Modeling the asymmetric evolution of a mouse and rat-specific microRNA gene cluster intron 10 of the Sfmbt2 gene

Micro RNAs (miRNA's) are 19 to 22 nt long, non-coding, single-stranded RNAs that can fine-tune the expression of protein-encoding genes [1, 2]. One example is the post-transcriptional repression of mRNA targets involving the so called miRNA "seed" which is nt 2-8 of the mature miRNA which recognizes complementary bases in the 3'untranslated region of the mRNA target [3]. miRNA genes form primary transcripts that are converted by Drosha to miRNA precursors of 70-90 nt length. Processing of miRNA precursors into mature miRNA's is catalyzed by the RNA processing enzyme Dicer [4, 5]. Processing enzymes recognize the secondary hairpin-structure of the miRNA precursor [6]. Most miRNA precursors have indeed a distinctive stem-loop structure that is often highly conserved among distant species and that is used to distinguish them from other small RNA classes.

The exact number of miRNA genes, collective expression of which makes the miRNA repertoire of an organism, is not known and the question how new miRNA genes arise is an interesting and insufficiently studied problem in evolutionary biology. The conserved hairpin structure in miRNA precursors was applied in two primary miRNA identification methods: directional cloning and computational identification [7, 8]. Consequently, newly evolved, not-conserved miRNAs were likely to be overlooked by these methods. Furthermore, some miRNA gene candidates accumulated sequence mutations that, over time, either led to mature miRNA genes or to gene inactivation [9]. Both actions make it difficult today to trace back the origin of miRNA genes. Using a deep sequencing approach, a large group of evolutionarily young miRNA genes was discovered in Drosophila [9]. In this study, a high birth rate of new miRNA genes was described (12 new miRNA genes per million years). The main sources for gaining miRNA genes in plants are based on miRNA gene duplication events and local inverted duplication events of short segments from protein-coding genes [10]. An alternative source of short inverted sequence segments could be based on transposons as these often carry terminal inverted repeats or as they can insert at short distance from their origin, resulting in an inverted gene arrangement [11]. The involvement of transposons in the birth of clusters of miRNA genes might be underestimated as computational miRNA detection methods were designed to exclude transposon sequences [8]. However, recent analyses in several mammalian species [8, 12, 13] indicated that a number of miRNA gene clusters were derived from repetitive elements. This may have contributed to "leaps" in the expansion of the miRNA repertoire in placental mammals [14].

Approximately, 40% of the mouse genome is composed of repetitive elements, the largest part of which is interspersed between or within genes. Two examples are simple repeats and B1 elements (short interspersed nuclear elements (SINE), an Alu-like family of retrotransposons in rodents). Each occupies a similar fraction of the mouse genome (2.3% vs. 2.7%) [15]. Mouse B1 elements are rodent-specific retrotransposons (length approximately 140-bp) that may have originated from reverse transcription of ancestral 7SL RNA [16]. Transposable elements like B1 in mice and Alu in humans were previously thought to be either harmless "junk" or to influence functional genes negatively [17]. More recently, however, it was proposed that retrotransposons can actually contribute to gene function and genome evolution, either by promoting local duplications, inversions or deletions events or by altering gene expression [18, 19]. Simple repeats are present in micro- and, minisatellites which are short tandem repeats of one repetitive unit [20] with a characteristic length and sequence. For instance, in microsatellites, the repeat unit is 1 to 6 bp long. It has been shown that microsatellites are characterized by a high sequence mutation rate (10^-6 to 10^-2 per bp per generation) [21]. This makes microsatellites an adaptable reservoir for genetic changes, in addition to the phenomenon that numerous host genes evolved from retrotransposons [22, 23].

