Conservation and evolution of miRNA regulatory programs in plant development

Summary of recent advances

Over the past two years, microarray technologies, large-scale small RNA and whole genome sequencing projects, and data mining have provided a wealth of information about the spectrum of miRNAs and miRNA targets present in different plant species and the alga Chlamydomonas. Such studies have shown that a number of key miRNA regulatory modules for plant development are conserved throughout the plant kingdom, suggesting that these programs were critical to the colonization of land. New genetic and biochemical studies of miRNA pathways in Arabidopsis, the spatiotemporal expression patterns of several conserved miRNAs and their targets, and the characterization of mutations in Arabidopsis and maize have begun to reveal the functions of these ancient miRNA-regulated developmental programs. In addition to these conserved miRNAs, there are many clade and species-specific miRNAs, which have evolved more recently and whose functions are currently unknown.

Introduction

MicroRNAs (miRNAs) are 19-24 nt RNAs that direct the post-transcriptional silencing of transcripts of target genes with high complementarity to the miRNA (reviewed in [1-3]). The source and targets of miRNAs distinguish them from other classes of small RNAs. MiRNAs originate from longer, non-protein-coding RNAs that have the ability to form hairpins. The miRNA sequence is found within the stem-loop, paired with a partially complementary sequence known as the miRNA*. A Dicer-like enzyme, primarily DCL1 in plants, cleaves twice within the stem-loop, releasing the miRNA:miRNA* duplex [4]. DCL1 activity is facilitated by the nuclear-localized RNA-binding protein HYL1 [5-7]. According to the prevailing model, miRNA exporters, including HASTY, export the duplex to the cytoplasm [8]. Within the cytosol, it is unwound, and the mature miRNA interacts with an Argonaute protein (AGO1, and possibly other redundant AGO proteins) and a complementary target mRNA in a silencing complex [9,10]. AGO1 initiates silencing in plants primarily via target cleavage within the miRNA complementary site [9-11], but miRNAs can also act via translational repression [12-14].

Beyond their origin from a hairpin precursor, miRNAs are unique from most other classes of small RNAs (sRNAs) because they act in trans on non-self RNAs. The only other sRNAs that specifically silence non-self mRNAs are the trans-acting siRNAs (ta-siRNAs), which originate from long, non-protein-coding transcripts that are themselves targets of miRNAs [15,16]. Following miRNA-directed RNA cleavage, the products are transformed into double-stranded molecules by RDR6, and cleaved in phase by DCL4 to yield ta-siRNAs [15-18].

The Ambros and Ruvkun labs first identified miRNAs as regulators of developmental timing in Caenorhabditis elegans; the lin-4 and let-7 miRNAs silence genes important for the transition between different stages of larval development [19-21]. Interestingly, the initial miRNAs discovered in plants largely regulate transcription factors and other genes involved in plant development [11,22,23]. Plant miRNAs have also been shown to target genes involved in abiotic and biotic stress responses [24-28], hormone signaling [28,29], metabolism [30], general transcription [31], and the miRNA machinery itself [32,33]. Sequencing of additional miRNAs will likely find many more roles.

The centrality of miRNAs in the regulation of plant development is further supported by the severe developmental phenotypes of loss-of-function alleles of genes involved in their biogenesis and function. Hypomorphic alleles of DCL1 reduce the production of most miRNAs [34] and display a range of developmental phenotypes, including defects in leaf morphology, axillary meristem maintenance, flowering time, inflorescence determinancy, floral organ patterning, and ovule development, while null alleles are embryonic lethal [35-37]. Mutations of AGO1 cause genes targeted by miRNAs to be upregulated or ectopically expressed, and also produce a decrease in the abundance of some miRNAs [32,38-40]. AGO1 hypomorphic alleles have defects in lateral organ polarity and leaf and flower morphology [32,38]. Null AGO1 alleles are seedling lethal, often due to a defect in meristem maintenance, and these phenotypes are worsened when combined with mutations in other AGO genes, suggesting possible functional redundancy [32,37-39,41].

Over the past few years, many labs have been working to identify and catalog the miRNA component of the plant transcriptome using microarrays, high-throughput sRNA cloning, and data mining of whole genome and EST sequences. While these studies initially focused on Arabidopsis, researchers quickly began to explore the sRNA make-up of other species, advancing our understanding of the conservation, evolution, and diversity of miRNA functions in plants. This work has shown that a number of miRNAs and their targets are conserved throughout land plants. These conserved miRNA modules are involved primarily in developmental regulation and are typically more highly expressed than non-conserved miRNAs, which have more diverse roles. The ancient origin of these developmental programs suggests that the emergence of these pathways may have been critical to the colonization of land.

