# Source paper archive

- Source title: PRC2-mediated H3K27me3 uncouples PHE1 binding from transcriptional activation at imprinted genes
- Author(s): Bente, H., Höhne, M., Köhler, C.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.19.767895v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.19.767895v1.full.pdf
- PDF SHA-256: 1b29d3b1994b71b6f0816570536f0ffab572098be90f572f3e2ea777b263c60f
- Converter: pymupdf4llm 0.0.27
- Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version.

---

&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


1 **PRC2-mediated H3K27me3 uncouples PHE1 binding from transcriptional activation at**


2 **imprinted genes**


3 Heinrich Bente &#91;1&#93;, Melanie Höhne &#91; 1&#93;, Claudia Köhler &#91;1*&#93;


4 1 Department of Plant Reproductive Biology and Epigenetics, Max Planck Institute of


5 Molecular Plant Physiology, Potsdam 14476, Germany


         6 Author for correspondence: Claudia Köhler, Koehler@mpimp-golm.mpg.de


7


8 **Abstract**


9 Genomic imprinting results in parent-of-origin-dependent gene expression and plays an


10 important role in endosperm development. The type I MADS-box transcription factor


11 PHERES1 (PHE1) has been implicated in the regulation of imprinted genes, particularly


12 paternally expressed genes (PEGs). Silencing of the maternal alleles of PEGs is mediated


13 by repressive histone modifications, including H3K27me3 established by the Polycomb


14 Repressive Complex 2 (PRC2). Here, we established an adapted low-input CUT&amp;Tag


15 protocol to profile the allele-specific genome-wide binding of PHE1 in the endosperm and


16 compared the resulting binding profiles with previously published ChIP-seq data. Our


17 CUT&amp;Tag analysis substantially recapitulated the previously identified PHE1-binding


18 landscape while requiring considerably less input material. Consistent with earlier findings,


19 PHE1 binding was enriched at RC/Helitron transposable elements and imprinted genes,


20 specifically PEGs. Interestingly, most PHE1-bound PEGs exhibited binding on both maternal


21 and paternal alleles despite H3K27me3 enrichment on their maternal alleles. Perturbation of


22 H3K27me3 deposition did not alter the predominantly biallelic PHE1 occupancy observed at


23 these loci, supporting that PHE1 binding can persist in an H3K27me3-associated chromatin


24 environment. In contrast, analysis of gene expression in _fertilization independent seed_ ( _fis_ )


25 _prc2_ mutants impaired in H3K27me3 establishment caused increased expression of PHE1


26 targets and altered their parent-of-origin-specific expression patterns. Our findings reveal


27 that biallelic transcription factor occupancy can coexist with parent-of-origin-specific


28 transcriptional regulation, highlighting the importance of chromatin context in determining the


29 transcriptional output of PHE1-bound genes.


30


1


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


31 **Introduction**


32 The endosperm is a nourishing tissue that supports embryo growth. It is formed after double


33 fertilization, whereby one sperm cell fuses with the central cell to form the endosperm, while


34 the other fertilizes the egg to develop into the embryo. In _Arabidopsis thaliana_ (hereafter


35 referred to as Arabidopsis), fertilization of the homodiploid central cell by a haploid sperm


36 cell gives rise to the triploid endosperm, characterized by a 2:1 maternal-to-paternal (2m:1p)


37 genome ratio. Maintaining this ratio is essential for normal endosperm development, as


38 deviations, such as those resulting from crosses between parents of different ploidy levels,


39 frequently leads to seed abortion (Dilkes et al., 2008, Scott et al., 1998).


40 The endosperm is also a major site of genomic imprinting, an epigenetic phenomenon that


41 results in parent-of-origin-specific gene expression (Gehring et al., 2011, Hsieh et al., 2011,


42 Batista and Köhler, 2020). Depending on the parental origin of the expressed allele,


43 imprinted genes are classified as maternally expressed genes (MEGs) or paternally


44 expressed genes (PEGs). The non-expressed allele is typically silenced by repressive


45 epigenetic modifications, collectively referred to as the ‘imprint’. The imprint varies between


46 the affected parental allele, whereby MEGs are typically characterized by reduced DNA


47 methylation on the maternal allele, as a result of DNA demethylation of small repetitive


48 regions in the central cell (Gehring et al., 2009, Gehring et al., 2011, Pignatta et al., 2014,


49 Hsieh et al., 2011). On the contrary, the majority of PEGs are marked by trimethylation at


50 lysine 27 of histone H3 (H3K27me3) on their maternal allele, mediated by the activity of the


51 FERTILIZATION INDEPENDENT SEED (FIS) Polycomb Repressive Complex 2 (PRC2) in


52 the central cell (Moreno-Romero et al., 2016, Mozgova et al., 2015, Zhang et al., 2014).


53 The M-type MADS-box transcription factor PHERES1 (PHE1) has previously been shown to


54 bind to numerous key regulators of endosperm development, including many PEGs) (Batista


55 et al., 2019. Although loss of PHE1 and its paralog PHE2 does not cause detectable


56 phenotypic abnormalities in diploid seeds, the _phe1 phe2_ double mutant suppresses triploid


57 seed abortion (Batista et al., 2019), supporting a causal role for increased PEG expression


58 in triploid seed abortion and a role of PHE1 in activating of these genes (Kradolfer et al.,


59 2013, Wolff et al., 2015, Huang et al, 2017). Previous work proposed that PHE1 binds to


60 both parental alleles of PEGs, indicating that allele-specific H3K27me3 prevents activation of


61 the maternal alleles despite PHE1 presence. However, this conclusion was based on the


62 analysis of two PEGs, limiting its general applicability. A genome-wide investigation of allele

63 specific PHE1 binding was previously hampered by the large amounts of developmental


64 stage-specific hybrid material required for ChIP-seq, the gold-standard method for mapping


65 transcription factor binding sites at the time. To overcome this limitation, we adapted the


2


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


66 Cleavage under target and tagmentation (CUT&amp;Tag) protocol (Kaya-Okur et al., 2019),


67 which requires substantially less input material than conventional Chip-seq. Using samples


68 comparable to those analyzed in previous Chip-seq studies, we demonstrate that our


69 CUT&amp;Tag protocol identified a largely overlapping set of PHE1 targets, including a high


70 proportion of PEGs. Allele-specific analysis revealed that PHE1 binds to both parental alleles


71 of the majority of its targets genome-wide, including PEGs. We further show that this biallelic


72 binding occurs independently of H3K27me3 deposition, whereas H3K27me3 prevents


73 transcriptional activation of the maternal alleles, despite PHE1 binding. Together, our


74 findings provide new insights into the interplay between MADS-box transcription factors and


75 epigenetic imprinting mechanisms in regulating seed development. Moreover, our adapted


76 CUT&amp;Tag protocol provides a powerful approach for profiling genome-wide transcription


77 factor binding in tissues with severely limited sample availability, such as the endosperm.


