Báo cáo khoa học: New human and mouse microRNA genes found by homology search pptx

These approaches predict that the human genome might contain 200–255 miRNA genes Keywords antisense transcript; genomic localization; human genome; microRNA; mouse genome; mRNA degradati

by homology search

Michel J Weber

Laboratoire de Biologie Mole´culaire Eucaryote, UMR5099, CNRS and Universite´ Paul Sabatier, IFR109, Toulouse, France

MicroRNAs (miRNA) are
22 nucleotide-long RNAs

that function in translational repression by base

pair-ing with their target mRNA in a variety of

pluricellu-lar organisms They originate from long precursors

(pri-miRNA) that, in animals, are cleaved by the

Dro-sha endonuclease in the nucleus [1] to give
70

nuc-leotide-long miRNA precursors (pre-miRNAs) with a

characteristic hairpin structure In the plants, excision

of pre-miRNAs is performed by DCL1, a Dicer

homo-logue [2,3] Following the export of pre-miRNAs to

the cytoplasm by Exportin-5 [4,5], the loop region of

the hairpin is removed by the Dicer endonuclease to

produce a short double-stranded RNA (dsRNA), a

strand of which, corresponding to the mature miRNA,

is predominantly incorporated in the RNA-induced

silencing complex (RISC) The RISC complex either inhibits translation elongation or triggers mRNA deg-radation, depending upon the degree of complementar-ity of the miRNA whith its target (for reviews, see [6–8])

Since the seminal identification of the first miRNAs, Caenorhabditis elegans lin-4 and let-7, using genetic approaches [9,10], hundreds of miRNAs have been characterized experimentally using various cloning strategies in plants, C elegans, Drosophila melanogas-ter, zebrafish, pufferfish, mouse, rat and human [8] More recently, algorithms have been developed to identify putative precursor miRNAs from sequenced genomes [11,12] These approaches predict that the human genome might contain 200–255 miRNA genes

Keywords

antisense transcript; genomic localization;

human genome; microRNA; mouse

genome; mRNA degradation; RNA

interference

Correspondence

M Weber, LBME, 118 route de Narbonne,

31062 Toulouse Cedex, France

Fax: +33 5 61 33 58 86

Tel: +33 5 61 33 59 56

E-mail: weber@ibcg.biotoul.fr

(Received 4 June 2004, revised 24 August

2004, accepted 7 September 2004)

doi:10.1111/j.1432-1033.2004.04389.x

Conservation of microRNAs (miRNAs) among species suggests that they bear conserved biological functions However, sequencing of new miRNAs has not always been accompanied by a search for orthologues in other spe-cies I report herein the results of a systematic search for interspecies ortho-logues of miRNA precursors, leading to the identification of 35 human and

45 mouse new putative miRNA genes MicroRNA tracks were written to visualize miRNAs in human and mouse genomes on the UCSC Genome Browser Based on their localization, miRNA precursors can be excised either from introns or exons of mRNAs When intronic miRNAs are anti-sense to the apparent host gene, they appear to originate from ill-character-ized antisense transcription units Exonic miRNAs are, in general, nonprotein-coding, poorly conserved genes in sense orientation In three cases, the excision of an miRNA from a protein-coding mRNA might lead

to the degradation of the rest of the transcript Moreover, three new exam-ples of miRNAs fully complementary to an mRNA are reported Among these, miR135a might control the stability and⁄ or translation of an alter-native form of the glycerate kinase mRNA by RNA interference I also dis-cuss the presence of human miRNAs in introns of paralogous genes and in miRNA clusters

Abbreviations

miRNA, microRNA; pre-miRNAs, miRNA precursors; RISC, RNA-induced silencing complex; snoRNA, small nucleolar RNA.

