Using the small RNA cloning method described by Lagos-Quintana et al. 31, cDNA libraries were constructed from RNA samples of mouse mammary glands collected at different stages (virgin 8 weeks, gestation 2-, 6- and 18-days, lactation 4-days and involution 1-day). A total of 455 clones were obtained and sequenced, of which 169, corresponding to 120 non redundant sequences, had inserts ranging between 18 and 28 nt in length (Additional file 1). Most of the remaining clones contained inserts smaller than 17 nt and were not further analysed. The 120 non redundant sequences were compared with known miRNAs by searching the miRBase database (release 12.0). Nine sequences matched with mature miRNAs, 3 of them perfectly with mouse let-7b (MG030), let-7c (MG070) and human miR-923 (MG041) respectively, 3 with a 1-nt difference with mouse miR-126-5p (MG034), miR-429 (MG018) and human miR-1268 (MG068), and 3 have partial homologies (Additional file 1). Human miR-923 is 100% homologous with the mouse genomic sequence. The above-mentioned 3 sequences that perfectly matched with known miRNAs were thus considered as corresponding to true miRNAs and not further analysed. Thirty-three sequences have partial homologies with 1 or several precursors of different species, among them 17 have partial homology with the miRNA strand, 8 with the miRNA* strand and 8 with the loop (Additional file 1).

One important criteria that distinguishes miRNA from other endogenous small RNA is the ability of their genomic flanking sequences to adopt a hairpin structure with the mature miRNA properly positioned within one of its strand in order to be excised during Dicer processing 32. The 117 sequences cloned were aligned with the mouse genome using BLAST, 31 of them were not found to be homologous. They were discarded; and considered as being cloning artefacts. For some of them we cannot exclude that the mismatches observed are due to RNA editing 33. Among the 86 homologous sequences, 15 showed homology with a single location, whereas the others could be located in several genomic regions, leading overall to a total of 441 localisations. To assess which of the 441 regions correspond to potential miRNA genes, their secondary structures were studied using the RNA folding program MFOLD 34. One hundred and fourteen chromosomal regions, which correspond to 40 cloned sequences, could form a stable hairpin structure. Among the 114 chromosomal regions, 57 were annotated as ribosomal RNA or small nuclear RNA and were discarded. The 57 remaining corresponded to 33 cloned sequences.

Their chromosomal locations and sequence alignments revealed that several cloned sequences were partially overlapping and could derive from the same precursor. In our study 5 families, corresponding to 13 sequences, were detected (Figure 1). Recently, the presence of variants has been reported after the characterization of cDNA libraries of small RNAs from porcine fibroblast cells 35 and from bovine adipose tissue and mammary gland 29. The 3'-end variants may thus arise from preferential degradation at the 3' end or from imprecise processing of miRNA precursors by Dicer, thereby generating miRNAs with differing 3' ends 3637. However, we cannot exclude the possibility that these miRNA variants originate from multiple genomic loci. The functional specialization of miRNA variants is still unknown. The 33 cloned sequences could derive from 41 precursors (Additional file 2). In fact one cloned sequence could derive from several precursors (Additional file 2, for example MG141 could be issue from 3 precursors). In some cases one cloned sequence possess i) 2 identical precursors localised in 2 different chromosomes (Additional file 2, MG016-13 and -16), ii) 3 identical precursors clustered in one chromosome (MG055) or iii) 2 similar precursors (with 1 or 3 different nt) in two different chromosomes (Additional file 2, MG009/MG037/MG056/MG066-06 and -17, MG016-04 and -1 or MG141-12 and -15, for example). Overall the precursors are distributed on all the chromosomes but chromosomes 5, 10 and Y (Figure 2). Chromosome 1 appears to contain more mammary gland miRNA genes than any other chromosome. Our result are in agreement with published results 3839 that showed that miRNA genes are distributed among all the chromosomes but chromosome Y. Twenty-six of the precursors are localised in intergenic regions, 14 in genes (intron: 11, 5'UTR: 1, 3'UTR: 1 and exon: 1) and 1 corresponds to one miRNA non-fully characterised (ENSMUSG00000076325) (Additional file 2). Thus the majority of miRNA genes are part of intergenic sequences (62%) as observed by Ro et al. 40. By analysing the proximity of the 57 chromosomal locations in the mouse genome, 4 miRNA clusters were observed (Table 1). The miRNA genes are in the same cluster if they are less than 1000 bp apart on the same chromosome 29. This physical proximity is consistent with recent reports of miRNA clustering within the human genome 41. Among the 4 clusters, 2 correspond to the association of two new precursors, MG016-01 and MG016-16, with 2 miRNAs already described mmu-miR-689-1 and mmu-miR-689-2, respectively (Table 1). The cluster MG055 on chromosome 1 has three repeats of the same precursor.