78 **Results**


79 **CUT&amp;Tag works reliably for an endosperm-expressed transcription factor**


80 The transcription factor PHE1 is transiently expressed in the endosperm for only approx. 5


81 days after pollination (DAP) (Weinhofer et al., 2010). This temporally and spatially restricted


82 expression makes it difficult to generate sufficient amounts of time-resolved input material for


83 traditional methods such as ChIP-seq for the identification of transcription factor binding


84 sites. To overcome this challenge, we have adapted the recently published CUT&amp;Tag


85 protocol (Kaya-Okur et al., 2019), proposed to require only a fraction of previously needed


86 input material. To evaluate the performance of CUT&amp;Tag for PHE1-GFP compared to ChIP

87 seq, we generated CUT&amp;Tag libraries from the same transgenic lines at comparable time


88 points of two DAP, as had been used in previous work (Batista et al., 2019). A genome-wide


89 analysis of our newly generated CUT&amp;Tag data of _PHE1::PHE1-GFP_ containing plants data


90 revealed good Spearman correlation between the samples from independent biological and


91 methodological replicates for both PHE1-GFP and anti-H3 sample (Supplementary


92 Fig. S1A). To compare the CUT&amp;Tag target enrichments in comparison to ChIP-seq, we first


93 observed well-known PHE1 targets manually in a genome browser using three independent


94 replicates, in comparison to anti-H3 samples (for CUT&amp;Tag) or input samples (for ChIP-seq)


95 as background control (Fig. 1A). The upstream regions of known PHE1 target genes such as


96 _YUCCA10 (YUC10)_ _AGAMOUS-LIKE 23 (AGL23)_ and _HAIKU2 (IKU2)_ showed stronger


97 signal from anti-GFP CUT&amp;Tag samples (dark green), but not in anti-H3 CUT&amp;Tag (dark


98 grey), revealing enrichment for well-known targets (Fig. 1A). Importantly, CUT&amp;Tag


99 performed on wild-type siliques containing no _PHE1::PHE1-GFP_ transgene did not show


100 enrichment of known PHE1 target regions (Fig. 1A). Compared to ChIP-seq data (Batista et


3


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


101 al., 2019) the signal enrichment was detectable in the same promotor region, however,


102 CUT&amp;Tag led to a much more prominent enrichment (Fig. 1A). A stronger target-enrichment


103 through CUT&amp;Tag had previously also been reported for histone modifications in comparison


104 to ChIP-seq (Kaya-Okur et al., 2019, Abbasova et al., 2025). To obtain only high confidence


105 CUT&amp;Tag peaks, we adopted the ENCODE project pipeline including the use of permissive


106 peak calling followed by pairwise irreproducible discovery rate (IDR) peak ranking


107 (Supplementary Fig. S1B), which resulted in 1090 peaks. Different replicates showed good


108 to moderate Spearman correlation over these 1090 regions, ranging from 0.53 to 0.76


109 depending on replicate combination (Supplementary Fig. S1C). In contrast, transgene-free


110 samples resulted in 10 enriched regions, of which only 8 overlapped with the 1090 PHE1


111 target regions (Supplementary Fig. S1D). The fraction of fragment in peaks (FFiPs) over


112 these 1090 regions was approximately five-fold higher in anti-GFP libraries from


113 _PHE1::PHE1-GFP_ containing samples, compared to Col-0 derived anti-GFP samples


114 (Supplementary Fig. S1E). Annotation of PHE1 targets from both methods led to unique


115 target genes, in total 994 for CUT&amp;Tag and 1749 for ChIP-seq (Fig. 1B), with a significant

116 number of 562 PHE1 target genes (Hypergeometric test p &lt; 2.2*10 &#91;-16&#93; ) were shared between


117 CUT&amp;Tag and ChIP-seq (Fig. 1C). The discrepancy in identified target-regions is a


118 consequence of the higher number of CUT&amp;Tag replicates compared to the previous ChIP

119 seq data, allowing the more stringent peak selection inspired by the ENCODE IDR workflow.


120 We performed feature annotation of the CUT&amp;Tag-identified 1090 PHE1 target sides, which


121 showed that 91.47% of PHE1 target sites are located in promoter regions (Fig. 1D, -1.5 to


122 +0.5 kb relative to TSS), as previously shown for transcription factors (Yu et al., 2016,


123 Heyndrickx et al., 2014, O'Malley et al., 2016). Similar annotation of the ten target regions


124 from wild-type samples, showed that only five of those falling into the same promoter window


125 (Supplementary Fig. S1E). Enriched regions identified by CUT&amp;Tag displayed similar


126 features as those identified by ChIP-seq experiments, including enrichment for transposable


127 elements (TEs), specifically for RC/Helitrons (Batista et al., 2019). CUT&amp;Tag confirmed that


128 half (52%) of PHE1 target regions overlap with TEs, of which 66% belong to the RC/Helitron


129 superfamily (Fig. 1E). Thus, 34% of all PHE1 target sides overlap with RC/Helitrons,


130 corresponding to previous ChIP-seq experiments (Batista et al., 2019). Lastly, as PHE1 is a


131 MADS box transcription factor, we tested if MADS motives were prevalent in the 994 target


132 sides. We found enrichment for multiple MADS motives, among them the previously


133 identified SVP MADS motif (Fig.1 F). Overall, the in-depth comparison between CUT&amp;Tag


134 and ChIP-seq performed in comparable tissues, revealed that both methods could recover


135 promoter-target regions with good reproducibility, while CUT&amp;Tag library preparation only


136 required a fraction of the input. Nevertheless, less peaks were recovered with CUT&amp;Tag


4


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


137 after stringent analysis, but these peaks were more prominent compared to peaks identified


138 by ChIP-seq.


139 **Imprinted regulation of PHE1 targets is a consequence of allele-specific H3K27me3**


140 In the previous ChIP-seq experiment, many PHE1 targets were imprinted genes (Batista et


141 al., 2019). Among the PHE1 targets established using CUT&amp;Tag, we also found a


142 substantial number of imprinted genes, specifically PEGs. We opted to verify the enrichment


143 of imprinted genes in our CUT&amp;Tag data using a list of high confidence imprinted genes,


144 appearing in at least two independent studies referred to as ‘core’ imprinted genes (Del


145 Toro-De Leon et al., 2024). This list consists out of 268 ‘core’ MEGs as well as 103 ‘core’


146 PEGs, we simply refer to as MEGs or PEGs respectively. CUT&amp;Tag revealed that only a few


147 MEGs (6/268, 2.24%) were targeted by PHE1, whereas a significant number of 20 PEGs

148 (hypergeometric test p = 2.6*10 &#91;-11&#93; ), thus 19.42% of PEGs were PHE1 targets (Fig. 2A). This


149 high number of targeted PEGs raised the question if PHE1 targeting itself exerts allele


150 specific binding. To investigate allele-specificity, we made use of single nucleotide


151 polymorphism between the natural Arabidopsis accessions Col-0 and Landsberg _erecta_


152 (L _er_ ). We used L _er_ as maternal parents whereas _PHE1::PHE1-GFP_ transgenic plants in Col

153 0 background were used as pollen donor. To allow comparability with publicly available


154 imprinting data, F1 siliques were harvested at 4DAP and subjected to CUT&amp;Tag sequencing.


155 However, PHE1 PEG targets did not show allele-specific PHE1 binding across all three


156 replicates (Fig. 2B), supporting our previous observations that were based on two analyzed


157 PEGs (Batista et al., 2019). Similarly, the identified MEG targets showed mainly biallelic


158 PHE1 occupancy (Fig. 2B), in contrast to previously detected predominantly maternal


159 binding (Batista et al., 2019). However, the two previously analyzed MEGs were not


160 detected in the CUT&amp;Tag data, leaving the possibility that allele-specific binding of PHE1


161 may occur at some MEGs. Browershots of PHE1 PEG targets confirmed the biallelic signal


162 in the PHE1 target region (Fig. 2C). Consistent with the trend for PEGs/MEGs, the majority


163 of non-imprinted PHE1 targets were bound on the maternal and paternal alleles with an


164 average maternal ratio of 0.72, close to the expected biallelic ratio of 0.66 (2m:1p) in the


165 endosperm (Fig. 2D). Similarly, the maternal ratio in defined promoter windows (-1500 bp


166 TSS +500 bp) of all PHE1 target genes showed a mean maternal ratio of 0.76


167 (Supplementary Fig. S2). The strong biallelic binding of PHE1 contrasts with the paternally


168 biased expression of its target genes, raiding the question how maternal alleles remain


169 silenced despite PHE1 binding.


170 We previously found that PEGs are associated with maternal-specific H3K27me3 (Moreno

171 Romero et al., 2019, Moreno-Romero et al., 2016), suggesting that the parent-specific


5


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


172 presence of H3K27me3 is decisive for allele-specific expression of PHE1 target genes rather


173 than allele-specific binding of PHE1. We tested this hypothesis by analyzing the allele

174 specific presence of H3K27me3 at PHE1 PEG target genes and found indeed that the


175 maternal alleles of PHE1 PEG targets are enriched for H3K27me3 (Fig. 2E), suggesting that


176 the presence of H3K27me3 determines whether PHE1 binding leads to expression of the


177 targeted allele. We did not detect allele-specific H3K27me3 for the 6 MEG PHE1 targets,


178 consistent with previous work showing that imprinting of most MEGs is independent of allele

179 specific H3K27me3 (Fig. 2F) (Moreno-Romero et al., 2019, Moreno-Romero et al., 2016).


