development.

    Page 6, last paragraph: "Because the endosperm is triploid, in these comparisons there are 3 (wild-type), 2 (pat nrpd1+/-), 1 (mat nrpd1+/-) and 0 (nrpd1-/-) functional NRPD1 alleles in the endosperm. However, NRPD1 is a paternally expressed imprinted gene in wild-type Ler x Col endosperm and the single paternal allele contributes 62% of the NRPD1 transcript whereas 38% comes from the two maternal alleles (Pignatta et al., 2014). Consistent with paternal allele bias in NPRD1 expression, mRNA-Seq data shows that NRPD1 is expressed at 42% of wild- type levels in pat nrpd1+/- and at 91% of wild-type levels in mat nrpd1+/- (Supplementary Table 6)".

    I would think this would really complicate the analysis. Should all of the dosage values include NPRD1 imprinting values? That is to say, expressed in terms of expression values? This is also a bit confusing. The maternal copies together express 38% of the transcript, so why isn't the mat nrpd1 at 68%, rather than 91%? In any event, given this imprinting and differences in dosage of the male and female it appears that two variables, parental origin and expression levels are being compared. Since 91% is awfully close to 100%, are the mat pat comparisons really just comparing low with nearly normal expression of NRPD1? And actually, given that, the outsized effect of the mat nrpd1 +/- is even more striking.

    Response: We included the details of dosage rather than imprinting values because the potential for buffering of expression upon loss of one allele could not be discounted. Indeed, we do find that the endosperm transcriptome buffers against the loss of the maternal or paternal alleles (Supplementary Table 6). The reviewer is correct in pointing out that the outsized effect of mat nrpd1+/- on gene expression is even more striking, and strongly supports our view that these effects are parental rather than endospermic.

    To reduce confusion in this section, we removed the details about 38% maternal allele transcripts obtained from our previous study, and instead report only the observed values from this study (which are also consistent with the previously reported paternally-biased expression of NRPD1 in endosperm).

    Page 4, Line 16. I'm afraid it's still a bit difficult to understand what was being compared what in this section. Please clarify.

    __Response: __The authors in this previously published study compared sRNAs obtained from wild-type whole seeds (which consists of three different tissues, including endosperm) with mutant endosperm. We are pointing out that the difference in tissue composition makes the effect of nrpd1 mutation hard to disentangle from the tissue differences between the two genotypes.

    Page 5, Line 5. I'm sure this is fine, but it's not entirely clear what is from the previously published paper and what is reanalysis here. All the crosses and measurements were made then, but not organized in this way?

    Response: This data was indeed previously published. In that analysis, we had pooled results from different crosses and calculated significance between genotypes using chi-square tests. During a later study (Satyaki and Gehring, 2019), we realized that we were losing information by ignoring the seed abortion values per cross. So, a reanalysis of that data on a cross by cross basis allowed us to find strong evidence for maternal and paternal effects.

    Page 6, Line 26. This is an excellent dosage series, but it's complicated by imprinting. So it's not 3, 2, 1, 0 effective copies. If we set the paternal copy at ~1 and each maternal at ~0.1, then it's 1.2 (wild type), 0.20 (pat nrpd1+/-), 1 (mat nrpd1+/-), and 0 (nrpd1-/-).

    Response: At the genomic DNA level, its 3, 2,1 and 0 doses. The reviewer’s comment on the transcriptional dose is not clear to us. Based on measured gene expression levels, relative wild-type *NRPD1 *transcriptional dose =1, pat nrpd1+/- is 0.42, and mat *nrpd1+/- *is 0.91.

    Page 6, line 31. Is the main thing we are comparing the difference between expression at 42% verses 91% of wild type?

    Response: We are using the small RNA-seq data alongside the mRNA-seq data to argue that loss of mat and pat nrpd1+/- have no impact on overall Pol IV activity in endosperm (as measured by small RNA production). A *nrpd1 *heterozygous endosperm has almost the same small RNA profile as a wild-type endosperm. Thus any effects seen in the endosperm, including the effects on mRNA expression described later in the manuscript, are likely parental rather than zygotic endospermic effects.

    Page 7, line 11. So, the overall effect in either direction on smRNA gene targets was really quite small, and I'm guessing the effect on gene expression was even smaller.

    Response: The effects of loss of maternal or paternal Pol IV on sRNAs was indeed small (Fig. 1/Fig. S3). Effect of loss of maternal Pol IV on gene expression was substantially large and distinct from the relatively small impacts observed upon loss of paternal Pol IV (Fig. 3) This observation supports the view that Pol IV mediates parent-of-origin effects on gene expression.

    Page 7, line 17. I take it that it is this difference, rather than the overall numbers that is of interest.

    __Response: __Correct. The lack of a relationship between sRNAs impacted upon loss of mat and pat nrpd1 is additionally suggestive of parent-of-origin effects

    Page 9, line 2. Really interesting, since one might expect that these are methylated loci that would be expected to be fed into any existing embryo maintenance methylation pathway. Surprising that they are maintained independently.

    Response: It is indeed surprising that Pol IV activity in parents can have different impacts on sRNAs in the endosperm. It should be noted though, that as described in Erdmann et al 2017 and in this paper later on, many endosperm sRNA loci are in fact not associated with endosperm DNA methylation. In addition, sRNA loci that are dependent on paternal Pol IV activity are more likely to be associated with DNA methylation than are sRNA loci associated with maternal Pol IV activity. These points have been described in Figure S8.

    Page 9, line 22. Proportion of total imprinted genes? Did the mutant obviate/enhance the imprinting?

