Transposable element‐mediated DNA methylation of the NAC20 and NAC26 promoters led to a maternal effect on grain filling

TL;DR

This study shows that transposable elements cause DNA methylation in rice genes NAC20 and NAC26, leading to a maternal effect on grain filling in high-latitude Japonica rice.

Contribution

The novel finding is that TE insertions and DNA methylation mediate maternal expression of NAC20/26, affecting grain filling in specific rice varieties.

Findings

NAC20/26 show allele-specific maternal expression linked to transposable element insertions in Japonica rice.

TE deletions reduce DNA methylation and restore biallelic expression of NAC20/26.

High-latitude Japonica varieties carry TEs associated with the maternal effect on grain filling.

Abstract

Parent‐of‐origin effects are usually caused by selective expression of maternal or paternal alleles. Although genome‐wide studies suggest that imprinted gene expression occurs primarily in the endosperm in plants, detailed studies of allele‐specific gene expression and its associations with parent‐of‐origin phenotypes are scarce. NAC20 and NAC26 (NAC20/26 hereafter), a pair of tightly linked NAC‐family transcription factors, redundantly regulate grain filling and albumin accumulation in rice endosperm. Here, we show that NAC20/26 exhibited allele‐specific maternal expression, and the floury endosperm phenotype of the nac20/26 double mutant was inherited with a maternal effect. Further studies showed that the imprinted NAC20/26 expression and floury endosperm phenotype with a maternal effect are associated with insertions of two TEs in NAC20/26 of two Japonica rice varieties, but not in two Indica ones examined. The maternal NAC20/26 expression was associated with elevated DNA methylation in their paternal DMRs, and deletions of those TEs by gene editing led to decreased methylation in these DMRs, and biallelic NAC20/26 expression. Geographical analyses showed that Japonica varieties with high‐latitude origins examined carried these TEs. These results establish that TE‐mediated DNA methylation lead to grain filling with a maternal effect in high‐latitude Japonica rice varieties, which may associate with northward expansion of rice during domestication.

The rice transcription factor genes NAC20 and NAC26 regulate grain filling and show allele‐specific maternal expression due to transposable element insertions and paternal hypermethylation in their 5' upstream regions. Geographical analyses showed that japonica varieties that originated at high latitudes carried these transposable elements.

Figures

NAC20/26 regulate rice grain filling with an allele‐specific maternally effect (A) Reciprocal crosses to show that the floury endosperm phenotype in nac20/26 ZH11 was inherited with a maternal effect. Note that the phenotype was observed in self‐pollinated nac20/26 ZH11, or when nac20/26 ZH11 was pollinated by ZH11. Scale bar = 1 mm. (B) Allelic expression analyses of NAC20 and NAC26 in reciprocal crosses between ZH11 and nac20/26 ZH11, when ZH11 was used either as female (left) or male (right), to show that NAC20 and NAC26 maternal alleles are predominantly expressed. Clones sequenced ≥ 30. (C) Detections of parental NAC20 and NAC26 transcripts (distinguished by one SNP between NJ6 and ZH11) in endosperm at 9 DAP after reciprocal crosses between ZH11 and NJ6. Note that both NAC20 and NAC26 showed maternal expression when ZH11 was used as the pollen donor (NJ6 × ZH11), but biallelic expression when NJ6 was used as the pollen donor (ZH11 × NJ6). Red dashed boxes indicate SNPs. (D) Reciprocal crosses to show that the floury endosperm phenotype of nac20/26 was only observed after self‐pollination and when nac20/26 NJ6 was pollinated by ZH11 (nac20/26 NJ6 × ZH11), not in NJ6 × nac20/26 NJ6 and nac20/26 NJ6 × NJ6, nor nac20/26 ZH11 × NJ6. n > 20; Scale bar = 1 mm. (E) Endosperm phenotypes in grains produced after reciprocal crosses between NIP and nac20/26 NIP (upper), and between 9311 and nac20/26 9311 (lower). Note the maternal effect observed in nac20/26 NIP. Scale bar = 1 mm. (F) Transcripts of parental NAC20 and NAC26 alleles (distinguished by one SNP between two parents) in endosperm after reciprocal crosses between NIP and NJ6 (left), and between ZH11 and 9311 (right). Note the biallelic expression observed when NJ6 and 9311 were used as pollen donors (lower), not NIP and ZH11 as pollen donors (middle).

Differential DNA methylation is associated with allele‐specific expression of NAC20 and NAC26 in endosperm (A) Bisulfite sequencing to show that NAC20 and NAC26 genomic DNAs, extracted from endosperm at 9 DAP, showed decreased hypomethylation (red lines) when compared to those in embryos. Green squares represent transposable elements. P1~P4 are primers for the detection of transposable elements (P1 and P2 for NAC20; P3 and P4 for NAC26). (B) Local bisulfite sequencing showed that DMRs of NAC20 (−494 to −205 bp, left) and NAC26 (−475 to −186 bp, right) showed lower levels of methylation in endosperm(EN) examined at 9 and 21 DAP in ZH11 for all three methylation contents (CG, CH, and CHH) when compared with those in embryos (EM). Numbers of clones sequenced ≥ 10. (C) Local bisulfite sequencing of DMRs in genomic DNA extracted from 9 DAP endosperm after the reciprocal crosses between ZH11 and NJ6. Note that when the ZH11 allele was sequenced, higher methylation was observed when ZH11 was used as a pollen donor, but when NJ6 alleles were sequenced, low levels of methylation were observed irrespective of whether NJ6 was used as male or female. Numbers of clones sequenced ≥ 20. (D) Local bisulfite sequencing in ZH11, NJ6, NIP, and 9311 endosperm examined at 9 DAP to show that DMRs in two Japonica varieties (ZH11 and NIP) were hypermethylated (CG especially), while two Indica varieties (NJ6 and 9311) were hypomethylated. Numbers of clones sequenced ≥ 10.