[13] However, a comprehensive list of these sequences

is still not available

It was recognized that many miRNAs are

evolu-tionarily conserved, some of them from worm to

human [14] MiRNA genes have been characterized

experimentally from a variety of organisms and

might have orthologues in other species, suggesting a

powerful method to predict the existence of new

miRNA genes I report here on the results of a

sys-tematic search for potential human orthologues of

mouse miRNAs deposited in the miRNA Registry

[15] and, vice-versa, of potential mouse orthologues

of known human miRNAs In addition, I searched

for human orthologues of recently identified rat

miRNAs [16] This led me to identify potential

orthologues of miRNAs that were described

previ-ously in one species, but not in the other After

inclusion of these new data, the total number of

human and mouse miRNAs deposited at the miRNA

Registry now approaches the theoretical number of

255 predicted by Lim et al [13]

Using this information, I wrote custom tracks that

allow for the localization of the miRNA genes in the

human and mouse genomes These are now available

on the UCSC Genome Browser [17] Using this tool, I

systematically determined the position and orientation

of miRNA genes relative to known transcriptional

units, examined the conservation of miRNA gene

local-ization between the human and mouse genomes, and

made a comprehensive list of miRNA clusters This

search led to several testable hypotheses concerning

the transcription of miRNA genes, and to the

predic-tion of new mRNA targets

Results and Discussion

New potential human and mouse microRNA

precursors

The entire set of human and mouse precursor and

mature miRNA sequences from the miRNA Registry

(version 2.2) was submitted to a BLAT search against

the human genome The results (in BED format) were

exported to excelTMto generate a table with 2873

ent-ries A similar BLAT search was performed against the

mouse genome, generating an excelTMtable of a

sim-ilar size These tables were then ranked according to

chromosome number and chromosome position and

filtered for perfect and near-perfect matches The

cor-responding sequences were subsequently examined for

a potential hairpin structure with mfold, and the

results were compared to those of known miRNAs

from the miRNA Registry

I only considered, as valid candidates, those poten-tial miRNA precursors that conformed to the empir-ical criteria proposed by Ambros et al [18] in particular, a hairpin structure of the lowest free energy, as predicted by mfold, and a minimum of 16 nucleotides of the mature miRNA engaged in Watson– Crick or G⁄ U base pairings (criterion C) The method used for searching the new miRNAs ensured their phy-logenetic conservation (criterion D) Moreover, the detection of the
22-nucleotide mature forms by Nor-thern blot (criterion A) and⁄ or their identification in a cDNA library (criterion B) were checked for the sequences deposited in the Rfam2.2 miRNA Registry (Wellcome Trust Sanger Institute) for at least one species In most cases, sequences of the new mature miRNAs were perfectly conserved between human and mouse, but differed by one nucleotide in few cases (see for example, mir-155) Further validation of these can-didates was performed on the basis of additional cri-teria, such as conservation of the host gene (see below)

or position relative to known miRNA clusters

This led to the identification of 60 new potential miRNA precursors (15 for human and 45 for mouse) that were made available before publication in Rfam version 3 of the miRNA Registry of the Wellcome Trust Sanger Institute (Table S1) Moreover, with the collaboration of S Griffiths-Jones (Wellcome Trust Sanger Institute), the names of several miRNAs depos-ited at the miRNA Registry have been changed, so that orthologous precursors, based on conservation of both sequences and synteny, have similar names in both human and mouse This was particularly important when a mature miRNA had multiple, closely related, precursors (see for example 9-2, mmu-mir-138-12 and mmu-mir-199a-1)

Moreover, the coordinates of both new and previ-ously described miRNA precursors were used to write custom tracks that allowed their localization in the human and mouse genomes on the UCSC Genome Browser A similar track was also written for the gen-ome of C elegans

New human miRNA precursors predicted

by homology with rat microRNAs

I searched for potential human orthologues of recently cloned rat miRNAs [16] The best BLAT hits were examined for conservation of both synteny in human and rodent genomes and the presence of a stem-loop structure This allowed one to propose 20 new human microRNA precursors (Table 1)

151 was identified in the mouse [19], and miR-151* in the rat [16] In both cases, the same predicted

Name hsa-miR-322 *hsa-miR-323 *hsa-miR-324-5p *hsa-miR-324-3p hsa-miR-325 *hsa-miR-326 *hsa-miR-328 hsa-miR-329 *hsa-miR-330 *hsa-miR-331 *hsa-miR-335

Note Dif

Same intron of the host gene of hsa-mir

Name *hsa-miR-337 *hsa-miR-338 *hsa-miR-339 *hsa-miR-340 *hsa-miR-342 hsa-miR-345 hsa-miR-346 *hsa-miR-1