    Response: We have modified the manuscript to describe effects on imprinted genes: “ The expression of imprinted genes is known to be regulated epigenetically in endosperm. In mat nrpd1+/- imprinted genes were more likely to be mis-regulated than expected by chance (hypergeometric test p-15) – 15 out of 43 paternally expressed and 45 out of 128 maternally expressed imprinted genes were mis-regulated in mat nrpd1+/- while two maternally expressed imprinted genes but no paternally expressed imprinted genes were mis-regulated in pat *nrpd1+/- *(Table S6).” We have also added a new supplementary figure (Fig. S6) that describes the impacts of NRPD1 loss of imprinted gene expression.

    Page 9, line 27. How could 2) occur in the homozygous mutant?

    Response: Loss of *NRPD1 *may impact gene expression in both parents. When the nrpd1-/- mutant endosperm is investigated, we are also examining the consequences of the inheritance of these disrupted gene expression states. We refer to this as epistatic interactions of mat and pat nrpd1.

    Page 10, line 9. Interesting!

    Response: We strongly agree!

    Page 10, line 11. Is this 2.7 versus 2.18 significant because it's statistically significant, or because it's conceptually significant?

    Response: We are pointing out that the 2.7-fold value is quite similar to the predicted value of 2.18-fold, which is arrived at by simply summing the effects of mat *nrpd1 *and pat nrpd1. This is a conceptually significant point.

    Are the examples in 3D representative, or the most convincing examples? And a big difference in ROS1 is of some concern, since that may well be expected to affect imprinting indirectly. I know I'm being picky here, but the pattern is so intriguing I'd be worried about confirmation bias.

    Response: The examples in 3D are representative for those genes with significant changes in expression in both mat and pat nrpd1, and other genes also behave similarly. The antagonistic effect described for 3D can also be observed as a much broader trend affecting hundreds of genes to varying extents in Fig 3C and 3E-H. The concern about ROS1 is not clear to us but we agree that an effect of ROS1 may be one way that NRPD1 controls gene expression.

    Page 10, line 18. Ok, but 0.123 is a pretty subtle negative correlation. Although I do appreciate that it clearly is not a positive correlation. If I'm understanding correctly, this was the "aha" moment, because it's exactly what one might expect of NRPD1 from the mother and father or working at cross purposes. But the numbers are getting awfully small here.

    Response: It is unclear how to calibrate our expectations of effect sizes considering that our study is the first (to our knowledge) to make such a measurement involving gene expression in parental conflict. A review of the few empirical examples of parental conflict’s impact on seeds shows that parental conflict may drive small changes in seed size (Brandvain and Haig, 2018).

    The evolution of quantitative traits maybe driven by selection for large effects at a small number of loci and/or by selection of small effects at a large number of loci. In a similar vein, parental conflict can impact seed phenotypes either via large effects at a few loci or via small effects at a large number of loci. Our analysis described in Fig 3D-H can fit either possibility. Large effects can be found at a few loci such as SUC2 and *PICC *(Fig. 3D). Smaller antagonistic effects can be found at hundreds of loci as shown in Figure 5A. The negative correlation described in this figure can be observed even upon dropping the genes that show a statistically significant differential expression in both mat and pat *nrpd1+/- *(slope after dropping genes significantly mis-regulated in both mat and pat nrpd1+/- * is -0.126). *In summary, a correlation of -0.123 strongly supports the existence of a widespread antagonistic regulatory effect.

    Page 10, line 29. The point simply being that that other phenomenon is also significant even if the differences are that large?

    Response: We are pointing out that the magnitude of the effects we see are similar to that observed for phenomenon such as dosage compensation.

    Page 12. So, there is no effect on cleavage and no obvious effect on flanking siRNA clusters. The suspense is building...

    Page 12, line 24. And not in potential regulatory regions? CNSs?

    Response: We did not identify a significant enrichment for differentially methylated regions in regulatory regions. We used the relative distance function in bedtools (https://bedtools.readthedocs.io/en/latest/content/tools/reldist.html) to calculate the relationship between the genomic location of DMRs and genomic location of a differentially expressed gene. This analysis was chosen as it does not make a priori assumptions about the size of the regulatory region of a gene. A broad association between DMRs and differentially expressed genes would be indicated by a frequency far greater than 0.02. We show the results of this analysis in Fig. S8F; we find no evidence for significant enrichment of DMRs in the regulatory regions of differentially expressed genes.

    Page 12, line 28. I guess it depend on whether or not the changes are in regulatory sequences no immediately apparent as part of the gene, doesn't it?

    Response: We examined DNA methylation over genes here because in endosperm, unlike in other tissues, many small RNAs are genic. Moreover, DNA methylation within the gene may control transcript abundance (Eimer et al., 2018; Klosinska et al., 2016). We have also examined regulatory regions adjacent to genes in Fig S8F and found no effect.

    Line 13, line 2. Not sure it's that important, but couldn't you chop all of these genes in half and see if they are no longer significant collectively?

    Response: We do not think that this will provide a useful insight.

    Page 14, line 15. I'm afraid I'm getting confused here with the terms cis and trans here. Just to be clear, cis means a direct effect of small RNAs that are dependent on NRPD1 on a gene and trans means anything else? But in this context, it's not clear that is what is meant. Do you mean that gene expression is determined and preset in the gametophyte? What are the levels of expression of NRPD1 in the two gametophytes?

    Response: The reviewer’s interpretation of cis and trans is correct. However, the cis imprints may be preset in gametophytes or in the sporophytic tissues that surround or give rise to the gametophyte. Pol IV is known to be active either in gametophyte or in related sporophytic tissues in both the mother and the father(Kirkbride et al., 2019; Long et al., 2021; Olmedo-Monfil et al., 2010).

    Page 14, line 19. Prior to fertilization?

    Response: Yes, that is the idea. As described in the manuscript, Pol IV activity in either the parental sporophyte or gametophyte prior to fertilization could impact gene expression in the endosperm after fertilization.

