Genomic imprinting, methylation and molecular evolution of maize Enhancer of zeste (Mez) homologs


(fax +1 612 625 1738; e-mail


Imprinted gene expression refers to differential transcription of alleles depending on their parental origin. To date, most examples of imprinted gene expression in plants occur in the triploid endosperm tissue. The Arabidopsis gene MEDEA displays an imprinted pattern of gene expression and has homology to the Drosophila Polycomb group (PcG) protein Enhancer-of-zeste (E(z)). We have tested the allele-specific expression patterns of the three maize E(z)-like genes Mez1, Mez2 and Mez3. The expression of Mez2 and Mez3 is not imprinted, with a bi-allelic pattern of transcription for both genes in both the endosperm and embryonic tissue. In contrast, Mez1 displays a bi-allelic expression pattern in the embryonic tissue, and a mono-allelic expression pattern in the developing endosperm tissue. We demonstrate that mono-allelic expression of the maternal Mez1 allele occurs throughout endosperm development. We have identified a 556 bp differentially methylated region (DMR) located approximately 700 bp 5′ of the Mez1 transcription start site. This region is heavily methylated at CpG and CpNpG nucleotides on the non-expressed paternal allele but has low levels of methylation on the expressed maternal allele. Molecular evolutionary analysis indicates that conserved domains of all three Mez genes are under purifying selection. The common imprinted expression of Mez1 and MEDEA, in concert with their likely evolutionary origins, suggests that there may be a requirement for imprinting of at least one E(z)-like gene in angiosperms.


The contribution of the maternal and paternal genomes to the transcriptome may be different even though their primary DNA sequence is often nearly identical. Genes that show mono-allelic transcription are referred to as imprinted if differential expression depends upon the parent of origin. Genomic imprinting represents a unique example of differential gene regulation of identical alleles in the same nucleus. Numerous studies have documented the existence of genes that display an imprinted pattern of gene expression in both mammals and plants (Gehring et al., 2004; Grossniklaus, 2005; Morison et al., 2005). In mammals, mechanisms exist that control genomic imprinting such that both paternal-specific and maternal-specific loci are expressed (Wrzeska and Rejduch, 2004), and normal embryonic development strictly requires contributions from both genomes.

Imprinting has been documented for at least 83 loci in mammals (Morison et al., 2005); in contrast, there are relatively few examples of imprinting in plants. Most characterized examples of imprinting in plants have been documented in the triploid endosperm tissue. Endosperm tissue is responsible for providing nourishment to the developing embryo; however, it makes no genetic contribution to the next generation. Mutations in essential imprinted genes are expected to show parent-of-origin-dependent phenotypes, although such phenotypes may result from other mechanisms as well (Dilkes and Comai, 2004; Grossniklaus et al., 1998). In Arabidopsis thaliana, the maternal-effect phenotypes of the medea (mea), fertilization-independent endosperm (fie) and fertilization-independent seed2 (fis2) mutants (Castle et al., 1993; Chaudhury et al., 1997; Grossniklaus et al., 1998; Ohad et al., 1996) suggest that these genes might be imprinted. However, because all these genes are already expressed in the embryo sac prior to anthesis, it is difficult to discriminate between maternal transcripts that are produced and stored prior to fertilization and allele-specific expression due to genomic imprinting (Grossniklaus, 2005). Allele-specific transcription analysis in combination with the detection of nascent transcripts in endosperm nuclei has shown that MEA is a maternally expressed imprinted gene (Baroux et al., 2006; Kinoshita et al., 1999; Vielle-Calzada et al., 1999). FIS2 and the flower-timing gene FWA show a maternal mono-allelic expression pattern (Jullien et al., 2006a; Kinoshita et al., 2004), but because these genes are already expressed prior to fertilization their imprinting status has not been unambiguously resolved. Their regulation by the DNA methyltransferase MET1 and the DNA glycosylase DEMETER (DME), however, indicates that these are imprinted loci (Jullien et al., 2006a; Kinoshita et al., 2004). There is no conclusive evidence for the imprinting of FIE, and a study by Yadegari et al. (2000) found bi-allelic expression of both the paternal and maternal FIE alleles during endosperm development. The phenotype of the msi1 mutant also suggests that MSI1 may be imprinted; however, allele-specific expression has not yet been investigated (Guitton et al., 2004; Köhler et al., 2005). To date, the only paternally expressed imprinted gene identified in plants is PHERES1 (PHE1), which is directly regulated by MEA (Köhler et al., 2003, 2005). While the MEA and FIS2 genes are required for normal seed development (Grossniklaus et al., 1998; Luo et al., 2000), and PHE1 plays a role in seed abortion in hybrids (Josefsson et al., 2006), FWA is not essential for normal seed development.

Recent studies in Arabidopsis have begun to reveal the mechanism underlying the imprinted expression of MEA, FWA and FIS2. DNA methylation (mediated by MET1 at the FWA/FIS2 loci and DDM1 and MET1 at the MEA locus) plays a critical role in the regulation of all three loci (Jullien et al., 2006a; Kinoshita et al., 2004; Vielle-Calzada et al., 1999). More specifically, the DNA glycosylase DME is required for expression of the maternal allele of MEA, and a mechanism involving the excision of methylated cytosines from the MEA locus has been proposed to lead to its activation (Choi et al., 2002; Gehring et al., 2006). A functional DME allele is also required for expression of the maternal copies of FWA and FIS2 (Jullien et al., 2006a; Kinoshita et al., 2004). A current model for the regulation of all three loci involves the methylation of both parental alleles by MET1 during vegetative growth. During male gametogenesis, MET1 maintains paternal methylation patterns and the inactive expression state persists. However, expression of DME in the female gametophyte removes the DNA methylation, and this is thought to result in expression of the maternal allele in the central cell (Gehring et al., 2006; Jullien et al., 2006a). However, the fact that the paternal MEA allele is not activated in the absence of MET1 (Luo et al., 2000) and that the maternal MEA allele is active in the embryo in the absence of DME activity, suggests that the imprinted regulation of MEA is more complex. Indeed, it was recently reported that the MEA protein is involved in regulating its own imprinted expression through two distinct mechanisms specific to the maternal and paternal MEA alleles (Baroux et al., 2006; Gehring et al., 2006; Jullien et al., 2006b).