TE‐mediated DNA methylation is necessary and sufficient for maternal imprinted NAC20 and NAC26 expression in rice endosperm (A) PCR amplifications to show the presence of TEs (with longer fragments) in the NAC20 and NAC26 promoters of ZH11 and NJ6, but not in NIP and 9311 (shorter fragments). (B) PCR amplification using primers in (Figure S3), followed by gel electrophoresis, to show that in the NAC20/26 ΔTE plant, TEs in both NAC20 and NAC26 were removed. M, molecular weight marker. (C) Local bisulfite sequencing to show greatly decreased CG, CHG, and CHH methylation in both endosperm and embryo in NAC20‐DMR and NAC26‐DMR of NAC20/26 △TE plants, when compared with those in the wild type (ZH11). (D) Photo showing that when NAC20/26 ΔTE was used as a pollen donor to pollinate nac20/26 ZH11 (right), no floury endosperm was observed. nac20/26 ZH11 pollinated by ZH11 was used as a control (left). Scale bar = 1 mm. (E) Allelic expression analyses showing that in endosperm of nac20/26 ZH11 × NAC20/26 ΔTE , both maternal (M, nac20/26 ZH11) and paternal NAC20 and NAC26 alleles (P, NAC20/26 ΔTE ) were expressed. Numbers of clones sequenced ≥ 30. (F) A model for TE‐mediated DNA methylation in paternal alleles is necessary and sufficient to maintain the maternally imprinted expression of NAC20/26 in the endosperm and to regulate grain filling with a maternal effect.

Distribution of TEs in the NAC20 and NAC26 promoters is associated with differentiation of Japonica and Indica subspecies and northward expansion of rice cultivation (A–B) Statistics for distributions of TEs located in NAC20 and NAC26 promoters in the 174 rice varieties according to the Indica and Japonica subgroup (A), and planting area in China (B). No TEs in the NAC20/26: The TE ORSgTEMT00100531 was not detected in the 5' upstream region of NAC20 and the retro‐TEs (ORSiTEMT00T00140) were not detected in the 5' upstream region of NAC26 (bule solid circle); TE in the NAC20 or NAC26: The TE ORSgTEMT00100531 was detected in the 5' upstream region of NAC20 or the retro‐TEs ORSiTEMT00T00140 were detected in the 5' upstream region of NAC26 (orange solid circle); and TEs in the NAC20/26: The TE ORSgTEMT00100531 was detected in the 5' upstream region of NAC20 and the retro‐TEs ORSiTEMT00T00140 were detected in the 5' upstream region of NAC26 (red solid circle).

Tables

Table 1. Genetic analyses in nac20/26 ZH11

Female × malea	Endosperm phenotypes		Total	Expected segregation ratio	P values	χ2 testsb
	Wild type	Floury
ZH11 × ZH11	30	0	30	‐	‐	NA
ZH11 × nac20/26 ZH11	30	0	30	‐	‐	NA
nac20;nac26 ZH11 × nac20;nac26 ZH11	0	30	30	‐	‐	NA
nac20;nac26 ZH11 × ZH11	0	30	30	‐	‐	NA
nac20/NAC20;nac26 ZH11 × ZH1	16	21	37	1:1	0.411 > 0.05	NS
nac20;nac26/NAC26 ZH11 × ZH11	22	18	40	1:1	0.527 > 0.05	NS

Table 2. Segregation analyses of nac20 and nac26 mutants in different backgrounds

Lines	Endosperm phenotypes		Total	Expected segregation ratio	P values	χ2 testsa
	Wild type	Floury
nac20 ZH11	124	0	124	‐	‐	NA
nac26 ZH11	143	0	143	‐	‐	NA
nac20/26 ZH11	0	147	147	‐	‐	NA
nac20;nac26/NAC26 ZH11	115	103	218	3:1	< 0.0001	Yes
				1:1	= 0.416 > 0.05	NS
nac20/NAC20;nac26 ZH11	79	98	177	3:1	< 0.0001	Yes
				1:1	= 0.151 > 0.05	NS
nac20/NAC20; nac26/NAC26 ZH11	124	56	180	15:1	< 0.0001	Yes
nac20/NAC20; nac26/NAC26 NJ6	334	23	357	15:1	= 0.88 > 0.05	NS

Full text

INTRODUCTION

Traits with a parent‐of‐origin effect are usually conferred by maternally or paternally imprinted gene expression, which is associated with asymmetric epigenetic modifications such as DNA methylation and histone modifications between maternal and paternal alleles (MacDonald, 2012; Satyaki and Gehring, 2017). Imprinting genes have been found predominantly in placentae of mammals (Bartolomei and Hanna, 2020) and endosperm of plants (Köhler and Grossniklaus, 2005; Gehring, 2013). Both tissues are post‐fertilization subsidiary structures, with a primary function of nourishing embryos. According to the parental conflict theory, imprinted genes—paternally expressed genes (PEGs) act as enhancers and maternally expressed genes (MEGs) as inhibitors for progeny growth—are believed to mediate maternal resource allocation among offspring (Haig and Westoby, 1989; Haig, 2004). One example supporting this theory is that reduced expression of insulin‐like growth factor 2 (Igf2), a paternally expressed gene in mice, led to a 40% reduction in fetal growth (DeChiara et al., 1990). Disruption of the Igf2 imprinting region led to biallelic Igf2 transcription, resulting in fetal and placental overgrowths (MLeighton et al., 1995).

Endosperm is a triploid tissue derived from fertilized central cell, and is the major site of imprinting genes identified in plants (Lopes and Larkins, 1993; Li and Berger, 2012). The seed‐specific PRC2 family gene MEDEA (MEA) was identified as the first maternally expressed gene (MEG) that suppresses endosperm development before fertilization (Grossniklaus et al., 1998; Kiyosue et al., 1999; Luo et al., 1999). Similarly, another gene in the PRC2 complex, Fertilization‐Independent Seed 2 (FIS2), is also a MEG, and the fis2 mutant showed precocious endosperm development as well (Luo et al., 1999), suggesting a critical role of the PRC2 complex in repressing endosperm development before fertilization, which in a way supports the parental conflict theory. However, maternally expressed MEG1 in maize that encodes a cysteine‐rich small peptide regulates the formation of a basal endosperm transfer layer (BETL), and the knockdown of MEG1 expression resulted in compromised BETL formation and reduced starch content in kernels (Costa et al., 2012), suggesting that MEG1 facilitates rather than restricts nutrient allocation to the offspring. In rice, knockouts of some MEGs led to increased amount of small starch granules, and reduced grain and embryo sizes (Yuan et al., 2017). These studies indicate that MEGs are instrumental in facilitating the development of embryos and endosperm, which in a way challenges the parental conflict theory.

