Identification of Novel Genes
Coding for Small Expressed
RNAs
Mariana Lagos-Quintana, Reinhard Rauhut, Winfried Lendeckel,
Thomas Tuschl*
In Caenorhabditis elegans, lin-4 and let-7 encode 22- and 21-nucleotide (nt)
RNAs, respectively, which function as key regulators of developmental timing.
Because the appearance of these short RNAs is regulated during development,
they are also referred to as small temporal RNAs (stRNAs). We show that many
21- and 22-nt expressed RNAs, termed microRNAs, exist in invertebrates and
vertebrates and that some of these novel RNAs, similar to let-7 stRNA, are
highly conserved. This suggests that sequence-specific, posttranscriptional reg-
ulatory mechanisms mediated by small RNAs are more general than previously
appreciated.
Two distinct pathways exist in animals and
plants in which 21- to 23-nt RNAs function
as posttranscriptional regulators of gene ex-
pression. Small interfering RNAs (siRNAs)
act as mediators of sequence-specific mRNA
degradation in RNA interference (RNAi) (1–
5), whereas stRNAs regulate developmen-
tal timing by mediating sequence-specific
repression of mRNA translation (6 –11).
siRNAs and stRNAs are excised from dou-
ble-stranded RNA (dsRNA) precursors by
Dicer (12–14 ), a multidomain ribonuclease
III protein, thus producing RNA species of
similar sizes. However, siRNAs are be-
lieved to be double-stranded (2, 5, 12),
whereas stRNAs are single-stranded (8).
We previously developed a directional
cloning procedure to isolate siRNAs after
processing of long dsRNAs in Drosophila
melanogaster embryo lysate (2). Briefly, 5⬘
and 3⬘ adapter molecules were ligated to the
ends of a size-fractionated RNA population,
followed by reverse transcription polymerase
chain reaction (PCR) amplification, con-
catamerization, cloning, and sequencing.
This method, originally intended to isolate
siRNAs, led to the simultaneous identifica-
tion of 16 novel 20- to 23-nt short RNAs,
which are encoded in the D. melanogaster
genome and are expressed in 0- to 2-hour
embryos (Table 1). The method was adapted
to clone RNAs in a similar size range from
HeLa cell total RNA (15), which led to the
Department of Cellular Biochemistry, Max Planck In-
stitute for Biophysical Chemistry, Am Fassberg 11,
D-37077 Go¨ttingen, Germany.
*To whom correspondence should be addressed. E-
mail: ttuschl@mpibpc.gwdg.de
Fig. 1. Expression of
miRNAs. Representa-
tive examples of
Northern blot analy-
sis are depicted (21).
The position of 76-nt
val-tRNA is indicated
on the blots; 5S rRNA
serves as a loading
control. (A) Northern
blots of total RNA
isolated from staged
populations of D.
melanogaster, probed
for the indicated
miRNA. E, embryo; L,
larval stage; P, pupa;
A, adult; S2, Schnei-
der-2 cells. (B) North-
ern blots of total RNA
isolated from HeLa
cells, mouse kidneys,
adult zebrafish, frog
ovaries, and S2 cells, probed for the indicated miRNA.
Fig. 2. Genomic orga-
nization of miRNA
gene clusters. The
precursor structure is
indicated as a box,
and the location of
the miRNA within the
precursor is shown in
black; the chromo-
somal location is also
indicated to the right.
(A) D. melanogaster miRNA gene clusters. (B) Human miRNA gene clusters. The cluster of
let-7a-1 and let-7f-1 is separated by 26,500 nt from a copy of let-7d on chromosomes 9 and
17. A cluster of let-7a-3 and let-7b, separated by 938 nt on chromosome 22, is not illustrated.
R EPORTS
www.sciencemag.org SCIENCE VOL 294 26 OCTOBER 2001 853
identification of 21 novel human micro-
RNAs (Table 2), thus providing further evi-
dence for the existence of a large class of small
RNAs with potential regulatory roles. Because
of their small size, and in agreement with the
authors of two related papers in this issue (16,
17), we refer to these novel RNAs as micro-
RNAs (miRNAs). The miRNAs we studied
are abbreviated as miR-1 to miR-33, and the
genes encoding miRNAs are named mir-1 to
mir-33. Highly homologous miRNAs are re-
ferred to by the same gene number, but followed
by a lowercase letter; multiple genomic copies
of a mir gene are annotated by adding a dash
and a number.
The expression and size of the cloned, en-
dogenous short RNAs were also examined by
Northern blotting (Fig. 1 and Tables 1 and 2).
For analysis of D. melanogaster RNAs, total
RNA was prepared from different developmen-
tal stages, as well as from cultured Schneider-2
(S2) cells, which were originally derived from
20- to 24-hour D. melanogaster embryos (18)
(Fig. 1 and Table 1). miR-3 to miR-7 are ex-
pressed only during embryogenesis and not at
later developmental stages. The temporal ex-
pression of miR-1, miR-2, and miR-8 to miR-13
was less restricted. These miRNAs were ob-
served at all developmental stages, and signifi-
cant variations in the expression levels were
sometimes observed. Interestingly, miR-1,
miR-3 to miR-6, and miR-8 to miR-11 were
completely absent from cultured S2 cells,
whereas miR-2, miR-7, miR-12, and miR-13
were present in S2 cells, therefore indicating cell
type–specific miRNA expression. miR-1, miR-
8, and miR-12 expression patterns are similar to
those of lin-4 stRNA in C. elegans, as their
expression is strongly up-regulated in larvae and
sustained to adulthood (19). miR-9 and miR-11
are present at all stages but are strongly reduced
in the adult, which may reflect a maternal con-
tribution from germ cells or expression in one
sex only.