180 Also, consistent with previous observations, PHE1-binding sites on paternal MEG alleles


181 exhibited increased CG DNA methylation; however, due to the limited number of PHE1

182 targeted MEGs, this difference was not statistically significant (Supplementary Fig. S3A). In


183 contrast, PEGs targeted by PHE1 exhibited increased CG methylation on paternal alleles


184 and, consistent with previous findings, increased CHG methylation on maternal alleles


185 (Supplementary Fig. S3B) (Moreno-Romero et al., 2019). Together, these results suggest


186 that PHE1 binds the majority of its targets biallelically, including imprinted genes. For PEGs,


187 paternal-allele specific expression is therefore likely maintained through H3K27me3

188 mediated silencing of the maternal allele.


189 **PHE1 binding is compatible with allele-specific H3K27me3**


190 Publicly available endosperm-specific expression data (Moreno-Romero et al., 2019)


191 showed that the average PHE1 target expression is increased in comparison to all annotated


192 genes in Araport11, and slightly increased compared to expressed genes in the endosperm


193 (RPKM &gt;5, Fig. 3A). Furthermore, PHE1 targets had slightly, but significantly higher levels of


194 H3K27me3 compared to all endosperm expressed genes (Fig. 3B). This enrichment of


195 H3K27me3 compared to H3 was not only prevalent in the gene body region, but also in the


196 promoter region where PHE1 binds (Fig. 3C). To explore a possible effect of H3K27me3 on


197 PHE1 binding, we used kmeans clustering to obtain three clusters around the midpoint of all


198 994 PHE1 targets in a +-1kb interval. This resulted in cluster 1 containing 189 PHE1 target


199 sites enriched in H3K27me3, cluster 2 with 410 PHE1 target sides with lower levels of


200 H3K27me3, and cluster 3 with 395 genes having the least H3K27me3 (Fig. 3D). Cluster 1


201 contained 16 of the 20 PHE1 PEG targets, whereas the remaining four PEGs belonged to


202 cluster 2 (Fig. 3E), consistent with previous work revealing that PEGs are marked by


203 H3K27me3 (Moreno-Romero et al., 2019, Moreno-Romero et al., 2016). Most H3K27me3 in


204 cluster 1 is present on the maternal alleles (Fig. 3F), which was a general trend and not


205 restricted to PEGs and applied for both directions of the crosses (Supplementary Fig. S4A).


206 Similar as for PEGs (Supplementary Fig. S4B-C) the presence of H3K27me3 reduced


207 expression of the maternal alleles, biasing genes located in cluster 1 significantly for paternal


6


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


208 expression, whereas genes in clusters 2 and 3 were biallelically expressed (Fig. 3G).


209 Nevertheless, genes in cluster 1 were biallelically bound by PHE1 (Fig. 3H), despite the


210 presence of maternal H3K27me3 on PHE1 target sites.


211 To further test whether the presence of H3K27me3 on the maternal alleles does not interfere


212 with PHE1 binding, we tested PHE1 binding in seeds of _mea/+_ mutants, where we


213 introduced the _PHE1::PHE1-GFP_ through pollen. _Mea_ is a maternal gametophytic mutant,


214 with 50% of seeds from a _mea/+_ plant being depleted for MEA function and abort


215 (Grossniklaus et al., 1998). We compared the allele-specific PHE1 binding at known PHE1


216 target sites (Fig. 2). (Fig. 4A). Consistent with the data showing PHE1 binding in the


217 presence of H3K27me3 (Fig 3D), PHE1-binding on the maternal alleles was not substantially


218 increased in the _mea_ mutant (Fig. 4A-C). While the ratio of bound alleles was significantly


219 different for genes in cluster 1, there was no pronounced change towards PHE1 binding of


220 the maternal allele (Fig. 4A). This suggests that while _mea/+_ indeed influences PHE1


221 binding, this affects both alleles not just the maternal one. One possible explanation could be


222 that increased expression of _PHE1_ in the _mea_ mutant (Köhler et al., 2003) leads to


223 increased expression of PHE1 targets.


224 To test this hypothesis, we analyzed publicly available endosperm transcriptome data from


225 plants deficient in FIE, a core component of the FIS-PRC2, which also comprises MEA, FIS2


226 and MSI1 (Hsieh et al., 2011, Köhler et al., 2003). Many PHE1 target genes were


227 upregulated in _fie_ compared to wild type (Fig. 4D). The upregulation of PHE1 target genes in


228 _fie_ was accompanied by stronger expression of the maternal alleles, consistent with


229 H3K27me3 being predominantly located on maternal alleles (Fig. 4E). This was also


230 reflected by PHE1 target genes located in cluster 1 (having highest H3K27me3) showing the


231 strongest expression changes in _fie_, followed by genes in cluster 2, whereas cluster 3


232 showed only minor upregulation compared to wild type (Fig. 4F). These data suggest that


233 genes with higher H3K27me3 enrichment exhibit stronger transcriptional upregulation in FIS

234 PRC2 mutants, suggesting that the extent of H3K27me3 enrichment is associated with the


235 magnitude of PRC2-dependent transcriptional repression. PHE1 targeted PEGs were among


236 the most highly upregulated genes in _fie_ and were significantly higher expressed compared


237 to all other imprinted genes and also non-imprinted PHE1 target genes (Fig 4G). Notably,


238 PHE1-targeted PEGs, which showed the strongest transcriptional upregulation, also


239 exhibited the highest levels of H3K27me3 enrichment on their maternal alleles compared


240 with the other gene categories. Interestingly, the strong upregulation of PHE1 targeted PEGs


241 was not accompanied by increased PHE1 targeting to the maternal alleles (Fig. 4A-C). This


242 suggests that PHE1 binds both parental alleles, but H3K27me3 prevents transcriptional


243 activation of the maternal alleles, which become derepressed upon loss of FIS-PRC2


7


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


244 activity. Collectively, these results demonstrate that PHE1 binds to both alleles of several


245 H3K27me3-marked genes. H3K27me3 depletion results in increased expression of PHE1


246 target genes, predominantly driven by enhanced maternal allele expression, consistent with


247 the preferential deposition of H3K27me3 on maternal alleles in the endosperm (Moreno

248 Romero et al., 2016). Together, these findings suggest that H3K27me3 restricts PHE1

249 mediated transcriptional activation, particularly on maternal alleles, thereby contributing to


250 the maintenance of parent-of-origin-specific gene expression.


251 **Discussion**


252 Here we report CUT&amp;Tag as a quick and low-cost alternative for genome-wide monitoring of


253 TF binding sites from limited amount of tissue, like the Arabidopsis endosperm. Using this


254 technique, we were able to monitor allele-specific binding of the PHE1, a key regulator of


255 PEGs in Arabidopsis and to compare this data with our previously established ChIP-seq data


256 of PHE1 from the same tissue (Batista et al., 2019). Compared to ChIP-seq, CUT&amp;Tag


257 comes with several advantages that are particularly relevant when working with limited


258 amounts of tissue. CUT&amp;Tag protocols are substantially faster than ChIP-seq, require less


259 input material, and generally result in better signal to noise output, obtained from less deep


260 sequenced samples. Due to the shorter and hence faster protocols, CUT&amp;Tag-seq, as well


261 CUT&amp;RUN-seq are slowly replacing ChIP-seq, especially for the analysis of histone marks in


262 mammalian cell systems (Kaya-Okur et al., 2019, Bartosovic et al., 2021), but also in plants


263 (Zheng and Gehring, 2019, Tao et al., 2020, Del Toro-De Leon et al., 2024).