In maize, two distinct types of imprinting exist: allele-specific imprinting and gene-specific imprinting (Messing and Grossniklaus, 1999). Allele-specific imprinted phenotypes of kernel pigmentation are conferred by several alleles of R and B, two maize loci that regulate anthocyanin production (Kermicle and Alleman, 1990; Selinger and Chandler, 2001). The imprinting of R appears to be a result of increased transcription of the maternal allele rather than silencing of the paternal allele (Kermicle and Alleman, 1990). Chaudhuri and Messing (1994) documented reciprocal differences in the accumulation of zein proteins in maize due to the imprinted expression of the controlling locus dzr1, and there is evidence for imprinting of the α-tubulin genes (Lund et al., 1995) as well. These examples of imprinting are allele-specific, not gene-specific, in that only certain alleles of these loci are imprinted.

At the transcriptional level, there is currently evidence for gene-specific imprinting of two maize genes, ZmFie1 (Danilevskaya et al., 2003) and Nrp1 (Guo et al., 2003). There are two maize orthologs of the Arabidopsis FIE gene, ZmFie1 and ZmFie2 (Gutierrez-Marcos et al., 2006; Springer et al., 2002). ZmFie1 is imprinted in the endosperm in a manner similar to that of the Arabidopsis MEA gene, in that paternal transcripts cannot be detected throughout seed development (Danilevskaya et al., 2003). The other maize ortholog, ZmFie2, displays a delayed activation of the paternal allele in the endosperm such that expression is imprinted in early endosperm tissue but is bi-allelic at later time points (Danilevskaya et al., 2003). Similar to ZmFie2, there is also evidence for delayed paternal activation of the gene Meg1 (Gutierrez-Marcos et al., 2004). Expression of the maize transcription factor Nrp1, from the No-apical-meristem gene family, is restricted to the endosperm tissue, and only transcripts derived from the maternal allele are detected (Guo et al., 2003). Differentially methylated regions (DMRs) have been identified for ZmFie1 and ZmFie2 which correlate with maternal expression of these two genes in maize endosperm (Gutierrez-Marcos et al., 2006).

MEA is homologous to the Drosophila Polycomb group (PcG) gene Enhancer of Zeste (E(z)) (Grossniklaus et al., 1998). Plant PcG proteins are thought to function in multi-protein complexes involved in maintaining transcriptional repression of certain genes during development (Goodrich et al., 1997). MEA is one of three Arabidopsis genes that are related to the DrosophilaE(z) gene: the others are CURLY LEAF (CLF), and SWN (EZA1) (Goodrich et al., 1997; Grossniklaus et al., 1998; Preuss, 1999). The maize genome also encodes three homologs of the DrosophilaE(z) gene: Mez1, Mez2 and Mez3 (Springer et al., 2003). The goal of this study was to determine whether the maize E(z)-like genes Mez1, Mez2 and Mez3 display an imprinted pattern of gene expression in the endosperm.


Imprinted expression of Mez1 in the endosperm

The possibility of imprinted expression patterns for the maize E(z) homologs was tested using an allele-specific RT-PCR assay (Figure 1). CAPS markers were identified that allowed the allele-specific analysis of Mez1, Mez2 and Mez3 expression in various maize genotypes (Table S1). The allele-specific expression pattern of Mez1 and Mez2 could be investigated in tissues generated from the B73 and Mo17 genotypes. Embryo and endosperm were dissected from seeds of ears derived from B73 self-pollinations, Mo17 self-pollinations, B73 (♀) × Mo17 (♂) pollinations and Mo17 (♀) × B73 (♂) pollinations. Total RNA was extracted from endosperm and embryo tissue isolated 13 days after pollination (DAP). The primers used for the allele-specific RT-PCR expression analysis flank introns to avoid the possibility of contamination by amplification of genomic DNA. Each amplified PCR product was digested with the restriction enzyme AluI.

Figure 1.

 (a) Mez1 genomic structure. The 13 084 bp of Mez1 is indicated by the line with the position of exons indicated by blank boxes. The position of a CPG island is indicated above the sequence. The position of the Alu I site that was used for CAPS analysis (Alu I site present in B73 but absent in Mo17) is indicated by the asterisk.
(b) The allele-specific expression of Mez1 was monitored in the embryo and endosperm tissue derived from four genotypes; B73, B73 (female) × Mo17 (male), Mo17 (female) × B73 (male), and Mo17. The expression in the embryo is bi-allelic, while the endosperm shows mono-allelic expression of the maternal allele. The first lane in each image is the molecular weight marker (MWM).
(c) Mez2 allele-specific expression was monitored in the same samples as in (b). Mez2 expression was bi-allelic in both embryo and endosperm tissues.

The presence and intensity of both maternal and paternal Mez1 transcripts in the embryo of heterozygous F1 plants resulting from reciprocal crosses indicate that Mez1 expression is bi-allelic in the embryo (Figure 1b). Similarly, the presence of both maternal and paternal Mez2 transcripts can be seen in heterozygous embryo tissue (Figure 1c). The expression pattern of Mez2 in the endosperm is very similar to the pattern observed in the embryo. These findings indicate that Mez2 does not display allele-specific expression patterns in these two genotypes and is not imprinted in these tissues. In contrast, only the maternal allele of Mez1 was detected in the endosperm tissue of heterozygous plants from reciprocal crosses (Figure 1b). This indicates that Mez1 is regulated by genomic imprinting in maize endosperm. Analysis of tissue derived from other inbred lines suggests that Mez3 does not display an imprinted pattern of expression (Table 1).