Here, we highlight these recent data by examining the basic mechanisms of miRNA evolution and the phylogenetic distribution of these miRNA modules. We then compare two specific developmental miRNA pathways in Arabidopsis and maize to demonstrate the level of similarity in these programs between a core eudicot and a monocot. Finally, we discuss the roles of non-conserved, clade or species-specific miRNAs in plant biology.

Evolution of miRNA families and their targets in plants

The inverted duplication hypothesis of miRNA evolution

The current model for miRNA evolution in plants—the inverted duplication hypothesis—proposes that miRNA genes arose from inverted duplications of their target genes [42]. (See Figure 1.) According to this model, the mRNAs transcribed from these duplications form long, self-complementary hairpins that are cleaved by Dicer-like proteins into siRNAs. These siRNAs then initiate silencing of the original gene, as well as other closely related genes. Over time, these inverted duplications accumulate mutations that cause them to become highly divergent from their source except in small regions corresponding to the miRNA and miRNA* sequences. MiRNA gene families arise from tandem and segmental duplications of these founder miRNA genes, followed by further divergence and specification. Under the proper conditions, these duplications may generate miRNAs targeting new gene families, as is suggested for miR447 and miR856 in Arabidopsis [43]. Target genes and their gene families evolved in parallel through similar evolutionary events [42].

Figure 1.

The inverted duplication hypothesis of miRNA evolution. This model suggests that miRNAs evolve from inverted duplications of their target genes. When transcribed, such inverted duplications yield hairpin mRNAs that can be processed to siRNAs able to silence the target genes. Over time, mutations accumulate within the inverted duplication, and it becomes increasingly more divergent from the target gene, except within the regions that give rise to the miRNA and miRNA* sequences. As the gene evolves, the mRNA hairpin takes on a structure more like a miRNA precursor, and miRNAs are formed. Young miRNA genes still maintain significant sequence similarity to their target genes, while ancient miRNA genes are only homologous to the target genes within the miRNA and miRNA* sequences. In this figure, black is used to designate those portions of the genes and mRNAs identical to the original target genes, and red represents non-identical regions. The blue arrows specify the path of miRNA evolution.

The inverted duplication hypothesis is supported by the discovery of loci representing intermediate stages in this process [42-44]. Most miRNAs in multigene families are conserved across many species, and show very little homology to their target genes outside of the miRNA and miRNA* sequences, suggesting an ancient origin [42]. By contrast, several studies have shown that some non-conserved miRNAs contain high homology to target genes throughout the miRNA precursor [42-44]. One large-scale sRNA sequencing project in Arabidopsis revealed that 16 of 48 non-conserved miRNA precursors had fold-back arms with high complementarity to protein-coding genes [43]. These findings suggest that, while many miRNAs in plants have an ancient origin, miRNAs arise continually through inverted duplication of genes followed by subsequent evolution [42-44]. The effects of random mutations on the expression and specificity of the siRNAs generated by these duplications, and whether or not their new silencing activities are deleterious, advantageous, or neutral, will determine whether such inverted duplications evolve into true miRNAs [43].

Phylogenetic distribution of conserved miRNA modules in plants

The study of miRNA evolution in plants has benefited from microarray analysis using miRNA arrays, large-scale sRNA sequencing/cloning projects, as well as data mining of whole genome and EST sequences from a number of species. Bryophytes (mosses, liverworts, and hornworts) are the most ancient land plants, and some of the most exciting information about the evolution of miRNAs has come from recent studies of mosses, which are members of this family [45-50]. These studies have shown that at least 14 miRNA families are conserved between core eudicots and mosses, and, therefore, were likely present in the earliest common ancestor of land plants (Table I). Data from the completely sequenced genomes of Arabidopsis, rice, and poplar reveal that these conserved miRNA families generally have more members and are expressed at higher levels than non-conserved miRNAs, consistent with an earlier evolutionary origin [25,43,44,50-53].

Table 1.

The distribution of miRNA gene families in the plant kingdom.^a

^aShaded boxes denote the miRNA gene families that have been found to date in each taxon, with references in brackets. This list was compiled with assistance from miRBase, Version 9.1 [88,89].