264 Still, CUT&amp;Tag is rarely used to obtain genome-wide binding events of transcription factors,


265 and it is often recommended to use CUT&amp;RUN instead, as the general higher salt


266 concentrations in CUT&amp;Tag buffers, which are needed to dampen the DNA affinity of free


267 pA-Tn5, are supposed to prevent weak TF-DNA interactions (Kaya-Okur et al., 2020).


268 Indeed, our CUT&amp;Tag against PHE1-GFP identified about half the number of the previously


269 identified ChIP-seq targets (Batista et al., 2019); with about half of those identified targets


270 being shared between both methods. Similarly, a recent comparison of CUT&amp;Tag-seq- and


271 ChIP-seq-mediated profiling of histone modifications found about half of the target sites


272 overlapped between both methods (Abbasova et al., 2025), similar to the overlap found in


273 this study. Together with the significant overlap between CUT&amp;Tag and ChIP-seq targets,


274 the enrichment of MADS-box binding motifs and TEs, particularly RC/Helitrons, supports the


275 reliability of CUT&amp;Tag in identifying PHE1 binding sites. Consistent with previous PHE1


276 ChIP-seq experiments (Batista et al., 2019), we found a significant enrichment of imprinted


277 genes, particular PEGs, among PHE1 target genes identified by CUT&amp;Tag. The lower input


8


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


278 requirement of CUT&amp;Tag over ChIP-seq enabled us to generate allele-specific PHE1 binding


279 profiles, thus we were able to test if parent-of-origin-specific expression is accompanied by


280 allele-specific PHE1 binding. We were able to show that all by CUT&amp;Tag identified PEG


281 PHE1 targets, PHE1 localizes to both alleles, the actively transcribed paternal, but also to


282 the repressed maternal allele. These data support and extend our previous findings that


283 were based on two PEG PHE1 targets (Batista et al., 2019). PEGs in the Arabidopsis


284 endosperm are characterized by the presence of H3K27me3 on the maternal alleles


285 (Moreno-Romero et al., 2019, Moreno-Romero et al., 2016). We found that PHE1 binds to


286 the repressed maternal alleles of PEGs, but also of other, not imprinted genes marked by


287 H3K27m3 on their maternal alleles and surrounding the PHE1 binding sites. Repressive


288 H3K27me3 found at PEGs are established in the central cell by the FIS-PRC2 and inherited


289 to the endosperm (Gehring, 2013, Rodrigues and Zilberman, 2015). Since _PHE1_ is


290 expressed after fertilization (Köhler et al., 2005), it is unlikely that PHE1 itself recruits the


291 FIS-PRC2, while we cannot rule out a maintenance function. Possibly the function of PHE1


292 depends on different interaction partner, as MADS box TFs show a broad interaction matrix


293 with each other (Bemer et al., 2010, de Folter et al., 2005).


294 Consistent with our finding that PHE1 shows binding to regions marked by H3K27me3, we


295 found that depletion of this modification in _mea_ mutants did not impact on PHE1 binding.


296 However, the maternal alleles of PHE1 targets became activated upon H3K27me3 depletion,


297 revealing that the presence of H3K27me3 prevents activation of the maternal alleles despite


298 PHE1 binding. Interestingly we observed a positive connection between H3K27me3


299 enrichment at PHE1 targets and the magnitude of gene activation following PRC2 loss.


300 Previous work established that the level of H3K27me3 required to maintain transcriptional


301 silencing depends on promoter activity, with transcriptionally poised promoters becoming


302 active despite retaining levels of H3K27me3 (Jadhav et al., 2020). What determines the


303 level of H3K27me3 remains to be elucidated, but it suggests that there is a connection


304 between the potential transcriptional activity of a gene and the level of H3K27me3 required


305 to keep it silenced.


306 In conclusion, we provide a CUT&amp;Tag protocol applicable for small amount of tissues that


307 works reliable as confirmed by comparing our data with previously published ChIP-seq data.


308 The high signal-to-noise ratio typical for CUT&amp;Tag allowed us to reduce the starting material


309 substantially, enabling us to generate allele-specific TF binding profiles from early


310 endosperm tissue, further the understanding of PHE1 transcriptional control and its


311 relevance for seed development. These findings bridge transcription factor function with


9


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


312 epigenetic regulation, offering new insights into how MADS-box proteins connect with


313 imprinting mechanisms to control seed development.


314 **Methods**


315 **Growth condition**


316 The Col-0 _PHE1::PHE1-GFP_ line was previously described by Batista et al. (2019). Seeds


317 were sterilized by immersing them in 70% ethanol for 10 minutes, followed by 100% ethanol


318 for 5 minutes. The ethanol was then removed, and the seeds were air-dried under sterile


319 conditions. Sterilized seeds were sown on ½ strength Murashige and Skoog medium


320 containing 1% sucrose and 0.68% agar. Stratification was performed in darkness at 4°C for


321 two days. Seedlings were grown on plates for ten days under long-day conditions (16 hours


322 of light at 21°C and 8 hours of darkness at 18°C with a light intensity of 150 µmol/s/m²)


323 before being transferred to soil. Following transfer, plants were grown under long-day


324 conditions with a light intensity of 150 µmol/s/m² and 70% relative humidity.


325 For experiments on homozygous plants siliques from self-fertilized Col-0 _PHE1::PHE1-GFP_


326 plants were removed, leaving only open flowers corresponding to a time point of 1-2 days


327 after pollination (DAP). The next day, siliques were harvested, weighed, flash-frozen in liquid


328 nitrogen, and stored at -70°C. Material from approximately 2-3 plants were combined per


329 replicate, each replicate consisted out of 250 mg of siliques.


330 For crossing experiments, five-week-old plants were emasculated and pollinated 48 hours


331 later, as indicated. Siliques were harvested at 4 DAP, flash-frozen in liquid nitrogen, and


332 stored at -70°C until nuclei isolation.


333 **Nuclei Isolation and CUT&amp;Tag**


334 Nuclei isolation was performed using approximately 250 mg of siliques per sample and


335 replicate, utilizing the GentleMACS (Miltenyi Biotec) system as previously described


336 (Moreno-Romero et al., 2017). A modified Honda buffer, in which MgCl ₂ was replaced by 0.5


337 mM spermidine, was used to prevent carryover and potential activation of the transposase in


338 subsequent steps. Following filtration and centrifugation for 6 minutes at 1600 g, nuclei were


339 washed with 4 ml of CUT&amp;Tag Wash150 buffer (150 mM NaCl, 20 mM HEPES-Na pH 7.5,


340 0.5 mM spermidine, 1x protease inhibitor) and centrifuged for 3 minutes at 600 g and 4°C.


341 The supernatant was discarded, and the nuclei pellet was washed once with 2 ml of


342 Wash150 buffer, followed by another centrifugation at 600 g for 3 minutes at 4°C. The


343 resulting nuclei pellet was resuspended in 1000 µl of Wash150 buffer with 0.01% digitonin


344 (Dig150 buffer) and centrifuged in a 2 ml tube before being resuspended in X*100 µl of


10


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


345 antibody buffer (Dig150 buffer + 2 mM EDTA, pH 8) in 1.5 ml tubes. Since nuclei


346 immobilization by ConA beads was unsuccessful in several attempts (data not shown),


347 possibly due to the presence of starch or sugar, all steps requiring buffer exchange or


348 washing were done by centrifugation for 3 min at 600 g and 4°C (600 g,3’, 4°C) if not


349 indicated otherwise. For each 100 µl, 2 µl of primary antibody was added (anti-GFP ab290,


350 Abcam; anti-H3, H9289-200UL, Sigma-Aldrich). Samples were gently mixed by flicking,


351 collected by a quick centrifugation for 30 seconds at 600 g, and incubated overnight at 4°C


352 on a nutator. Following incubation, the tubes were centrifuged (600 g,3’, 4°C), and the


353 supernatant was discarded. Then, 100 µl of Dig150 buffer containing 1:100 secondary


354 antibody (anti-Rabbit-IgG, ABIN101961, Antibodies-Online) was added, and the nuclei pellet


355 was resuspended by pipetting.


356 Nuclei were incubated on a nutator for 30 min and subsequently centrifuged (600 g,3’, 4°C),


357 the supernatant were removed and the pellet washed twice with each 300 ul Dig150 buffer.