Table 1.   Imprinting of Mez1, Mez2 and Mez3 in maize hybrids
B73 × Mo17A619 × W64aB84 × B57Oh43 × IL14HB57 × B79
Mez1ImprintedImprintedImprintedImprintedNot polymorphic
Mez2Bi-allelicBi-allelicNot testedNot polymorphicNot tested
Mez3Not polymorphicNot polymorphicBi-allelicBi-allelicBi-allelic

Unlike many of the other imprinted genes detected to date in maize (e.g. R and B), the presence of a reciprocal pattern of expression in both of the F1 heterozygotes suggested that Mez1 imprinting is common to both inbred alleles, not to a specific allele of only one inbred line. We proceeded to test for imprinting of Mez1, Mez2 and Mez3 in a set of five different reciprocal crosses (Table 1). In some cases, specific genes were non-polymorphic and the imprinting status could not be assessed. Mez1 was consistently imprinted in the endosperm tissue of all crosses analyzed, while neither Mez2 nor Mez3 were found to display imprinted expression. These data suggest that the imprinting of Mez1 is not an allele-specific imprinting phenomenon but is an example of gene-specific imprinting common to all alleles tested.

Confirmation of imprinting using Sequenom allele-specific expression assays

A secondary technique was used to monitor allele-specific expression in endosperm tissues derived from B73 × Mo17 crosses. Sequenom technology uses mass spectrometry to distinguish the primer extension products of two alleles and calculate the relative proportion of the two alleles in a sample (Jurinke et al., 2005). Allele-specific assays based on single nucleotide polymorphisms (SNPs) were developed for ZmFie1, Mez1, Zmet3 and Mez2. Previous studies have documented imprinting for ZmFie1 (Danilevskaya et al., 2003), while there is no previous evidence (or expectation) that Zmet3 displays an imprinted pattern of expression (Cao et al., 2000). For Mez1, three separate assays that test the relative frequency of distinct SNPs were developed. The fractions of transcripts derived from the B73 allele were determined for B73 × Mo17 and Mo17 × B73 cDNA as well as mixes of cDNA derived from B73 and Mo17 inbred lines (Figure 2a,b). The results from the Sequenom analysis are fully consistent with maternal-specific expression of Mez1 and ZmFie1 in endosperm tissue of maize at both 13 and 19 DAP. As expected, Mez2 and Zmet3 display bi-allelic, non-imprinted patterns of gene expression. Both of these genes display expression patterns consistent with the maternal:paternal dosage ratios.

Figure 2.

 Analysis of allele-specific gene expression using Sequenom.
Sequenom assays (MALDI-TOF detection of SNP primer extension products) were designed for B73/Mo17 SNPs in Zmet3, Mez2, ZmFie1 and three different SNPs in Mez1, (Mez1a, Mez1b and Mez1c).
(a) Allele-specific expression was performed using cDNA extracted from endosperm tissue from B73 × Mo17 (B×M) and Mo17 × B73 (M×B) crosses at 13 DAP. In addition, we mixed cDNA extracted from B73 and Mo17 in 1:1 (M+B), 1:2 (M+2B) and 2:1 (2M+B) ratios as controls to demonstrate linear detection of both alleles in endosperm tissue.
(b) The same set of cDNA samples were also analyzed in endosperm tissue at 19 DAP.

Mez1 imprinting is maintained throughout endosperm development

Results obtained by Vielle-Calzada et al. (2000) suggest that the entire paternal genome of Arabidopsis thaliana is transcriptionally inactive during early endosperm and embryo development. Different genes showed different times of reactivation during endosperm and embryo development. A similar observation has recently been made in maize (Grimanelli et al., 2005). Moreover, both ZmFie2 and Meg1 exhibit delayed reactivation of the paternal allele (Danilevskaya et al., 2003; Gutierrez-Marcos et al., 2004). Expression from the paternal allele of ZmFie2 is delayed by 5–10 DAP relative to the maternal allele (Danilevskaya et al., 2003). It is possible that the parent-of-origin effects on Mez1 expression that we observed in 13 DAP endosperm are due to delayed activation of the paternal genome, not gene-specific imprinting regulation of the Mez1 locus. In order to further explore the possibility that Mez1 expression is due to a delayed activation of the paternal genome, we determined the allele-specific expression of Mez1 at multiple time points during endosperm development (Figure 3a,b). Only transcripts corresponding to the maternal allele of Mez1 were detected in all endosperm tissue tested from 8 until 27 DAP, suggesting that Mez1 shows imprinted expression throughout seed development.

Figure 3.

 Allele-specific expression of Mez1 during endosperm development. Endosperm tissue was harvested at different times during development and used for allele-specific analysis of Mez1. The first lane is derived from B73 homozygous endosperms at 13 DAP while the second lane is derived from Mo17 homozygous endosperms at 13 DAP. These lanes illustrate the expected patterns produced by mono-allelic expression of the B73 or Mo17 allele.
(a) Analysis of Mez1 allele-specific expression in tissue harvested from a cross of Mo17 (female) and B73 (male) at 8, 10, 13, 17, 20, 23 and 27 DAP.
(b) Tissue was harvested at the same time points from a cross of B73 (female) and Mo17 (male).
(c) A different set of primers was used to amplify genomic DNA isolated at 8, 10, 13, 17, 20, 23 and 27 DAP to detect whether both alleles are present in the genomic DNA at all time points.

Allele-specific expression of Mez1 is not caused by allele-specific DNA degradation

There is a formal possibility that allele-specific degradation (Yerk et al., 1993) or sequence elimination of the paternal allele, not allele-specific transcription, could cause the presence of only maternal Mez1 transcripts. To test this possibility, DNA was isolated from endosperm tissue, and intron primers were used to amplify the genomic sequence surrounding the CAPS marker. Using PCR followed by restriction digestion, we were able to detect similar levels of DNA from both parental alleles throughout endosperm development (Figure 3c).