    Page 14, line 27. Do you mean driven by, or just associated with?

    Response: In response to the comment, we have replaced the phrase “driven by” with “due to” for increased clarity. In wild-type, DOG1 is predominantly expressed from the paternal allele. In mat nrpd1+/-, the paternal allele is somewhat upregulated but the maternal allele, which is almost silent in wild-type, is highly expressed in mat nrpd1+/-.

    Page 15, line 26. And this is really the issue. The primary conclusion, backed up by the lack (I'm assuming) of phenotypic differences between mat NRPD1 -/+ and pat NRPD1 +/- suggests that the observed differences in expression are not particularly important, unless the exceptional cases are informative.

    Response: We are not sure whether the reviewer means “issue” in a negative, neutral, or positive light. Seed phenotypes are often subtle and we have not examined phenotypic differences in sufficient detail to comment.

    Page 15, line 12. Yes, but I'm not at all clear what the mechanism for this is.

    Response: We have tested and falsified multiple hypotheses to explain how Pol IV can regulate gene expression in endosperm. Considering the complex genetics and the difficulty of isolating endosperm, we have concluded that this is a matter for a future study. The point of this study is the discovery of Pol IV’s parental effects.

    Page 15, line 23. Since this is a very small subset of genes, are these genes that you might expect to play a role in parental conflict?

    Response: The functions of most genes in endosperm remain unknown. However, some have a likely role in conflict. *SUC2 *is antagonistically regulated by parental Pol IV (Fig. 3D). SUC2 transports sucrose, the key form of carbon imported into seeds from the mother (Sauer & Stolz, 1994).

    Page 15, line 33. Indeed, these could be the informative exceptions.

    Response: We believe the reviewer means that the identify of strongly antagonistically regulated genes may be informative in terms of thinking about these results in the context of parental genetic conflict, which we agree with.

    Page 15, line 29. Hardly surprising, given that the paternal copy of NRPD1 is expressed at a higher level than the maternal copies, is it?

    Response: It is actually somewhat surprising since we show in Fig. 2 that the sRNA production in mat and pat nrpd1+/- are comparable to that of wild-type. The higher contribution of NRPD1 from the paternal copy does not really explain the methylation differences

    Page 16, line 1. So this is what you mean by in cis. Presetting?

    Response: The reviewer’s previous interpretation of cis (acting directly at a target gene) is correct. However, the cis imprints may be preset in gametophyte or in the sporophytic tissues that surround or give rise to the gametophyte. Pol IV is known to be active in gametophytes and in related sporophytic tissues in both the mother and the father.

    These are intriguing results that would benefit from a test of the hypothesis by comparing these result with those obtained in an outcrossing plant species.

    Response: We agree that it would interesting and informative to perform similar experiments in an outcrossing species. However, loss of NRPD1 or other components of the the RdDM pathway have dramatic effects on gametogenesis in outcrossing species (Grover et al., 2018; Wang et al., 2020), preventing the detection of subtle parent-of-origin effects on seed development. Additionally, this would be a separate study.

    Reviewer #3

    We thank the reviewer for their comments.

    • Expression of NRPD1 was 42% of WT in paternal nrpd1 and 91% of WT in maternal nrpd1, yet throughout the paper the effect of maternal nrpd1 was far stronger than paternal nrpd1. The authors may also want to confirm that protein levels follow the same pattern, in case protein degradation or post-transcriptional regulation may play a role.

    Response: We show in Fig. 2 that sRNA production in mat and pat nrpd1+/- are similar to wild-type endosperm. This strongly suggests that *NRPD1 *protein is produced at functionally equivalent levels in wild-type, mat and pat *nrpd1+/-. *The finding that mat nrpd1+/- has a stronger effect on gene expression and small RNAs, despite having higher levels of *NRPD1 *transcript in endosperm, is consistent with our conclusion that the effects we are observing in heterozygous endosperm are due to *NRPD1 *action before fertilization.

    P. 9 line 1 - this only seems to be true for maternal ISRs, not paternal ISRs; this claim should be narrowed.

    Response: Accordingly, we have modified the text here to : “In summary, these results indicate that most maternally and some paternally imprinted sRNA loci in endosperm are dependent on Pol IV activity in the parents and are not established de novo post-fertilization.”

    A small number of sRNA loci become highly depleted in maternal nrpd1 but not paternal nrpd1 (Fig. 1D, F, Fig. 2C) - are these siren loci?

    Response: This is an interesting question. Siren loci have not been defined in Arabidopsis but are described as loci with high levels of sRNAs in ovules, seed coat, endosperm and embryo (Grover et al., 2020). Loci losing sRNAs in maternal nrpd1+/- include a large number of maternally expressed imprinted sRNAs (mat ISRs). We do not know if mat ISR loci are expressed in the ovule. In Erdmann et al (2017), we excluded loci that were also expressed in the seed coat from mat ISRs. Thus, these loci meet only some of the conditions for being defined as siren loci.

    Fig. 2 suggests that many of the downregulated sRNA regions in maternal nrpd1 are maternally biased to begin with. Related, are genic sRNAs more likely to be affected by maternal or paternal nrpd1 than non-genic or TE sRNAs?

    Response: As described in Fig. 1B and S3, loss of maternal NRPD1 has more impacts on the sRNA landscape. As a percentage of total loci, genes are more likely to be affected than TEs.

    For the sRNA loci shown in Fig. 2C, how is % maternal affected in maternal vs. paternal nrpd1? These ISRs are normally maternal or paternal biased, does this change in maternal or paternal nrpd1?