The evolutionary conservation of the 33 new miRNAs (25 distinct miRNAs + 8 variants; Table 2) was studied by comparison of their sequences with the human, rat, dog, monkey, chicken, Danio rerio, monodelphis and Ornithorhynchus anatinus genome sequences. Only 3 were conserved in all the species studied. Six were not detected in any other species (Table 2). If the phylogenetic conservation of the miRNA sequence is one of the criteria established by Ambros et al. 32 to characterize miRNA, it is important to recall that for some species the sequencing of their genome is not complete enough to detect all the sequences cloned here. It remains that the bioinformatics' analysis performed in this study has allowed identifying mouse specific miRNA.

miRNA expression is generally examined to better understand their physiological function. In this study, the tissue expression profile was used to assess the mammary gland specificity of the new miRNAs. In addition these analyses provide additional evidence for the identification of a bona fide miRNA. Expression of the 33 new miRNAs cloned in this study was analysed using an adapted RT-PCR described by Shi and Chiang 42 and Ro et al. 43 and detailed in the Methods' section. Using this approach, we could detect expression for 22 of them in at least one of the 4 analysed tissues (lactating or involuting mouse mammary gland, brain, muscle and liver, Figure 3 and Table 3). Based upon the multi-tissues expression profiles, these new miRNAs have been grouped into 4 categories: undetected (11 miRNAs), ubiquitously expressed (10 miRNAs), mammary specific (4 miRNAs), and expression in several but not all tissues (8 miRNAs). Among the latter category, one is not expressed in the mammary gland stages used for the RT-PCR experiment (lactation and involution).

The expression profile is the same for the 3 miRNAs present in the cluster MG113/MG130; MG141. The characterization of the undetected miRNAs was completed by studying their expression in some supplementary mammary gland stages (virgin 8 weeks, gestation day-18 or lactation day-4). Among the 11 miRNAs, only 2 have been detected (MG004 and MG123, data not shown). Among the 9 miRNAs undetected 3 correspond to variants of different families. We cannot exclude that these 3 variants are cloning artefact or that this lack of detection is due to individual variation for such miRNA families. In a same family all the variants are not expressed in the same tissues. A first analysis has allowed the identification of 4 mammary gland specific new miRNAs, to confirm this preliminary result their expression has been studied by RT-PCR on more tissues (heart, intestine, ovary, lung, spleen and kidney). Finally they were found not to be mammary gland specific: MG013 is expressed in all the tissues except kidney and heart; and MG056 is expressed in all the tissues except in ovary and kidney; MG119 have been detected in all tissues. MG009 was only found to be expressed in spleen and mammary gland.

In 11 cases the miRNA is detected in mammary gland in involution but not in lactation, a result in agreement with the expression profile of known miRNAs at different stages of the mammary gland biology obtained in our team (unpublished results). However, no correlation has been observed between the tissue-distribution of the miRNA expression (Table 3) and their evolutionary conservation (Table 2), or between their mammary expression profile (Table 3) and their conservation in mammals (Table 2). And the 6 miRNAs present only in the mouse genome are expressed in the 4 tissues tested. If in this study no mammary gland specific miRNA have been identified, several new miRNAs which are not ubiquitously expressed have been cloned.

The bioinformatics' analysis allowed the detection of potential precursors, but these results could not determine if these hairpin structures will be matured by RNase III, Drosha in the nucleus and Dicer in the cytoplasm. The use of expression vectors in cell culture, allowing the synthesis of hairpin structures that are matured into miRNAs, has already been demonstrated 4445. To test the functional validity of the precursors obtained by bioinformatics' analysis, we expressed them in transfected cells and checked for the presence of the mature miRNAs. The constructs carrying the precursors (MG008-X and MG053-01) were transiently transfected in COS-7 cells and the expression of the mature miRNAs (MG008 and MG053) was studied by Northern blot analysis (Figure 4). As a control the precursor of a known miRNA (let-7b) was used (data not shown). This approach could be used to validate precursor(s) of miRNA.

The tissues were isolated from FVB/N mice. The mammary glands were collected at different stages: virgin 8 weeks, gestation 2-, 6- and 18-days, lactation 3- and 4-days and involution 1- and 3-day. For all pregnancy samples, day 0 of pregnancy is the day we observed the vaginal plug. Day 1 of involution was designated as 24 h after the removal of the pups. All mouse manipulations were done following the French Commission de Génie Génétique recommendations.

For the library construction, small RNAs were isolated using mirVana^TMP miRNA isolation kit (Ambion) according the manufacturer's instructions. Small RNA cloning was performed as described by Lagos-Quintana et al. 31 without the step of concatemerization of the PCR products. Briefly, 500 μg of RNA were size fractionated using 15% denaturing polyacrylamide gel electrophoresis (PAGE). Excised gel bands were homogenized in NaCl 0.3 M overnight at 4°C to solubilize RNA. 3' adapter (5' phosphorylated) was ligated to the RNA fraction in presence of T4 RNA ligase (Amersham-Pharmacia). The mix was size fractionated using 15% PAGE and a 5' adapter was ligated. The sample was again size fractionated using 15% PAGE before to be reverse transcribed using primer complementary to the 3' adapter sequence and PCR amplified using primers on both adapters. Amplified products were cloned using pGEM-T vector system I (Promega). Clones with inserts were sequenced.