Sequence analysis of the Mez1 promoter

For the imprinted FWA gene, a SINE element located in FWA has been identified as an important cis-acting determinant of imprinting regulation (Lippman et al., 2004). In contrast, transposons and repeats found at the MEA locus, which show differential methyaltion patterns (Gehring et al., 2006), are not required for imprinting (Spillane et al., 2004). To search for cis-acting regulatory sequences mediating imprinting at the Mez1 locus, we obtained sequence upstream of Mez1 through analysis of a BAC clone (b0165E14) containing the Mez1 gene. The Mez1 genomic sequence (AY422167) includes approximately 2.8 kb of upstream sequence. No known retrotransposons or other known repetitive sequences were identified in this region. A simple sequence repeat (SSR) (containing 23 TA repeats) is found at bp −1662 to −1617.

Danilevskaya et al. (2003) suggested that the presence of two CpG islands could potentially be involved in ‘marking’ a gene for imprinting in plants, based on the observation of two CpG islands in the ZmFie1 coding sequence. However, using the same methods as Danilevskaya et al. (2003), only a single CpG island, near the start of transcription, was detected in the imprinted Mez1 sequence (Figure 1a). No other CpG islands were detected in this sequence.

DNA methylation analysis of the Mez1 5′ upstream region

Bisulfite sequencing was used to determine the methylation status of the Mez1 5′ proximal sequence. DNA was isolated from B73 × Mo17 F1 hybrid endosperms and treated with sodium bisulfite, which results in the conversion of unmethylated cytosines to uracil but does not affect methylated cytosines. A set of primers was used to amplify regions from the (TA)23 SSR through to the first exon of Mez1 (Figure 4a). In addition, primers that target the region 5′ of the (TA)23 SSR and exon 9 near the middle of the Mez1 gene were also used. In most cases, the same primers could be used to amplify the B73 and Mo17 alleles; however, for some regions we needed to design B73- and Mo17-specific primers due to sequence polymorphisms. The resulting PCR products for each of these regions were cloned and sequenced. For most of the regions, we were able to determine whether each of the individual clones was maternally or paternally derived based upon B73/Mo17 polymorphisms. The relative levels of maternal and paternal methylation are shown in Figure 4(b,c). There is very little methylation present in the sequences near the transcription start site of Mez1. However, a 556 bp differentially methylated region (DMR) is located from −677 to −1232 bp relative to the transcription start site. This region displays very high levels of methylation in the CpG and CpNpG context on the paternal allele of Mez1, but low levels of methylation on the maternally inherited Mez1 alleles (Figure 4). In contrast, asymmetric cytosine residues showed very little methylation for either maternal or paternal Mez1 alleles. There was no evidence for differential methylation in the region 5′ of the (TA)23 SSR or in exon 9 of the Mez1 gene (data not shown). These regions both display virtually no methylation.

Figure 4.

 Methylation status of the Mez1 promoter.
DNA was isolated from B73 (female) × Mo17 (male) endosperm tissue and treated with sodium metabisulfite. Following conversion, primers were used to amplify different regions of the Mez1 promoter. The PCR products were cloned and sequenced to determine the methylation patterns of the forward strand of DNA. In most cases, B73/Mo17 sequence polymorphisms allowed the unambiguous determination of the parent of origin for each bisulfite clone.
(a) Diagram of the Mez1 promoter. The regions amplified by PCR and cloned are shown in orange, with the number of paternal and maternal clones analyzed shown above and below the lines, respectively. The differentially methylated region (DMR) is outlined in red. The primers used for verification are shown as green arrows, and the location of the BstUI site is also shown.
(b, c) The percentage methylation at each CpG (red), CpNpG (blue) and CpNpN (gray) is plotted for the maternal and paternal alleles. Cytosines for which no data were obtained are shown below the line. Almost no asymmetric methylation was detected in the regions that were tested.

A methylation-sensitive PCR assay that determines the methylation status of a methyl-sensitive BstUI site was developed to confirm this differentially methylated domain. Mock and BstUI digests were performed using genomic DNA from inbred and reciprocal F1 hybrid endosperm tissue. PCR amplification was performed using the mock- and BstUI-digested DNA to assess whether this site was methylated (amplification following BstUI digest indicates that this site was methylated and protected from digestion). In order to determine whether methylation was present on one or both of the two alleles, the amplified DNA (representing the methylated fraction) was digested with HaeIII, which differentiates the B73 and Mo17 alleles (Figure 5a). Analysis of the reciprocal F1 hybrid samples indicates that the paternal Mez1 allele is protected from cleavage by BstUI and is therefore methylated (Figure 5a,b). The ability to amplify both alleles following BstUI digestion of DNA isolated from F1 hybrid immature ear tissue indicates bi-allelic methylation of this site in vegetative tissues (Figure 5). We also tested for methylation at a second BstUI site that lacked methylation according to our bisulfite analysis. The failure to amplify a product following BstUI digests of genomic DNA confirms the lack of methylation at this site (data not shown).

Figure 5.

 . Confirmation of differential methylation in maternal and paternal alleles by methyl-sensitive PCR.
(a) In order to confirm the presence of differential methylation of the maternal and paternal alleles in the endosperm, we employed a combination of methyl-sensitive PCR and CAPS analysis. Primers (horizontal arrows) that flank a BstUI site that is predicted to be methylated on the paternal allele and demethylated on the maternal allele were used to perform PCR on mock-digested and BstUI-digested genomic DNA that was derived from homozygous and reciprocal heterozygous endosperm DNA samples. Following PCR amplification, the DNA was digested with HaeIII which has an additional site in the Mo17 allele that is absent in the B73 allele.
(b) The mock digest shows that both alleles can be amplified and detected using this system.
(c) Following digestion with BstUI, we were able to determine that, in the reciprocal hybrids, only the paternal allele was protected from digestion and amplified. This result confirms the paternal specific methylation of this site. In addition, the same assays were performed using DNA isolated from immature ear tissue and provide evidence for bi-allelic methylation in vegetative tissues.