In plants, DNA methylation of cytosines in CG, CHG, and CHH contexts (H = A, T, or C) commonly occurs in promoter regions, where it generally acts to repress gene transcription (Zhang et al., 2018). Genome‐wide surveys showed that the paternal genome in endosperm is hypermethylated, whereas the maternal genome is hypomethylated (Ibarra et al., 2012; Rodrigues et al., 2013; Zhang et al., 2014). Active DNA demethylation in the central cell is believed to be the origin of the maternal hypomethylation and subsequent activated expression of MEGs in endosperm (Choi et al., 2002). Transposable elements (TEs) comprise a large proportion of plant genome; most of them are negatively correlated with the fitness of the host, and plants have evolved epigenetic defense mechanisms to silence their expression (Slotkin and Martienssen, 2007; Lisch, 2009). HOMEDOMAIN GLABROUS3 (HDG3) is an allele‐specific paternal imprinted gene in Arabidopsis, which is linked to DNA methylation at a Helitron TE sequence located upstream of the transcriptional start site (Gehring et al., 2009), and methylation of the sequence within Helitron TE is sufficient to promote HDG3 switch to imprinted expression from non‐imprinted expression (Köhler et al., 2018). FWA is a MEG in Arabidopsis endosperm that depends on DNA methylation in the FWA promoter (Kinoshita et al., 2004), which is comprised of two direct repeats contained in a sequence related to a SINE retroelement (Kinoshita et al., 2006). Nevertheless, despite the absence of the repeat structure within the FWA promoter in the Arabidopsis halleri, the FWA still displays maternal imprinted expression in the endosperm (Ecker et al., 2008). In maize, two novel MEGs have been identified, and their emergence is attributed to Mu transposon insertions occurring in the upstream regions of these genes, suggesting that TEs are associated with imprinted gene expression (Li et al., 2023). TE‐mediated gene silencing also influences expression of adjacent genes, causing spreads of DNA methylation from TEs (Ahmed et al., 2011). Thus, although genome‐wide analysis showed that the expression of a large number of imprinted gene in endosperm is evolved from TE‐mediated DNA methylation (Gehring et al., 2009; Pignatta et al., 2014; Rodrigues et al., 2021), detailed studies of imprinted genes and their functions are insufficient.

Adaptive evolution is driven by natural selection, a process where those individuals that excel in survival and reproductive success are more likely to pass on their genes to future generations, thereby shaping the genetic composition of the population over time to increase genetic variants (Casacuberta and González, 2013). TE plays a relevant role in adaptation because of their ability to generate genetic variants of great magnitude within genome, and this genetic variability leads to alterations in gene expression and regulatory networks, ultimately resulting in plant diversity (Negi et al., 2016; Bourque et al., 2018; Ramakrishnan et al., 2022). Arabidopsis is postulated to have its center of origin in the Balkan region, and expands mainly along the east–west axis (Lee et al., 2017). The variations in TEs accumulation contribute to the adaptation during the Arabidopsis expansion (Li et al., 2018; Jiang et al., 2024). Maize originated in the low‐latitude Balsas River Valley of tropical south‐western Mexico, and subsequently dispersed across the globe through a range of latitude adaptations (Piperno et al., 2009). Maize flowering time is a key agronomic trait for adaptation to latitude and altitude (Romero Navarro et al., 2017). Some loci are associated with maize flowering time that have been well characterized, and found to contribute to latitude adaptation, such as Vegetative to generative transition 1 (Vgt1) (Salvi et al., 2007; Ducrocq et al., 2008), ZmCCT9 (Huang et al., 2017), and ZmCCT10 (Hung et al., 2012). Interestingly, these loci are associated with the regulatory roles of TEs (Salvi et al., 2007; Hung et al., 2012; Huang et al., 2017). Heading Date 1 (Hd1) is the first cloned rice gene that plays a role in the photoperiodic regulation of flowering, and is homologous to CONSTANS (CO) in Arabidopsis (Yano et al., 2000). Hd1 possesses numerous natural variants and is a significant target for domestication of flowering time diversity. In high‐latitude regions, some natural nucleotide varieties in Hd1 have been selected to promote flowering (Zhang et al., 2015). γ‐ray mutagenesis resulted in the insertion of a sequence into an intron of Hd1, leading to changes in the timing of heading and flowering, and the inserted sequence is known as mPing TE (Yano et al., 2000; Jiang et al., 2003). Asian cultivated rice comprises two subspecies: Indica (Oryza sativa ssp. Indica) and Japonica (Oryza sativa ssp. Japonica). Indica rice cultivars are primarily distributed in the southern tropical and subtropical regions with low latitudes, whereas Japonica rice cultivars, having undergone nearly ten thousand years of artificial domestication and selection, have been cultivated gradually northward in temperate high‐latitude regions (Mackill and Lei, 1997). Through the analysis of expression data from 208 Asian cultivated rice samples, it was found that TE insertions are associated with changes in the expression of many genes related to rice domestication, and some of these insertions already existed in wild ancestors of rice and have undergone different selection pressures in Indica and Japonica subspecies (Castanera et al., 2023). Intriguingly, the copy number of mPing TEs varies in Japonica rice from different planting regions, with a significantly higher copy number in temperate varieties when compared to tropical ones (Jiang et al., 2003). These studies suggest that changes in gene expression caused by TE insertions are involved in domestication and adaptation of rice to high latitudes.

NAC20 and NAC26 (NAC20/26 hereafter), a pair of NAC‐family transcription factors, redundantly regulate grain filling and albumin accumulations in the endosperm of rice (Wang et al., 2020; Wu et al., 2023). In this study, we show that both NAC20/26 are allele‐specific MEGs that regulate grain filling in Zhonghua 11 (ZH11) and Nipponbare (NIP) of Japonica subspecies with a maternal effect. The allele‐specific maternal expression is associated with hypermethylation and silencing of its paternal allele. Deletions of TEs in 5′ upstream regions of NAC20/26 by genome‐editing activated expression of their paternal alleles of NAC20/26 and, consequently, losses of maternally regulated grain filling. We also showed that these TE insertions in NAC20/26 are associated with northward expansion of Japonica rice subspecies.

