Imprinting but not cytonuclear interactions determines seed size heterosis in Arabidopsis hybrids

Abstract

The parent-of-origin effect on seeds can result from imprinting (unequal expression of paternal and maternal alleles) or combinational effects between cytoplasmic and nuclear genomes, but their relative contributions remain unknown. To discern these confounding factors, we produced cytoplasmic–nuclear substitution (CNS) lines using recurrent backcrossing in Arabidopsis (Arabidopsis thaliana) ecotypes Col-0 and C24. These CNS lines differed only in the nuclear genome (imprinting) or cytoplasm. The CNS reciprocal hybrids with the same cytoplasm displayed ∼20% seed size difference, whereas the seed size was similar between the reciprocal hybrids with fixed imprinting. Transcriptome analyses in the endosperm of CNS hybrids using laser-capture microdissection identified 104 maternally expressed genes (MEGs) and 90 paternally expressed genes (PEGs). These imprinted genes were involved in pectin catabolism and cell wall modification in the endosperm. Homeodomain Glabrous9 (HDG9), an epiallele and one of 11 cross-specific imprinted genes, affected seed size. In the embryo, there were a handful of imprinted genes in the CNS hybrids but only 1 was expressed at higher levels than in the endosperm. AT4G13495 was found to encode a long-noncoding RNA (lncRNA), but no obvious seed phenotype was observed in lncRNA knockout lines. Nuclear RNA Polymerase D1 (NRPD1), encoding the largest subunit of RNA Pol IV, was involved in the biogenesis of small interfering RNAs. Seed size and embryos were larger in the cross using nrpd1 as the maternal parent than in the reciprocal cross, supporting a role of the maternal NRPD1 allele in seed development. Although limited ecotypes were tested, these results suggest that imprinting and the maternal NRPD1-mediated small RNA pathway play roles in seed size heterosis in plant hybrids.

Seed size heterosis is affected by imprinting but not cytoplasmic–nuclear interactions in Arabidopsis hybrids.

Introduction

Seed size is important to plant evolution and crop production and often shows heterosis. Heterosis is a widely observed phenomenon in which the offspring show greater growth and fitness than either or both parents. Heterosis has been applied in agriculture to generate higher yielding and more resilient crops (Duvick 2001; Lippman and Zamir 2007; Chen 2013; Schnable and Springer 2013; Springer and Schmitz 2017). The phenomenon was systematically studied by Charles Darwin, who observed that cross-pollinated plants displayed increased growth compared to self-pollinated plants (Darwin 1876). And many studies and theories have attempted to explain the molecular basis that underlie this phenomenon (Lippman and Zamir 2007; Chen 2013; Schnable and Springer 2013; Springer and Schmitz 2017). An interesting observation is the parent-of-origin effect on heterosis where the offspring of the reciprocal crosses show varying degrees of heterosis (Waters et al. 2011; Miller et al. 2012; Groszmann et al. 2014; Ng et al. 2014). In animals, mules (horse × donkey, by convention, the maternal parent is listed first in a genetic cross) are larger and stronger than hinnies (donkey × horse) (McLean et al. 2019). The underlying mechanism for this is poorly understood. It involves imprinting (unequal expression of paternal and maternal alleles) in the nuclear genome and/or maternal effect from the cytoplasm. In Arabidopsis thaliana intraspecific hybrids, the parent-of-origin effect on biomass heterosis is mediated by the RNA-directed DNA methylation (RdDM) pathway that modulates levels of CHH (H = A, T, or C) methylation in the promoter of Circadian Clock Associated1 (CCA1) (Ng et al. 2014), a central regulator of plant circadian clock (Greenham and McClung 2015). This supports the notion that imprinting mediates the parent-of-origin effect on hybrid vigor (Vu et al. 2013; Ng et al. 2014). However, in the conventional reciprocal crosses, maternal effect cannot be ruled out simply because the effects of imprinted gene expression in the nuclear genome and cytoplasmic genomes are confounded in the reciprocal hybrids.

The cytoplasm consists of mitochondrial and plastid genomes that affect plant growth and development. In wheat (Triticum aestivum L.) alloplasmic lines, cytoplasmic genomes from Aegilops affect flowering time and fertility in wheat (Kihara 1982). Additional studies of these lines have found that substitution of the cytoplasm is associated with a variety of changes in the expression of genes involved in photosynthetic components, mitochondrial electron transport chain, and retrograde signaling to the chloroplast (Crosatti et al. 2013). In maize (Zea mays L.), substitution of maize cytoplasm with cytoplasm from the wild progenitor teosinte leads to a decrease in yield, and cytonuclear interactions can change plant height in F₂ populations (Edwards et al. 1996; Tang et al. 2013). In Arabidopsis, cytonuclear interactions contribute to changes in the metabolome and growth (Douglas et al. 2015; Joseph et al. 2015). In cybrids generated by haploid inducer line in 6 genotypes (Bur, C24, Col-0, Ler, Sha, WS-4, and ELy), certain plasmotypes, particularly that of the accessions Bur and Ely, show most epistatic effects (Flood et al. 2020). Many of these cybrids have a strong effect on photosynthetic phenotypes (Joseph et al. 2015; Flood et al. 2020), as cytoplasmic genomes play an important role in photosynthesis and growth.

Imprinted genes in the nuclear genome may also play a role in the parent-of-origin effect (Botet and Keurentjes 2020). According to the parental conflict theory, maternally expressed imprinted genes may limit the growth of offspring and conserve more maternal resources, whereas paternally expressed imprinted genes tend to drive growth of the offspring and promote the allocation of maternal resources toward this growth (Haig 2000, 2013). In plants, imprinted genes are known to be involved in regulating endosperm proliferation and seed growth. Studies in maize, rice (Oryza sativa L.), and Arabidopsis have identified a cross-specific imprinting phenomenon (Waters et al. 2013; Pignatta et al. 2014). For example, HDG3 is a paternally expressed imprinted gene in the majority of Arabidopsis accessions, but not in Cvi (Pignatta et al. 2014). These imprinted genes in reciprocal hybrids likely contribute to seed development.

Many imprinted genes and their roles in seed development have been studied in Arabidopsis (Gehring et al. 2011; Hsieh et al. 2011; Raissig et al. 2013; Pignatta et al. 2014; Fort et al. 2017). However, the contribution of cytoplasmic genomes and imprinted genes in the nuclear genomes to the seed development has not been separated in these studies.