Phylogenetic relationships of imprinted and non-imprinted E(z)-like genes

Maize and Arabidopsis both contain three E(z)-like homologs (Springer et al., 2003). While alignments and phylogenetic analyses based on the protein sequences of these genes suggest that MEA is highly divergent, there is evidence that MEA is a relatively recent duplicate of SWN (EZA1). This evidence is based upon the location of these two genes in collinear regions of the Arabidopsis genome and phylogenetic analyses based upon synonymous sequence changes (C. Spillane, S. Laouielle, K. Wolfe and U. Grossniklaus, unpublished data). An alignment was performed beginning with the protein sequence and then selecting a conserved region for performing DNA alignments for the maize, Arabidopsis, rice, Sorghum and poplar E(z)-like genes. The DNA alignment was then analyzed using MEGA to obtain a bootstrapped neighbor-joining tree based on synonymous or non-synonymous changes (Figure 6). Interestingly, the two genes with evidence for imprinting, MEA and Mez1, do not show close relationships with one another. Instead, both genes are more closely related to genes that do not show an imprinted pattern of gene expression.

Figure 6.

 Relationships among plant E(z)-like genes.
A 684 bp nucleic acid sequence corresponding to bp −46 to −730 of the Mez1 (AF443596) coding sequence was aligned with the following sequences using Clustal X: Drosophila Enhancer of zeste (U00180), MEDEA (AF060485), CLF (Y10580), EZA1 (AF100163), Mez2 (AF443597), Mez3 (AF443598), Sdg711 (AP003044), Sdg718 (AF407010). Neighbor-joining methods were used to analyze synonymous changes using MEGA version 2.1. Bootstrap analysis was performed using 500 replicates and values for all nodes supported at greater than 80% are shown.

Molecular evolutionary analysis of Mez genes

A major evolutionary model for the evolution of imprinting is the genetic conflict theory, which proposes that imprinting can arise where there is an antagonistic conflict between maternally and paternally inherited genes regarding resource allocation to the progeny (Haig and Westoby, 1989; Moore and Haig, 1991). In such instances, paternally derived genes are under selection to extract resources from mothers, whereas maternally expressed genes are under selection to oppose this. It is proposed that the antagonistic selection pressures will favor imprinting at such loci. Where such a conflict occurs between proteins derived from maternal or paternal alleles, it would be expected that the protein would undergo rapid evolution manifested as positive Darwinian selection (Smith and Hurst, 1998). In mammals, there has been little evidence to date for such antagonistic co-evolution at imprinted loci (Burt and Trivers, 1998; McVean and Hurst, 1997), although positive selection has been detected for the signal sequence of the imprinted Igf2r gene (Smith and Hurst, 1998). We tested for evidence of positive selection by calculating the ratio of the number of non-synonymous substitutions per non-synonymous site (Dn) to the number of synonymous substitutions per synonymous site (Ds): Dn/Ds < 1 is consistent with purifying selection, Dn/Ds = 1 is consistent with evolutionary neutrality, and Dn/Ds > 1 is indicative of positive selection. The Dn/Ds ratio calculated for the entire gene sequence was <1 for all three genes (Mez1 = 0.353138, Mez2 = 0.361675, Mez3 = 0.264317). A sliding window analysis did not provide evidence for any of the three genes having evolved in response to positive Darwinian selection (Figure 7).

Figure 7.

Dn/Ds sliding window analysis of Mez1, Mez2 and Mez3 from alignment of Zea mays and Sorghum bicolor homologs.
Window lengths of 10% of the overall alignment length were taken for each alignment, and the window was moved by half its length to produce the data points plotted. The conserved domains were taken from previous studies (Springer et al., 2002). Mez1 shows a high peak in region 460–553; on closer examination using a window size of 10 codons we found Dn/Ds ratios > 1 in a total of three windows (255–265, 480–490 and 530–540); these sites are not located in the previously described domains. Mez2 shows a number of peaks of elevated Dn/Ds in regions 308–397 (includes a small section of the EZD2 domain) and 440–529 (unknown domain); on closer examination using a window size of 10 codons we found Dn/Ds ratios > 1 in a total of three windows: 175–185 (EZD1 domain); 505–515 (unknown domain); 635–645 (region between NLS domain and Cys-rich domain). Analyses using a window size of 10 codons in length are too difficult to visualize here. Mez3 shows the greatest Dn/Ds ratio (1.11) in region 352–441, which does not fall into any of the described domains (Springer et al., 2002).


Genomic imprinting is the differential expression of an allele dependent on which parent it was inherited from. In plants, verified instances of imprinting have predominantly been documented in the triploid endosperm tissue. In Arabidopsis, the best studied example of genomic imprinting is the MEA gene. MEA encodes an E(z)-like SET domain protein. SET domain proteins are putative histone methyltransferases, and, on the basis of sequence similarity and chromatin immunoprecipitation (ChIP) assays, MEA most likely encodes a histone H3 lysine 27 methyltransferase (Gehring et al., 2006; Makarevich et al., 2006; Springer et al., 2003). The Arabidopsis genome contains two genes closely related to MEA,CLF and SWN (EZA1) (Goodrich et al., 1997; Preuss 1999). No genetic evidence for genomic imprinting has been observed for CLF and SWN, the other two E(z)-like genes present in the Arabidopsis genome (C. Spillane and U. Grossniklaus, unpublished results). We have previously characterized three maize E(z)-like genes, Mez1, Mez2 and Mez3 (Springer et al., 2002). Here, we demonstrate that the Mez1 gene is imprinted in maize endosperm whereas the Mez2 and Mez3 genes exhibit bi-allelic expression. We further demonstrate that Mez1 is imprinted in multiple genetic backgrounds, and therefore likely represents gene-specific rather than allele-specific imprinting.

