Abiotic stress-induced changes in duplicate gene expression in a natural and synthetic allopolyploid
In this study, we assayed the expression of 30 gene pairs under five abiotic stress treatments and in three organ types for a total of c. 450 gene pair/stress/organ combinations, plus untreated controls, in G. hirsutum. We found that a large majority of the duplicate gene pairs assayed (23 of 30 pairs; 77%) showed changes in the relative contributions of the homeologs to the transcript pool under one or more abiotic stress treatment and in one or more organ type. Among the 23 gene pairs that showed expression changes under abiotic stresses, 11 had biased expression level differences of > 20% (e.g. 60 : 40 to 80 : 20) compared with untreated plants. There were five gene pairs in which the relative contribution of the At homeolog increased after one stress, compared with untreated plants, but decreased after another stress. This phenomenon is analogous to duplicate gene pairs in which one copy is more highly expressed in some organ types and the other copy is more highly expressed in other organ types, referred to as ‘complementary expression’ (Ganko et al., 2007) or ‘quantitative subfunctionalization’ (Force et al., 1999; Duarte et al., 2006). Thus, this study indicates that stress-specific complementary expression (or quantitative subfunctionalization) has occurred for some duplicate gene pairs in polyploid cotton. (Quantitative subfunctionalization contrasts with qualitative subfunctionalization in that qualitative refers to no expression of one copy in one organ type (or tissue, cell type, condition, etc.) and no expression of the other copy in another organ type (Force, 1999). We think that the term ‘complementary expression’ provides a clearer description of the phenomenon reported here.)
In some cases, the effects of the abiotic stress treatments on duplicate gene expression varied by organ type. Many of the duplicate gene pairs examined in this study showed organ-specific responses to stress treatments (that is, At : Dt expression levels changed after stress treatment in only one or two of the examined organ types). Among these genes, four pairs showed reciprocal changes in At : Dt expression levels among different organs under the same stress treatment, that is, in one organ there was an increase in At : Dt expression and there was a decrease in another organ after the same abiotic stress treatment. Organ-specific responses to stress treatments were not unexpected because homeologous gene expression patterns in polyploids often vary by organ type (reviewed in Adams, 2007).
Overall, it appears that changes in the contributions of homeologous genes to the total transcript pool (transcriptome) in response to abiotic stresses is a common phenomenon in polyploid cotton. Considering that a large majority of the genes examined were abiotic stress regulated in cotton, or homologs of genes regulated by at least one abiotic stress in A. thaliana, 77% of duplicate gene pairs showing changes might not be generally applicable when applied to the whole G. hirsutum transcriptome, but our findings are probably applicable to stress-regulated genes. In the only previous study of the effects of abiotic stress treatments on duplicate gene expression in a polyploid plant, qualitative subfunctionalization in response to abiotic stress conditions was discovered (Liu & Adams, 2007). In that study, only one homeolog of the alcohol dehydrogenase gene AdhA was expressed in hypocotyls under water submersion treatment and only the other copy was expressed in hypocotyls under cold stress treatment (Liu & Adams, 2007). In this study, we did not find any cases of qualitative subfunctionalization, but we did find many cases of changes in the amount of expression of one duplicate relative to the other in response to abiotic stress treatments.
Our findings of altered duplicate gene expression levels and complementary expression in response to abiotic stress treatments demonstrate that abiotic stresses can have a major influence on the expression of duplicated genes, which may aid in the preservation of duplicated genes over evolutionary time. All of the duplicate gene pairs examined in G. hirsutum have multiple amino acid sequence differences (Table S5), although it is not known which of these amino acid differences affect function. Nevertheless, it is possible in a few cases that the product of one of the two homeologs performs better under abiotic stress conditions, and thus there has been selection for a higher expression level of that homeolog relative to the other homeolog, caused by mutations in cis-regulatory elements. The alteration of the expression patterns of some types of gene, such as transcription factors, in response to one or more abiotic stress conditions could affect the expression levels of downstream genes that they regulate. In that regard, the altered gene expression patterns of some genes in response to stress conditions could be the result of the process just mentioned, and not a result of selection on the genes themselves. Alternatively, there may not be any selection acting on some of the genes or their regulators and the expression level changes in response to stress conditions may be neutral, or of no functional consequence, rather than being adaptive.
In addition to finding gene pairs in G. hirsutum whose comparative expression levels change in response to abiotic stresses, we also found that high salt, cold and heat stress treatments had effects on At : Dt expression levels in a synthetic cotton allopolyploid. This finding shows that the phenomenon is not limited to natural allopolyploid species. There are no previous studies of homeologous gene expression changes in response to abiotic stress treatments in synthetic polyploids. However, the results from the synthetic allopolyploid are analogous to those of a previous study of allelic expression in diploid maize hybrids, which showed changes in allelic expression ratios of four genes in response to drought stress (Guo et al., 2004).
