Nucleotide sequence variation can potentially alter protein function if it occurs in the coding region and causes a qualitative difference. Alternatively, regulatory allelic variants can affect the level of gene expression and result in quantitative variants. Both types of allelic variation may produce functional changes and affect the phenotype of an organism. Naturally occurring allelic diversity in plants has been suggested to be an important genetic component for phenotypic variation (Doebley and Lukens, 1998; Buckler and Thornsberry, 2002). In contrast with induced mutations that eliminate or cause a large reduction of a functional gene product, naturally occurring allelic variation modulates gene products and may be an underlying mechanism for quantitative trait variation (Tanksley, 1993; Yano and Sasaki, 1997; Mackay, 2001). A large amount of nucleotide sequence data has been generated as facilitated by recent advances in DNA sequencing technology. The sequence data reveals that nucleotide sequence variation widely exists within a species. The maize (Zea mays) genome has an especially high level of DNA sequence polymorphism, approximately an order of magnitude higher than that in humans (Sunyaev et al., 2000; Bhattramakki et al., 2002; Buckler and Thornsberry, 2002; Ching et al., 2002). The types of allelic variants range from single nucleotide polymorphisms (SNPs), insertion/deletions (InDels), to large variants in structure involving several kilobases of DNA fragments (Fu and Dooner, 2002). These allelic variants may play a role in gene regulation and affect the expression level. However, the relationship of allelic expression differences resulting from changes in a regulatory region and the resultant phenotype is poorly understood because of its complexity and the lack of efficient methodology (Cowles et al., 2002; Glazier et al., 2002).

Recent studies have shown that allelic variation of expression level occurs frequently in mammals. It has been reported in mouse hybrids (Cowles et al., 2002) and humans (Yan et al., 2002) that alleles of nonimprinted autosomal genes are not expressed equally at the transcript level, and such allelic expression variation has been suggested as a regulatory mechanism involved in genetic diseases. Therefore, the difference in the regulatory regions of a gene could affect the pattern of gene expression and cause phenotypic variation. The importance of this allelic expression variation in determining phenotypic diversity has been reviewed recently by Knight (2004). Virtually nothing is known regarding such allelic variation of gene expression at the transcript level in maize. The high level of allelic variants in maize could lead to a high level of allelic expression variation. This hypothesis can be tested by comparing the two alleles in the maize hybrid where the alleles are exposed to a common genetic and environmental context.

Furthermore, for nearly a century, geneticists and breeders have known that the progeny of crosses between unrelated individuals often exhibit hybrid vigor or heterosis, a phenomenon of the superior performance of hybrid progeny in comparison with their inbred parents (Shull, 1908). The genetic difference between the inbred parents presumably contributes to the genetic basis of heterosis. The allelic sequence differences between inbreds have been characterized by analysis of SNPs or InDels (Helentjaris et al., 1985; Smith and Smith, 1992). Very little is known regarding the allelic differences at the transcript level in the hybrid.

In this study, we used RT-PCR to measure the proportion of transcript contributed by each parental allele expressed in the maize hybrid. We further analyzed the allele-specific transcript variation of hybrids of recent versus past development. Because yield gains of modern hybrids are primarily attributable to the genetic improvement in tolerance to both biotic and abiotic stresses (Janick, 1999; Duvick, 2001), we sought to compare allelic expression in a modern and an old hybrid and hybrids under high planting density and drought stress. Results from this study showed the predominance of allelic expression variation at the accumulated transcript level in maize hybrids, different allelic expression patterns between the two hybrids analyzed, and allelic variation in responding to environmental stresses.

Figure 1. Schematic Illustration of Allele Expression Analysis Using the WAVE dHPLC System and Genomic PCR as Control.

Although we designed the PCR primers at the conserved regions between alleles, we conducted genomic DNA PCR analysis with hybrid S1/NS1 for four genes to test for any PCR amplification bias for either allele, which may confound the allele expression data. The greatest deviation from the expected allelic ratio of 1.0 in the hybrid genomic DNA PCR analysis was 0.85 and was not significantly different from 1.0 (Table 1). Therefore, the allelic transcript ratios generated from the cDNA PCR are not likely because of any amplification bias. Subsequently, we used an allelic ratio cutoff of 0.85 (or 1.18, the inverse of 0.85) as a threshold for allelic transcript ratios deviating significantly from 1.0. We also tested whether the allelic transcript ratio was confounded by amplification of gene family members during RT-PCR. We used the PCR primer sequences in BLAST (Altschul et al., 1990) searches against the Pioneer/DuPont EST database to confirm that the primer sets did not have any significant matches to other genes. Amplicon sequencing and WAVE chromatograms confirmed that only one PCR product was amplified with each primer set.