To discern these confounding effects, we generated cytoplasmic–nuclear substitution (CNS) lines in Arabidopsis. The nuclear genome of the maternal parent is replaced by 6 generations of backcrossing with the recurrent (paternal) parent, while the cytoplasm remains the same. These CNS lines are also known as alloplasmic lines in wheat (Kihara 1951) or Brassica (Bonhomme et al. 1992), and are often used to generate cytoplasmic–nuclear male sterility lines in plants, due to the incompatibility between cytoplasmic and nuclear genomes. Here, we created a pair of CNS lines using the Arabidopsis Col and C24 ecotypes through 6 generations of backcrossing, and the nuclear genome was 99.99% identical to that in the maternal parent. The seedling phenotypes of CNS lines resembled the paternal parent. Interestingly, the parent-of-origin effect on seed size was found only in the reciprocal CNS crosses where the cytoplasm is the same. On the contrary, in the reciprocal crosses with the same nuclear genome combination, the parent-of-origin effect disappeared. We further characterized the imprinted gene expression using laser-capture microdissection (LCM) in the reciprocal crosses with the same cytoplasm. Despite only 1 pair of genotypes being tested, the results indicate a role of imprinted genes of the nuclear genome in the seed size heterosis in the CNS reciprocal hybrids. Moreover, the seed size difference is also associated with the small RNA pathway involving NRPD1.

Results

Generation and genotype conformation of CNS lines through a backcrossing scheme

Using a backcross scheme (Fig. 1A), we obtained the CNS lines Col(C24C) by backcrossing the A. thaliana accessions Col with pollen from C24C for 6 generations (expected homozygosity of 99.98%) and the CNS lines C24(ColC) by backcrossing C24 with pollen from ColC for 5 generations (∼99.95% of homozygosity). We genotyped multiple plants of each CNS line using simple sequence repeats and selected 4 lines (2 of each CNS line), 2 parental lines, and 1 F₁ hybrid (C24 × Col) for sequencing to determine percentage of homozygosity (Fig. 1B, Supplementary Fig. S1, A and B). Perfectly mapped sequence reads onto TAIR10 genome were used to call variants from other genotypes using GATK's HaplotypeCaller (Van der Auwera et al. 2013). A total of 591,232 SNPs were unique to either the Col or the C24 genome and present as heterozygous in the F₁ (C24 × Col). We found that >99.9% of the SNPs were homozygous for the genome in the Col(C24C) and C24(ColC) CNS lines, respectively (Fig. 1B and Supplementary Fig. S1, A and B), as expected after 5 to 6 generations of backcrossing.

Figure 1.

Generation of cytoplasmic–nuclear substitution (CNS) lines and genomic analysis. A) Backcrossing scheme used to generate the CNS lines, Col(C24C) and C24(ColC) lines, in which the nucleus in respective parental lines is replaced by C24C and ColC, respectively. C refers to the CCA1:LUC reporter in the C24 or Col line. B) SNPs called in the CNS lines that are unique to either the C24 or Col genome, confirming the conversion of the nuclear genome. Heterozygous F1(C24 × ColC) is shown as a control. SNPs are grouped into 100 kb bins. Gray circles indicate the position of the centromere. C) Typical rosette images from the parental and CNS A. thaliana lines at 21 d after sowing. Scale bar = 1 cm for all images.

The chloroplast and mitochondrial genomes were analyzed by PCR using known polymorphic markers to confirm retention of the cytoplasmic genomes. PCR was used to amplify a polymorphic region upstream of the atp6-2 gene in the mitochondrial genome (Forner et al. 2008) and a polymorphic region between rbcL and accD in the chloroplast genome (Azhagiri and Maliga 2007). Using these polymorphic markers, both the Col(C24C) and C24(ColC) lines were confirmed to have retained their original maternal mitochondrial and plastid genomes, respectively (Supplementary Fig. S1, C and D). Taken together with the nuclear genome sequencing results, the Col(C24C) and C24(ColC) CNS lines were converted as expected for further analyses.

CNS lines resembled the lines with the same nuclear genome, and the seed size variation depends on the nuclear genome

Using the CNS lines, we investigated changes in phenotypic traits that are caused by different combination of nuclear and cytoplasmic genomes. We found that seedling phenotypes resembled that of their paternal parent. For example, Col(C24C) resembled the C24 accession, and C24(ColC) resembled the Col-0 parent (Fig. 1C). There was no difference in flowering time (Supplementary Fig. S1E), rosette diameter (Supplementary Fig. S2A), or seed size (Supplementary Fig. S2, B and C) between the CNS lines and their paternal parents. This is in agreement with the notion that most cytoplasmic swaps did not change these phenotypes (Flood et al. 2020).

In previous studies, C24 × Col hybrids displayed increased biomass vigor as compared to the Col × C24 hybrids (Groszmann et al. 2014; Ng et al. 2014; Miller et al. 2015). There was 20% to 30% difference in mature seed size between the Col × C24 and C24 × Col hybrids (Miller et al. 2012; Groszmann et al. 2014). However, these studies cannot discern the effect of cytoplasmic and nuclear genomes on the seed size. Using the CNS lines, we generated pairs of reciprocal hybrids that have either the same cytoplasmic genomes or the same nuclear genome to separate these effects in the reciprocal hybrids (Fig. 2A). In the CNS hybrids Col(C24C) × Col and Col × C24C, which have the same Col cytoplasmic genomes, seed size was ∼20% larger in the Col(C24C) × Col cross than in the reciprocal cross Col × Col(C24C); this resembled the seed size difference between the conventional reciprocal hybrids (C24C × Col and Col × C24C) (Fig. 2B and Supplementary Fig. S2B). On the contrary, no difference in the mature seed size was observed between the reciprocal hybrid crosses Col(C24C) × Col and C24C × Col with the same nuclear (C24 × Col) genome combination but different cytoplasmic genomes (Col vs. C24).

Figure 2.

Parent-of-origin effects of seed size in CNS and conventional reciprocal hybrids. A) The genotypic differences between conventional hybrids, CNS hybrids with fixed cytoplasm, and CNS hybrids with fixed imprinting. B) Mature seeds are larger in the C24C × Col than in the Col × C24C cross. This difference was observed in CNS reciprocal hybrids with fixed cytoplasm, but not in the CNS reciprocal hybrids with fixed imprinting. Scale bar = 1 mm for all images. C) Micrographs of cleared seeds at 7 d after pollination (DAP), showing the embryo proper (EP), peripheral endosperm (PEN), and seed coat (SC). Scale bar = 100 µm for all images.