    Response: We assess the allelic bias of ISRs only when they have at least ten reads in the genotypes being compared. In mat nrpd1+/-, most mat ISRs lose almost all their reads (Fig. 2) and we can assess allelic bias only at 107/366 mat ISRs. As seen in the Rev. comment. Fig1, these 107 lose their maternal bias. In pat *nrpd1+/-, *loci with maternally biased sRNAs show somewhat increased expression (Fig 2E) but do not show an appreciable change in maternal bias (Figure Review 1). All paternal ISRs do not show any dramatic impacts on allelic bias in mat or pat *nrpd1+/-. *We have not added this additional datapoint to our paper because we were worried that the paper was becoming too dense – a concern also voiced by reviewer 1. However, we can add this to the manuscript if the reviewer prefers.

    • Might have missed this, but I didn't see the gene ontology results (p9 line 16) shown anywhere? Would like to see significance values, fold enrichments, etc. In particular, the group of paternal nrpd1 up-regulated genes seems too small to have much confidence for GO enrichment analysis.

    Response: We have added a Supplementary Table 7 with outputs of GO analyses.

    • I would suggest expanding the analysis in Fig. 3D-H to explore whether the additive model is more predictive of nrpd1-/- expression levels than other potential models (epistatic, etc.) in general at all genes, or only at the subsets of genes shown, independently of whether the effects are large enough to pass the arbitrary significance cutoffs used in E-H. Identifying specifically which genes do and don't follow this additive pattern could help dissect mechanism. For example, genes following this pattern might share a TF binding site for a TF that is regulated by Pol IV.

    Response: While we are interested, we currently cannot explore other models such as epistasis as this would require knock-down of NRPD1 in the endosperm and we plan to do this as part of a future study.

    13 line 26 - how do changes in CG methylation in maternal or paternal nrpd1 compare to changes in dme or ros1? Do either set of DMRs significantly overlap dme or ros1 DMRs? Could some of these be explained by changes in ROS1 expression, since ROS1 is a Pol IV target?

    Response: Yes. It’s entirely possible that a subset of observed gene expression changes are linked to changes in ROS1 expression. However, there are no comparable methylation data for ROS1 in the endosperm. A potential role for ROS1 has been discussed on Page 11, line 4. Comparison with DMRs in the *dme *endosperm is difficult. dme mutant endosperm has low non-CG methylation (Ibarra et al., 2012). We have unpublished data showing that the expression of genes involved in RNA-directed DNA methylation (RdDM) is reduced in the dme endosperm. It is therefore difficult to understand if and how DME-mediated demethylation may impact RdDM.

    P. 10 line 3 - is the overlap of 36 out of 51 genes unlikely to occur by chance

    Response: A hypergeometric test indicates that this is indeed significant. We have added it to text on Page 9, line 34.

    In sRNA and mRNA-seq libraries, what was the overall maternal/paternal ratio in each library? Did loss of Pol IV affect this?

    The graphs above show the maternally derived fraction of mRNA and sRNA libraries for different genotypes. Please note that the Ler nrpd1 mutant was generated by backcrossing Col-0 nrpd1+/- into Ler. Some Col-0 regions remain in this background and are called “hold-outs”. Reads mapping to these hold-outs have been excluded while calculating the maternal fraction of each library described in the graph above. We cannot confidently judge if the overall maternal fraction of the mRNA transcriptome is affected by loss of NRPD1 as we likely need more replicates. However, we find that loss of all NRPD1-dependent sRNAs (as in the *nrpd1 *null mutant) leaves behind sRNAs that roughly reflect the genomic 2:1 ratio.

    P. 9 line 22 - how many paternally and maternally expressed imprinted genes were considered? Were imprinted genes statistically more likely to be misregulated in mat nrpd1?

    Response: We considered 128 maternally and 43 paternally expressed genes that had been previously been identified as imprinted in Col x Ler crosses (Pignatta et al 2014). We have modified the manuscript to describe effects on imprinted genes: “ The expression of imprinted genes is known to be regulated epigenetically in endosperm. In mat nrpd1+/- imprinted genes were more likely to be mis-regulated than expected by chance (hypergeometric test p-15) – 15 out of 43 paternally expressed and 45 out of 128 maternally expressed imprinted genes were mis-regulated in mat nrpd1+/- while two maternally expressed imprinted genes but no paternally expressed imprinted genes were mis-regulated in pat *nrpd1+/- *(Table S6). “ We have also added a supplementary figure (Figure S6) that focuses on genic mRNA imprinting in NRPD1 heterozygotes and homozygous mutants.

    References cited in the response

    Brandvain, Y., & Haig, D. (2005). Divergent Mating Systems and Parental Conflict as a Barrier to Hybridization in Flowering Plants. The American Naturalist, 166(3), 330–338. https://doi.org/10.1086/432036

    Brandvain, Y., & Haig, D. (2018). Outbreeders pull harder in a parental tug-of-war. Proceedings of the National Academy of Sciences, 115(45), 11354–11356. https://doi.org/10.1073/pnas.1816187115

    Eimer, H., Sureshkumar, S., Singh Yadav, A., Kraupner-Taylor, C., Bandaranayake, C., Seleznev, A., Thomason, T., Fletcher, S. J., Gordon, S. F., Carroll, B. J., & Balasubramanian, S. (2018). RNA-Dependent Epigenetic Silencing Directs Transcriptional Downregulation Caused by Intronic Repeat Expansions. Cell. https://doi.org/10.1016/j.cell.2018.06.044

    Grover, J. W., Burgess, D., Kendall, T., Baten, A., Pokhrel, S., King, G. J., Meyers, B. C., Freeling, M., & Mosher, R. A. (2020). Abundant expression of maternal siRNAs is a conserved feature of seed development. Proceedings of the National Academy of Sciences of the United States of America, 117(26), 15305–15315. https://doi.org/10.1073/pnas.2001332117

    Grover, J. W., Kendall, T., Baten, A., Burgess, D., Freeling, M., King, G. J., & Mosher, R. A. (2018). Maternal components of RNA ‐directed DNA methylation are required for seed development in Brassica rapa. The Plant Journal, 94(4), 575–582. https://doi.org/10.1111/tpj.13910