There was little correspondence between stress-induced changes in duplicate gene expression in synthetic and natural allopolyploids. One possibility is that there were similar stress-induced expression changes on allopolyploidy, followed by considerable divergence in the expression patterns in the natural polyploid G. hirsutum, perhaps caused by selection in response to environmental stresses. Another possibility is that the phenomenon is not caused by merger of the A and D genomes in a common nucleus of the allopolyploid cotton per se. Instead, the process may be largely random in terms of which genes are affected in the synthetic polyploidy, and probably caused by the various molecular processes that take place on allopolyploidization; these processes are discussed below. Previous studies have shown stochastic changes in duplicate gene expression levels in newly synthesized polyploids, albeit not in response to abiotic stresses (Wang et al., 2004). Even if stress-induced expression changes are random after allopolyploidy, there could be subsequent divergence caused by selection in response to environmental stresses or other factors.
The mechanisms and causes of gene silencing and expression level changes of homeologs in polyploids are starting to become known, but are still relatively poorly understood. They include cis- and trans-regulatory variation, changes in regulatory hierarchies and epigenetic changes such as cytosine methylation, histone modifications and small RNAs (reviewed in Osborn et al., 2003; Riddle & Birchler, 2003; Chen & Ni, 2006). Nothing is currently known about the causes and mechanisms of expression changes in duplicated genes in response to abiotic stresses in polyploids.
Mutations in cis-regulatory elements in which stress-induced transcription factors bind, sometimes referred to as stress-responsive elements, may account for homeolog expression changes in natural polyploids. Changes in the expression of trans-factors that regulate gene expression, particularly in response to abiotic stresses, may also contribute to the expression changes shown in this study. An often-discussed cause of gene expression changes in polyploids is the reuniting of diverged regulatory hierarchies on allopolyploidy from the two parental species, which may alter levels and patterns of gene expression (Osborn et al., 2003; Riddle & Birchler, 2003). In the case of expression changes in response to abiotic stress conditions, it is possible that regulator–sequence interactions in stress-regulated pathways may be altered in a synthetic allopolyploid, or that duplicated regulatory pathways diverge from each other to result in altered gene expression in response to stress in a natural allopolyploid.
There are several possible epigenetic mechanisms for the altered expression of duplicated genes in response to abiotic stress. Changes in cytosine methylation have been shown in several allopolyploid systems, as reviewed in the Introduction section. DNA methylation pattern alterations can occur in response to abiotic stresses (reviewed in Urano et al., 2010) and the patterns of cytosine methylation can affect gene expression (reviewed in Liu et al., 2010). Thus, it is possible that cytosine methylation pattern changes between the two duplicates in a pair may alter the expression levels of one or both genes in response to abiotic stress conditions. In the case of the cotton allopolyploids, divergent methylation patterns between the homeologs could be a factor affecting abiotic stress-responsive expression in G. hirsutum, but it is unlikely to be a factor in the synthetic allopolyploid because synthetic cotton allopolyploids have been shown to experience few cytosine methylation changes (Liu et al., 2001). Other epigenetic changes have been shown to occur in allopolyploids and affect gene expression, including histone methylation and histone acetylation (Wang et al., 2004; Ni et al., 2009). Alterations in histone modifications can occur in response to abiotic stresses (reviewed in Kim et al., 2010), and thus it is a probable candidate mechanism for stress-regulated changes in allopolyploids. Another possibility is the nonadditive expression of small RNAs, which has been shown in Arabidopsis allopolyploids (Ha et al., 2009), that could trigger changes in the amount of microRNA (miRNA) degradation of the homeologs. Abiotic stresses can result in changes in small RNA expression patterns and can affect gene regulation by miRNAs (reviewed in Sunkar et al., 2007).
Duplicate gene expression and responses to abiotic stresses
This study adds to the literature on the expression responses of duplicated genes to abiotic stresses. Expression patterns of genes duplicated by an ancient polyploidy event during the evolution of the Arabidopsis lineage, in response to abiotic stresses, were examined in three relatively recent studies. Kim et al. (2005) found, in a microarray study, that 117 gene pairs showed significant expression responses to oxidative stress by both genes in the duplicate pair, with some of the genes having distinct and sometimes opposite expression patterns. Ha et al. (2007) showed many examples of divergence in expression patterns between the duplicated genes in response to nine abiotic stress conditions in their analysis of the microarray dataset of Kilian et al. (2007). Zou et al. (2009) examined the expression of duplicates derived from the ancient polyploidy event and tandem duplicates with the nine abiotic stress microarray dataset (Kilian et al., 2007) using an ancestral state estimation approach. Numerous putative cases of partitioning of stress responsiveness between the duplicates were inferred. The above studies indicate the long-term evolutionary changes in responses to abiotic stresses by duplicated genes, whereas the present study examined expression in response to abiotic stresses on a shorter term evolutionary time scale of c. 1.5 million yr for G. hirsutum (Senchina et al., 2003) and in a synthetic allopolyploid.
Not only can abiotic stresses affect the expression patterns of duplicated genes, but biotic stresses can also have such effects. For example, expression assays of two peroxidase homeologs in Brassica napus showed differential expression responses of the homeologs to pathogen infection (Zhao et al., 2009). In addition, the studies of anciently duplicated genes in Arabidopsis, mentioned above (Ha et al., 2007; Zou et al., 2009), included microarray data from biotic stresses in addition to abiotic stresses, with data from the two types of stresses analyzed together.