Evolution of imprinted expression patterns

It is intriguing that, while there is evidence for imprinting of E(z)-like genes in both maize and Arabidopsis, these specific imprinted genes are not orthologs in these two species. Phylogenetic analysis of the plant E(z)-like genes suggests that the imprinted genes Mez1 and MEA probably diverged prior to the split of monocots and dicots. The common imprinting of MEA and Mez1, and lack of imprinting of CLF, SWN, Mez2 and Mez3, suggests either that (a) both MEA and CLF were imprinted in progenitor species and that some monocots have since lost the MEA ortholog while CLF has lost its imprinted expression pattern, or (b) convergent evolution of imprinting has occurred such that different members of the E(z)-like gene family have become imprinted in both monocots and dicots. We favor the second, more parsimonious, explanation. This question can best be resolved by studying the allele-specific expression pattern of the E(z)-like genes in several other monocot and dicot species.

There is significant copy-number polymorphism of the E(z) homologs in both dicots and monocots, such that the number of genes in this family varies between rice (two genes), sorghum (two), maize (three), Arabidopsis (three) and poplar (four). Most animal species have a single copy of E(z), while mice have two copies of an E(z) homolog. The extent and plasticity of such duplication of the E(z)-like genes in plants could be consistent with the Mez gene family undergoing neo- and sub-functionalization in terms of their roles during plant development (Ohno, 1970). Such divergence of functions due to gene duplication may also extend to divergence in terms of the imprinted status of E(z) homologs in plants. Our phylogenetic analysis suggests that Mez1 is a probable homolog of CLF, yet Mez1 is imprinted in maize but CLF is not imprinted in Arabidopsis. Conversely, while MEA is the likely homolog of Mez2, MEA is imprinted in Arabidopsis, yet Mez2 is not imprinted in maize. Further analysis of the imprinting status of E(z) homologs in a wider range of dicots and monocots will allow us a greater understanding of the role of gene duplication in the evolution of genomic imprinting (Walter and Paulsen, 2003).

While there have been suggestions that imprinted genes could be subject to adaptive or diversifying selection (Smith and Hurst, 1998), we did not detect strong evidence to suggest that Mez1 is subject to adaptive selection. A Dn/Ds analysis based on alignments of maize and sorghum sequences indicated that there is considerable variation in selection pressures/evolutionary rates across the Mez1, Mez2 and Mez3 homologs, with a Dn/Ds > 1 peak in the Mez3 gene (Figure 7). For all three Mez genes, there is evidence of purifying selection for the SET, SANT and EZD1 domain regions.

Cis-acting sequences that control imprinting

The identification of additional examples of imprinting in plants allows further testing of the emerging models of imprinting regulation. Recent studies have documented a role for DNA methylation in the control of imprinted gene expression patterns (Gutierrez-Marcos et al., 2006; Jullien et al., 2006a; Kinoshita et al., 2004; Luo et al., 2000; Vielle-Calzada et al., 1999; Xiao et al., 2003). It has been proposed that allele-specific loss of DNA methylation, induced by DME, plays a critical role in imprinting (Gehring et al., 2006). We have documented the presence of bi-allelic methylation of a region of the Mez1 5′ proximal sequence in vegetative tissues but mono-allelic methylation patterns in the endosperm tissue. These findings are similar to those described for MEA in Arabidopsis at late stages of seed development (Gehring et al., 2006).

An important unresolved question regarding imprinted gene expression revolves around understanding the cis-acting sequences that determine imprinted expression. We were able to compare the sequence of the Mez1 promoter, and specifically the DMR, with the sequence of promoters of other imprinted genes in plants. In our analysis, we did not detect any primary sequence motifs common to the promoters of the known imprinted genes in Arabidopsis and maize. In addition, we did not detect the double CG island noted by Danilevskaya et al. (2003) at the ZmFie1 locus. However, it is possible that our search window of 2766 bp was too limited and needs to include more distal sequences. In animals, some of the imprinting regulatory elements can occur at loci >100 kb distal from the transcribed region (Ferguson-Smith and Surani, 2001), and the regulation of paramutation in maize can involve sequences over 100 kb away from the gene itself (Stam et al., 2002). Alternatively, as the primary motifs responsible for imprinted transcription have not been defined, it could be that our search algorithms could not detect motifs within the region studied. These motifs may be defined by features other than primary sequence. Further experimentation is necessary to map additional cis-acting sequences involved in the imprinted transcription of Mez1.

The majority of examples of imprinting in plants share the common features of maternal-specific expression in endosperm tissue. These imprinted genes can be divided into two classes based on whether they are also expressed in other plant tissues. FWA, ZmFie1 and Nrp1 are all imprinted in the endosperm and not expressed at detectable levels in other plant tissues (Danilevskaya et al., 2003; Guo et al., 2003; Kinoshita et al., 1999). However, the MEA and Mez1 genes are expressed in vegetative tissues including the embryo as well as in endosperm tissue. Assuming a plant imprinting model of DNA methylation as the default state and allele-specific demethylation in the endosperm would suggest that these genes might be methylated and expressed in vegetative tissues. Indeed, there is evidence for bi-allelic methylation and expression of Mez1 in immature ear tissue. It will be interesting to monitor the developmental time course of DNA methylation patterns for these genes and how this methylation affects their expression. Future experiments will elucidate the role of DNA methylation and histone modifications in controlling imprinting.

Experimental procedures

Plant materials

Maize inbred lines were grown using standard conditions. Leaf tissue was harvested for DNA collections. Reciprocal crosses and self-pollinations were performed. The ears were harvested at 7, 10, 13, 17, 20, 23 and 27 DAP. Endosperm and embryo tissue was collected by dissection. Tissue derived from multiple kernels from the same ear was pooled prior to performing RNA or DNA extractions.