Note Dif

nucleotides hsa-miR-1

precursor encodes both mature miRNAs from its 50

(miR-151*) and 30 (miR-151) portions This also holds

true for hsa-mir-151, although the mature form of

miR-151 differs from the rodent sequence by one

nuc-leotide This conservation reinforces the hypothesis

that, in mammals, the same precursor gives rise to both

miR-151 and miR-151* From the predicted hairpin

structure of the precursor, the energies of hybridization

of the four nucleotides located at the 50 end of

miR-151* and miR-151 are 1.7 and 0.1 kcalÆmol)1,

respect-ively As ‘RISC assembly favors the siRNA strand

whose 50end has a greater propensity to fray’ [20], it is

expected that miR-151* will be more abundant than

miR-151 Indeed, miR-151 was cloned only once

among 913 miR sequences [19] Although, to my

know-ledge, neither miR-151, nor miR-151* were cloned in

human, it is unlikely that the single base substitution

compared to rodent sequence (at nt 10 of miR-151)

could alter the balance between the two miRs

For certain rodent miRNA precursors, a BLAT

search in the human genome produced no matches

However, a likely orthologue could be found by exploring the most likely (i.e of conserved synteny) portion of human genome for conserved sequences; e.g mmu-mir-345 resides upstream of the AK047628 RefSeq gene Its human orthologue was found upstream of C14orf69, the best BLAT hit for AK047628 Such identification was made straightfor-ward by the examination of the ‘Human⁄ Chimp ⁄ Mouse⁄ Rat ⁄ Chicken Multiz Alignments & PhyloHMM Cons track’ of the human UCSC Genome Browser Similarly, hsa-mir-329 and -322 were identified on the basis of their conserved stem-loop structure and conserved position relative to other miRNAs or Ref-Seq genes However, the presumptive mature miRNAs hsa-miR-329 and -322 differ from their mouse and rat orthologues by four nucleotides (Fig 1A,B) Most of these changes retained base-pairing in the precursor miRNA by forming G⁄ U interactions or resided in un-paired positions (data not shown) Consequently, the folding free energy calculated using mfold was either little affected (mir-329) or even decreased in the case

Fig 1 Alignment of microRNA sequences from mammalian genomes (A–C) Alignment of the sequences from mammalian mir-329 (A), mir-322 (B) and mir-346 (C) Abbreviations: hsa, Homo sapiens; mmu, Mus musculus; rno, Rattus norvegicus; pan, Pan troglodytes (D) Alignment of the sequences of rodent mir-350 with the corresponding sequences from human and chimpanzee genomes Rodent sequences were retrieved from The miRNA Registry Human sequences were retrieved from the human genome sequence by examination

of highly conserved regions in the syntenic segments The sequence of pan-mir-350 was obtained by BLATing the human sequence against the chimpanzee genome The sequences of mir-329 and mir-298 are 100% conserved between human and chimpanzee The sequences of mature microRNAs are boxed The antisense miR boxes indicate the portion of the hairpin precursor structure that is base-paired with the miR, as predicted by MFOLD

of mir-322 (DG ¼)46.6 for human and )41.0 kcalÆ

mol)1for mouse)

The same conclusion holds for hsa-mir-346, where

the mature miR sequence differed from those of mouse

and rat by two and three nucleotides, respectively

(Fig 1C) In this case, the folding free energy of the

human and rat precursors remained comparable ()46.4

and)50.6 kcalÆmol)1, respectively)

Mutations in mature miARNs, particularly in their

30 portion, are compatible with their function as

trans-lational repressors [21] The same study, however,

revealed that G:U wobble pairing in the 50 region of

the miRNA had detrimental effects that could not be

predicted on the basis of changes in the free energy of

annealing with the target mRNA Therefore,

hsa-mir-329, hsa-mir-322 and hsa-mir-346 require further

experimental validation to be considered as bona fide

miRNA precursors

It was however, surprising that, in the predicted

hairpin structures of mir-329, -322 and -346, the

anti-sense sequence of the miR was more conserved than

the miR itself (Fig 1A–C) Significantly, a mutation in

the antisense sequence of miR-322 was accompanied by

a compensatory change in the miR sequence, so that a

G⁄ C base pair in the mouse precursor was replaced by

an A⁄ U in the human one The single change in the

antisense sequence of miR-329 occurred at an unpaired

position in both human and rodent precursor hairpin

structures As it is difficult to conceive that

evolution-ary pressure might be higher on the antisense than on

the sense strand [8], this may suggest that the antisense

strand was cloned accidentally in certain cases [16]