358 Afterwards, the pellet was eluted in 60 ul CUT&amp;Tag Dig300 buffer (300 mM NaCl, 20 mM


359 HEPES-Na pH 7.5, 0.5 mM spermidine, 1x protease inhibitor, 0.01% Digitonin) including 3 ul


360 pAG-Tn5 enzyme (15-1017, EpiCypher). Nuclei were incubated on a nutator for 1h at room


361 temperature, followed by another centrifugation and two rounds of washing with CUT&amp;Tag


362 300 wash buffer. Hereafter, pellet was eluted in 100 ul Tagmentation buffer (300 mM NaCl,


363 20 mM HEPES-Na pH 7.5, 0.5 mM spermidine, 1x protease inhibitor, 0.01% Digitonin, 10


364 mM MgCl 2 ) followed by one hour incubation at 37°C in a thermoblock in which tagmentation


365 of target DNA fragments is happening. After tagmentation nuclei were collected by


366 centrifugation the supernatant removed and the pellet washed with 200 ul TAPS buffer (10


367 mM TAPS, 2 mM EDTA pH 8) and centrifuged. Afterwards, supernatant was removed and


368 10 ul of SDS Release buffer (SDS, TAPS, EDTA) was added. Nuclei were vortexed for 10’’


369 and incubated for 1h at 58°C. Afterwards, 30 ul of SDS-quench buffer (0.67% Triton-X100 in


370 H2O) was added, together with each 4 ul 10 mM i5 and i7 index primers. The solution was


371 transferred to PCR tubes and amplified using Phusion Polymerase (F534L, Thermo Fisher)


372 for 17 cycles. The resulting libraries were purified with 1.3x SPRIselect beads (B23318,


373 Beckman) and eluted in 24 ul dH2O. Library integrity and concentration was determined


374 using TapeStation D5000 ScreenTapes. Sequencing was performed on NextSeq1000


375 (Illumina) in 50 bp paired-end mode or in 150 bp paired-end mode for allele-specific


376 experiments. For non-allele specific samples, it was aimed for 20 Mio paired-end reads, for


377 allele specific samples 50 Mio reads were aimed for.


378 **Bioinformatic analysis**


11


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


379 Raw fastq reads were trimmed and quality filtered using ‘TrimGalore’ (Version 0.6.10)


380 (https://github.com/FelixKrueger/TrimGalore) using the parameters ‘-q 20 --stringency 3 -

381 length 32 –paired’. Trimmed reads were mapped to a the TAIR10 genome using ‘Bowtie2’


382 (Version 2.5.2)(Langmead and Salzberg, 2012) using the options ‘-I 10 -X 700 --no-unal -

383 no-mixed --no-discordant --phred33’ in the case of _PHE1::PHE1-GFP_ samples and Col-0


384 isogenic wildtype samples. Alignment files were converted to bam files and quality filtered (-q


385 20) and only alignments kept that align to the nuclear chromosomes using ‘Samtools’


386 (Version 1.19.2)(Li et al., 2009). Filtered alignments were used for subsequent analysis.


387 Genome tracks where generated using deeptools ‘bamCoverage’ (Version 3.5.4.post1)


388 (Ramírez et al., 2014) with the options -bs 50 --effectiveGenomeSize 119481543 -

389 normalizeUsing RPGC –skipNAs. Log2 normalization with of H3K27me3 to H3 was achieved


390 using deeptools ‘bamCompare’ using the options --operation log2 -bs 10. Sample correlation


391 matrices were obtained using either deeptools ‘multiBamSummary bins’ with the option -bs


392 5000 for genome wide analysis, or ‘multiBamSummary BED-file’ for target site specific


393 correlation. Spearman correlation was obtained using deeptools ‘plotCorrelation’ with the


394 options --corMethod spearman --whatToPlot scatterplot/heatmap --skipZeros –plotNumbers.


395 Deeptools was run in a Python environment (Version 3.12.0)(Python Software, 2026).


396 Quantification of signal in genomic intervals was achieved via bedtools map (Version 2.30)


397 (Quinlan and Hall, 2010) using the -o mean operator.


398 Prior peak calling and to account for differences in sequencing depth between biological


399 replicates, aligned CUT&amp;Tag libraries were randomly downsampled to the sequencing depth


400 of the replicate with the lowest number of properly paired alignments. The number of


401 properly paired alignments was determined using SAMtools (samtools view -c -f 0x2). For


402 each replicate, the required sampling fraction was calculated relative to the replicate with the


403 lowest read count, and BAM files were randomly downsampled using samtools view -b -s


404 with a fixed random seed of 42. Peaks were called using MACS2 callpeak (Version 2.2.9.1)


405 (Zhang et al., 2008) with the parameters ‘--format BAMPE --gsize 1.119e8 -q 1e-3 in


406 narrowPeak format for reproducibility analysis. Reproducibility peaks between biological


407 replicates was assessed using the irreproducible discovery rate (IDR)(Li et al., 2011)


408 framework (IDR v2.0.4.2). IDR analysis was performed for all pairwise combinations of


409 biological replicates using the corresponding MACS2 narrowPeak files, with peaks ranked


410 according to their p-value. Reproducible peaks were selected using a local IDR threshold of


411 &lt;0.05. Peaks passing this IDR reproducibility threshold were retained from the pairwise


412 comparisons and combined to generate a final consensus set of high-confidence


413 reproducible binding sites. Coordinates of the consensus peaks were used to annotate


12


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


414 target regions with genic features using the package “ChIPseeker”(Yu et al., 2015) in R


415 Version (4.3.3)(R Core Team, 2020).


416 Overlap with TE superfamilies from the Araport11 annotation


417 (https://doi.org/10.6084/m9.figshare.30383395) as well as with ChIP-seq PHE1-GFP peak


418 (Batista et al., 2019) was determined using Bedtools intersect (Version 2.30)(Quinlan and


419 Hall, 2010) using the -u option.


420 PHE1 DNA-binding motifs were identified from 994 _PHE1::PHE1–GFP_ CUT&amp;Tag peak


421 regions associated with genes using HOMER’s findMotifsGenome.pl (Version 5.1) (Heinz et


422 al., 2010) function, using the default settings. To determine possible allele-specificity we


423 performed CUT&amp;Tag against PHE1-GFP from siliques at 4DAP in three independent


424 replicates. Crossing were done using either Ler-0 wild type or _pistillata_ maternal plants to


425 generate the needed input of about 250 mg of siliques/replicate. All plants were pollinated


426 with different individuals of the afore described Col-0 _PHE1::PHE1-GFP_ line. CUT&amp;Tag


427 procedure was performed as described above, and subsequent sequencing of libraries was


428 performed in 150 bp paired-end mode to increase the coverage of possible polymorphisms.


429 Trimming of reads hereafter was done as described before. For allele specific analysis


430 samples were mapped against a genome previously masked for SNPs between Col-0


431 (TAIR10) and Ler-0 using Bowtie2 as described above. The same subsequent quality

432 filtering steps were applied as described above. For visualization and to increase the


433 sensitivity of detecting parentally biased binding events, allele-specific replicates were


434 downsampled to match the fragment count of the replicate with the fewest fragments and


435 subsequently merged. Allele discrimination was done with ‘SNPsplit’ (Version 0.6.0) using


436 the –paired parameter (Krueger and Andrews, 2016). For subsequent analysis only reads


437 specific to either one of the two genomes were used further, non-assigned of ambiguously

438 assigned reads were discarded. Reads in within the genomic coordinates of the afore


439 defined 994 target regions associated with genes were counted using ‘featureCounts’ from


440 the Subread package (Version 2.0.6) (Liao et al., 2014) with the options “-T 8 -p -O -s 0”. To


441 avoid low coverage allele-specific reads we only considered those with &gt;= 5 informative


442 reads (reads that span a SNP) for each of the targets. Maternal ratio for those regions were


443 calculated by Ler-reads/(Ler-reads + Col-reads). Cumulative CPM was calculated by the


444 number of counts for a certain allele divided by counts from both allele i.e.(Col-reads/(Col

445 reads+Ler-reads)*1000000.