RESULTS

NAC20/26 regulate grain filling in ZH11 with a maternal effect

NAC20/26 double‐knockout mutants generated in ZH11 (named nac20/26 ^ZH11^ hereafter) showed floury endosperm, caused by compromised grain filling due to decreased accumulation of both starch and albumins (Wang et al., 2020; Wu et al., 2023). Constructs of pUBI:NAC20 and pUBI:NAC26 to overexpression NAC20 and NAC26, respectively, were developed and transformed into ZH11. A floury endosperm phenotype was observed in transgenic plants carrying either pUBI:NAC20 or pUBI:NAC26 (Figure S1A), and qRT‐PCR analyses showed that NAC20/26 expression were dramatically reduced in endosperm (Figure S1B), suggesting the occurrence of co‐suppressions in these transgenic plants (Napoli et al., 1990), and the expression levels of NAC20 and NAC26 are tightly associated with grain filling.

When nac20/26 ^ZH11^ was either self‐pollinated or pollinated by ZH11 (written as nac20/26 ^ZH11^ × ZH11 for female × male, and for all crosses hereafter), all grains produced showed floury endosperm (Figure 1A; Table 1). However, in ZH11 × nac20/26 ^ZH11^, all grains produced were wild type‐looking (Figure 1A; Table 1). In crosses of either nac20/NAC20;nac26 ^ZH11^ × ZH11 (NAC20 was heterozygous) or nac20;nac26/NAC26 ^ZH11^ × ZH11 (NAC26 was heterozygous), grains produced showed 1:1 segregations for normal and floury endosperm (Table 1). The same segregation ratio was also observed in grains produced from either self‐pollinated nac20/NAC20;nac26 ^ZH11^ or nac20;nac26/NAC26 ^ZH11^ (Table 2). These results together suggest that NAC20/26 redundantly regulate grain filling in ZH11 with a gametophytic maternal effect.

*NAC20/26

regulate rice grain filling with an allele‐specific maternally effect

(A) Reciprocal crosses to show that the floury endosperm phenotype in nac20/26 ZH11 was inherited with a maternal effect. Note that the phenotype was observed in self‐pollinated nac20/26 ZH11, or when nac20/26 ZH11 was pollinated by ZH11. Scale bar = 1 mm. (B) Allelic expression analyses of NAC20 and NAC26 in reciprocal crosses between ZH11 and nac20/26 ZH11, when ZH11 was used either as female (left) or male (right), to show that NAC20 and NAC26 maternal alleles are predominantly expressed. Clones sequenced ≥ 30. (C) Detections of parental NAC20 and NAC26 transcripts (distinguished by one SNP between NJ6 and ZH11) in endosperm at 9 DAP after reciprocal crosses between ZH11 and NJ6. Note that both NAC20 and NAC26 showed maternal expression when ZH11 was used as the pollen donor (NJ6 × ZH11), but biallelic expression when NJ6 was used as the pollen donor (ZH11 × NJ6). Red dashed boxes indicate SNPs. (D) Reciprocal crosses to show that the floury endosperm phenotype of nac20/26 was only observed after self‐pollination and when nac20/26

NJ6 was pollinated by ZH11 (nac20/26 NJ6 × ZH11), not in NJ6 × nac20/26 NJ6 and nac20/26 NJ6 × NJ6, nor nac20/26 ZH11 × NJ6. n > 20; Scale bar = 1 mm. (E) Endosperm phenotypes in grains produced after reciprocal crosses between NIP and nac20/26 NIP (upper), and between 9311 and nac20/26 9311 (lower). Note the maternal effect observed in nac20/26 NIP. Scale bar = 1 mm. (F) Transcripts of parental NAC20 and NAC26 alleles (distinguished by one SNP between two parents) in endosperm after reciprocal crosses between NIP and NJ6 (left), and between ZH11 and 9311 (right). Note the biallelic expression observed when NJ6 and 9311 were used as pollen donors (lower), not NIP and ZH11 as pollen donors (middle).*

Table 1: Genetic analyses in nac20/26 ZH11

NAC20 and NAC26 are allele‐specific MEGs

Traits with maternal effects are usually controlled by MEGs (Autran et al., 2005). To address if NAC20 and NAC26 are MEGs, allelic expression analyses were performed by sequencing individual clones created from PCR‐amplified NAC20 and NAC26 cDNAs prepared from endosperm excised at 9 d after pollination (DAP) after reciprocal crosses between ZH11 and nac20/26 ^ZH11^. Results showed that NAC20/26 maternal alleles were predominantly detected in endosperm cDNAs, irrespective of whether nac20/26 ^ZH11^ was used as male or female (Figures 1B, S2), suggesting that both NAC20 and NAC26 are MEGs in ZH11 endosperm. We then examined NAC20/26 expression using a single‐nucleotide polymorphism (SNP) difference in their coding regions of ZH11 and NJ6 (Nanjing 6, an Indica variety) after reciprocal crosses. Sanger sequencing performed in PCR products amplified from endosperm of reciprocal crosses showed that, in the cross of NJ6 × ZH11, only maternal NJ6 allele expression of NAC20/26 were detected, whereas in the cross of ZH11 × NJ6, both maternal ZH11 and paternal NJ6 expression were detected (Figures 1C, S2). As controls, the same analyses were performed in the reported biallelic expressed gene, NAC23, the maternally expressed imprinted gene, NAC24, and the paternally expressed imprinted gene, NAC129, in endosperm of reciprocal crosses between ZH11 and NJ6 (Cheng et al., 2021) (Figure S3).

Additional nac20/26 double mutants were generated in another Japonica variety NIP (named nac20/26 ^NIP^), and two Indica varieties of NJ6 (named nac20/26 ^NJ6^) and 9311 (named nac20/26 ^9311^) using CRISPR/Cas9‐based gene editing (Figure S4A). All of them showed the floury endosperm phenotype as nac20/26 ^ZH11^ (Figure S4B–D). Reciprocal crosses of these mutants with their background varieties showed a maternal effect in nac20/26 ^NIP^ (Figures 1E, S2; Table 2), but not in nac20/26 ^NJ6^ and nac20/26 ^9311^ (Figures 1D, E, S2; Table 2), suggesting that the maternal effect occurs only in two Japonica varieties. Allelic expression analyses performed in reciprocal crosses between NIP and NJ6, and between ZH11 and 9311, confirmed that paternal NAC20/26 alleles were expressed in endosperm when NJ6 and 9311 were used as pollen donors, but not when ZH11 and NIP were used as pollen donors (Figure 1F), suggesting that the maternal expression of NAC20/26 are allele‐specific for two Japonica varieties.