Table 1. Allele-Specific Transcript Ratio in Hybrid S1/NS1 Grown in Different Environments

We analyzed gene expression using immature ears, prepollination of hybrid 3394 (S1/NS1), which were grown in three environments (see Methods). We first selected a random set of genes for this study from the Pioneer/DuPont EST databases. We then obtained partial sequences of these genes for S1 and NS1 alleles. Fifteen genes were selected for allelic expression analysis based on the criteria of presence of allele sequence polymorphisms resolvable by the WAVE dHPLC system and expression in immature ear tissue. The selection of these genes was therefore random in a sense that it involved little biological input from the perspective of this study. It is important to note that only partial regions of these genes (not the entire gene) were analyzed to find allele-specific polymorphism as a marker (see supplemental data online). Very often, the allelic sequence polymorphisms were located at the 3′ untranslated region. The relative allele expression in the hybrid is measured as a ratio of the transcript amount from the NS1 parental allele relative to that from the S1 parental allele. An equal allele expression in the hybrid would give a transcript ratio of NS1:S1 = 1.0. Allele-specific transcript ratios for nine genes expressed in this hybrid deviated from 1.0 (P < 0.05) in at least one environment, indicating unequal transcript accumulation between the two alleles (Table 1). The allelic variation ranged from near equal expression (biallelic) to only one allele (monoallelic) being expressed at a detectable level.

When comparing hybrids that were grown in different environments, we found five genes exhibited significant environmental variation in their allelic ratio (Table 1). The most dramatically affected genes were LTP (Lipid transfer protein) and ARDA (Auxin repressed dormancy associated protein). The LTP gene showed monoallelic expression in one environment but biallelic expression in others (Figure 2A). These genes are known to respond to stress (Treviño and O'Connell, 1998; Park and Han, 2003; Yubero-Serrano et al., 2003). Based on our microarray expression profiling experiments (our unpublished data), we know that these two genes also showed significantly reduced expression under abiotic stress in maize immature ear tissue. The allelic transcript data indicates that different alleles of LTP and ARDA responded differentially in different environments. Because only one biological sample was analyzed in each environment, some of the environmental variation could be confounded with variation between plants, although each sample consisted of three pooled individual plants.

Figure 2. WAVE Chromatograms Showing Allelic Variation in Transcript Accumulation in Hybrids.

Allele expression variation in different tissue types and developmental stages was examined with four genes, Ribo-S29 (Ribosomal S29 protein), Histone H4, GGT (Geranyl geranyl transferase), and LTP and in the following three tissues: primary ear, secondary immature ears (prepollination), and seedlings of the hybrid. We sampled these four genes based on their detectable expression in all three tissues. The Ribo-S29 gene demonstrated equal allelic expression. Ribo-S29 and LTP did not show significant difference in different tissues. GGT and Histone H4 exhibited developmental variation in allele expression (i.e., the allelic difference was greater in the primary immature ear compared with the other two tissues) (Figure 2A, Table 2). The primary ear is developmentally more advanced compared with the secondary immature ear and seedling. Although with a limited number of genes and tissues, the results may suggest a differential regulation of the two parental alleles in these tissues and a trend toward increased allele expression difference as development progresses.

Table 2. Allelic Variation in Different Tissues of Hybrid S1/NS1

Table 3. Allele-Specific Transcript Ratio in Hybrids of Reciprocal Cross

Seven of the 15 genes analyzed in hybrid S1/NS1 had resolvable allele sequence polymorphisms in hybrid S2/NS2. Allele-specific expression of these seven genes was analyzed in seedlings of both hybrids grown in the greenhouse (see Methods). Three biological replications were sampled; each consisted of three individual plants. Distinct gene expression patterns were found between the two hybrids (Figures 2C and 2D). Whereas the modern, improved hybrid S1/NS1 expressed both alleles of all 15 genes, the less improved hybrid S2/NS2 expressed only a single allele in four of the seven genes tested. The monoallelic expression in the S2/NS2 hybrid was not because of the absence of expression in one parent because the gene was expressed in both inbred parents.

Figure 3. Allele Expression Variation in Response to Density and Drought Stresses.

Table 4. Allele-Specific Transcript Ratio in Hybrids under Density and Drought Stresses

Allelic difference in the accumulated transcript level could result from various regulatory mechanisms, including transcriptional and posttranscriptional regulations. Differential mRNA degradation, for instance, is a posttranscriptional mechanism that controls transcript abundance. Gene regulation by RNA degradation appears to be limited to only a small fraction of genes and involves specific functional categories in Arabidopsis thaliana (Gutiérrez et al., 2002). Although RNA degradation is a possible mechanism, such regulation would involve regions with allelic sequence variation responsible for the allele specificity of the RNA degradation. Although other possibilities cannot be excluded, presence of allelic regulatory variants, such as cis-acting sequence polymorphisms, might cause allelic difference in transcriptional regulation. Variants of these regulatory regions may respond differently to developmental and environmental cues and result in allele-specific expression in different tissue types, different environments, or stress response.