    Ibarra, C. A., Feng, X., Schoft, V. K., Hsieh, T.-F., Uzawa, R., Rodrigues, J. A., Zemach, A., Chumak, N., Machlicova, A., Nishimura, T., Rojas, D., Fischer, R. L., Tamaru, H., & Zilberman, D. (2012). Active DNA Demethylation in Plant Companion Cells Reinforces Transposon Methylation in Gametes. Science, 337(6100), 1360–1364. https://doi.org/10.1126/science.1224839

    İltaş, Ö., Svitok, M., Cornille, A., Schmickl, R., & Lafon Placette, C. (2021). Early evolution of reproductive isolation: A case of weak inbreeder/strong outbreeder leads to an intraspecific hybridization barrier in Arabidopsis lyrata. Evolution, 75(6), 1466–1476. https://doi.org/10.1111/evo.14240

    Kirkbride, R. C., Lu, J., Zhang, C., Mosher, R. A., Baulcombe, D. C., & Chen, Z. J. (2019). Maternal small RNAs mediate spatial-temporal regulation of gene expression, imprinting, and seed development in Arabidopsis. Proceedings of the National Academy of Sciences, 116(7), 2761–2766. https://doi.org/10.1073/pnas.1807621116

    Klosinska, M., Picard, C. L., & Gehring, M. (2016). Conserved imprinting associated with unique epigenetic signatures in the Arabidopsis genus. Nature Plants, 2, 16145. https://doi.org/10.1038/nplants.2016.145

    Lafon-Placette, C., Hatorangan, M. R., Steige, K. A., Cornille, A., Lascoux, M., Slotte, T., & Köhler, C. (2018). Paternally expressed imprinted genes associate with hybridization barriers in Capsella. Nature Plants, 4(6), 352–357. https://doi.org/10.1038/s41477-018-0161-6

    Long, J., Walker, J., She, W., Aldridge, B., Gao, H., Deans, S., Vickers, M., & Feng, X. (2021). Nurse cell­–derived small RNAs define paternal epigenetic inheritance in Arabidopsis. Science, 373(6550). https://doi.org/10.1126/science.abh0556

    Olmedo-Monfil, V., Durán-Figueroa, N., Arteaga-Vázquez, M., Demesa-Arévalo, E., Autran, D., Grimanelli, D., Slotkin, R. K., Martienssen, R. A., & Vielle-Calzada, J.-P. (2010). Control of female gamete formation by a small RNA pathway in Arabidopsis. Nature, 464(7288), 628–632. https://doi.org/10.1038/nature08828

    Raunsgard, A., Opedal, Ø. H., Ekrem, R. K., Wright, J., Bolstad, G. H., Armbruster, W. S., & Pélabon, C. (2018). Intersexual conflict over seed size is stronger in more outcrossed populations of a mixed-mating plant. Proceedings of the National Academy of Sciences, 115(45), 11561–11566. https://doi.org/10.1073/pnas.1810979115

    Sauer, N., & Stolz, J. (1994). SUC1 and SUC2: two sucrose transporters from Arabidopsis thaliana; expression and characterization in baker’s yeast and identification of the histidine-tagged protein. The Plant Journal, 6(1), 67–77. https://doi.org/10.1046/j.1365-313X.1994.6010067.x

    Wang, Z., Butel, N., Santos-González, J., Borges, F., Yi, J., Martienssen, R. A., Martinez, G., & Köhler, C. (2020). Polymerase IV Plays a Crucial Role in Pollen Development in Capsella. The Plant Cell, 32(4), 950–966. https://doi.org/10.1105/tpc.19.00938

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Short Summary:

    In this study, Satyaki and Gehring investigate the role of RNA Pol IV in Arabidopsis endosperm, focusing on parent-of-origin-specific functions and potential mechanisms. Using a combination of gene expression, sRNA profiling, and DNA methylation data from reciprocal crosses, they find that maternal loss of Pol IV has distinct, and in some cases opposite, effects on gene expression compared to paternal loss of Pol IV. This is also true to a lesser extent for sRNAs and DNA methylation, consistent with the function of RNA Pol IV in driving 24nt sRNA production and targeted DNA methylation through the RdDM pathway. DNA methylation was more strongly affected by paternal Pol IV loss while expression was much more affected in maternal Pol IV loss. Surprisingly, the authors consistently find no evidence that the minor changes sRNA production or DNA methylation in maternal/paternal nrpd1/+ heterozygotes are correlated with gene expression changes in either heterozygote. However, while the mechanism remains unclear, evidence presented here that maternal and paternal Pol IV can have opposite, additive effects on gene/sRNA expression and phenotype (rescue of paternal excess crosses) is convincing and an interesting finding, potentially consistent with the idea that Pol IV helps mediate parental conflict in endosperm.