A number of recent findings suggest that the biology of miRNA's and of retrotransposons may be connected at several levels. One starting point is the idea that RNA interference is a mechanism that protects genomes against an unacceptably high, self-destructive retrotransposon activity [24, 25]. Interestingly, the RNA interference pathway and the maturation and action of miRNAs share some processing proteins and both mediate endonucleolytic cleavage [26–28]. Furthermore, the evolutionary growth of an Alu-rich primate miRNA cluster [29] and a marsupial miRNA cluster [30] were proposed to be driven by local duplication events involving retrotransposons. For the miRNA genes in the human chromosome 19 cluster, the encoded seed sequence was found to be complementary to the most conserved parts of Alu RNA [29].

In the present study we have analyzed the sequence of a large miRNA cluster which is specifically located in intron 10 of the Sfmbt2 gene on chromosome 2 of the mouse (from now on abbreviated as C2MC). This cluster is very rich in B1 retrotransposons and microsatellites, which are ordered in a regular manner. In rats the same repeat elements are present, but small changes in DNA sequence may have resulted in less-stable hairpins and failure to form mature miRNA from precursors. In all other mammals an ortholog of C2MC could not be detected. On basis of this analysis we propose that B1 retrotransposons have contributed to rapid growth of the mouse C2MC miRNA cluster, which further illustrates the connection between retrotransposition and the evolution of a miRNA repertoire in mammals.

In this study, we analyzed the repeat elements of C2MC, a mouse specific miRNA cluster in a large intron of a protein-encoding gene. The encoded protein SFMBT2 (Scm-like with four mbt domains 2) belongs to the polycomb protein family which is implicated in embryonic development [33]. Interestingly, the Sfmbt2-gene is imprinted and primarily expressed from the paternal allele in early embryos [34]. This makes another example of a large cluster of non-coding small RNA genes in regions of imprinted DNA [35, 36]. In the analysis we performed in this paper, evidence was found for the idea that B1 transposons together with two self-complementary microsatellites, have generated a miRNA-containing cassette that has duplicated many times during a relatively short time of evolution. This idea fits well into an emerging concept that retroposon-driven local expansion of miRNA clusters can contribute to the miRNA repertoire of mammals [37, 38]. In our analysis, the smallest repeat unit bearing one miRNA gene (MSHORT1) is composed of a 3' fragment of a B1F3 element, two microsatellites and additional sequence (Figure 2). The two microsatellites overlap with the miRNA gene of MSHORT1 and could have created a stable hairpin due to base pair complementarity. Perfect hairpins were proposed to be a source of de-novo generated plant miRNAs [10].

One can argue about the point whether or not the encoded small non-coding RNAs in C2MC are "real" miRNA's. In general, miRNA's are defined by a combination of five criteria for both expression and biogenesis (for review see [39]). Previous work has indicated that the miRNAs of C2MC fulfill three of these criteria [40–42]: (i) detection of a distinct expressed mature ~22 nt RNA transcript; (ii) the identification of its precise genomic match; (iii) the dependence on Dicer for miRNA production [39]. What is lacking for the C2MC mRNA's is phylogenetic conservation of the ~22 nt miRNA sequence and its predicted fold-back precursor [39]. Phylogenetic conservation may be exist, however as suggested by other miRNAs that are classified together with C2MC miRNA's into families 466 and 467) [43]. But one should be cautious about this as, we found that this family classification does largely depend on the microsatellite related origin of these miRNAs. While miRNAs of C2MC do contain a combination of (TG)n and (CA)n microsatellite, the other species' miRNAs grouped in these miRNA families generally involve a (TA)n microsatellite (Additional File 2 Table S4). Indeed, when we analyzed a 300 nt sequence with the microsatellite content masked, the miRNA genes of C2MC showed no sequence similarity to other miRNAs the 466 and 467 families. Therefore, the miRNAs of C2MC might have a different origin, and should perhaps be grouped in a separate miRNA family. An interesting detail is that mouse miRNAs of families 466 and 467 that do not belong to C2MC could not be confirmed in an experimental miRNA evaluation approach [42]. The last criterion for miRNA evaluation describes structural requirements for a hairpin precursor that contains the ~22 nt miRNA sequence. In this criterion, the hairpin must be the folding alternative with the lowest free energy and should not contain large internal loops or bulges, particularly not asymmetric bulges [39]. We have arguments that the small RNA's from C2MC are different from piRNAs (Piwi protein-associated small RNA's). First, piRNA gene clusters are present in intergenetic regions [44], while the small RNA's in C2MC are encoded in an intron of a protein-encoding gene. Second, the mature small RNAs of C2MC have a length of 22 nt, which is characteristic of miRNAs, and smaller as the length of piRNAs (24 nt to 31 nt) [45, 46].