Differential DNA methylation is associated with maternal‐specific expression of NAC20/26

Maternal allele‐ and paternal allele‐specific gene expression often involve parent‐of‐origin specific DNA methylation and/or histone modifications (Pignatta et al., 2014; Klosinska et al., 2016). By examinations of whole‐genome bisulfite sequencing data (Liu et al., 2018), we observed that, in ZH11, 5' upstream differentially methylated regions (DMRs) for NAC20 (NAC20‐DMR) and NAC26 (NAC26‐DMR; red lines) were hypomethylated in the endosperm, but hypermethylated in the embryo (Figure 2A). Bisulfite sequencing performed in the ZH11 embryo and endosperm showed that decreased methylation levels were observed in NAC20‐DMR and NAC26‐DMR in endosperm excised at 9 and 21 DAP (Figure 2B; EN), when compared with those in embryos, for all three methylation contexts (CG, CHG, and CHH; Figure 2B; EM). Next, we used SNPs that were able to distinguish parental alleles to examine *NAC20‐*DMR and NAC26‐DMR methylation levels in endosperm excised at 9 DAP after reciprocal crosses between ZH11 and NJ6. Results showed that, when ZH11 alleles were examined, hypermethylation (both CG and CHG) was observed for both DMRs when ZH11 was used paternally, while hypomethylation was observed when ZH11 was used maternally (Figure 2C, left). In contrast, when NJ6 alleles were examined, both maternal and paternal alleles of both DMR showed almost no methylation (Figure 2C, right), suggesting that NJ6 genomes of these DMRs were hypomethylated. We further examined methylation levels in NAC20‐DMR and NAC26‐DMR in endosperm and leaves of ZH11, NIP, NJ6, and 9311, and results showed that DNA methylation levels in NJ6 and 9311 in these two DMRs were very low, which were much higher in NIP and ZH11 (Figures 2D, S5). The results were consistent with the whole‐genome bisulfite sequencing data from various rice varieties (Figure S6). Collectively, these findings together suggest that the allele‐specific expression of NAC20/26 in two Japonica varieties are derived from elevated DNA methylation in their paternal alleles.

*Differential DNA methylation is associated with allele‐specific expression of

NAC20

and

NAC26

in endosperm

(A) Bisulfite sequencing to show that NAC20 and NAC26 genomic DNAs, extracted from endosperm at 9 DAP, showed decreased hypomethylation (red lines) when compared to those in embryos. Green squares represent transposable elements. P1~P4 are primers for the detection of transposable elements (P1 and P2 for NAC20; P3 and P4 for NAC26). (B) Local bisulfite sequencing showed that DMRs of NAC20 (−494 to −205 bp, left) and NAC26 (−475 to −186 bp, right) showed lower levels of methylation in endosperm(EN) examined at 9 and 21 DAP in ZH11 for all three methylation contents (CG, CH, and CHH) when compared with those in embryos (EM). Numbers of clones sequenced ≥ 10. (C) Local bisulfite sequencing of DMRs in genomic DNA extracted from 9 DAP endosperm after the reciprocal crosses between ZH11 and NJ6. Note that when the ZH11 allele was sequenced, higher methylation was observed when ZH11 was used as a pollen donor, but when NJ6 alleles were sequenced, low levels of methylation were observed irrespective of whether NJ6 was used as male or female. Numbers of clones sequenced ≥ 20. (D) Local bisulfite sequencing in ZH11, NJ6, NIP, and 9311 endosperm examined at 9 DAP to show that DMRs in two Japonica varieties (ZH11 and NIP) were hypermethylated (CG especially), while two Indica varieties (NJ6 and 9311) were hypomethylated. Numbers of clones sequenced ≥ 10.*

Removals of TEs adjacent to NAC20/26 led to biallelic NAC20/26 expression

DNA hypermethylation usually occur in TEs and their adjacent regions (Gehring et al., 2009; Song and Cao, 2017). Genomic survey in ZH11 revealed the presence of a miniature inverted‐repeat transposable element (MITE)‐family TE (ORSgTEMT00101054) in the 5' upstream region of NAC20 and two tandem repeated retro‐TEs (ORSiTEMT00T00140) in the 5' upstream region of NAC26 (green boxes, Figure 2A). We examined distribution of those TEs in NJ6, NIP, and 9311 by PCR using primers flanking these TEs (Figure 2A), and results showed that they are present in 5' upstream regions of NAC20/26 in NIP (Japonica), but not in two Indica subspecies of NJ6 and 9311 (Figures 3A, S7). To address if hypermethylation in NAC20‐DMRs and NAC26‐DMRs in these Japonica varieties are caused by the presence of these TEs, we concurrently removed these TEs in ZH11 using CRISPR/Cas9‐based gene editing, guided by dual targets. Among the transgenic lines obtained, homozygous ones with both ORSgTEMT00101054 and ORSiTEMT00T00140 were removed (named NAC20/26 ^ ΔTE ^ hereafter; Figures 3B, S8). The grain of NAC20/26 ^ ΔTE ^ had no detectable phenotype compared to ZH11 (Figure S9). Examinations of NAC20/26 ^ ΔTE ^ plants and their progeny showed that, as expected, DNA methylation levels were greatly reduced in NAC20‐DMR and NAC26‐DMR in both embryos and endosperm (Figure 3C). DNA methylation levels were measured in a region within the RISBZ1 gene body across five rice genotypes: ZH11, NJ6, NIP, 9311, and NAC20/26 ^ ΔTE ^, which was used as the control (Figure S10). In the cross of nac20/26 ^ZH11^ × NAC20/26 ^ ΔTE ^ created, no floury endosperm was observed in the grains produced (Figure 3D). Sequencing of cDNA prepared from endosperm of nac20/26 ^ZH11^ × NAC20/26 ^ ΔTE ^ confirmed that paternal NAC20/26 alleles were expressed (Figure 3E). These results together demonstrated that TE‐mediated DNA hypermethylation in paternal NAC20‐DMR and NAC26‐DMR result in silencing of NAC20/26 paternal alleles, and consequently, the inheritance of the floury endosperm phenotype of nac20/26 ^ZH11^ with a maternal effect (Figure 3F).