    Major Comments/suggestions:

    • Expression of NRPD1 was 42% of WT in paternal nrpd1 and 91% of WT in maternal nrpd1, yet throughout the paper the effect of maternal nrpd1 was far stronger than paternal nrpd1. The authors may also want to confirm that protein levels follow the same pattern, in case protein degradation or post-transcriptional regulation may play a role.
    • P. 9 line 1 - this only seems to be true for maternal ISRs, not paternal ISRs; this claim should be narrowed.
    • A small number of sRNA loci become highly depleted in maternal nrpd1 but not paternal nrpd1 (Fig. 1D, F, Fig. 2C) - are these siren loci? Fig. 2 suggests that many of the downregulated sRNA regions in maternal nrpd1 are maternally biased to begin with. Related, are genic sRNAs more likely to be affected by maternal or paternal nrpd1 than non-genic or TE sRNAs?
    • For the sRNA loci shown in Fig. 2C, how is % maternal affected in maternal vs. paternal nrpd1? These ISRs are normally maternal or paternal biased, does this change in maternal or paternal nrpd1?
    • Might have missed this, but I didn't see the gene ontology results (p9 line 16) shown anywhere? Would like to see significance values, fold enrichments, etc. In particular, the group of paternal nrpd1 up-regulated genes seems too small to have much confidence for GO enrichment analysis.
    • I would suggest expanding the analysis in Fig. 3D-H to explore whether the additive model is more predictive of nrpd1-/- expression levels than other potential models (epistatic, etc.) in general at all genes, or only at the subsets of genes shown, independently of whether the effects are large enough to pass the arbitrary significance cutoffs used in E-H. Identifying specifically which genes do and don't follow this additive pattern could help dissect mechanism. For example, genes following this pattern might share a TF binding site for a TF that is regulated by Pol IV.
    • P. 13 line 26 - how do changes in CG methylation in maternal or paternal nrpd1 compare to changes in dme or ros1? Do either set of DMRs significantly overlap dme or ros1 DMRs? Could some of these be explained by changes in ROS1 expression, since ROS1 is a Pol IV target?
    • P. 10 line 3 - is the overlap of 36 out of 51 genes unlikely to occur by chance?

    Minor Comments:

    • In sRNA and mRNA-seq libraries, what was the overall maternal/paternal ratio in each library? Did loss of Pol IV affect this?
    • P. 9 line 22 - how many paternally and maternally expressed imprinted genes were considered? Were imprinted genes statistically more likely to be misregulated in mat nrpd1?

    Significance

    Significance:

    PolIV is a plant-specific polymerase that functions part of the plant-specific RNA-directed DNA methylation pathway, which has been very well characterized in Arabidopsis. Mutations in PolIV were previously shown to rescue paternal excess crosses when inherited paternally (Erdmann et al. 2017), and this study extended that finding to show that maternal vs. paternal loss of Pol IV has opposite effects on seed survival in paternal excess crosses. Only one other example (met1) of opposed paternal vs. maternal effects on seed development is known, making Pol IV a useful tool for studying why and how these effects occur. As the authors note, the dominant theory on 'why' involves Pol IV mediating parental conflict over resource allocation in the seed, and the opposite effects of Pol IV maternal/paternal loss at some genes support this hypothesis. The 'how' remains unclear, although this study eliminates several possibilities, and the most likely remaining model is that Pol IV parent-of-origin specific effects occur mostly in trans. Future work can build on these findings to identify the mechanism by which Pol IV achieves these parent-of-origin specific effects.

    My background is mostly in plant epigenetics and genomics.

    Referee Cross-commenting

    The other referee comments seem fair, and I have not comments at this time.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    This manuscript provides evidence that a loss of either the maternal or paternal copy of NRPD1 have different, and sometime opposite, effects on the accumulation of small RNAs and on expression of a subset of genes, with a loss of the maternal copy having more substantial effects. The manuscript is well written, and the conclusions, as far as they go, are justified by the data, which are effectively presented. The overall effect is subtle but informative and according to the authors support a parental conflict model for imprinting. The experiments failed to find a smoking gun in the form of a mechanism to explain how or why the maternal and paternal alleles have different effects and the explanation for a lack of clear phenotypic differences was reasonable, but untested. I would have like to see it tested by looking in a plant species that is outcrossing and highly polymorphic. However, I do appreciate that the observation that the male and female alleles can have distinct effect when mutant is an important clue. My specific comments below may reflect confusion on my part, rather than real issues. If that is that I hope that confusion can aid in clarifying what are in places quite subtle points.

    Specific comments:

    Page 6, last paragraph: "Because the endosperm is triploid, in these comparisons there are 3 (wild-type), 2 (pat nrpd1+/-), 1 (mat nrpd1+/-) and 0 (nrpd1-/-) functional NRPD1 alleles in the endosperm. However, NRPD1 is a paternally expressed imprinted gene in wild-type Ler x Col endosperm and the single paternal allele contributes 62% of the NRPD1 transcript whereas 38% comes from the two maternal alleles (Pignatta et al., 2014). Consistent with paternal allele bias in NPRD1 expression, mRNA-Seq data shows that NRPD1 is expressed at 42% of wild- type levels in pat nrpd1+/- and at 91% of wild-type levels in mat nrpd1+/- (Supplementary Table 6)".

    I would think this would really complicate the analysis. Should all of the dosage values include NPRD1 imprinting values? That is to say, expressed in terms of expression values? This is also a bit confusing. The maternal copies together express 38% of the transcript, so why isn't the mat nrpd1 at 68%, rather than 91%? In any event, given this imprinting and differences in dosage of the male and female it appears that two variables, parental origin and expression levels are being compared. Since 91% is awfully close to 100%, are the mat pat comparisons really just comparing low with nearly normal expression of NRPD1? And actually, given that, the outsized effect of the mat nrpd1 +/- is even more striking.

    Page 4, Line 16. I'm afraid it's still a bit difficult to understand what was being compared what in this section. Please clarify.

    Page 5, Line 5. I'm sure this is fine, but it's not entirely clear what is from the previously published paper and what is reanalysis here. All the crosses and measurements were made then, but not organized in this way?

    Page 6, Line 26. This is an excellent dosage series, but it's complicated by imprinting. So it's not 3, 2, 1, 0 effective copies. If we set the paternal copy at ~1 and each maternal at ~0.1, then it's 1.2 (wild type), 0.20 (pat nrpd1+/-), 1 (mat nrpd1+/-), and 0 (nrpd1-/-).

    Page 6, line 31. Is the main thing we are comparing the difference between expression at 42% verses 91% of wild type?