^bThe microarray approach used could not distinguish between miR159 and miR319 [46].

MiRNAs have also been identified in the unicellular green alga Chlamydomonas reinhardtii, demonstrating that they are not unique to multicellular plants [54]. Interestingly, none of the miRNAs discovered in Chlamydomonas are present in either moss or vascular plants [54]. As plants are thought to have evolved from green algae, this observation suggests that conserved miRNAs in plants may have been important for multicellularity and the colonization of land. It will be important to analyze the sRNAs of other green algae—particularly members of the Characeae, which are more closely related to plants—to further explore these ideas.

Recently, two miRNAs—miR854 and miR855—present in both plants and animals have been identified [31]. These miRNAs target a group of RNA-binding proteins, and are expressed in both Arabidopsis and several animal species, but not fungi. Potential family members have been identified in rice, but the distribution of these miRNAs in other species is not known [31].

A phylogenetic analysis of the target genes for several ancient miRNA families has shown that these target genes are also conserved (Table 2), implying that the functions of these miRNAs have been broadly maintained throughout plant evolution. For example, a conserved SBP protein target of miR156 was cloned from the moss Physcomitrella patens and shown to be cleaved within the predicted target site [45]. In addition, an analysis of AP2-like genes, which are targets of miR172, has found that the miR172 target sequence is largely conserved even in gymnosperm species [55]. The most striking example of miRNA module conservation is for miR165/166, negative regulators of class III HD-ZIP transcription factors. The Bowman lab has demonstrated that there is nearly perfect conservation of the miRNA target site in these genes for all land plants analyzed, including P. patens [56,57]. This sequence is not conserved in the multicellular green alga Chara, however [57].

Table 2.

Genes targeted by conserved miRNAs.

MiRNA family	Target gene family	Protein class	Function
156/157	SPL	Transcription factor	Developmental timing
158	PP2		Unknown
160	ARF	Transcription factor	Auxin response, leaf and root development, floral organ identity
165/166	HD-ZIPIII	Transcription factor	Meristem maintenance, vascular development, lateral organ polarity
167	ARF	Transcription factor	Auxin response
170/171	SCL	Transcription factor	Root development
172	AP2	Transcription factor	Developmental timing, floral organ identity
319	TCP	Transcription factor	Leaf development
390	TAS3 ( ta-siRNAs act on ARF)	Transcription factor	Auxin response, developmental timing, lateral organ polarity
395	sulfate transporter		Stress response
408	Multiple laccases, plantacyanin		Stress response
414	Unknown
418	Unknown
419	Unknown

Conserved miRNA-target gene modules gain new functions during evolution. For example, the miR169 gene family, which targets HAP2 transcription factors, is conserved from ferns to eudicots. In alfalfa, miR169 targets a HAP2 gene important for root nodulation, a function that specifically arose in the legume clade [58]. MtHAP2-1 and miR169 have nodule-specific, complementary expression patterns, and thus provide an interesting example of the co-evolution of genes with related functions [58]. Target gene duplication has been particularly important in the expansion of the developmental roles of the miR165/166 program [56,57]. In the ancestor of land plants, the miR165/166 module was likely important for tip growth, as seen in bryophytes [57]. With time, the target gene family expanded, and the module subsequently gained roles in vascular patterning and lateral organ polarity [59-62]. In Populus this module is also important for secondary growth [63].

Some conserved miRNAs may also have gained unique targets over time. For example, the stress-responsive miR395 has a tissue-specific expression pattern in moss [47], suggesting that this miRNA may have a developmental function in some organisms. In Arabidopsis and other species, miR390 has been shown to target TAS3 [17]. Although there is no evidence for the existence of TAS3 in P. patens, miR390 targets a different ta-siRNA precursor in this species—PpTAS4 [49]. PpTAS4 is postulated to target AP2-like transcription factors [49]. Remarkably, in both Arabidopsis and P. patens, a defect in the biogenesis of ta-siRNAs affects the timing of developmental transitions. In Arabidopsis, mutations in ZIP, RDR6, SGS3, or DCL4 (which block the biosynthesis of TAS3), cause a precocious transition to the adult vegetative phase [15,16,18,64,65], whereas a knockout of the RDR6 orthologue in P. patens (which blocks the production of siRNAs from PtTAS4) accelerates gametophore formation [49].