446 Publicly available mRNA-seq data was reanalyzed by downloading relevant samples from


447 the corresponding archives. Raw reads were trimmed using Trim Galore (Version 0.6.10)


448 (Krueger, 2015) with the options ‘--fastqc --clip_R1 5 --three_prime_clip_R1 20’. Trimmed


13


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


449 reads were mapped against the TAIR10 genome or for allele-specific analysis against a Col

450 Ler masked version of the TAIR10 genome using Hisat2 (Version 2.2.1)(Kim et al., 2019)


451 with the ‘–dta’ parameter set. Reads were counted using ‘featureCounts’ (Version 2.0.6)


452 (Liao et al., 2014) with the options ‘-T 10 -t exon -g gene_id’. Downstream expression


453 analysis was performed with DESeq2 (Love et al., 2014). Allele-specific analysis was done


454 as described above for CUT&amp;Tag derived fragments using &gt;5 informative reads/gene.


455 Statistical analysis was conducted using R (Version 4.3.3. and version 4.4.2) (R Core Team,


456 2020) as indicate per plot and plots were generated using R base or the ‘ggplot2’ package


457 (Wickham, 2009). Final arrangement of Figures was done with Affinity Designer 2 (Version


458 2.1.1; Serif Europe).


459 **Data availability**


460 CUT&amp;Tag data generated for this study was uploaded to Gene Expression Omnibus


461 (https://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE355283. Allele-specific


462 transcriptome data from 4DAP was taken from (Moreno-Romero et al., 2019). Allele-specific


463 expression data in _fie_ and wild type background and wild type from (Hsieh et al., 2011)


464 accession number GSE24644. Allele-specific H3K27me3 and DNA methylation data was


465 reanalyzed from (Del Toro-De Leon et al., 2024) available under GEO accession number


466 GSE260822.


467 **Acknowledgements**


468 This work was financially supported by the Max Planck Society. Writing and coding for this


469 manuscript was assisted provided by an AI-powered chat service, hosted by the Göttingen


470 University's Gesellschaft für wissenschaftliche Datenverarbeitung (GWDG). ChatGPT


471 (OpenAI, GPT-6) was used to assist with manuscript language editing and the development


472 and debugging of computational analysis scripts. The authors independently verified the


473 resulting analyses and take full responsibility for the content of the manuscript.


**References**


DILKES, B. P., SPIELMAN, M., WEIZBAUER, R., WATSON, B., BURKART-WACO, D., SCOTT, R. J. &amp;

COMAI, L. 2008. The Maternally Expressed WRKY Transcription Factor Controls Lethality in
Interploidy Crosses of Arabidopsis. Plos Biology, 6, 2707–2720.
SCOTT, R. J., SPIELMAN, M., BAILEY, J. &amp; DICKINSON, H. G. 1998. Parent-of-origin effects on seed

development in. Development, 125, 3329–3341.
GEHRING, M., MISSIRIAN, V. &amp; HENIKOFF, S. 2011. Genomic Analysis of Parent-of-Origin Allelic

Expression in Seeds. Plos One, 6.
HSIEH, T. F., SHIN, J. Y., UZAWA, R., SILVA, P., COHEN, S., BAUER, M. J., HASHIMOTO, M., KIRKBRIDE,

R. C., HARADA, J. J., ZILBERMAN, D. &amp; FISCHER, R. L. 2011. Regulation of imprinted gene


14


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


expression in Arabidopsis endosperm. Proceedings of the National Academy of Sciences of
the United States of America, 108, 1755–1762.
BATISTA, R. A. &amp; KÖHLER, C. 2020. Genomic imprinting in plants-revisiting existing models. Genes

Dev, 34, 24–36.
GEHRING, M., BUBB, K. L. &amp; HENIKOFF, S. 2009. Extensive Demethylation of Repetitive Elements

During Seed Development Underlies Gene Imprinting. Science, 324, 1447–1451.
PIGNATTA, D., ERDMANN, R. M., SCHEER, E., PICARD, C. L., BELL, G. W. &amp; GEHRING, M. 2014. Natural

epigenetic polymorphisms lead to intraspecific variation in Arabidopsis gene imprinting.
Elife, 3.
MORENO-ROMERO, J., JIANG, H., SANTOS-GONZALEZ, J. &amp; KOHLER, C. 2016. Parental epigenetic

asymmetry of PRC2-mediated histone modifications in the Arabidopsis endosperm. EMBO J,
35, 1298–311.
MOZGOVA, I., KÖHLER, C. &amp; HENNIG, L. 2015. Keeping the gate closed: functions of the polycomb

repressive complex PRC2 in development. Plant Journal, 83, 121–132.
ZHANG, M., XIE, S. J., DONG, X. M., ZHAO, X., ZENG, B., CHEN, J., LI, H., YANG, W. L., ZHAO, H. N.,

WANG, G. K., CHEN, Z. L., SUN, S. L., HAUCK, A., JIN, W. W. &amp; LAI, J. S. 2014. Genome-wide
high resolution parental-specific DNA and histone methylation maps uncover patterns of
imprinting regulation in maize. Genome Research, 24, 167–176.

KAYA-OKUR, H. S., WU, S. J., CODOMO, C. A., PLEDGERS, E. S., BRYSON, T. D., HENIKOFF, J. G.,

AHMAD, K. &amp; HENIKOFF, S. 2019. CUT&amp;Tag for efficient epigenomic profiling of small
samples and single cells. Nature Communications, 10.
WEINHOFER, I., HEHENBERGER, E., ROSZAK, P., HENNIG, L. &amp; KÖHLER, C. 2010. H3K27me3 Profiling

of the Endosperm Implies Exclusion of Polycomb Group Protein Targeting by DNA
Methylation. Plos Genetics, 6.
BATISTA, R. A., MORENO-ROMERO, J., QIU, Y., VAN BOVEN, J., SANTOS-GONZALEZ, J., FIGUEIREDO,

D. D. &amp; KÖHLER, C. 2019. The MADS-box transcription factor PHERES1 controls imprinting in
the endosperm by binding to domesticated transposons. Elife, 8.
ABBASOVA, L., URBANAVICIUTE, P., HU, D., ISMAIL, J. N., SCHILDER, B. M., NOTT, A., SKENE, N. G. &amp;

MARZI, S. J. 2025. CUT&amp;Tag recovers up to half of ENCODE ChIP-seq histone acetylation
peaks. Nat Commun, 16, 2993.
YU, C. P., LIN, J. J. &amp; LI, W. H. 2016. Positional distribution of transcription factor binding sites in

Arabidopsis thaliana. Sci Rep, 6, 25164.
HEYNDRICKX, K. S., VAN DE VELDE, J., WANG, C., WEIGEL, D. &amp; VANDEPOELE, K. 2014. A functional

and evolutionary perspective on transcription factor binding in Arabidopsis thaliana. Plant
Cell, 26, 3894–910.
O'MALLEY, R. C., HUANG, S. C., SONG, L., LEWSEY, M. G., BARTLETT, A., NERY, J. R., GALLI, M.,

GALLAVOTTI, A. &amp; ECKER, J. R. 2016. Cistrome and Epicistrome Features Shape the
Regulatory DNA Landscape. Cell, 166, 1598.
DEL TORO-DE LEON, G., VAN BOVEN, J., SANTOS-GONZALEZ, J., JIAO, W. B., PENG, H.,

SCHNEEBERGER, K. &amp; KOHLER, C. 2024. Epigenetic and transcriptional consequences in the
endosperm of chemically induced transposon mobilization in Arabidopsis. Nucleic Acids Res.
MORENO-ROMERO, J., DEL TORO-DE LEÓN, G., YADAV, V. K., SANTOS-GONZÁLEZ, J. &amp; KÖHLER, C.