This seed size difference was established early during seed development. At 7 d after pollination (DAP), the seed size was ∼20% larger in Col(C24C) × Col and in C24 × Col than in the Col × Col(C24C) or Col × C24 crosses (Fig. 2C and Supplementary Fig. S2C). The seed size was similar between the reciprocal hybrids where nuclear genome combination was the same. This result indicates that imprinting alone determines seed size and endosperm development.

Imprinted genes identified in the endosperm between CNS reciprocal hybrids

To investigate how imprinted gene expression changes during seed development, we isolated embryos and endosperm from seeds at 6 DAP using LCM (Belmonte et al. 2013; Ando et al. 2021). These LCM samples were used for mRNA-seq analysis to identify imprinted genes using a generalized linear model (Wyder et al. 2019).

Using 214,274 informative SNPs from the Col(C24C) line relative to the Col reference, we evaluated allelic expression of 15,134 loci in the endosperm (Supplementary Fig. S3, A and B). Our data set was also free of seed coat RNA contamination (Supplementary Fig. S3, C and D), a problem present in several previous studies of imprinted genes (Schon and Nodine 2017). In the conventional reciprocal crosses, we identified 123 maternally expressed genes (MEGs), and 108 paternally expressed genes (PEGs) (Fig. 3A and Supplementary Data Set 1). In reciprocal crosses with the same cytoplasm, we found 104 MEGs and 90 PEGs (Fig. 3B and Supplementary Data Set 1). The overlap between the 2 sets of data was very high, and 99 MEGs and 89 PEGs, respectively, were shared in both conventional and fixed-cytoplasm reciprocal hybrids (Fig. 3, C and D). However, very few imprinted genes were identified in the reciprocal hybrids with fixed cytoplasm but not in the conventional reciprocal crosses. These data indicate that the cytoplasmic genomes play a minor role in imprinted gene expression. Moreover, the imprinted genes that were specific to the conventional or reciprocal crosses with fixed cytoplasm tend to show weaker expression levels than MEGs and PEGs shared in both sets of crosses (Supplementary Fig. S4).

Figure 3.

Comparative analysis of imprinted genes in the CNS reciprocal and conventional hybrids. A) Expression plots of maternal and paternal reads (mean of 2 biological replicates) from the conventional reciprocal crosses (C24C × Col, y axis, and Col × C24C, x axis). Red, blue, and gray spots indicate maternally expressed genes (MEGs), paternally expressed genes (PEGs), and not imprinted genes, respectively. B) Expression plots of maternal and paternal reads (mean of 2 biological replicates) from the CNS reciprocal crosses with fixed cytoplasm [Col(C24C) × Col, y axis, and Col × C24C, x axis]. Red, blue, and gray spots indicate MEGs, PEGs, and not imprinted expression, respectively. C and D) Overlap of MEGs C) or PEGs D) in the conventional hybrids and CNS hybrids. Source data: Supplementary Data Set 1.

Gene Ontology (GO) analysis found that the genes upregulated in the Col × C24C cross relative to the Col(C24C) × Col cross were enriched for cell wall modification (GO:0042545) and carbohydrate metabolic process (GO:0005975) (Supplementary Fig. S5). They included many pectin methylesterase genes, which were found to be upregulated in the linear cotyledon stage of seed development (Wolff et al. 2015). It is likely that the Col × C24C seeds are developing slightly faster than the reciprocal cross, resulting in the increased expression of these genes that typically are upregulated around 6 DAP. The delay of entering this phase in the reciprocal cross Col(C24C) × Col may prolong cellularization, allowing the seeds to grow larger.

Our study identified approximately half of the PEGs and 2/3 of the MEGs, respectively (Fig. 4), compared to a combination of all imprinted genes identified from 3 studies (Gehring et al. 2011; Hsieh et al. 2011; Pignatta et al. 2014). This degree of variation is not unexpected, as there is a considerable amount of variation between imprinted genes identified among different studies (Klosinska et al. 2016; Schon and Nodine 2017; Wyder et al. 2019). This may be related to CNS lines and different ecotypes used, contamination of the endosperm with seed coat (Schon and Nodine 2017), and/or different statistical methods and cutoff values used for data analysis (Wyder et al. 2019).

Figure 4.

Comparative analysis of the imprinted genes identified in this and other studies. A and B) Overlap between PEGs identified in all (total) 3 published studies (Gehring et al. 2011; Hsieh et al. 2011; Pignatta et al. 2014) and the PEGs identified in this study in the conventional reciprocal hybrids A) and CNS reciprocal hybrids B). C and D) Overlap between the MEGs identified in all (total) 3 published studies (see above) and those identified in this study in the conventional reciprocal hybrids C) and CNS reciprocal hybrids D).

Cross-specific imprinting in the endosperm between the CNS reciprocal hybrids

Epigenetic variation contributes to changes in gene expression across Arabidopsis accessions (Kawakatsu et al. 2016), as the presence of epialleles in certain accessions can eliminate or establish imprinting during seed development (Hsieh et al. 2011; Pignatta et al. 2014). Seed size variation between Arabidopsis accessions could be related to cross-specific imprinting in reciprocal hybrids (Pignatta et al. 2014; Pignatta et al. 2018). Cross-specific imprinted genes lack imprinting in 1 of the 2 reciprocal crosses between some ecotypes.

Here, we identified 11 genes that showed biased expression in 1 cross, but not in the reciprocal cross (Supplementary Fig. S6), indicating an epiallele present in only 1 parental accession (Fig. 5 and Supplementary Data Set 2), also known as cross-specific imprinted genes (Pignatta et al. 2018). Three MEGs lacked maternal bias in the crosses using C24 as the maternal parent (Fig. 5B), 3 PEGs lost paternal bias in the crosses using C24 as the paternal parent (Fig. 5C), and 4 MEGs lacked maternal bias in the crosses using Col as the maternal parent (Fig. 5D).

Figure 5.

Analysis of cross-specific imprinted genes in the endosperm. A) There were 58 epialleles in Col and C24, 11 of which were identified as imprinted genes in previous studies (Gehring et al. 2011; Hsieh et al. 2011; Pignatta et al. 2014). B) Three imprinted genes displaying loss of maternal-specific expression (MSE) in C24 × Col and Col(C24) × Col. Red and green indicate paternal and maternal reads, respectively. C) Three imprinted genes displaying loss of paternal-specific expression (PSE) in Col × C24. D) Five imprinted genes displaying loss of MSE in Col × C24. E) Patterns of CHH (brown), CG (green), and CHG (purple) methylation on HDG9 in Col, CVI, Ler, and C24 from the 1,001 Epigenomes Project. Source data: Supplementary Data Set 2.