Conservation of developmental miRNA pathways in Arabidopsis and maize

Lateral organ polarity

Studies in Arabidopsis and maize have revealed a high level of conservation of developmental miRNA pathways between core eudicots and monocots. Lateral organ outgrowths are thought to arise by the juxtaposition of adaxial and abaxial polarities; in the absence of both fates, radialized organs are formed (reviewed in [66]). The miR165/166 genes are important for establishing and maintaining abaxial polarity. They are expressed on the abaxial side of leaf primordia, restricting the expression of HD-ZIPIII genes to the adaxial surface [60]. Semi-dominant mutations in the Arabidopsis HD-ZIPIII genes PHB, PHV, and REV, and the maize REV homologue Rld1 were all found to contain mutations in the miR165/166 binding site that block miRNA-directed cleavage [59-62]. These mutations produce an expanded domain of HD-ZIPIII expression, resulting in an adaxialized phenotype [59-62]. It has recently become clear that the target of miR390, TAS3, is also important for leaf polarity. In contrast to miR165/166, TAS3 is expressed on the adaxial surface of the leaf. This expression pattern likely restricts that of TAS3 targets—ETT/ARF3 and ARF4 in Arabidopsis and ARF3a in maize—to the abaxial domain [67,68]. Nogueira and colleagues found that the maize gene leafbladeless1 is a homologue of SGS3 in Arabidopsis and, like sgs3, blocks the production of TAS3 ta-siRNAs [68]. They also discovered that leafbladeless1 has an expanded domain of miR165/166 expression, suggesting that TAS3 helps to establish an adaxial fate by repressing the expression of these miRNAs [68].

Phase transitions

Flowering plants have three major post-embryonic phases of development— juvenile vegetative, adult vegetative, and reproductive. The temporal regulation of two key miRNA modules is important for these phase transitions. In Arabidopsis, the expression of the transcription factors SPL3, SPL4, and SPL5 increases during shoot development [69-71]. This expression pattern is specified post-transcriptionally by miR156, a miRNA that silences 10 members of the SPL gene family and is expressed in the opposite developmental pattern—highest in the young shoot and decreasing over time [14,71]. Overexpression of a miR156-insensitive SPL3 speeds the juvenile-to-adult transition, as well as flowering time, while overexpression of miR156 delays both transitions [14,71]. The delayed vegetative phase change phenotype of the dominant Corngrass1 mutation in maize is also attributable to the overexpression of miR156 [72], and there is evidence that this defect also explains the delayed phase change phenotype of Teopod1 and Teopod2 in maize (Mee Yeon Park and R.S.P., unpublished results). Consistent with this result, miR156 and its SPL targets have the same temporal expression pattern in maize as they do in Arabidopsis [72].

The miR156 module may be regulating a second miRNA pathway important for vegetative phase change. miR172 expression increases during the juvenile-to-adult transition in both Arabidopsis and maize [13,73], and a decrease in the expression of its target, Glossy15, is associated with vegetative phase change in maize [73]. The involvement of both miR156 and miR172 in vegetative phase change, and the observation that they are expressed in opposite temporal patterns, raises the possibility that the miR156 module regulates the expression of miR172. This idea is supported by the fact that miR172 expression is suppressed in Corngrass1 [72]. Additional studies will be required to determine the basis for this regulation.

Clade and species-specific miRNAs

All of the species subjected to high-throughput sRNA sequencing have been found to possess non-conserved miRNAs [22,24,25,30,43-45,47,48,52,53,74-77]. These findings indicate that miRNAs arise continually, probably through the kinds of evolutionary events discussed above [42,43]. Clade and species-specific miRNAs target mRNAs that have a wider range of functions than the targets of conserved miRNAs. For example, only about 1/9 of a newly discovered set of non-conserved miRNAs in Arabidopsis target transcription factors; instead, their targets include genes involved in metabolism, signal transduction, protein modification, and RNA or carbohydrate binding [43]. It will be interesting to learn how far these miRNAs extend beyond Arabidopsis, and their roles in the biology of this species.

Conclusions

The discovery of conserved miRNA-target modules is a major step towards identifying core pathways in plant development. These discoveries have been facilitated by the high degree of similarity between plant miRNAs and their targets, and by the use of high-throughput sequencing strategies. Future studies of the functions of these modules will not only provide new insights into the mechanisms of plant development, but may also offer new lessons about the factors that drive the co-evolution of regulatory genes and their targets.