The significance of gene expression differences to phenotypic variation has been reported in plants. Two genes controlling quantitative traits have previously been isolated, teosinte branched1 in maize (Wang et al., 1999) and the tomato (Lycopersicon esculentum) fruit weight fw2.2 gene (Cong et al., 2002). The phenotypic variation controlled by both genes is regulated by variation in expression, either in the level or timing of expression rather than from protein coding differences between alleles. Recent studies in mouse hybrids (Cowles et al., 2002) and humans (Yan et al., 2002; Lo et al., 2003) have proposed that such allelic expression differences may be a regulatory mechanism responsible for genetic diseases and human variability. If allelic variation in the regulatory region were responsible for the allele-specific transcript differences observed in this study, such regulatory allelic variants would exist to a great extent in natural populations. By this assumption, our results would be consistent with the finding of numerous trans-acting factors that modulate gene expression, which are hypothesized to be an important mechanism for quantitative variation (Guo and Birchler, 1994). Because in the hybrid the two alleles are exposed to the common trans-acting regulatory factors, the allelic sequence variation of the cis-regulatory elements may cause a target gene to interact or bind differentially with the transcriptional factors, thus resulting in differential transcription between alleles. However, more direct evidence is needed to determine the regulatory mechanism of the allelic expression differences. Whether such allelic variation indeed causes any phenotypic variation remains to be tested.

We examined the possibility that a parent-of-origin effect is responsible for the allele-specific expression variation observed in this study. The allele-specific transcript ratio for a majority of genes tested was not affected whether the allele was inherited maternally or paternally. Therefore, parental effect did not play a significant role in allele-specific transcript variation. However, two genes, ARDA and ZmPIP1-3, showed significant differences in their allele-specific transcript accumulation upon maternal versus paternal transmission, exhibiting a maternally preferred expression. The maternal effect could be because of cytoplasmic effects or genomic imprinting. Both possibilities need to be tested. Genomic imprinting in sporophytic tissue of plants has not been reported to date. However, using the RT-PCR/dHPLC method, we recently discovered an imprinted gene in the maize endosperm (Guo et al., 2003).

Recent studies in maize (Fu and Dooner, 2002; Song and Messing, 2003) show that some genes can be present in one inbred but missing in the other and suggest that the complementary or synergetic effect from such allelic diversity may be the underlying mechanism of heterosis. Although consistent with this concept of allelic complementation in hybrids, our data also suggest that a gene present in one inbred allele may not always be expressed or expressed in the correct spatial or temporal pattern to fulfill the complementation role to the other inbred allele in the hybrid. It is important to note that complementation at the level of gene regulation and the level of protein function is not mutually exclusive.

The differential expression between the alleles could potentially result in hybrids surpassing the inbred parents in expression in different dimensions, such as (1) expression level, (2) expression timing/duration, and (3) response to developmental and environmental cues. It has been reported in human gene expression that complementation of alleles produces expression patterns beyond the range of the parents in spatial and temporal patterns (Rockman and Wray, 2002). It remains to be tested whether such expression indeed occurs in maize hybrids.

In summary, this study shows that the allelic expression variation occurred frequently in maize hybrids. The data suggest that the two parental alleles in maize hybrids may be regulated differentially during plant development and in response to environmental signals. Although only a small number of genes were analyzed using one each of the hybrid, distinct allelic expression patterns were found between a modern and an old hybrid. This work demonstrates that the maize hybrid is an excellent system to study allele expression variation because alleles are compared within the same genotype of a hybrid and equally affected by genetic background or environmental factors.

The high planting density and drought stress experiments were conducted in the greenhouse for hybrids 3394 and 3306. For the density experiment, plants were grown in three replicated flats at a density of 20 plants per flat (11 × 21 inches) for the control and 250 plants per flat for the high-density stress treatment. Tissue was collected for RNA expression analysis at the V4 leaf stage. At this stage, the high-density plants displayed stress symptoms, including senescence of the first leaf and yellowing of the second leaf. Control plants appeared normal. For the drought experiment, we grew hybrids at a density of 50 plants per flat in three replicated flats. At the V4 leaf stage, irrigation was withheld for drought stress, whereas the control was irrigated regularly. The stressed plants started leaf rolling 4 d after the last irrigation. Tissue was sampled on the fifth day, 4 h after all plants showed leaf rolling and reached permanent wilting. From each replicate, three whole seedlings were pooled to make one RNA sample for gene expression analysis, and three biological replicates were sampled. All tissue samples were frozen immediately in liquid N₂ and stored at −80°C.

The authors thank Xiaofeng Yang, Suling Zhao, Jo Dieter, Erin McLaughlin, Guo-Hua Miao, Dave Ritland, Lynn Heetland, Shifu Zhen, and Susan Nilles for support with this work. The authors specially thank Olga Danilevskaya for valuable discussions and Michael Muszynski for critical reading of the manuscript.