The epiallelic variation in cross-specific imprinting could be related to DNA methylation. We examined 11 genes using the data from the 1,001 Epigenomes Project (Kawakatsu et al. 2016). Surprisingly, only 1 gene HDG9, encoding a homeodomain glabrous (HDG) transcription factor, showed allelic methylation patterns (Fig. 5E). The 5′ region of HDG9 was differentially methylated, and this methylation was lost in C24. In the crosses where HDG9 is imprinted, the maternal allele is demethylated, while the paternal allele remains methylated. This change in methylation is coincident with the loss of imprinting observed in the Col × C24C hybrid. The 5′ methylation could silence the paternal allele, and loss of this methylation in C24 associated with increased HDG9 expression levels in the cross using C24 as the paternal parent.

HDG transcription factors are involved in endosperm development (Pignatta et al. 2018) and act antagonistically with AINTEGUMENTA-LIKE (AIL) transcription factors to regulate cell proliferation in several plant tissues (Horstman et al. 2015). AILs generally promote cell proliferation, while HDGs restrict it. Thus, a change in imprinting in HDG9 could contribute to the seed size differences that were observed between the reciprocal hybrids involving C24. HDG3 is a cross-specific imprinted gene in reciprocal crosses involving the Cvi accession (Pignatta et al. 2018), and it is possible that a similar mechanism underlying HDG9 contributes to the seed size difference in the Col × C24 reciprocal crosses. To test this, we generated HDG9 knockouts in a C24 and Col-0 background using CRISPR/Cas9 and a cassette of the 2 gRNAs (Xing et al. 2014) and the pHEE401E construct (Wang et al. 2015). These lines were used to test the effect of HDG9 on seed size (Supplementary Fig. S7, A and B). Knockout of HDG9 in C24, but not in Col-0 led to an increase in mature seed size (Supplementary Fig. S7, C and D), suggesting the effect of the HDG9 knockout is limited to the cross-specific epiallele in the C24 ecotype.

Cytoplasmic genomes contribute little to the seed size heterosis

Seed size was similar in the reciprocal crosses where the nuclear genome combination (imprinting) is the same or fixed, suggesting little contribution of the cytoplasmic genomes to the parent-of-origin effect on seed size heterosis. Compared to the differentially genes in the reciprocal crosses with the same cytoplasmic genomes (Supplementary Fig. S8A), only 18 differentially expressed genes in the endosperm were identified when the imprinting were fixed (Supplementary Fig. S8B and Supplementary Data Set 3). Moreover, these differentially expressed genes belonged to those coded by the cytoplasmic genome itself, such as cytochrome oxidase subunit C (COX3), NAD(P)H quinone oxidoreductase K and NAD(P)H quinone oxidoreductase C, which were upregulated in the lines or crosses carrying a C24 cytoplasm. This is consistent with allelic variation between Col and C24, as reported (Forner et al. 2005), and the cox3 locus generates additional COX3 transcript isoforms associated with the C24 cytoplasm.

Imprinted genes identified in the embryo between the CNS reciprocal hybrids

In contrast to the endosperm, relatively few genes were differentially expressed in the reciprocal crosses of the embryo (Supplementary Data Set 4). In the embryo of conventional reciprocal crosses, we identified 7 MEGs and 3 PEGs (Fig. 6A and Supplementary Data Set 5). In the embryo of the reciprocal crosses with the same cytoplasm, we identified 4 MEGs and 1 PEG (Fig. 6B and Supplementary Data Set 5). Again, these imprinted genes in the embryo overlapped between reciprocal crosses of the conventional lines and CNS lines (Fig. 6C). Notably, all MEGs and 1 of the PEGs identified in the embryo were expressed at higher levels in the endosperm than in the embryo (Supplementary Fig. S6). This could be due to a potential contamination from the endosperm tissues in the embryo (Schon and Nodine 2017). However, our LCM approach has eliminated the contamination in the embryo tissues (Supplementary Fig. S3).

Figure 6.

Analysis of the paternally expressed gene AT4G13495 in the embryo. A) Expression plots of maternal and paternal reads (mean of 2 biological replicates) from the conventional reciprocal crosses (C24C × Col, y axis, and Col × C24C, x axis). Red, blue, and gray spots indicate maternally expressed genes (MEGs), paternally expressed genes (PEGs), and not imprinted genes, respectively. B) Expression plots of maternal and paternal reads (mean of 2 biological replicates) from the fixed-cytoplasm reciprocal crosses [Col(C24C) × Col, y axis, and Col × C24C, x axis]. Red, blue, and gray spots indicate MEGs, PEGs, and not imprinted expression, respectively. C) Overlap of the imprinted genes between CNS hybrids and conventional hybrids. D) Genomic feature of the locus AT4G13495. The lncRNA overlaps with 2 miRNAs, miR850 and miR863a. Green, gray, and yellow indicate exons, introns, and TEs, respectively. Source data: Supplementary Data Set 5. E) Expression validation of the imprinted lncRNA AT4G13495 with Sanger sequencing. Ratios indicate the mean ratio of the Col allele to the C24 allele (n = 3). Source data: Supplementary Data Set 5.

Only 1 PEG (AT4G13945) in the embryo displayed higher expression levels in the embryo than in the endosperm (Supplementary Fig. S8). AT4G13945 encodes a long-noncoding RNA (lncRNA) and has been identified as imprinted gene in the embryo of triploid Col × C24 crosses (Fort et al. 2017) and in the endosperm of reciprocal Col × Ler hybrids (Gehring et al. 2011). Our data suggest that AT4G13945 expression is embryo specific, with an expression level that is 10.9-fold higher in the embryo than in the endosperm at 6 DAP (Supplementary Fig. S8C). It is likely that the presence of 3 transposable elements that lie within this locus inadvertently led to the imprinting of the full transcript in the embryo (Fig. 6C).

