In this study we addressed the functional impact of alternative splicing in plants using a comparative analysis approach. According to the ascribed role of AS as a mechanism for expanding proteome diversity, a gene can become polymorph in its expression through an additional splice variant that encodes a different, yet functional protein. When the principle and alternative protein isoforms each confer a selective advantage, both are likely to be retained during evolution. This can be achieved through either retention of the AS-induced polymorphism or through gene-duplication followed by sub-functionalization of the duplicates [22
]. Not only different protein isoforms but also AS-mediated regulation of gene expression through the production of PTC containing isoforms that are destined to be degraded by the NMD-pathway can provide a selective advantage. The exact AS event is not necessarily required to remain the same for both the production of protein isoforms and the production of PTC containing transcripts isoforms. In this study, we investigated to what extent AS induced polymorphism as well as AS-mediated gene-regulation are conserved between orthologous genes of the dicot Arabidopsis
and the monocot rice and orthologous genes of the two monocots maize and rice. The comparison was focused on the functional outcome of the AS events rather than on the AS events themselves. The number of cases in which orthologs in Arabidopsis
and rice contained AS variants that are likely targets for NMD was roughly half that of ortholog pairs with variants that can be translated into proteins. Only a very small number of cases were found in which AS events in both orthologs resulted in similar modifications to the principle protein product. Interestingly, the fractions of ortholog pairs that have AS events on different positions and those that have similar or different event types on the same position were quite similar in both the NMD and translated subsets of the Arabidopsis
-rice orthologs. Because the isoforms in the NMD subset would never function as protein, this similarity suggests that a considerable fraction of the putatively conserved AS events are not the result of functional conservation. Similar values for the distributions mentioned above were observed in the translated subset of the maize-rice orthologs. This similarity further strengthens the notion that conservation of AS events is not the result of the preservation of functional AS induced protein polymorphism. Although the fraction of orthologs that have similar AS-event types on orthologous intron-positions leading to modifications at similar sites is higher in the maize-rice comparison than that observed for the Arabidopsis
-rice orthologs in the translated as well as the NMD set, these numbers are still remarkably low. In the NMD subset of the maize-rice orthologs the fractions of different- and similar event types on orthologous intron-positions are very similar. In summary, the results suggest that conservation of AS events as the result of the selective advantage of retaining the ability to produce multiple protein isoforms through AS is not frequent, even at short evolutionary distances.
Previous studies on animal data (see introduction) have revealed similarities between genes undergoing AS in various species. The observed patterns are not the result of ortholog-level conservation, which is evident from the low numbers of conserved AS events between, for instance, human and mouse [23
]. However, the patterns do point to the existence of mechanistic features that are associated to- or reflected in the AS process. One such mechanistic feature in animals is the high frequency of exon skipping events which is the consequence of the exon definition mechanism through which splice-sites are recognized [24
]. In contrast, the high frequency of intron retention events in plants is the consequence of the intron definition mechanism through which many plant introns are recognized [25
In contrast to animals, protein function-oriented studies of genome-wide AS patterns in plants are scarce. Motivated by the results obtained from computational analysis of the impact of AS on protein function in animals, we have applied a number of those analyses on plant data. The first analysis involved the distribution of Pfam domains over constitutively and alternatively spliced genes using a similar method as previously performed on human data [16
]. Three domains were significantly underrepresented in alternatively spliced genes in all three species. Genes encoding these domains share at least two properties that have previously been shown to be correlated with AS frequencies. First, the genes belong to highly expanded gene families in plants [26
]. Recent studies in animal species have shown that the incidence of AS is inversely correlated to gene family size [22
However, such an inverse correlation has not been found in Arabidopsis
and rice [31
]. These conflicting results may be the consequence of the different methods that have been used for delineating gene families. Second, genes encoding these domains have relatively low average numbers of introns in all three species (data not shown). It has previously been shown that the incidence of AS is positively correlated with the number of introns [15
Only one domain, the RRM domain, was significantly overrepresented in alternatively spliced genes in all species. This RNA recognition domain, like the domains mentioned above, is found in a large number of genes, many of which are involved in the regulation of the splicing process [32
]. However, in contrast to the three domains that were underrepresented, the average number of introns harbored by genes encoding the RRM domain is higher than the average number of introns in all genes in all three species and this might in part explain the observed data (data not shown).
Zooming in on the sequence level, it has been shown in four animal species that AS occurs less frequently within than outside the boundaries of predicted protein domains [17
]. We observed a similar trend in all three plant species. This result, together with those presented by Kriventseva and coworkers [17
], suggests that in general AS events within the boundaries of protein domains are less favored in evolution.
It has been reported that domains with particular functions are more frequently targeted by AS than expected by chance (reviewed in [33
]). Not only the location of introns but also the location of AS events within a gene are to a certain degree related to sequence- and structural features of the encoded protein [34
] and it was hypothesized that particular domains are more prone to undergo AS within their boundaries as the result of such features. This possibility was tested by searching for domains that were significantly enriched with alternatively spliced rather than with constitutively spliced introns. Although a few domains were identified that were significantly enriched with alternatively spliced introns in individual species, none of such domains were found across all species. This result suggests that none of the studied domains has a detectable evolutionary propensity for undergoing AS within its boundaries.
The final analysis involved AS events that remove entire protein domains. The specific AS events that lead to the removal of the domains are not required to be located within the boundaries of the spliced-out domain. In human genes, various domains have been shown to be more often spliced-out than average [19
]. Here, we identified a number of cases in which AS events resulted in the removal of complete protein domains in all three plant species, but only one domain was found to be spliced out at a significantly elevated rate in both Arabidopsis
and maize. The top five molecular function-related GO terms of the spliced-out domains, comprising around 50% of all GO-term assignments, were the same in Arabidopsis
and rice while in maize only two of these five GO-terms comprised 45% of GO term assignments. Genes having these GO terms in Arabidopsis
have been shown to have elevated levels of AS [36
]. However, in that study only the AS levels of genes with nucleotide binding activities were found to be significantly elevated.
Interestingly, in a large fraction of the unique splicing patterns in all species, the spliced-out domain was a unit of a tandem repeat. A similar observation has been reported in a study that was conducted on data derived from the Swiss-Prot and PDB- databases [7
]. The authors suggested that AS within repeated regions of a protein is more likely to be tolerated. Their conclusion is supported by the observation that variation in repetitive regions has frequently been used in evolution for modifying the properties of proteins without invoking loss of their fold [37
]. It is also possible that the large fraction of domains that are spliced-out from tandem repeats is the consequence of the duplication events themselves as is seen for the frequent appearance of AS through tandem exon duplications [38
]. However, further research is needed to clarify the observed frequent removal of domains from a tandem repeat.