Small RNA sequences ranging between 18 and 25-nt in length obtained from the libraries were multi-aligned by CLUSTAL W software 47 to exclude redundant sequences 48. The distinct sequences were used to search in miRBase 49 with BLASTN to identify conserved miRNA 50. The sequences were mapped in the mouse genome from EnsEMBL mouse genome database 51. A fragment of ~260 nucleotides genomic sequence flanking the small RNA at both 5' and 3' ends was used for predicting the secondary structure of the miRNA precursor (stem-loop formations) using the MFOLD program (version 3.2) 3452. If a sequence including a small RNA formed a stem-loop, if the small RNA size ranged from 18–25 nt and had not been registered in the miRBase, we classified it as a new miRNA. The sequences of the new miRNA candidates and their precursors were subjected to a BLASTN search against NCBI genomes to estimate the species conservation.

Total RNAs were isolated from tissue samples and from COS cells using the RNA NOW kit (Biogentex inc.) according to the manufacturer's protocol with a small modification. The RNAs were precipitated at -20°C overnight by the addition of 3 volumes of ethanol. An equal-molar mix of total RNA from 3 different mice was used for reverse transcription.

Detection of miRNAs by RT-PCR has been realized after addition of linkers to the retro-transcribed RNA before amplification 4243. Briefly, 5 μg of total RNA were polyadenylated according to Ambion's protocol (polyA Polymerase, Ambion, Applied Biosystems, France). The polyadenylated RNAs were reverse transcribed with a polyT adapter composed of a polyT sequence and a universal oligonucleotide (5'CGAATTCTAGAGCTCGAGGCAGGCGACATGGCTGGCTAGTTAAGCTTGGTACCGAGCTCGGATCCACTAGTCC(T)₂₅(ACG)(ACGT)3'). Detection was achieved by PCR using a set of primers composed of the universal primer corresponding to the 5' end of the polyT-adapter (5'CGAATTCTAGAGCTCGAGGCAGG3') and a primer specific of the miRNA sequence (Additional file 3). Reaction products were separated on 2% agarose gel. The detection of miRNA let-7c has been used as positive control and was indeed detected in all analysed samples (data not shown).

For Northern blot analysis, 20 μg of total RNA were fractionated using a 15% PAGE, transferred to Hybond-N+ membrane (Amersham) by capillarity. Blots were hybridized overnight at 55°C with radioactively [γ-³²P]ATP labeled DNA oligonucleotide probe complementary to miRNA sequences in Phosphate buffer 53, washed twice with 2×SSC at 55°C, and exposed to Phosphor Screen and the StormScan software.

The precursor vectors contain the sequences design from MG008 and MG053 precursors (sequence in bold in Additional file 2). For MG053 precursor, the oligonucleotides sense (^5'CGGGATCCCGAGCGCCGAATCCCCGCCGCGCGTCGCGGCGTG^3') and antisense (^5'CGGGATCCCGGGTCTTCCGTACGCCACATTTCCCACGCCGCGACGCGCGGCGG^3') were annealed, filled with the Klenow fragment enzyme and digested by BamHI. The resulting fragment was inserted into the BamHI site of the pUHD10.3 plasmid 54. The MG008 precursor fragment was obtained by PCR on mouse genomic DNA using the primer sense (^5'CGGGATCCCGGTTTCAAAGTTTTGATAGGTTCTACGCATG^3') and antisense (^5'CGGGATCCCGGCTTCAGCTTTGACTTTCAGAGCACTGGG^3') and digested by BamHI before to be cloned into the BamHI site of the pUHD10.3 plasmid. The orientation of the insert was characterized by PCR using the primer sense of the precursor and the primer of the plasmid design in each side of the cloning site (primer rtTA/1: ^5'GATGCCCTGGAATTGACGAG^3' and primer glob/2: ^5'TATAACATGAATTTTTCAATAGCG^3'). In these vectors, the precursor is placed under the transcriptional regulation of the CMV promoter, already validated in COS-7 cell culture (personal communication).

COS-7 cell were cultured in Dulbecco's Modified Eagle's medium (DMEM) with addition of 10% fetal calf serum, 2 mM glutamine, penicillin 10 U/μl and streptomycin 100 U/μl at 37°C (5% of CO₂). The expression vectors were transfected using the jetPEI™ cationic reagent (PolyPlus transfection) following the manufacturer's instructions. Cells were harvested 48 h after transfection.