In a few cases, the homology search allowed

local-ization of human sequences similar to some rodent

miRNA precursors but that had accumulated

deleteri-ous mutations For example, the human orthologdeleteri-ous

sequence of rodent mir-350 could be localized in an

intron of the KARP-1-binding protein (KAB) gene

However, a nine base-pair deletion in the human and

chimpanzee genomes removed the first seven

nucleo-tides of the mature microRNA (Fig 1D) It this case,

it is noteworthy that the antisense strand accumulated

mutations, possibly due to a lack of selective pressure

after the inactivation of mir-350 by the deletion in the

mature miRNA sequence

Human and mouse miRNA precursors reside

in conserved regions of synteny

Using the UCSC Genome Browser, I examined

whether orthologous human and mouse miRNA

precursors reside in conserved synteny regions This

proved to be the case for miRNAs located in known

coding genes (see below) Many human miRNAs are located in introns of noncoding mRNAs In general in these cases, the mRNA was not or poorly conserved

in the mouse genome When human miRNAs were in nonconserved genes, or outside characterized genes, I examined the known flanking genes In all cases but three, human and mouse miRNAs were found to reside in conserved synteny regions The three appar-ent discrepancies in the position of a miRNA in the human and mouse genomes were 9-3,

hsa-mir-339 and hsa-mir-326 As detailed in Appendix S1, these cases most probably originate from errors in the present assembly of the mouse genome Orthologous human and mouse miRNAs thus reside either in introns of orthologous genes, and⁄ or in conserved synteny with surrounding genes

Human miRNA precursors that reside in introns

of known genes Eighty one human miRNA precursors were found to

be located in an intron of a known gene, or of a gene defined by a complete cDNA sequence, in the sense orientation (Table S2) It is however, important to note that human miRNAs that were classified as located outside of known genes might in fact reside in still uncharacterized splicing variants For example, hsa-mir-10b is located 972 nucleotides upstream of the HOXD4 gene (NM_014621) However, mmu-mir-10b resides in an intron of a long form of the mouse Hoxd4 pre-mRNA (NM_010469), but this alternative form has no documented human orthologue

According to current models, intronic miRNA pre-cursors that have the same orientation as their host gene might be produced upon cleavage of the spliced intron by the Drosha endonuclease In certain cases, the miRNA sequence is included in intronless ESTs that are often members of a cluster of overlapping, also intronless, ESTs It is striking that, often, no such EST cluster was found in the other introns of the miRNA host gene (e.g hsa-mir-103–2 and hsa-mir-98)

or only in adjacent introns Due to the uncertainty in the orientation of intronless ESTs, it was not possible

to assess the orientation of the clusters relative to that

of the host gene, and that of the microRNA Never-theless, this suggests that introns that host microRNAs might be particularly stable Alternatively, certain of these intronic miRNAs might be produced by unchar-acterized transcription units embedded in the same orientation in the apparent host gene However, this possibility appears unlikely, as to my knowledge, there

is only two examples of such a situation in the human genome [22,23]

In addition, 17 miRNAs were located in an intron

of a known gene, but in the antisense orientation

(Table S3) Among these, 10 were in a miRNA cluster

(see below) To determine how these miRNAs might

be generated, their genomic context was carefully

explored As shown in Table S3, several miRNAs in

this category are in fact in a transcription unit that has

an orientation opposite that of the apparent host gene

In particular, hsa-mir-302 lies within an intron of the

HDCMA18P gene in the antisense orientation, but in

an intron of ESTs BG207228 and BU565001 in the

sense orientation (Fig 2A) These two ESTs largely

overlap the HDCMA18P transcription unit in the

opposite orientation

Similarly, hsa-let-7d resides in the gene defined by

the BC045813 mRNA in the opposite orientation

(Fig 2B) In addition, it is located in a cluster of

50 unspliced ESTs that spans about 2-kb and

over-laps the intronless BC064349 mRNA The orientation

of this cluster is opposite that of the BC045813

mRNA, as shown by the polyA tails of the

BC064349 mRNA and of several ESTs (Fig 2B)