    Page 7, line 11. So, the overall effect in either direction on smRNA gene targets was really quite small, and I'm guessing the effect on gene expression was even smaller.

    Page 7, line 17. I take it that it is this difference, rather than the overall numbers that is of interest.

    Page 9, line 2. Really interesting, since one might expect that these are methylated loci that would be expected to be fed into any existing embryo maintenance methylation pathway. Surprising that they are maintained independently.

    Page 9, line 22. Proportion of total imprinted genes? Did the mutant obviate/enhance the imprinting?

    Page 9, line 27. How could 2) occur in the homozygous mutant?

    Page 10, line 9. Interesting!

    Page 10, line 11. Is this 2.7 versus 2.18 significant because it's statistically significant, or because it's conceptually significant? Are the examples in 3D representative, or the most convincing examples? And a big difference in ROS1 is of some concern, since that may well be expected to affect imprinting indirectly. I know I'm being picky here, but the pattern is so intriguing I'd be worried about confirmation bias.

    Page 10, line 18. Ok, but 0.123 is a pretty subtle negative correlation. Although I do appreciate that it clearly is not a positive correlation. If I'm understanding correctly, this was the "aha" moment, because it's exactly what one might expect of NRPD1 from the mother and father or working at cross purposes. But the numbers are getting awfully small here.

    Page 10, line 29. The point simply being that that other phenomenon is also significant even if the differences are that large?

    Page 12. So, there is no effect on cleavage and no obvious effect on flanking siRNA clusters. The suspense is building...

    Page 12, line 24. And not in potential regulatory regions? CNSs?

    Page 12, line 28. I guess it depend on whether or not the changes are in regulatory sequences no immediately apparent as part of the gene, doesn't it?

    Line 13, line 2. Not sure it's that important, but couldn't you chop all of these genes in half and see if they are no longer significant collectively?

    Page 14, line 15. I'm afraid I'm getting confused here with the terms cis and trans here. Just to be clear, cis means a direct effect of small RNAs that are dependent on NRPD1 on a gene and trans means anything else? But in this context, it's not clear that is what is meant. Do you mean that gene expression is determined and preset in the gametophyte? What are the levels of expression of NRPD1 in the two gemetophytes?

    Page 14, line 19. Prior to fertilization?

    Page 14, line 27. Do you mean driven by, or just associated with?

    Page 15, line 26. And this is really the issue. The primary conclusion, backed up by the lack (I'm assuming) of phenotypic differences between mat NRPD1 -/+ and pat NRPD1 +/- suggests that the observed differences in expression are not particularly important, unless the exceptional cases are informative.

    Page 15, line 12. Yes, but I'm not at all clear what the mechanism for this is.

    Page 15, line 23. Since this is a very small subset of genes, are these genes that you might expect to play a role in parental conflict?

    Page 15, line 33. Indeed, these could be the informative exceptions.

    Page 15, line 29. Hardly surprising, given that the paternal copy of NRPD1 is expressed at a higher level than the maternal copies, is it?

    Page 16, line 1. So this is what you mean by in cis. Presetting?

    Page 16, line 9. So ideally, one would want to look at a highly polymorphic out-crosser. I'm not suggesting that for this paper, but would this be a good test of the hypothesis? How about maize?

    Page 16, line 15. But the pat and mat heterozygotes looked the same. No differences in phenotype?

    Page 17, line 22. I'm confused, since aren't most 24 nt smRNAs dependent on POLIV (Figure S2)? Do you mean differentially regulated smRNAs? Expression of POLIV specifically in one or the other parent?

    Page 17, line 23. How are you defining important here? Important because at least in the female NPRD1 is not expressed in the central cell? But not important, since this mutant has no effect on phenotype except in an imbalanced cross.

    Page 18, line 13. For this reason, it would be nice to know much more about these genes. Mutant phenotypes, for instance. And how many of these have this feature conserved?

    Significance

    These are intriguing results that would benefit from a test of the hypothesis by comparing these result with those obtained in an outcrossing plant species.

    Referee Cross-commenting

    I agree that the other comments seem both fair and reasonable.

  5. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary:

    This study addresses key aspects of gene regulation in the developing endosperm of flowering plants. The endosperm is the product of the fusion of the (normally diploid) female central cell with one of the sperm cells, and is indispensable for nourishing the developing embryo, among other important functions. Evolutionary models predict that in flowering plants, the endosperm ought to be the tissue in which parental conflict over the allocation of (female) resources to progeny should manifest. Consequently, endosperm gene expression (including the phenomenon of genomic imprinting) and developmental trajectories have been studied from various perspectives, including the possibility of the fast build-up of reproductive barriers due to failing endosperm (and thus seed) development.

    More specifically, this study utilizes knock-out mutants of the NRPD1 gene, which codes for the largest subunit of RNA Polymerase IV (Pol IV), which is part of the RNA-directed DNA methylation (RdDM) pathway. It builds on previous work by the authors that suggested an important role for Pol IV in mediating allelic dosage in developing endosperm, that small interfering RNAs are produced from both paternally and maternally-derived alleles (Erdmann et al. 2017; contra purported claims by other labs), and that normally inviable seeds from paternal-excess crosses (2n x 4n) can be largely rescued by knocking out individual (paternal) components of the RdDM pathway (Satyaki & Gehring 2019).

    Here, Satyaki & Gehring characterize a variety of expression responses in reciprocal heterozygotes, i.e. products of crosses between a homozygous WT and a homozygous nrpd1 mutant (all with the Ler and Col-0 accessions of A. thaliana). The resulting heterozygotes differ in whether the maternal parent (mat nrpd1+/-) or the paternal parent (pat nrpd1+/-) contributed the nrpd1 allele. In addition, mRNA and sRNA expression was also assessed for the WT (+/+) and the homozygous nrpd1 lines (-/-).