Identification of SNPs for CAPS analysis

In order to monitor imprinting, it was necessary to identify polymorphisms within the coding regions of the Mez1, Mez2 and Mez3 genes. A region corresponding to the 3′ coding region and UTR from Mez1 and Mez3 was amplified and sequenced from a set of 24 maize inbred lines. DNA was extracted from leaf tissue using the Qiagen DNeasy plant mini kit (Valencia, CA, USA) according to the manufacturer's instructions. PCR reactions were performed in a 30 μl total volume containing approximately 50 ng of DNA, 5 pmol of each primer, 0.65 units of HotStarTaq polymerase (Qiagen), 3 μl of 10 x reaction buffer and 0.2 μl of 25 mm dNTPs. Primers used were mez1F33 and mez1R33 for Mez1, mez2_SF and mez2_SR for Mez2, and mez3_SF and mez3_SR for Mez3. All primer sequences are provided in Table S2. PCR conditions were as follows: 94°C for 15 min, 35 cycles of 94°C for 30 sec, 58°C for 30 sec, 72°C for 1 min, followed by 72°C for 10 min.

For each of the three genes, we identified sequence polymorphisms that allowed the development of a cleaved amplified polymorphism sequence (CAPS) assay. Sequence information from B73 and Mo17 identified a CAPS marker in Mez2. This CAPS marker was used to classify the remaining inbred lines (Table S1). These CAPS assays can be employed to detect the presence of both alleles in genomic DNA or cDNA samples. The Mez1 and Mez2 CAPS assays rely upon a polymorphism in an AluI restriction site, while the Mez3 CAPS assays utilizes a polymorphism in an HaeIII site.

CAPS analysis of allele-specific expression patterns

RNA was isolated from endosperm and embryo tissue using the Qiagen RNeasy plant mini kit according to the manufacturer's instructions. Contaminating DNA was removed by digestion with RQ1 DNase (Promega, Madison, WI, USA). Total RNA (5 μg) was mixed with 0.5 μg oligo(dT) (Promega) and heated to 70°C for 10 min followed by 1 min on ice. First-strand cDNA synthesis was performed by adding 6 μl 5 x reaction buffer, 0.5 μl RNasin, 3 μl 5 mm dNTPs and 1 μl M-MLV reverse transcriptase (Promega). This reaction was incubated at 42°C for 50 min, followed by 70°C for 15 min. The resulting cDNA was purified by phenol:chloroform extraction and ethanol precipitation. The cDNA was resuspended in 20 μl ddH2O. PCR reactions were performed as above, and the amplified DNA was ethanol-precipitated and resuspended in 20 μl H2O. Restriction digestions were performed by mixing 10 μl DNA, 2 μl 10 x reaction buffer, 0.2 μl BSA, 2 μl AluI or HaeIII and 5.8 μl H2O, and incubating at 37°C overnight. The digested products were separated by electrophoresis in a TBE gel containing 2.5% Metaphor (FMC Bioproducts, Rockland, ME, USA) and observed by ethidium bromide staining.

Analysis of genomic DNA CAPS markers

Genomic DNA was extraction from endosperm tissue using the Qiagen DNeasy plant mini kit according to the manufacturer's instructions. PCR amplification was performed as described above using the primers mez1F20 and mez1R9 (Table S2). The amplified DNA was processed in the same manner as the amplified cDNA products.

Mass spectrometry-based analysis of allele-specific expression

RNAs from endosperm tissues at 13 and 19 DAP were treated with DNAse prior to allele-specific expression analyses. cDNAs were synthesized from all three biological replicates of Mo17 × B73 and B73 × Mo17 hybrid RNAs. Mixed cDNAs were also synthesized from 2:1, 1:1 and 1:2 mixes of the biological replicates of Mo17 and B73 inbred RNAs. The cDNAs were reverse-transcribed using Superscript III reverse transcriptase, according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA).

PCR-based assays for allele-specific expression analyses based on SNPs were designed in collaboration with Sequenom (San Diego, CA, USA) (see Table S2 for PCR and extension primers for these assays). PCR and extension PCR reactions on cDNA and DNA templates were performed according to the manufacturer's specifications. Mass spectrometry quantification of allele ratios was performed at the University of Minnesota Genotyping Facility. Multiple measurements of the ratio of the two alleles were performed for each of the three biological replicates of mixed RNA and F1 RNAs.