These data thus strongly suggest that hsa-let-7d,

and possibly hsa-let-7f-1 and hsa-let-7a-1, are part of

an intronless transcript antisense to the BC045813 transcription unit

A third example is hsa-mir-142 that lies within an intron of the AK090885 gene in the antisense orienta-tion, but also in the antisense, intronless and polyaden-ylated AX721088 mRNA (1.6kb) Interestingly, a natural chromosome translocation fuses the 50 portion

of hsa-mir-142 to a truncated c-Myc gene in aggressive B-cell leukemia [19,24] This translocation most prob-ably fuses the AX721088 transcription unit to the c-Mycgene

In several cases, the miRNA was embedded in a cluster of intronless ESTs, the orientation of which could not be ascertained This however, cannot be con-sidered as indicative of an antisense transcript, as a similar observation was made for some miRNAs that reside in introns in the same orientation as the host gene (see above)

Taken together, these observations suggest that when miRNA genes are located in introns of known genes in the antisense orientation, they might in fact

be part of transcription units in opposite orientation

to the presumptive host gene In the case of

hsa-mir-302, the microRNA is clearly located in an intron of

Fig 2 Antisense intronic miRNAs (A) Localization of hsa-mir-302 This miRNA resides in an intron of the HDCMA18P gene in the antisense orientation but in an intron of ESTs BG207228 and BU565001 in the sense orientation In all figures, miRNA genes are colored in green when they reside on the upper strand, and in magenta when on the lower strand (B) Localization of hsa-let-7d This miRNA resides in an intron of pre-mRNAs BC045813 and BC036695 in the antisense orientation The last exon of these two mRNAs overlaps the intronless BC064349 mRNA in the antisense direction The asterisks indicate the positions of the polyA tails of mRNAs and ESTs The miRNA hsa-let-7d resides in a cluster of several intronless ESTs, only some of which are shown The orientation of this cluster is antisense to that

of the BC045813 mRNA, as indicated by the presence of a polyA tail in the sequence of mRNA BC0644349 and of two ESTs The corres-ponding transcription unit might also contain hsa-let-7a-1 and hsa-let-7f-1 in the sense orientation Figures 2 and 3 are adapted from windows

of the UCSC Genome Browser.

the antisense transcript (Fig 2A) In other cases, like

hsa-mir-142, hsa-mir-133a-1⁄ mir-1-2 and possibly

hsa-let-7d⁄ let)1-7f-1⁄ let)1-7a-1, the microRNA is located in

an exon of the antisense transcript

Human miRNA precursors that overlap with

exons of known genes or ESTs

After intersecting the new human miRNA table of the

UCSC Genome Browser with the chrN_EST and

chrN_mRNA tables, I examined miRNAs that are

included in, or overlap with, exons The miRNAs that

were only included in intronless ESTs were for the

most part not further examined due to the uncertainty

in EST orientation In addition, the localization of

sev-eral miRNAs in an exon possibly corresponded to

intron retention events (hsa126, mir-25 and

-mir-224) Accordingly, these miRNAs were classified as

intronic

Except for the four examples discussed below

(hsa-mir-135a-1, hsa-mir-99b, hsa-let-7e, hsa-mir-125a),

miRNAs that are embedded in, or overlap with exons

of known transcripts, are always in the same

orienta-tion The corresponding genes were generally

noncod-ing, except for hsa-mir-198, which resides in the

30-UTR of the FSTL1 (follistatin-like) gene, and for

hsa-mir-133a-2, located in the C20orf166 gene that

encodes a potential 117 amino acid protein In

addi-tion, hsa-mir-21 probably resides in the sense

orienta-tion in an alternative form of the 30-UTR of the

VMP1 gene, characterized by mRNA BC053563

Accordingly, mmu-mir-21 is located in the 30-UTR of

the mouse VMP1 ortholog, 4930579A11Rik, also in

the sense orientation

These three cases are particularly intriguing, as

exci-sion of the miRNA precursor from the host mRNA by

the Drosha endonuclease would probably trigger the

degradation of the rest of the mRNA A similar

mech-anism is documented in E coli, where RNase III

cleaves its own mRNA at a stem-loop structure and

triggers its degradation [25] Similarly, RNase III

cleaves the polycistronic metY-nusA-infB RNA, to

release the metY tRNA and initiate the decay of the

nusB-infB protein-coding mRNA [26] Whether such a

mechanism also operates in higher eucaryotes remains

speculative Of note, no rodent orthologue of

hsa-mir-198 could be found by a BLAT search in the mouse

and rat genomes

Four miRNAs (34b, 205,

hsa-mir-133a-2 and hsa-mir-99b) overlap a splicing site In the

three first cases, it is not clear how the miRNA

precur-sor is produced It might originate either from an

exon, or from an intron of uncharacterized alternative

forms of the host gene mRNA The case of hsa-mir-99b is further discussed below

MiRNAs complementary to expressed sequences and potential regulation of glycerate kinase gene expression by miR-135a