*TE‐mediated DNA methylation is necessary and sufficient for maternal imprinted

NAC20

and

NAC26

expression in rice endosperm

(A) PCR amplifications to show the presence of TEs (with longer fragments) in the NAC20 and NAC26 promoters of ZH11 and NJ6, but not in NIP and 9311 (shorter fragments). (B) PCR amplification using primers in (Figure S3), followed by gel electrophoresis, to show that in the NAC20/26

ΔTE plant, TEs in both NAC20 and NAC26 were removed. M, molecular weight marker. (C) Local bisulfite sequencing to show greatly decreased CG, CHG, and CHH methylation in both endosperm and embryo in NAC20‐DMR and NAC26‐DMR of NAC20/26

△TE plants, when compared with those in the wild type (ZH11). (D) Photo showing that when NAC20/26

ΔTE was used as a pollen donor to pollinate nac20/26 ZH11 (right), no floury endosperm was observed. nac20/26 ZH11 pollinated by ZH11 was used as a control (left). Scale bar = 1 mm. (E) Allelic expression analyses showing that in endosperm of nac20/26 ZH11 × NAC20/26

ΔTE , both maternal (M, nac20/26 ZH11) and paternal NAC20 and NAC26 alleles (P, NAC20/26

ΔTE ) were expressed. Numbers of clones sequenced ≥ 30. (F) A model for TE‐mediated DNA methylation in paternal alleles is necessary and sufficient to maintain the maternally imprinted expression of NAC20/26 in the endosperm and to regulate grain filling with a maternal effect.*

TEs in NAC20/26 are associated with northward expansion of Japonica varieties

NAC20/26 are endosperm‐specifical expressed genes in the Japonica varieties ZH11 (Wang et al., 2020; Wu et al., 2023), and NAC20/26 are expressed in the endosperm, not in the embryo among Japonica (NIP and KitaaKe) and Indica varieties (9311 and IR64) (Figure S11) (Rodrigues et al., 2021). qRT‐PCR was performed to detect expression of NAC20/26 in rice varieties of ZH11, NIP, NJ6, and 9311 in tissues of seedling, inflorescence, and endosperm. Results showed that NAC20/26 expression in NAC20/26 ^ ΔTE ^ were similar to those rice varieties that were barely detectable in these tissues, except in the endosperm (Figure S12), suggesting that TEs' deletions did not alter their expression patterns. Based on the fact that these TEs were only present in 5' upstream regions of NAC20/26 in two Japonica varieties examined, but not in two Indica varieties of NJ6 and 9311, we speculate that the TE insertions might be associated with the differentiation and distribution of Indica and Japonica subspecies. PCR analyses were performed in randomly selected 172 Japonica/Indica accessions (Table S1) to detect the presence of these TEs in NAC20/26. Results showed that the majority of Japonica cultivars carried TEs in either NAC20 (65.38%) or NAC26 (84.63%), or in both (53.85%), but were much lower in Indica varieties (14.58% in NAC20, 5.21% in NAC26, and 2.08% in both; Figure 4A), suggesting that the TEs in NAC20/26 promoter regions are positively associated with Japonica varieties. Subsequently, we examined geographical distributions where these cultivars were originally developed in China, and results showed that all varieties developed in Northeast China, with latitudes higher than 33°N, carried TEs in both NAC20 and NAC26 (Figure 4B, filled in red), whereas varieties located lower than 33°N were mixed with no TEs (Figure 4B, filled in blue), TEs in one of the NAC20/26 (Figure 4B, filled in green), or in both of them (Figure 4B, filled in red), suggesting that TE‐mediated allele‐specific expression of NAC20/26 are associated with northward expansion of Japonica subspecies during domestication and breeding.

*Distribution of TEs in the

NAC20

and

NAC26

promoters is associated with differentiation of

Japonica

and

Indica

subspecies and northward expansion of rice cultivation

(A–B) Statistics for distributions of TEs located in NAC20 and NAC26 promoters in the 174 rice varieties according to the Indica and Japonica subgroup (A), and planting area in China (B). No TEs in the NAC20/26: The TE ORSgTEMT00100531 was not detected in the 5' upstream region of NAC20 and the retro‐TEs (ORSiTEMT00T00140) were not detected in the 5' upstream region of NAC26 (bule solid circle); TE in the NAC20 or NAC26: The TE ORSgTEMT00100531 was detected in the 5' upstream region of NAC20 or the retro‐TEs ORSiTEMT00T00140 were detected in the 5' upstream region of NAC26 (orange solid circle); and TEs in the NAC20/26: The TE ORSgTEMT00100531 was detected in the 5' upstream region of NAC20 and the retro‐TEs ORSiTEMT00T00140 were detected in the 5' upstream region of NAC26 (red solid circle).*

DISCUSSION

MEGs are found primarily in embryo‐nourishing tissues such as placenta in mammals (Bartolomei and Hanna, 2020) and endosperm in plants (Gehring, 2013). Hundreds of imprinted MEGs have been identified in endosperm of Arabidopsis (Gehring et al., 2011; Hsieh et al., 2011), rice (Luo et al., 2011; Yuan et al., 2017; Chen et al., 2018), and maize (Zhang et al., 2011), but the functions of the vast majority of them and regulation of their imprinted expression have not been elucidated. Although associations of TE‐mediated DNA hypermethylation and silencing of neighboring genes have been reported for MEGs in several plant species (Hsieh et al., 2009; Zemach et al., 2010; Zhang et al., 2014), direct evidence supporting the causality remains scarce. In this study, we showed that TE‐mediated silencing of NAC20/26 paternal alleles is associated with their allele‐specific maternal expression in two Japonica rice varieties, and consequently, the inheritance of the floury endosperm phenotype of their nac20/26 double mutant with a maternal effect. Using reciprocal crosses and transcript sequencing analysis, we identified NAC20/26 as novel allele‐specific maternally expressed genes in endosperm of Japonica varieties examined. Previous studies failed to identify NAC20/26 as allele‐specific maternally expressed genes (Luo et al., 2011; Yuan et al., 2017; Chen et al., 2018). Further, we used gene editing to delete TEs located in 5' upstream regions of NAC20 and NAC26, to show that the allele‐specific maternal expression of NAC20/26 are caused by TE‐mediated hyper‐methylation and silencing of their paternal alleles.