Bisulfite sequencing

Endosperm tissue was dissected from B73 × Mo17 and Mo17 × B73 kernels at 18 DAP. DNA was isolated using the Qiagen DNeasy kit according to the manufacturer's protocol. DNA aliquots (2 μg) were digested with 10 units of BamHI and HindIII in a volume of 100 μl at 37°C overnight. Digested DNA was extracted with phenol:chloroform, precipitated with 100% ethanol supplemented with 2 μl tRNA, and resuspended in 20 μl H2O. DNA was denatured at 97°C for 1 min, snap-cooled on ice for 2 min and incubated with 1 μl 6.3 m NaOH (freshly prepared) for 30 min at 39°C. Bisulfite solution was prepared by dissolving 40.5 g sodium bisulfite in 80 ml H2O with slow stirring, adjusting the pH to 5.1 using freshly prepared 10 m NaOH, adding 3.3 ml 20 mm hydroquinone, adjusting the volume to 100 ml, and protecting from light, and 208 μl of this solution was added to DNA at 39°C. Samples were incubated for five cycles of 55°C for 3 h, 95°C for 5 min. Samples were cleaned using a Qiagen PCR purification kit according to the manufacturer's protocol (eluted in 100 μl Qiagen elution buffer). Then 5 μl 6.3 m NaOH was added, and the mixture was incubated at 37°C for 15 min. DNA was precipitated with 0.1 vol 10 m NH4OAc, 3 volumes of 100% ethanol and 2 μl tRNA (resuspended in 100 μl Qiagen EB), and 2 μl of bisulfite-treated DNA was used per PCR reaction. PCR components were 4 μl dNTPs, 5 μl 10 x ExTaq buffer (Takara, Otsu, Shiga, Japan), 1 μl reverse primer and 38 μl H2O. Primers used were mez1BSr7-F5 and mez1BSr7-R9M, mez1BSr7-F5 and mez1BSr7-R9B, mez1BSr6-F3M and mez1BSr6-R8, mez1BSr6-F3B and mez1BSr6-R8, mez1BSr5-F2M and mez1BSr5-R3, mez1BSr5-F2B and mez1BSr5-R3, mez1BSr4-F7 and mez1BSr4-R10, mez1BSr3-F6 and mez1BSr3-R9M, mez1BSr3-F6 and mez1BSr3-R9B, mez1BSr2-F4M and mez1BSr2-R8, mez1BSr2-F4B and mez1BSr2-R8 (Table S2). PCR conditions were as follows: 95°C for 5 min, followed by addition of 1 μl ExTaq, then five cycles of 95°C for 20 sec, 60°C for 3 min, 72°C for 3 min, addition of 1 μl forward primer, then 10 cycles of 95°C for 20 sec, 60°C for 1.5 min, 72°C for 2 min and 30 cycles of 95°C for 20 sec, 50°C for 1.5 min, 72°C for 2 min, followed by 72°C for 10 min. PCR products were excised from a low-melting point agarose gel and extracted using the Qiagen gel extraction kit according to the manufacturer's protocol (resuspended in 30 μl EB). The PCR product (2 μl) was cloned using a TA TOPO cloning kit (Invitrogen). Plasmid DNA was isolated, and the insert size was verified by PCR using M13F and M13R primers and sequenced.

Bisulfite verification

Genomic DNA was extracted as described above from B73, Mo17, B73 × Mo17 and Mo17 × B73 endosperm tissue and B73 × Mo17 and Mo17 × B73 immature ear tissue. DNA (500 ng) from each tissue was digested with 10 units BstUI or 1 μl glycerol (mock digest) in a total volume of 40 μl at 60°C for 4 h. The reaction was cleaned up using a Qiagen PCR purification kit according to the manufacturer's protocol (eluted with 50 μl EB). Aliquots (4 μl) of DNA were then PCR-amplified as described above using mez1F55 and mez1R52 primers (Table S2). PCR products were ethanol-precipitated and resuspended in 30 μl H2O. DNA was then digested with 20 units HaeIII at 37°C for 4 h. The digested products were separated by electrophoresis in a 2.5% Metaphor TBE gel and observed by ethidium bromide staining.

Sequence analysis

The amino acid and nucleic sequences for Mez1, Mez2, Mez2, Sdg711, Sdg718, Medea, Clf, Eza1 and E(z) were aligned using Clustal X version 1.83 (Thompson et al., 1997). These alignments were edited, and the phylogenetic relationships were analyzed using MEGA (Kumar et al., 2001). The neighbor-joining method was used and a bootstrap analysis with 500 replicates was performed. The Mez1, MEDEA, ZmFie1 and Sdg711 promoter sequences were tested for the presence of conserved non-coding sequences using the methods described by Kaplinsky et al. (2002).

DNA sequencing of sorghum Mez homologs

For DNA sequencing of sorghum Mez homologs, both cDNA clones and genomic DNA from Sorghum bicolor were used as templates. Genomic DNA of S. bicolor (breeding line IS3620C) was obtained from Patrick James Brown (Cornell University, Ithaca, NY, USA). cDNA clones of Mez homologs from S. bicolor (line Btx623, code A002) were obtained from Lee Pratt (University of Georgia, Athens, GA, USA). These cDNA clones were HS1-16-D08 and POL1-41-A10. The following primer pairs were used for amplification and sequencing of the Mez1 homolog from the HS1-16-D08 clone: mez1-F1 and mez1-R1; mez1-F2 and mez1-R2; mez1-F3 and mez1-R3; mez1-F4 and mez1-R4 (Table S2). The following primer pairs were used for amplification and sequencing of the Mez2/3 homolog from the POL1-41-A10 clone and S. bicolor genomic DNA: mez2/3-F1 and mez2/3-R1; mez2/3-F2 and mez2/3-R2; mez2/3-F3 and mez2/3-R3; mez2/3-F4 and mez2/3-R4 (Table S2).

Molecular evolutionary analysis

For each homologous sequence pair, the coding DNA sequence was translated into protein sequence, and an alignment was generated using Clustal W (Li, 1993; Thompson et al., 1994). Alignments generated and used in this analysis are available in Figure S1. Using the aligned protein sequence and the original unaligned nucleotide sequences, gaps were placed in the nucleotide sequences based on their corresponding amino acid position. For each alignment, the number of non-synonymous substitutions per non-synonymous site (Dn) and the number of synonymous substitutions per synonymous site (Ds) was calculated. The sliding window method implemented in CRANN (Creevey and McInerney, 2003) and the Li (1993) method for unbiased calculation of Dn and Ds were used. Values for Dn and Ds were calculated for overlapping sections/windows of the alignment by sliding half the window length in a left to right motion for the entire length of the alignment. For all alignments, an initial window size of 10% of the overall alignment length was used. An additional analysis using a window size of 10 codons in length was applied to the Mez1 and Mez2 alignments to further clarify findings in the Dn/Ds profile. The conserved domains represented in Figure 7 were taken from Springer et al. (2002).


The authors would like to acknowledge Peter Hermanson, Robert Stupar, Sarah Kerns and Virginia Zaunbrecher who performed nucleic acid extractions and provided technical assistance, and Olga Danilevskaya and Pedro Herman who helped perform the CpG island analysis. We are grateful to Steve Jacobsen for providing the bisulfite sequencing protocol. This work is supported by an NSF grant DBI-0421619 to S.M.K. and N.M.S., the University of Zürich, and by a Science Foundation Ireland (SFI) grant supporting C.S., M.J.O'C. and S.L.