2019. Epigenetic signatures associated with imprinted paternally expressed genes in the
Arabidopsis endosperm. Genome Biology, 20.
GROSSNIKLAUS, U., VIELLE-CALZADA, J. P., HOEPPNER, M. A. &amp; GAGLIANO, W. B. 1998. Maternal

control of embryogenesis by MEDEA, a Polycomb Group Gene in Arabidopsis. Science, 280,

446–450.
KÖHLER, C., HENNIG, L., BOUVERET, R., GHEYSELINCK, J., GROSSNIKLAUS, U. &amp; GRUISSEM, W. 2003.

Arabidopsis MSI1 is a component of the MEA/FIE Polycomb group complex and required for
seed development. EMBO J, 22, 4804–14.


15


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


BARTOSOVIC, M., KABBE, M. &amp; CASTELO-BRANCO, G. 2021. Single-cell CUT&amp;Tag profiles histone

modifications and transcription factors in complex tissues. Nature Biotechnology, 39, 825–

835.

ZHENG, X. Y. &amp; GEHRING, M. 2019. Low-input chromatin profiling in Arabidopsis endosperm using

CUT&amp;RUN. Plant Reproduction, 32, 63–75.
TAO, X. Y., FENG, S. L., ZHAO, T. &amp; GUAN, X. Y. 2020. Efficient chromatin profiling of H3K4me3

modification in cotton using CUT&amp;Tag. Plant Methods, 16.
KAYA-OKUR, H. S., JANSSENS, D. H., HENIKOFF, J. G., AHMAD, K. &amp; HENIKOFF, S. 2020. Efficient low
cost chromatin profiling with CUT&amp;Tag. Nat Protoc, 15, 3264–3283.
GEHRING, M. 2013. Genomic imprinting: insights from plants. Annu Rev Genet, 47, 187–208.
RODRIGUES, J. A. &amp; ZILBERMAN, D. 2015. Evolution and function of genomic imprinting in plants.

Genes Dev, 29, 2517–31.
KÖHLER, C., PAGE, D. R., GAGLIARDINI, V. &amp; GROSSNIKLAUS, U. 2005. The Arabidopsis thaliana

MEDEA Polycomb group protein controls expression of PHERES1 by parental imprinting. Nat
Genet, 37, 28–30.
BEMER, M., HEIJMANS, K., AIROLDI, C., DAVIES, B. &amp; ANGENENT, G. C. 2010. An atlas of type I MADS

box gene expression during female gametophyte and seed development in Arabidopsis.
Plant Physiol, 154, 287–300.

DE FOLTER, S., IMMINK, R. G., KIEFFER, M., PARENICOVA, L., HENZ, S. R., WEIGEL, D., BUSSCHER, M.,

KOOIKER, M., COLOMBO, L., KATER, M. M., DAVIES, B. &amp; ANGENENT, G. C. 2005.
Comprehensive interaction map of the Arabidopsis MADS Box transcription factors. Plant
Cell, 17, 1424–33.

JADHAV, U., MANIERI, E., NALAPAREDDY, K., MADHA, S., CHAKRABARTI, S., WUCHERPFENNIG, K.,

BAREFOOT, M. &amp; SHIVDASANI, R. A. 2020. Replicational Dilution of H3K27me3 in
Mammalian Cells and the Role of Poised Promoters. Mol Cell, 78, 141–151 e5.
MORENO-ROMERO, J., SANTOS-GONZALEZ, J., HENNIG, L. &amp; KOHLER, C. 2017. Applying the INTACT

method to purify endosperm nuclei and to generate parental-specific epigenome profiles.
Nat Protoc, 12, 238–254.
LANGMEAD, B. &amp; SALZBERG, S. L. 2012. Fast gapped-read alignment with Bowtie 2. Nat Methods, 9,


357–9.

LI, H., HANDSAKER, B., WYSOKER, A., FENNELL, T., RUAN, J., HOMER, N., MARTH, G., ABECASIS, G.,

DURBIN, R. &amp; PROC, G. P. D. 2009. The Sequence Alignment/Map format and SAMtools.
Bioinformatics, 25, 2078–2079.
RAMÍREZ, F., DÜNDAR, F., DIEHL, S., GRÜNING, B. A. &amp; MANKE, T. 2014. deepTools: a flexible

platform for exploring deep-sequencing data. Nucleic Acids Research, 42, W187–W191.
PYTHON SOFTWARE, F. 2026. Python Programming Language. Python Software Foundation.
QUINLAN, A. R. &amp; HALL, I. M. 2010. BEDTools: a flexible suite of utilities for comparing genomic

features. Bioinformatics, 26, 841–842.
ZHANG, Y., LIU, T., MEYER, C. A., EECKHOUTE, J., JOHNSON, D. S., BERNSTEIN, B. E., NUSBAUM, C.,

MYERS, R. M., BROWN, M., LI, W. &amp; LIU, X. S. 2008. Model-based analysis of ChIP-Seq
(MACS). Genome Biol, 9, R137.
LI, Q., BROWN, J. B., HUANG, H. &amp; BICKEL, P. J. 2011. Measuring reproducibility of high-throughput

experiments.
YU, G. C., WANG, L. G. &amp; HE, Q. Y. 2015. ChIPseeker: an R/Bioconductor package for ChIP peak

annotation, comparison and visualization. Bioinformatics, 31, 2382–2383.
R CORE TEAM, R. 2020. R: A language and environment for statistical computing. Citeseer.

HEINZ, S., BENNER, C., SPANN, N., BERTOLINO, E., LIN, Y. C., LASLO, P., CHENG, J. X., MURRE, C.,

SINGH, H. &amp; GLASS, C. K. 2010. Simple combinations of lineage-determining transcription
factors prime cis-regulatory elements required for macrophage and B cell identities. Mol
Cell, 38, 576–89.


16


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


KRUEGER, F. &amp; ANDREWS, S. R. 2016. SNPsplit: Allele-specific splitting of alignments between

genomes with known SNP genotypes. F1000Res, 5, 1479.
LIAO, Y., SMYTH, G. K. &amp; SHI, W. 2014. featureCounts: an efficient general purpose program for

assigning sequence reads to genomic features. Bioinformatics, 30, 923–930.
KRUEGER, F. 2015. Trim Galore!: A wrapper around Cutadapt and FastQC to consistently apply

adapter and quality trimming to FastQ files, with extra functionality for RRBS data.

Babraham Institute.

KIM, D., PAGGI, J. M., PARK, C., BENNETT, C. &amp; SALZBERG, S. L. 2019. Graph-based genome

alignment and genotyping with HISAT2 and HISAT-genotype. Nat Biotechnol, 37, 907–915.
LOVE, M. I., HUBER, W. &amp; ANDERS, S. 2014. Moderated estimation of fold change and dispersion for

RNA-seq data with DESeq2. Genome Biol, 15, 550.
WICKHAM, H. 2009. ggplot2: Elegant Graphics for Data Analysis. Ggplot2: Elegant Graphics for Data

Analysis, 1–212.


17


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Figure 1: CUT&amp;Tag reveals known PHE1 binding sites.** A) Browsershots of replicates of known
PHE1 targets as identified by Batista et al 2019. Tracks are shown from CUT&amp;Tag-seq in Col-0
_PHE1::PHE1-GFP_ containing plants and Col-0 transgene-free wild-type plants and ChIP-Seq tracks


18


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


from Batista et al. 2019. Shown are three independent replicates obtained from GFP (green) and
background (grey, H3 = dark grey, ChIP input = light grey) antibody samples. Reads where
normalized to reads per genomic content (RPGC), per track range indicated in brackets. B) Number
of annotated peaks in CUT&amp;Tag and ChIP-seq data. C) Overlap of uniquely annotated genes
between CUT&amp;Tag and ChIP-seq targets. Statistical significance was tested via hypergeometric test.
D) Piechart of CUT&amp;Tag peak feature annotations per ChIPseeker using a -1500 bp + 500 bp
promoter relative to TSS. E) Piedonut chart depicting the relative overlap of PHE1 CUT&amp;Tag targets
and transposable elements (TEs). F) Enrichment of MADS motives in uniquely annotated CUT&amp;Tag
peaks as determined by Homers findMotifsGenome.pl.