The mouse mmu-mir-135a-1 miRNA resides in the antisense orientation in an alternative, long form of the 30-UTR of the 6230410P16Rik gene, the ortho-logue of the human GLYCTK (glycerate kinase) gene (Fig 3A) Therefore, mmu-miR-135a is perfectly com-plementary to the alternative form of Glyctk mRNA and might regulate its stability by an siRNA mechan-ism Interestingly, mmu-mir-135a-1 also resides in an intron of the spliced AK051019 mRNA in the sense orientation; the latter might thus be the actual miRNA host gene These two transcriptional units are largely overlapping, so that several genomic segments are exonic on both strands (Fig 3A)

A similar situation holds for the human genome: hsa-mir-135a-1 is located 741-base pairs downstream

of the GLYCTK gene, in the antisense orientation This miRNA is embedded in a cluster of 10 overlap-ping intronless ESTs that might be part of a longer, alternative form of the GLYCTK 30-UTR (Fig 3B) This hypothesis is supported by the fact that several ESTs of this cluster (AI493054, AI380271, AW204878, AW207007 and BM555864) are polyadenylated In addition, hsa-mir-135a-1 resides, in the sense orienta-tion, in an intron of EST AI936688 Based on these observations, it is tempting to speculate that, in both mouse and human, mir-135a-1 is produced from an intron of the host gene (AK051019 and AI936688 in mouse and human, respectively) and can direct the degradation of a long form of the glycerate kinase mRNA by an RNA interference mechanism Accord-ingly, a switch from GLYCTK to AI936688 gene tran-scription would be accompanied by the production of hsa-miR-135a, which could base-pair with pre-existing GLYCTK long mRNAs and trigger their degradation This mechanism would thus block glycerate kinase production at both the transcriptional and transla-tional levels, while the shorter form of glycerate kinase mRNA, which results from the use of an alternative polyA site, would not be affected In addition, hsa-miR-135a could be produced from its second precur-sor, hsa-mir-135a-2

As noted above, hsa-let-7e and hsa-mir-125a are located in the first exon of mRNA AK125996, in the antisense orientation, whereas hsa-mir-99b overlaps the splicing donor site (Fig 3C) Interestingly, the first exon of mRNA AK125996 overlaps that of the antisense

mRNA AY358799 over 113 nucleotides Although the

overlapping region is outside of the miRNA-containing

segment, it is tempting to speculate that these three

miRNAs reside, in the sense orientation, in a longer form of the AY358799 mRNA, and might regulate the translation and⁄ or stability of the mRNA AK125996

Fig 3 Antisense exonic miRNAs (A) Localization of mmu-mir-135a-1 This miRNA resides in the antisense orientation in an alternative lon-ger form of the 3 0 -UTR of the mouse 6230410P16Rik mRNA This gene is the orthologue of the human GLYCTK gene, as shown by BLATing the sequence of the human protein (see BLAT track) The miRNA mmu-mir-135a-1 is also located in an intron of the antisense AK051019 mRNA, indicated in red (B) Localization of hsa-mir-135a-1 This miRNA resides in a cluster of ESTs downstream of the GLYCTK gene, in the opposite orientation The asterisks indicate the localization of the polyA tail of some of these ESTs In addition, hsa-mir-135a-1 resides in the sense orientation in an intron of EST AI936688 (indicated in red) (C) Localization of hsa-mir-99b, hsa-let-7e and hsa-mir-125a The latter two miRNAs are in the first exon of the AK125996 mRNA, in the antisense orientation, whereas hsa-mir-99b overlaps the splicing donor site Note that the AK125996 mRNA overlaps in the antisense direction with the BC041134 and AY358799 mRNAs.