AT4G13495 lncRNA overlapped with 2 miRNA loci, miR863 and miR850 (Fig. 6D). Both AT4G13495 and miR863 showed paternally-biased expression in the embryo (Fig. 6E). The lncRNA has been found to be involved in retrograde signaling to the chloroplasts (Habermann et al. 2020). In the vegetative tissues, AT4G13495 transcripts from Col × C24 and C24 × Col were biallelically expressed in rosettes (Supplementary Fig. S9E), using a previously published RNA-seq data set (Miller et al. 2015).

To test the effect of this lnRNA in seed size, we examined SALK T-DNA knockout lines (Supplementary Fig. S9A). In both T-DNA insertion lines, no transcript was detected by RT-qPCR (Supplementary Fig. S9, B and C). This lncRNA can moderate SERRATE expression through miR863 (Niu et al. 2016; Habermann et al. 2020). We evaluated the expression of miR863 and SERRATE in the mutants with RT-qPCR. Although miR863 was not expressed in the at4g13495 mutants, SERRATE expression remained unchanged in the mutants (Supplementary Fig. S9D) with no obvious effect on imprinted expression (Supplementary Fig. S9E). As a result, no obvious phenotype was observed during embryogenesis (Supplementary Fig. S9, F and G) or vegetative growth (Supplementary Fig. S9H) in these 2 lines. The exact function of this lncRNA during embryo and seed development needs further investigation.

Seed and embryo size heterosis is affected by NRPD1

NRPD1 encodes the largest subunit of plant-specific RNA Pol IV and a homolog of the RNA polymerase II subunit and is involved in biogenesis of small interfering RNAs (siRNAs) (Herr et al. 2005; Onodera et al. 2005). Moreover, a group of NRPD1-dependent or p4-siRNAs is maternally transmitted (Mosher et al. 2009). Biogenesis of these p4-siRNAs is dependent on ploidy and maternal genome dosage in the endosperm (Lu et al. 2012), and they regulate expression of the target including some imprinted genes in the endosperm (Lu et al. 2012; Kirkbride et al. 2019). To test the effect of NRPD1 on seed size, we compared mature seed size in the reciprocal crosses involving nrpd1 (Supplementary Fig. S10, A and B). The seed size between C24 × Col and C24 × nrpd1 crosses was similar, suggesting that the paternal NRPD1 allele is not associated with seed size vigor in the hybrids. However, the seed size was reduced statistically significant in the nrpd1 × C24 hybrid, compared to the reciprocal hybrid (C24 × nrpd1) or wild-type hybrid (C24 × Col), supporting that the maternal NRPD1 allele affects seed size (Lu et al. 2012; Kirkbride et al. 2019). Likewise, the embryo size was also larger in the nrpd1 × C24 hybrid relative to the reciprocal hybrid C24 × nrpd1 or C24 × Col (Supplementary Fig. S10C). Together, these results indicate a role for the maternal NRPD1 allele in seed and embryo size in hybrids.

Discussion

Maternal and paternal genomes are critical to the growth and development of offspring in sexually reproductive organisms like humans and flowering plants. However, the parent-of-origin effects in plants are confounded by the imprinting (nuclear genes) and cytonuclear genome interactions. Using the CNS lines, we can discriminate between the imprinting and cytoplasmic effects. Notably, cybrids can also be produced in various ecotypes (Flood et al. 2020) using haploid inducer techniques (Ravi and Chan 2010). The latter may introduce some side effects by the centromeric protein and transgenic approach, while CNS lines may retain a minute amount of residual heterozygosity. Nonetheless, through generating CNS reciprocal hybrids that had either fixed cytoplasmic genomes or fixed imprinting, we found that the parent-of-origin effect on seed size is largely dependent on imprinting, while cytoplasmic genomes play a minor role in gene expression and seed size variation.

The MEGs and PEGs identified in the CNS reciprocal crosses overlap with many imprinted genes that were identified in previous studies (Gehring et al. 2011; Hsieh et al. 2011; Pignatta et al. 2014), confirming the role of imprinting in seed size variation. The cross-specific imprinted genes also contribute to the parent-of-origin effect (Pignatta et al. 2014; Pignatta et al. 2018). For example, in Cvi, a loss of methylation at HDG3 removes imprinting of the gene resulting in biallelic expression, contributing to the small seed size of seeds with Cvi as the paternal parent (Pignatta et al. 2014). Similarly, HDG9 is biallelically expressed when C24 is used as the paternal parent, but is a MEG in the crosses involving Ler, Col and Cvi accessions. This lack of HDG9 imprinting in C24 is like the lack of imprinting at HDG3 in Cvi when Cvi is used as a parental parent (Pignatta et al. 2014; Pignatta et al. 2018). Both HDG9 and HDG3 are class IV homeodomain glabrous transcription factors that are likely involved in endosperm development. HDG3 is thought to regulate endosperm cellularization; a decrease in HDG3 expression leads to premature endosperm cellularization, limiting seed size and contributing to smaller seeds. We found that C24 is associated with larger seed size when it is used as the maternal parent and smaller seed size when it is used as the paternal parent. Knockout of HDG9 in C24 led to an increase in mature seed size, suggesting the effect of the HDG9 knockout is limited to the cross-specific epiallele in the C24 ecotype.

Methylated alleles such as HDG3 and HDG9 in imprinting are heritable in the hybrid seeds. For example, HDG3 is unmethylated in Cvi but methylated in Col-0, and the hybrid possessed the methylated copy of HDG3 probably from Col-0, and this scenario occurs in a number of other genes (Pignatta et al. 2014). The inheritance of methylated (imprinted) alleles in seeds may be different from those in seedlings (somatic tissues). The trans-acting methylation may occur in the hybrid seedlings as observed in Arabidopsis (Greaves et al. 2012) and maize (Cao et al. 2022) and heritable across 9 generations of backcrossing and self-pollination (Cao et al. 2022).

Using CNS lines in the reciprocal crosses, we found minor effects of the cytoplasmic genomes on seed size heterosis. A small set of genes that are affected by parent-of-origin effect are related to photosynthesis, such as NPQ and Φ_PSII, as previously reported (Flood et al. 2020). This is probably because imprinting is limited to the seed and endosperm development in plants and the parent-of-origin effect is not common in plant somatic tissues. In addition, the genetic diversity between cytoplasmic genomes may affect cytoplasmic–nuclear compatibility, leading to cytoplasmic–nuclear male sterility (CMS). For example, ecotypes such as Ely with a mutation in PsbA that affects PSII efficiency (El-Lithy et al. 2005) show greater phenotypic effects among the cytoplasmic swap lines (Flood et al. 2020), including changes in biomass, flowering time, seed size, photosynthetic efficiency, and CMS. Among the hybrids derived from the ecotypes with a low level of cytoplasmic diversity, nuclear gene and genomic interactions determine imprinting and seed size phenotypes.