In plants, it has been shown before that genomic DNAs in endosperm are hypomethylated, in comparison with hypermethylation in embryos and other vegetative tissues (Hsieh et al., 2009; Zemach et al., 2010; Zhang et al., 2014). NACs, one of the largest transcription factor families in plants, have undergone duplications during evolutions (Mohanta et al., 2020). Some of them acquire new functions, whereas others maintain redundant roles (Mohanta et al., 2020). We showed previously that NAC20/26 transcription factors regulate grain filling and albumin accumulations in rice endosperm by activating expression of 5 albumin genes (Wu et al., 2023). In maize endosperm, NAC128 and NAC130, orthologous to NAC20/26 in rice, also act redundantly in regulating starch biosynthesis and storage protein accumulations (Zhang et al., 2019). In this study, we showed that mutations of both NAC20/26 led to defective grain filling in all varieties tested. However, NAC20/26 allelic expression are different in different varieties. In Japonica varieties of NIP and ZH11, NAC20/26 are MEGs, but biallelic expression in Indica varieties of NJ6 and 9311. A MITE family TE is found in the 5' upstream region of NAC20 and two tandem retro‐TEs in the 5' upstream region of NAC26 in Japonica varieties of ZH11 and NIP. Interestingly, the insertions of these TEs were not observed in Indica varieties of NJ6 and 9311. These results showed that NAC20/26 are allele‐specific MEGs. As expected, maternal alleles of NAC20/26 5' flanking regions showed hypomethylation but paternal alleles followed with hypermethylation in the endosperm of two Japonica varieties, which was not observed in two Indica varieties examined, suggesting that the allelic‐specific expression of NAC20/26 in the endosperm of Japonica varieties are associated with hypomethylation and silencing of their paternal alleles, and demethylation and maintenance of the hypermethylation state of their maternal alleles.

Endosperm is the major site for imprinted gene expression (Gehring et al., 2004). Studies in Arabidopsis and rice suggest that endosperm DNA hypomethylation is driven by demethylation in central cells (Park et al., 2016). DEMETER (DME) is a maternally expressed DNA demethylase in Arabidopsis that removes DNA methylation in central cells of the female gametophyte to allow the maternal genome to be hypomethylated (Choi et al., 2002; Park et al., 2016). Although the DME ortholog has not been defined in rice, six putative DNA demethylase genes have been found (Liu et al., 2018). Among them, REPRESSOR OF SILENCING 1 (OsROS1) that is expressed in all tissues examined plays an important role in DNA demethylation (Liu et al., 2018; Kim et al., 2019). The OsROS1 knockout mutant showed lethality in both male and female gametophytes (Ono et al., 2012), implying its critical role in rice gametogenesis. A weak mutant allele of OsROS1, named ta2, showed increased DNA methylation in the endosperm when compared with the wild type, and multi‐cell layered aleurone and improved nutrition (Liu et al., 2018). However, the DNA methylation level of NAC20/26 was not decreased significantly in Osros1 compared to the wild type (Figure S13) (Liu et al., 2018). In fact, the methylation level at NAC20/26 was even decreased in the embryo and was inconsistent in the endosperm: NAC20 decreased, while NAC26 increased (Figure S13). OsDML4 (DEMETER‐LIKE4) is another DNA demethylase in rice. Knockout of OsDML4 can lead to chalky endosperm under high‐temperature treatment, accompanied by reduction of storage protein accumulation (Yan et al., 2022). NAC20/26 have been established as key regulators in the control of storage protein accumulation in rice (Wang et al., 2020; Wu et al., 2023). Whether NAC20/26 are direct targets of DML3 remains to be investigated.

We showed further that the allele‐specific maternal expression of NAC20/26 are caused by TE‐mediated DNA methylation in promoter regions of their paternal alleles, as removals of these TEs in Japonica rice varieties by gene editing led to activated expression of their paternal alleles in the endosperm and, consequently, bi‐allelic expression. These results suggest that TE insertions in NAC20/26 are necessary and sufficient for increased DNA methylation in NAC20/26 adjacent regions, and ultimately lead to allele‐specific maternal expression of these two genes in ZH11 and NIP endosperm. We thus provide direct evidence to show that the maternal NAC20/26 expression are caused by TE‐mediated DNA methylation and silencing of their paternal alleles, which consequently led to maternal expression of NAC20/26 and inheritance of the nac20/26 phenotype with a maternal effect (Figure 3F). The RNA‐directed DNA methylation (RdDM) pathway is essential for de novo methylation in plants and plays multiple roles in various biological processes, such as genome imprinting (Lisch, 2009; Zhang et al., 2018; Pal et al., 2024). To determine whether this pathway is required for the methylation of NAC20/26 mediated by nearby transposons, we analyzed the methylation level in mutants of key RdDM components (Osrdr2, fem2, fem3, and pol iv) (Wang et al., 2022; Xu et al., 2024). Compared to the wild type (WT), these mutants showed no significant reduction in NAC20/26 methylation levels, even in the CHH context (Figure S14), suggesting that this transposon‐mediated methylation occurs through an RdDM‐independent mechanism. The precise pathway underlying this process warrants further investigation.

Geographical analyses of rice cultivars across China showed that all varieties from northeast China, at latitudes higher than 33°N, carried TEs in both NAC20 and NAC26 (Figure 4B, filled in red), whereas varieties located in south China, at latitudes lower than 33°N, were mixed with either no TEs, TEs in one of NAC20/26, or in both of them, suggesting that Japonica varieties that develop in higher latitude areas (with moderate temperatures) show the presence of TEs in NAC20/26. It is plausible that these TE insertions occurred after the divergence of Japonica and Indica subspecies. Interestingly, 33°N as an important boundary line is closely correlated with the presence of different RFT1 (Rice Flowering Locus T1) variants in rice cultivars (Zhao et al., 2015). It is apparent that either natural or breeding selection has occurred in the trait. Whether the occurrence of TE‐mediated maternal expression of NAC20/26 is associated with taste preferences and/or environmental adaptation remains to be investigated. The discovery of epigenetically regulated gene expression with a maternal effect has motivated the development of numerous theories of their evolutionary origins and benefits, including the neutralist theory, the kinship theory, the sexual antagonism theory, and the maternal–offspring coadaptation theory (Patten et al., 2014). Although NAC20/26 are tightly linked redundant genes, the fact that their maternally imprinting expression are caused by two different types of TEs that mediate DNA methylation and the presence of these TEs is associated with northward expression of Japonica varieties seems to agree with the neutralist theory, which proposed that gene imprinting is a consequence of a depressed epigenetic state of the endosperm (Berger et al., 2012). Further studies are needed to decipher the exact role of imprinted expression of NAC20/26 in high‐attitude rice varieties.