Differences between MSHORT1 and RSHORT1 should explain why the mouse miRNA cluster C2MC contains more miRNAs than its rat counterpart. In comparison to MSHORT1, the right arm of the RSHORT1 hairpin shifted laterally by 10 nt and gained 3 asymmetrical bulges, while MSHORT1 had none (Figure 4E). An energy profile of primary miRNA transcripts is described for humans. There, the region of the lower hairpin stem towards the basal segments and the upper stem close to the Drosha cleavage site were identified as thermodynamic most stable [31]. This energy profile is accentuated by the entropy profiles of MSHORT1, but it is less consistent with the profile of RSHORT1 (Figure 4). Further, sequences flanking the hairpins were described as essential for efficient in vitro processing [4, 47]. However, RSHORT1 lacks this sequence on its 5 prime side (Additional File 1 Figure S3C). Together these differences might explain the low number of genes producing mature miRNAs in the rat cluster.

The results of BLAST analysis of MSHORT1 suggests a model for the evolution of C2MC (Figure 5). SHORT1 copies accumulated via a recombination-based mechanism in the common rodent ancestor of today's mice and rats (Figure 5A). These ancestral SHORT1 units may have been more similar to MSHORT1 than to RSHORT1. In fact, some repeat units in the cluster of modern rats resemble MSHORT1 instead of RSHORT1, while all repeat units in modern mice only resemble MSHORT1 (Additional File 2 Table S2 and S3). Thus, a truncation event resulting in the ~200nt RSHORT1 occurred in the rat after the speciation of mice and rats. We speculate that single nucleotide mutations and small deletion events in 10 neighboring MSHORT1 units, caused the formation of one larger unit MLONG1 (Figure 5B). MLONG1 was further multiplied in the C2MC center. This model is similar to what we proposed before in primates (27). In the mouse the individual miRNAs of C2MC have been duplicated together with a B1F3 retrotransposon.

What could be the biological implication of the growth of the C2MC miRNA gene cluster? One possibility is that such growth contributes to the well known expansion of a miRNA gene repertoire in mammals [14]. In order to regulate a great diversity of mRNA targets, a repertoire of "seeds" is needed. This requires a large set of miRNA genes. Local duplication events of miRNA genes in clusters are, therefore, to some extent analogous to a repertoire of variable immunoglobulin, T-cell receptor and Trypanosome variable surface glycoprotein genes [48–50]. A problem with this explanation is that the growth of the C2MC cluster has not yet generated new "seeds'" to regulate novel mRNA targets: the miRNAs from C2MC share the "seed" region within a group of five distinct miRNA clusters, present in mouse, human and zebra fish [51]. On the other hand, the miRNA genes may be evolutionary young and since they overlap with microsatellites, rapid evolution of the "seed" sequences may still occur. Another possibility of gene duplication is a dosage effect of miRNA's with the same seed. This is particularly interesting for imprinted genes such as Sfmbt2. More research is needed to evaluate the impact of miRNA concentration generated by the miRNA genes in C2MC and to compare this action with that other amplified miRNA clusters, e.g. on chromosome 19 in primates.

In summary, we analyzed C2MC, a large mouse miRNA cluster that expanded during the time since mice and rats diverged from each other. Our data indicate that a B1F3 retrotransposon has contributed to the generation of a tandemly repeated unit MSHORT1, which grouped into the 2.5 KB MLONG1 element. The latter led to rapid growth in the center of the cluster by further duplication events. Our study indicates that C2MC is a new example of the phenomenon in which lineage-specific SINE elements contribute to growth of the miRNA gene repertoire in a particular species.