The imprinted genes identified in study are largely void of seed coat contamination, a predominant problem as previously noted (Schon and Nodine 2017). Using CNS lines and LCM, we also found a few imprinted genes in the embryo. Many of the imprinted genes in the embryo are also expressed at higher levels in the endosperm with 1 exception. The gene AT4G13495 is strongly expressed in the embryo but weakly expressed in the endosperm. A previous investigation suggests that this lncRNA may be involved in retrograde signaling to the chloroplast (Habermann et al. 2020). Imprinting of this lncRNA is related to 3 adjacent TEs, while 2 miRNAs within the locus may control the expression of target genes. The role of this lncRNA in embryo and see development remains to be tested. Finally, the parent-of-origin effect on seed and embryo size is largely controlled by NRPD1 or p4 (Lu et al. 2012; Kirkbride et al. 2019). This is presumably because a group of p4-siRNAs may regulate expression of their target genes in both embryo and endosperm during seed development. It will be of interest to investigate if there is a crosstalk between small RNA-mediated pathway and imprinting to regulate embryo and seed development.

Materials and methods

Plant materials and growth conditions

We utilized 4 Arabidopsis (A. thaliana) lines to generate the CNS lines—Col-0 or Col Wt, C24, ColC with the transgene proCCA1:LUC [Col(CCA1:LUC)], and C24C with the transgene proCCA1:LUC [Col(CCA1:LUC)] (Ng et al. 2014). The CCA1:LUC was used for an intent of diurnal studies, which were not included in this study. To generate the CNS lines, we backcrossed Col-0 Wt with C24C over 6 generations, and C24 with ColC over 5 generations, respectively. To perform manual crosses, flowers were first emasculated by removing sepals, petals, and anthers from immature flowers, with pollination performed the following day. For example, the CNS line Col(C24C) resulted from the Col-0 Wt in which the nuclear genome was replaced by C24C. We obtained Col(C24C) and C24(ColC) CNS lines. The regular reciprocal crosses were derived from Col × C24C and C24 × Col combinations. The reciprocal crosses Col Wt × C24C and Col(C24C) × Col Wt were used to test imprinting effect on nuclear genes in the same cytoplasmic background. At the meantime, genetic crosses Col(C24C) × Col Wt and C24C × Col Wt were used to test the cytoplasmic effects on gene expression and phenotypes. To investigate the role of AT4G13495, 2 SALK T-DNA knockout lines, SALK 52375 and SALK 205712C, were obtained from the Arabidopsis Biological Resource Center.

The seeds were sterilized in 20% (v/v) bleach (Clorox) for 10 min. Following sterilization, seeds were washed 5 times with 1 mL sterile ddH₂0 and germinated on 0.5 MS media with 1% (w/v) sucrose and stratified in the dark at 4 °C for 48 h. Plates were then transferred to the 22 °C growth room for with 16 h of light and 8 h of dark per day to germinate the seeds. At 7 d after germination, seeds were transplanted onto a soil mix of 3 parts Pro-Mix Biofungicide to 1 part Field and Fairway. Soil was treated with 4 g of Miracle Gro Plant Food and 1 teaspoon of Gnatrol Biological Larvicide (Valent Biosciences LLC, Mitchell County, Iowa) per gallon of water at first watering. Bonide copper soap fungicide was sprayed as needed or weekly to prevent powdery mildew infection.

Sequencing the CNS lines

Rosette leaves were harvested from plants at 21 d after germination. Leaves were flash frozen in liquid nitrogen and stored at −80 °C. Genomic DNA was extracted from the leaves using a QIAGEN Plant DNEasy mini kit (Qiagen, Hilden, Germany). DNA was sheared to a target size of ∼450 bp fragments using a sonicator (Covaris, Woburn, Massachusetts), using 4 × 14 s cycles, an intensity of 4, and a 10% duty cycle, and purified using the QIAGEN PCR Purification kit. An aliquot of DNA (100 ng) was used to create libraries using the NEBNext Ultra II kit (New England Biolabs, Ipswich, Massachusetts). Libraries were sequenced at The University of Texas at Austin Genome Sequencing and Analysis Facility on a HiSeq 4000, with paired 150 bp reads and a total of ∼30 million reads per sample.

Bioinformatics for genome sequences

Reads were first trimmed using fastx (http://hannonlab.cshl.edu/fastx_toolkit/) and mapped to the TAIR10 genome using bowtie (Langmead and Salzberg 2012). The mapped reads were filtered for optical duplicates and base quality scores were recalibrated using GATK (Van der Auwera et al. 2013). Variants were called using GATK's HaplotypeCaller and filtered to identify SNPs that had a minimum of 10 reads supporting them. A reference set of SNPs for Col-C24 SNPs was created by identifying SNPs that were called as homozygous in Col and C24, as well as heterozygous in Col × C24. Homozygous variants were filtered using GATK's SelectVariants tool requiring that 100% of the reads supported the variant. Heterozygous variant calls were filtered using GATK's SelectVariants so that there was between 75% and 25% allele balance between the 2 calls to filter out spurious heterozygous calls. Variants were filtered for depth and allele balance in the CNS lines using the same criteria, and these were then filtered against the reference set of SNPs identified in the Col, C24 and F₁ lines. Variants per 100 kb window were counted and displayed using SnpEff (Cingolani et al. 2012).

Seed and embryo measurements

Plants were emasculated and then manually pollinated the following day. At 7 DAP, siliques were dissected, by slicing along the replum and removing seeds with forceps (Ando et al. 2023). Seeds were placed onto a 0.5 mL drop of water, which was removed and replaced with clearing solution (chloral hydate:30% glycerol, 1:1, w:v). Seeds were allowed to clear overnight and visualized on a Nikon Eclipse Ni compound light microscope using Nomarski optics. Seed area and embryo length were measured using ImageJ (Schindelin et al. 2012).

Rosette size measurements

Rosettes were photographed at 21 and 28 d after germination. The rosette area and diameter were measured using PlantCV software (Gehan et al. 2017), as previously reported (Miller et al. 2015).