Only two other vertebrate miRNAs have been

previ-ously shown to be fully complementary to a cellular

mRNA: mmu-mir-127 and mmu-mir-136 reside in the

intronless Rtl1 gene, in the opposite orientation

Whereas the Rtl1 gene is paternally expressed, the two

miRNAs are only expressed from the maternal

chro-mosome [27] This reciprocal imprinting suggests that

mmu-miR-127 and mmu-miR-136 regulate Rtl1 gene

expression by an siRNA mechanism This situation

probably also holds for the human genome, as both

hsa-mir-127 and hsa-mir-136 reside in the inverse

ori-entation within the presumptive human orthologue of

Rtl1 (XM_352144)

Therefore, this study uncovers new cases where a

miRNA might regulate the stability of a cellular

mRNA through an siRNA mechanism In all cases

dis-cussed above, the miRNA resides on the opposite

strand relative to its target mRNA This situation

dif-fers from that of miR-196a, which is fully

complement-ary to the 30-UTR of HOXB8 mRNAs, except for a

G⁄ U wobble [28] In that case, the miRNA gene resides

at distance from its target gene, although in the same

HOX locus

MicroRNAs in gene families

As shown in Table 2, many related human precursor miRNAs reside in corresponding introns of paralogous genes The mature miRNAs are either identical

(miR-15, miR-218), closely related (miR-199a and b, miR26a and b), or display significant homology (hsa-mir-148b and -152, hsa-mir-107, -103-1 and -103-2) Therefore, the sequences of these intronic miRNAs have been largely conserved after gene duplications, raising the possibility that their function might have been conserved as well This also suggests that addi-tional genes in the families shown in Table 2 might contain still uncharacterized miRNAs Intriguingly, hsa-mir-211 and -204 are located within an intron of the TRPM1 and TRPM3 genes, respectively, that bracket paralogous exons These localizations are con-served in the mouse genome However, these two miR-NAs have no sequence similarity In this case, it is thus possible that the presence of miRNAs in the introns of the TRPM1 and TRPM3 genes is posterior

to the expansion of the TRPM gene family This hypothesis is reinforced by the fact that the other members of the family, TRPM2 and TRPM4 through TRPM8, do not host known miRNAs The large dif-ference in the size of the introns of the TRPM1 and TRPM3 genes that host mir-211 and mir-204 (3 and

44 kb, respectively) also suggests extensive rearrange-ments posterior to gene duplication

It is also intriguing to note that hsa-mir-199b, -199a-1 and -199a-2 reside in introns of the DNM1, DNM2 and DNM3 genes, respectively, but in the opposite orientation (Table 2) As discussed above, the DNM genes are thus probably not the actual hosts for these miRNAs Indeed, hsa-mir-199a-2 is embedded in a large cluster of ESTs antisense to the DNM3 gene that also contains hsa-mir-214 (Table S3) Similarly, mmu-mir-199a-2 and mmu-mir-214 reside in the opposite orientation in an intron of the dynamin 3 gene (9630020E24Rik) and are embedded

in a large cluster of antisense mRNAs and ESTs (not shown) Although there is no clear evidence for tran-scripts antisense to DNM1 and DNM2 genes, these observations suggest that the conservation of highly related miRNAs in the DNM gene family might result from the expansion in the vertebrate genomes of antisense transcripts in addition to the DNM genes themselves

Clusters of microRNAs The existence of miRNA clusters has been already noted [14,29,30], but a precise definition of a cluster

Table 2 MicroRNA that reside in introns of paralogous gene

families.

hsa-mir-199a-1 DNM2

hsa-mir-199a-2 DNM3

hsa-mir-218–1 SLIT2 Slit homologs, axonal guidance b

hsa-mir-218–2 SLIT3

hsa-mir-103–2 PANK2

hsa-mir-103–1 PANK3

hsa-mir-211 TRPM1 Transient receptor potential

cation channels (melastatins)

d

complex, subunits zeta

e

a

mmu-mir-153 is located in the Ptprn2 gene No documented miRNA

resides in the mouse Ptprn gene b

No documented miRNA resides in the human SLIT1 gene In the October 2003 freeze of the mouse

gen-ome, the Slit2 gene, as evidenced by BLATing the human Slit2 protein,

comes in two parts, located on chr5_random (aa 133–538) and chr5 (aa

539–1529) Mmu-mir-218–1 resides in the chr5_random part c

hsa-mir-103–2 and mir-103–1 are closely related to hsa-mir-107 (87 and 91%

identity, respectively) d

hsa-mir-211 and )204 have no sequence homol-ogy.ehsa-mir-148b and )152 display significant homology (77% identity

over 66 nucleotides).