19


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Figure 2: Allele-specific CUT&amp;Tag reveals biallelic binding of PHE1 to the majority of its**
**targets** . A) Venn diagram showing the overlap of PHE1 targets, with "core PEGs" and "core MEGs"
as defined in Del Toro de León et al. 2024. Statistically significant overlaps were determined via
hypergeometric test. B) Maternal ratio of PHE1 binding on targeted core imprinted genes established
via CUT&amp;Tag on PHE1 target sides with sufficient numbers of informative reads (&gt;5 reads/SNP). C)
Browsershot of PHE1 PEG targets showing biallelic binding. Tracks were normalized to CPM
normalized to the combined allele-specific library size (cumulative CPM, see material and methods).
D) Histogram of maternal ratios from all PHE1 target sites with sufficient informative reads (&gt;5
reads/SNP). Targets with allele-bias are marked by color (maternal ratio &lt; 0.5, blue); for maternal
allele (maternal ratio &gt; 0.9, red); or bialleic PHE1 binding (maternal ratio &gt;=0.5 &amp; &lt;= 0.9, white). E)
Allele-specific log2 of H3K27me3 / H3 on PEGs targeted by PHE1 in endosperm nuclei 4 DAP in F1
Ler x Col. F) Allele-specific log2 of H3K27me3 / H3 on MEGs targeted by PHE1 in endosperm nuclei
4 DAP in F1 Ler x Col. Statistical difference was determined with two-sided Wilcoxon signed-rank test
p &lt; 0.05. Endosperm-specific H3K27me3 and H3 data are from Del Toro de León et al. 2024.


20


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Figure 3: PHE1 targets are enriched for H3K27me3 on gene bodies and promoter regions on**
**the maternal alleles** . A) Expression of PHE1 targets in comparison to all expressed genes (RPKM
&gt;5) and all annotated genes. Expression data are from INTACT purified endosperm nuclei at 4 DAP in
F1 Ler × Col seeds form Moreno Romero et al. 2019. B) Log2 normalized levels of total
H3K27me3/H3 levels at PHE1 target genes, at all expressed genes in the endosperm in F1 Ler x Col
crosses. C) Metagene plot over PHE1, targets and random genes in reciprocal crosses. D) Kmeans
clustering and heatmap over all PHE1 target sides in a window of 1 kb up- and downstream to midpoint of annotated peaks. E) Venn diagram depicting the overlap of PHE1 PEG target genes with
identified clusters. F) Parental-specific H3K27me3 in clusters as defined before. Statistical difference
was determined with two-sided Wilcoxon signed-rank test p &lt; 0.05.G) Maternal ratio of PHE1 target
gene expression per cluster at 4 DAP. Expression data from Moreno-Romero et al. 2019. H) Maternal
ratio of PHE1 binding by clusters. Wilcoxon rank-sum test was applied in panels A,B, G and H.
H3K27me3 data are from Del Toro-De León et al. 2024.


21


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Figure 4: PHE1 targets are repressed by H3K27me3** . A) Violin plots depicting the allele-specific
PHE1 binding in Ler x _PHE1::PHE1-GFP_ (Col) in afore established cluster 1 (A) cluster 2 (B) or
cluster 3 (C) at PHE1 target sides. Dashed lines represent the theoretical parental ratio of 2m:1p,


22


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


0.66D) Volcano plot of PHE1 targets genes identified by CUT&amp;Tag in FIE deficient seeds. E)
Rainclould plot of allele ratios from PHE1 target genes meeting allele criteria (&gt;5 reads per SNP, see
material and methods) in F1 endosperm of Ler x Col and F1 _fie_ x Col. Mean maternal allele ratios
were compared between genotypes using a two-sided Wilcoxon signed-rank test. F) Differential gene
regulation in H3K27me3 clusters as defined earlier.G) Log2-fold change between _fie_ and wt crosses
of PEGs, MEGs (PHE1 targets and all) as well as all target genes. Transcriptome data are from Hsieh
et al. 2011. Two-sided Wilcoxon signed-rank tests were used for panels A,B,C,E and two-sided
Wilcoxon rank-sum tests were applied in panels F and G; p &lt; 0.05. H) Modell of PHE1 binding at
PEGs in wild- type and H3K27me3 compromised conditions. Allele-specific expression is indicated by
arrows, thicker arrows indicate stronger expression. Paternal alleles are marked in blue, maternal in
red. Repressive H3K27me3 are illustrated by red upside-down lollipops.


23


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Supplementary Figure S1:** A) Heatmap of Spearman correlation of from CUT&amp;Tag samples used for
ChIP validation. Anti-H3 samples are labeled with “H3” anti-GFP samples with “GFP”, replicates are
indicated by “Rep”. B) Schematic representation of irreproducible discovery rate (IDR) workflow. For
this mapped GFP samples were down-sampled to the read number of the smallest replicate and


24


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


subsequently peaks were called for each replicate separately using MACS2 with a relaxed threshold
(p &lt; 0.001). All three pairwise replicate comparisons were subjected to IDR analysis, and peaks with a
global IDR ≤ 0.05 were retained. Overlapping reproducible regions were subsequently merged to
obtain the final consensus peak set. Numbers in brackets show the number of peaks meeting the
described thresholds per set. C) Scatterplot of Spearman correlation within the genomic coordinates
of reproducible peaks (n= 1090). D) Number of peaks detected in Col-0 _PHE1::PHE1-GFP_ samples
and transgene-free wild type samples, as well as the reciprocal overlap calculated by bedtools
intersect. Eight of the ten background peaks overlapped six distinct PHE1 peaks, as multiple
background peaks could overlap the same PHE1 region. E) Fraction of fragments in peaks (FFiPs) of
the three replicates form _PHE1::PHE1-GFP_ containing samples and the two transgene-free Col-0
replicates. FFiP was calculated over al 1090 peaks regions as defined above. F) Feature annotation
of the ten peaks from transgene-free samples done with ChIPseeker. Number of peaks in promoter
are plotted.


25


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Supplementary Figure S2:** Maternal ratio distribution of PHE1-GFP binding in defined promoter
windows of -1500bp to +500 bp relative to genes. A) Histogram of maternal ratios promoters of PHE1
target genes. B) Histogram of maternal ratios in all Araport11 annotated promoters in the afore
defined windows. Blue represents all promoters with a bias for paternal PHE1 binding (&lt; 0.5), red
maternal PHE1 binding (&gt;0.9). Dashed lines represent the theoretical allele distribution of 0.66 in the
endosperm. Shown are only promoters with &gt;= 5 informative reads, n, mean and median as indicated.


26


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Supplementary Figure S3:** Average DNA methylation per context in PHE1 target sites. DNA
methylation is shown in CG (A), CHG (B) and CHH (C) whereas H represents any nucleotide other
than G. Analysis was performed over all PHE1 targeted PEGs (n=20) or MEGs (n=6) respectively.
Orange represents the maternal allele ( ♀ ), blue the paternal alleles ( ♂ ). Statistical analysis between
same parental alleles between MEGs and PEGs was performed using unpaired Wilcoxon rank-sum
test, whereas different alleles from MEGs, or PEGs were tested via paired Wilcoxon signed-rank test.
Significant p-values are plotted above. DNA methylation data from combined Ler × Col replicates from
Del Toro-De León et al. 2024.


27


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.19.767895; this version posted November 20, 2026. The copyright holder for this preprint&#93;(https://doi.org/10.64898/2026.11.19.767895)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
&#91;available under a CC-BY-NC-ND 4.0 International license.&#93;(http://creativecommons.org/licenses/by-nc-nd/4.0/)


**Supplementary Figure S4:** A) Parental specific H3K27me3 in clusters as defined before, reciprocal
to Fig 3E. H3K27me3 data are from Del Toro-De León et al. 2024. Expression data is 4 DAP INTACT
purified endosperm from F1 Ler × Col crosses from Moreno-Romero et al. 2019. B) Maternal ratio of
expression. C) Total expression of target genes in RPKM. Statistical significance was determined
using unpaired Wilcoxon rank-sum test.


28