Mature seed area measurements

A set of 30 seeds from each line was harvested from mature siliques and photographed using a Nikon SMZ1500 microscope. Seed area was measured using ImageJ (Schindelin et al. 2012). Seeds/embryos were manually selected along their outlines, and pixel values of selected areas were automatically generated from ImageJ after calibration from scale bars of the original pictures. Approximately 300 seeds in each cross were fixed and pictures were taken. Only seeds/embryos that were well displayed in a picture (i.e. clearly fixed and oriented to show the whole image) were chosen to be measured. A total of 20 to 50 images were used to estimate seed and embryo volumes (mm), respectively. Student’s t-tests were used for statistical significance analysis.

Laser-capture microdissection (LCM)

LCM procedure followed a published protocol with 2 biological replicates (Ando et al. 2023). At 6 DAP, siliques were harvested by slicing along the replum and cut into 3 to 5 mm sections using a scalpel. The sections were immersed in a fixative of ice cold 75% (v/v) ethanol and 25% glacial acetic acid. Immersed sections were put under vacuum for 10 min to infiltrate sections with the fixative. Siliques were left immersed in the ethanol and acetic acid fixative overnight at 4 °C and fixed using a microwave-based fixation method (Takahashi et al. 2010; Ando et al. 2021), in which seeds are dehydrated using an ethanol series, followed by infiltration with butanol and then paraffin. Briefly, in the following day, the fixative was replaced with fresh solution, and seeds were then microwaved at 250 W for 15 min in an ice bath in a PELCO Biowave 34700 with a Cold Spot (Ted Pella) set to 37 °C. This was repeated 3 more times. Seeds were dehydrated using an ethanol series (70%, 70%, 80%, 90%, 100%, 100%, 50% butanol, 100% butanol) by replacing the existing solution with the ethanol and then microwaving the solution for 90 s at 350 W, with the Cold Spot set to 58 °C. The seeds were microwaved in a mixture of 50% molten Paraplast Plus paraffin (Millipore Sigma, St. Louis, Missouri) for 10 min at 250 W. This was followed by microwaving the seeds in 100% molten paraffin, again for 10 min at 250 W. Finally, the seeds were microwaved in fresh 100% molten paraffin, partially submersed in a 75 °C water bath, for 30 min at 250 W, replacing the paraffin each time. The silique segments were then embedded into paraffin blocks, and stored at 4 °C. Blocks were sectioned into 7 µm sections using a paraffin microtome and mounted onto PEN foil slides (ZEISS, Oberkochen, Germany). Slides were deparaffinized by immersion in xylenes for 2 min, followed by air-drying of the slide for 5 min. The seeds were dissected using a ZEISS PALM Laser Microdissection microscope (ZEISS, Oberkochen, Germany), isolating the embryos first, and the endosperm afterwards. Endosperm sections consisted of chalazal, micropylar, and peripheral endosperm. The cut tissue was collected onto the adhesive collection cap of LCM collection tubes (ZEISS, Oberkochen, Germany). RNA was isolated from LCM-isolated tissue using the RNAqueous Micro Kit RNA isolation kit (Invitrogen, Carlsbad, California). RNA quality and quantity were checked on a Bioanalyzer 2100 using the RNA Pico eukaryotic assay (Agilent, Santa Clara, California). gDNA was removed from RNA samples using RQ1 DNAse (Promega, Madison, Wisconsin). mRNA-seq libraries were prepared using 50 to 100 ng of RNA (with a RIN score greater than 5), using the NEB Ultra II RNA seq kit (New England Biolabs, Ipswich, Massachusetts), with the PolyA selection module. Libraries were sequenced at Genome Sequencing and Analysis Facility (GSAF) at The University of Texas at Austin, with a target of ∼20 million reads per sample. The libraries for the conventional reciprocal crosses were sequenced using single end 100 bp reads on an Illumina HiSeq 2500. One biological replicate of the reciprocals with fixed cytoplasm was sequenced using single end 75 bp reads on an Illumina NextSeq with a target of ∼30 million reads per sample, and the second was sequenced using single end 100 bp reads on an Illumina NovaSeq with a target of ∼20 million reads per sample.

Analysis of imprinted and differentially expressed genes

Reads were first trimmed with trimmomatic (Bolger et al. 2014) to remove low quality reads. High-quality reads were mapped onto the TAIR10 genome using STAR (Dobin et al. 2013) with the following settings: –outFilterMismatchNoverLmax 0.04 –outFilterMultimapNmax 20 –alignIntronMin 25 –alignIntronMax 3000. Reads were filtered to identify uniquely mapped reads using samtools (-q 60 setting) (Barnett et al. 2011). For differential expression analysis, reads overlapping each gene were counted using HTseq with the union and reverse stranded settings. Gene annotations from Araport11 (Cheng et al. 2017) were used to identify gene loci. Statistical analysis of differential expression was performed using EdgeR, using linear models to identify differentially expressed genes (Robinson et al. 2010; McCarthy et al. 2012). Genes that showed a greater than 2-fold change and a Benjamini–Hochberg adjusted P-value of less than 0.05 were classified as differentially expressed between the samples. To identify imprinted genes, bam files of uniquely mapped reads were processed using a previously published python script that discriminated variants using bam files. The read counts from the called variants and a bed file indicating the location of known SNPs that overlap with genes were used to estimate the number of reads corresponding to the paternal and maternal origins at each gene (Wyder et al. 2019). The EdgeR protocol utilizing linear models to identify imprinted genes was used for statistical analysis. Genes with a Benjamini–Hochberg adjusted P-value of less than 0.05 were identified as imprinted genes for further analysis.

GO enrichment analysis was performed using the TopGO package (https://bioconductor.org/packages/release/bioc/html/topGO.html) using the weight01 algorithm. Significantly enriched GO terms were defined as having a P-value lower than 0.05.

Analysis of cross-specific imprinted genes

Cross-specific imprinting of genes was identified using the workflow outlined in a published paper (Pignatta et al. 2014). Briefly, log₂(maternal/paternal) read ratios were calculated and plotted for all pairs of reciprocal crosses. These were then filtered for genes that fell within a Euclidian distance of 1 from x = 1 or y = 1, which identifies genes that lack parental bias in at least 1 parent. A normalized parental bias factor was calculated for all remaining genes, and genes within the top 5% of parental bias were retained. Genes that had the same parental bias in both replicates, and fell within the top 5% parental bias, were classified as having cross-specific expression in 1 reciprocal cross, but not in the other. These genes were compared with the imprinted genes identified previously (Gehring et al. 2011; Hsieh et al. 2011; Pignatta et al. 2014) to identify cross-specific imprinted genes, which display imprinting in crosses with other ecotypes, but lacked imprinting in only 1 of the reciprocal crosses (Col × C24 or C24 × Col).