    Key findings of this work are that the loss of Pol IV in maternal and paternal parents has different consequences for endosperm gene expression, some of which appear to be antagonistic. In other words, the presence of a functioning Pol IV in the mother and father have parent-of-origin effects on the resulting endosperm. Furthermore, one parent's copy of NRPD1 was found to be sufficient for the production of most Pol IV-dependent sRNAs, yet with a fairly small number of mostly non-overlapping loci losing sRNAs upon loss of either maternal or paternal NRPD1. Pol IV activity in the father and mother is shown to have distinct impacts on the endosperm's DNA methylation landscape.

    Interestingly, while the proportion of mis-regulated genes seems small in both heterozygotes, it is much more restricted in pat nrpd1+/-. Jointly, the authors' results suggest that paternal and maternal Pol IV are genetically antagonistic and that their effects on endosperm transcription in heterozygotes is established before fertilization.

    Major Comments:

    I believe that this is a very sound and authoritative study. The analysis of all data seems appropriate and robust, and many connections between the data (and subsets of data) and their possible interpretations have been considered. In fact, in the massive Results section, some interpretations are supported by cited references (this is not meant as a critique). However, I wonder about the length of the Results section, and the balance between it and the relatively short Discussion section. It is difficult for me to nail down any part of Results that might be shortened, as I could not find clear redundancies. I also think that the level of speculation is absolutely warranted, and I did not find excessive claims being made to this or that end. Rather, I suggest to broaden the perspective somewhat (in their Discussion; see below under Significance), which might allow people with a less mechanistic perspective to grasp the potential relevance of this work for non-model plant systems studied mostly by evolutionary geneticists.

    One aspect that might warrant more scrutiny is the mapping of sRNA reads to the reference genome. I found the short section of this (M&M section, page 20, lines 23-25) to be too brief. It is not clear to me which of ShortStack's v3 weighting scheme the authors used, which is relevant for multi-mapping reads (see NR Johnson et al. 2016, G3). In addition, it is not mentioned whether zero mismatches were allowed. Perhaps this is described in more detail in Erdmann et al. (2017), but even if so, it deserves to be clarified here.

    All in all, I find this work to be meticulously presented and the data to be thoughtfully interpreted. The major conclusions seem to be convincing and adequate, given the underlying data. I have no qualms about replication issues, nor about statistics.

    Minor Comments:

    The manuscript is well-written and concise, despite the length of the Results section. The verbal clarity and absence of typos or grammatical issues is superb. I did find some of the Figures to be somewhat "un-intuitive", in the sense that it takes acute concentration for an outsider (of sorts) to gather and interpret the underlying data. This is probably due to the many cross-comparisons of differences between two genotypes on one axis and those of a different pair of genotypes on the other axis. I am not sure how this issue can be ameliorated (nor whether this is really necessary); however, from a technical point of view, all Figures and Suppl. Figures are really well-done.

    The list of references seems adequate in terms of citing relevant (both older and very recent) publications. However, almost all cited papers concern Arabidopsis or other model species; I suggest to consider adding a few relevant studies on non-Brassicaceae (whether considered model taxa or not), in conjunction with my suggestion (in Significance) to potentially broaden the scope by searching for natural phenomena that also involve parent-of-origin effects on endosperm/seed development. Curiously, many of the references are "incomplete" in the sense of stopping with the journal's name, then stating the doi, i.e. they lack volume numbers and page/article numbers. This should be harmonized throughout.

    Significance

    Significance:

    While part of the earlier data from Erdmann et al. (2017) were re-analyzed in the present study, the vast amount of data are new and concern the expression consequences at the diploid level (2n x 2n crosses), and thus may prove to be more relevant for future comparisons with non-model flowering plants, either for normal intraspecific seed development or (partly) failing crosses between slightly diverged evolutionary lineages. In my view, this study presents a significant advance in understanding the downstream consequences (endosperm mRNA and sRNA expression levels, levels and patterns of DNA methylation) of a perturbed Pol IV expression in both parents or the female vs. male parent. Much of the emphasis in the field has been on paternal-excess crosses, within the larger realm of the "triploid block" or the reproductive barriers between plants of different ploidy (typically diploids x tetraploids in both cross directions).

    The fact that the molecular consequences of disabled Pol IV in one or both parents were assessed in balanced crosses (and not interploidy crosses) may allow an easier connection to natural phenomena such as (partial) hybrid seed lethality between closely related lineages of flowering plants, where parent-of-origin effects have emerged in the recent literature, both at the phenotypic level of endosperm/seed growth and at the molecular level (perturbed imprinting in the endosperm, mRNA and sRNA expression levels, etc.). It thus might prove worthwhile to screen recent papers in Solanum, Mimulus, and Capsella to evaluate the possibility of "connections" between the current data and recent, admittedly more descriptive, findings in diverse taxa that don't offer the same genomic resources as Arabidopsis.

    What the current version of this work already does is relating the finding of partly antagonistic influences of pat nrpd1+/- versus mat nrpd1+/- on endosperm mRNA expression to evolutionary models championed by D. Haig ("parental conflict", "kinship theory"). My above suggestions would strengthen such connections and likely would broaden the appeal of this work to scientists with diverse backgrounds outside the core expertise of plant molecular/developmental biologists. In any case, I see the prime scientific audience as the latter group, but I see potential to intrigue people with a more evolutionary background.

    Field of expertise:

    population genomics, hybrid seed lethality, speciation, genomic imprinting, evolutionary models.

    Referee Cross-commenting

    I agree that the other referee comments (while being quite complementary to mine due to differences in main expertise) seem both fair and reasonable.

  6. Version published to 10.1101/2021.03.24.436774v2 on bioRxiv
  7. Version published to 10.1101/2021.03.24.436774v1 on bioRxiv