MATERIALS AND METHODS

Plant materials and growth conditions

The varieties of rice (Oryza sativa, L.) used in this study, including two Japonica varieties of ZH11 and NIP, two Indica varieties of NJ6 and 9311, and 172 cultivars collected in China, were grown in either the experimental field of the Institute of Botany, Chinese Academy of Sciences, or growth rooms at 30°C ± 2°C (d) and 22°C ± 2°C (night), and 12 h light (410 μmol m^−2^s^−1^)/12 h dark, with 60%~80% humidity.

RNA extraction, reversed transcription, quantitative real‐time PCR (RT‐qPCR), and allelic expression analysis

Total RNAs from reciprocal crossed endosperm were extracted using TRIzol Reagent (15596018, Invitrogen), and cDNA synthesis was performed using the FastQuant RT Kit (KR106, Tiangen) according to the manufacturers' instructions. Transcripts of NAC20/26 were amplified individually from those cDNA using gene‐specific primers and then cloned into the pEASY‐T1 vector (CT101‐01, TransGen Biotech). The recombinant constructs were transformed into DH5α competent cells and cultured on the LB (Luria–Bertani) solid medium with Kanamycin, colonies were randomly picked for sequencing, and the resulting data were subjected to statistical analysis. qRT‐PCR was performed using Bio‐Rad iTaq™ Universal SYBR®Green reagent and the CFX‐Connect™ PCR program, with rice ubiquitin gene as an internal control. The primers used are listed in Table S2.

Genetic transformation and gene editing

To generate double‐knockout mutants in NAC20/26, CRISPR/Cas9‐based gene editing was performed as described (Ma et al., 2015). To delete all TEs in promoter regions of NAC20/26, dual sgRNAs targeting flanking sites of TEs located upstream of NAC20/26 were inserted into the CRISPR/Cas9 binary vector (Ma et al., 2015) and transformed into ZH11, NIP, NJ6, and 9311 using Agrobacterium‐mediated transformation. The primers used are listed in Table S2.

Bisulfite sequencing

Genomic DNAs were isolated from different tissues using the CTAB method (Doyle, 1987). The EZ DNA Methylation‐Gold kit was used for bisulfite treatment following the manufacturer's instructions (D5005, ZYMO Research). The bisulfite‐treated DNA, serving as the template, was amplified by specific primers. PCR products were purified, ligated into the pEASY‐T1 vector (CT101‐01, TransGen Biotech), and transformed into DH5α competent cells. A subset of colonies was randomly selected for sequencing, and data were analyzed using the online platform http://katahdin.mssm.edu/kismeth (Gruntman et al., 2008). The primers utilized are listed in Table S2.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

C.M.L. designed the research; M.W.W. performed most of the experiments; R.L., W.W.T., M.M.C., and J.L.L. helped with genetic analyses; J.L. analyzed some data; J.D.Z., T.Y., and H.C. helped with some experiments; and M.W.W. and C.M.L. wrote the manuscript. All authors read and approved the paper.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article: http://onlinelibrary.wiley.com/doi/10.1111/jipb.70126/suppinfo

Figure S1. Co‐suppressions of NAC20 and NAC26 expression and floury endosperm were observed in pUbi:NAC20 and pUbi:NAC26 transgenic plants Figure S2. Parental genome DNA of NAC26 were detected in the endosperm of reciprocal crosses between ZH11 and nac20/26 ^ZH11^, NIP and nac20/26 ^NIP^, 9311 and nac20/26 ^9311^, and ZH11 and NJ6 Figure S3. A biallelic expressed gene, a maternally expressed gene (MEG), and a paternally expressed gene (PEG) were detected in the endosperm of reciprocal crosses between ZH11 and NJ6 Figure S4. Double mutants of nac20/26 in NJ6, NIP, and 9311 backgrounds showed floury endosperm Figure S5. Differential DNA methylation in DMRs of NAC20 (left) and NAC26 (right) was observed in leaves among Japonica rice cultivars (ZH11 and NIP), Indica rice cultivars (NJ6 and 9311), and NAC20/26 ^ △TE ^. Numbers of DNA fragments sequenced ≥ 8 Figure S6. Snapshots in the Integrated Genome Browser showing DNA methylation levels at NAC20 and NAC26 in the embryo (EN) and endosperm (EM) of NIP, 9311, and IR64 (Indica) Figure S7. Alignments of sequences of PCR products amplified by P1, P2, P3, and P4 primers for NAC20 from Japonica (ZH11 and NIP) and Indica rice varieties *(*NJ6 and 9311) Figure S8. Alignments of sequences of PCR products amplified by P1/P2 primers for NAC20 and P3/P4 primers from ZH11 and NAC20/26 ^ △TE ^

Figure S9. Whole (upper) and sectioned grains (lower), to show that NAC20/26 ^△TE^ had no detectable phenotype Figure S10. DNA methylation in the RISBZ1 was observed in leaves among Japonica rice cultivars (ZH11 and NIP), Indica rice cultivars (NJ6 and 9311), and NAC20/26 ^ △TE ^

Figure S11. The graphs illustrate the RNA‐seq‐based expression levels of NAC20/26 in the embryo (EM) and endosperm (EN) of four rice cultivars: NIP, KitaaKe (KIT), 9311, and IR64 Figure S12. The expression levels of NAC20 and NAC26 were measured in the rice seedling (Sd), inflorescence (Inf), and 9 DAP endosperm (En) among ZH11, NIP, NJ6, and 9311 and among Japonica rice cultivars (ZH11 and NIP), Indica rice cultivars (NJ6 and 9311), and NAC20/26 ^ △TE ^ by RT‐qPCR Figure S13. Snapshots in the Integrated Genome Browser showing DNA methylation levels in the NAC20 and NAC26 of ZH11 and Osros1 endosperm Figure S14. Snapshots in the Integrated Genome Browser showing DNA methylation levels in the NAC20 and NAC26 of NIP and Osrdr2, and TP (Taipei), fem3 (Osnrpe1), and pol iv seedling

Table S1. Information of 172 rice varieties

Table S2. Primer sequences used in this study