RT-qPCR validation of gene expression

RNA was extracted from 3 pools of 10 seedlings at 7 d after germination per sample. The seedlings were flash frozen in liquid nitrogen, and then ground to a fine powder in a QIAGEN TissueLyser LT (Qiagen, Hilden, Germany). RNA was extracted from the powder using TRIzol reagent (Invitrogen, Carlsbad, California), and resuspended in DEPC-treated ddH₂O. An aliquot of RNA (1 µg) was treated with RQ1 DNAse (Promega, Madison, Wisconsin), and 500 ng of RNA was then reverse transcribed using the Omniscript RT kit (Qiagen, Hilden, Germany). The cDNA from this reaction was used as template for qPCR, using the FastStart Universal SYBR Green Master Mix (Roche, Indianapolis, Indiana). The following PCR cycle was used for all RT-qPCR experiments: 50 °C, 120 s preincubation and a 50 °C, 600 s secondary preincubation followed by the following 2-step amplification cycle: 40 cycles of 95 °C for 15 s, followed by 60 °C for 60 s. The LightCycler 96 system (Roche, Indianapolis, Indiana) was used for qPCR, and relative expression levels of each gene were calculated using the LightCycler software and ACT7 (AT5G09810) as a control. Controls without reverse transcriptase were performed for each sample, and 2 technical replicates were performed for each of the 3 biological replicates per sample. Primers used for RT-qPCR can be found in Supplementary Data Set 6.

Stem-loop RT-qPCR for analysis of miRNA relative expression

Stem-loop reverse transcription and qPCR was adopted from a published protocol (Varkonyi-Gasic 2017). The U6 snRNA used as a control for relative expression of miR863. Briefly, RNA was extracted from 3 pools of 10 seedlings at 7 d after germination per sample. The seedlings were flash frozen in liquid nitrogen, and then ground to a fine powder in a QIAGEN TissueLyser LT (Qiagen, Hilden, Germany). RNA was extracted from the powder using TRIzol reagent (Invitrogen, Carlsbad, California), and resuspended in DEPC-treated ddH₂O. A total of 50 ng of RNA was used in a pulsed reverse transcription reaction performed using SuperScript III RT (Invitrogen, Carlsbad, California), with separate reactions for U6 snRNA and miR863. Reverse transcription was carried out at 16 °C for 30 min, followed by 60 cycles of 30 °C for 30 s, 42 °C for 30 s, and 50 °C for 1 s. The reaction was then inactivated by incubation at 85 °C for 5 min. cDNA was then used as a template for qPCR using the Luna Universal Probe Master Mix (New England Biolabs, Ipswich, Massachusetts) with the Universal Probe Library #21 probe (Roche, Indianapolis, Indiana). qPCR was performed with one 95 °C incubation for 5 min, followed by 50 2-step cycles of 95 °C for 5 s and 60 °C for 10 s. The LightCycler 96 system (Roche, Indianapolis, Indiana) and its associated software were used to perform qPCR and quantification of the results. Controls without reverse transcriptase were performed for each sample, and 2 technical replicates were performed for each of the 3 biological replicates per sample. Primers used for RT-qPCR can be found in Supplementary Data Set 6.

Knockout of HDG9 using CRISPR-Cas9

Two gRNAs were designed to target HDG9 (AT5G17320) using the CRISPR-PLANT tool (https://www.genome.arizona.edu/crispr/instruction.html) and the CRISPR-P v. 2.0 tool (http://crispr.hzau.edu.cn/CRISPR2/). A construct containing Cas9 and a cassette of the 2 gRNAs (Supplementary Data Set 6) was assembled using the Golden Gate assembly protocol outlined in Xing et al. (2014), using the pHEE401E construct from Wang et al. (2015). The constructs were used to transform Agrobacterium tumefaciens GV3101, which was used to transform the Col and C24 accessions, respectively, by floral dip (Clough and Bent 1998).

Accession numbers

RNA sequencing data from this article were deposited under the GenBank/EMBL Gene Expression Omnibus accession number GSE249562.

Author contributions

V.J. and Z.J.C. conceived the research, analyzed the data, and wrote the paper. X.S. performed the experiments.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. Genomic analysis of CNS lines showing nuclear genome conversion while retaining their original mitochondrial and plastid genomes.

Supplementary Figure S2. Seedling growth vigor observed between reciprocal CNS hybrids with fixed cytoplasmic genomes, but not in reciprocal CNS hybrids with fixed imprinting.

Supplementary Figure S3. Tissue contamination was absent in the samples collected via LCM.

Supplementary Figure S4. Weak expression of the imprinted genes only in the conventional reciprocal crosses or in reciprocal CNS hybrids with fixed cytoplasm.

Supplementary Figure S5. Gene Ontology (GO) analysis of imprinted genes.

Supplementary Figure S6. Analysis of cross-specific imprinted genes using LCM samples.

Supplementary Figure S7. CRISPR/Cas9-edited HDG9 lines and seed phenotypes.

Supplementary Figure S8. Differentially expressed genes in the endosperm or embryo in the reciprocal CNS hybrids and endosperm-expression of embryonic imprinted genes.

Supplementary Figure S9. Analysis of imprinted AT4G13495 locus consisting of 2 miRNAs in T-DNA insertion lines.

Supplementary Figure S10. Embryo and seed size are affected by NRPD1.

Supplementary Data Set 1. Imprinted genes in the endosperm.

Supplementary Data Set 2. RNA-seq reads of cross-specific imprinted genes.

Supplementary Data Set 3. Differentially expressed genes between the reciprocal hybrids in the endosperm.

Supplementary Data Set 4. Differentially expressed genes between the reciprocal hybrids in the embryo.

Supplementary Data Set 5. Imprinted genes in the embryo.

Supplementary Data Set 6. Lists of primers used in this study.

Funding

The financial support for this work was partly provided by the National Institutes of Health (GM109076) to Z.J.C.

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• HDG3 Gramene: AT2G32370

• HDG3 Araport: AT2G32370

• AT4G13495 Gramene: AT4G13495

• AT4G13495 Araport: AT4G13495

• CCA1 Gramene: AT2G46830

• CCA1 Araport: AT2G46830

• NRPD1 Gramene: AT1G63020

• NRPD1 Araport: AT1G63020