The mir-3 to mir-6 genes are clustered (Fig.
Fig. 3. Predicted precursor structures of D. melanogaster miRNAs. RNA
secondary structure prediction was performed using mfold version 3.1
(32) and manually refined to accommodate G/U wobble base pairs in the
helical segments. The miRNA sequence is underlined. The actual size of
the stem-loop structure is not known experimentally and may be slightly
shorter or longer than represented. Multicopy miRNAs and their corre-
sponding precursor structures are also shown.
R EPORTS
26 OCTOBER 2001 VOL 294 SCIENCE www.sciencemag.org854
Fig. 4. Predicted precursor structures of human miRNAs. For legend, see Fig. 3.
R EPORTS
www.sciencemag.org SCIENCE VOL 294 26 OCTOBER 2001 855
2A), and mir-6 is present as triple repeat with
slight variations in the mir-6 precursor sequence
but not in the miRNA sequence itself. The ex-
pression profiles of miR-3 to miR-6 are highly
similar (Table 1), which suggests that a single
embryo-specific precursor transcript may give
rise to the different miRNAs or that the same
enhancer regulates miRNA-specific promoters.
Several other fly miRNAs are also found in
gene clusters (Fig. 2A).
The expression of HeLa cell miR-15 to miR-
33 was examined by Northern blotting using
HeLa cell total RNA, in addition to total RNA
prepared from mouse kidney, adult zebrafish,
Xenopus laevis ovary, and D. melanogaster S2
cells (Fig. 1B and Table 2). miR-15 and miR-16
are encoded in a gene cluster (Fig. 2B) and are
detected in mouse kidney, adult zebrafish, and
very weakly in frog ovary, which may result
from miRNA expression in somatic ovary tissue
rather than in oocytes. mir-17 to mir-20 are also
clustered (Fig. 2B) and are expressed in HeLa
cells and adult zebrafish, but undetectable in
mouse kidney and frog ovary (Fig. 1 and Table
2), and therefore represent a likely case of tis-
sue-specific miRNA expression.
The majority of vertebrate and invertebrate
miRNAs identified in this study are not related
by sequence, but a few exceptions do exist and
are similar to results previously reported for
let-7 RNA (8). Sequence analysis of the D.
melanogaster miRNAs revealed four such in-
stances of sequence conservation between inver-
tebrates and vertebrates. miR-1 homologs are
encoded in the genomes of C. elegans, C. brigg-
sae, and humans and are found in cDNAs from
zebrafish, mice, cows, and humans. The expres-
sion of mir-1 was detected by Northern blotting
in total RNA from adult zebrafish and C. el-
egans, but not in total RNA from HeLa cells or
mouse kidney (Table 2) (20). Interestingly, al-
though mir-1 and let-7 are both expressed in
adult flies (Fig. 1A) (8) and are both undetected
in S2 cells, only let-7 is detectable in HeLa cells.
This represents another case of tissue-specific
expression of an miRNA and indicates that
miRNAs may play a regulatory role not only in
developmental timing but also in tissue specifi-
cation. miR-7 homologs were found by database
searches of the mouse and human genomes and
of expressed sequence tags (ESTs). Two mam-
malian miR-7 variants are predicted by se-
quence analysis in mice and humans and were
detected by Northern blotting in HeLa cells and
adult zebrafish, but not in mouse kidney (Table
2). Similarly, we identified mouse and human
miR-9 and miR-10 homologs by database
searches but only detected mir-10 expression in
mouse kidney.
The identification of evolutionarily related
miRNAs, which have already acquired multiple
sequence mutations, was not possible by stan-
dard bioinformatic searches. Direct comparison
of the D. melanogaster miRNAs with the hu-
man miRNAs identified an 11-nt segment
shared between D. melanogaster miR-6 and
HeLa miR-27, but no further relationships were
detected. It is possible that most miRNAs only
act on a single target and therefore allow for
rapid evolution by covariation. Highly con-
served miRNAs may act on more than one
target sequence and therefore have a reduced
probability for evolutionary drift by covariation
(8). An alternative interpretation is that the sets
of miRNAs from D. melanogaster and humans
are fairly incomplete and that many more mi-
RNAs remain to be discovered, which will
provide the missing evolutionary links.
lin-4 and let-7 stRNAs were predicted to be
excised from longer transcripts that contain
stem-loop structures about 30 base pairs in
length (6, 8). Database searches for newly iden-
tified miRNAs revealed that all miRNAs are
flanked by sequences that have the potential to
form stable stem-loop structures (Figs. 3 and 4).
In many cases, we were able to detect the pre-
dicted precursors (about 70 nt) by Northern
blotting (Fig. 1). Some miRNA precursor se-
quences were also identified in mammalian
cDNA (EST) databases (21), indicating that pri-
mary transcripts longer than 70-nt stem-loop
precursors also exist. We never cloned a 22-nt
RNA complementary to any of the newly iden-
tified miRNAs, and it is as yet unknown how the
cellular processing machinery distinguishes be-
tween an miRNA and its complementary strand.
Comparative analysis of the precursor stem-
loop structures indicates that the loops adjacent
to the base-paired miRNA segment can be lo-
cated on either side of the miRNA sequence
(Figs. 3 and 4), suggesting that neither the 5⬘ nor
the 3⬘ location of the stem-closing loop is the
determinant of miRNA excision. It is also un-
likely that the structure, length, or stability of the
precursor stem is the critical determinant be-
cause the base-paired structures are frequently
imperfect and interspersed by G/U wobbles and
less stable, non–Watson-Crick base pairs such
as G/A, U/U, C/U, and A/A. Therefore, a se-
quence-specific recognition process is a likely
determinant for miRNA excision, perhaps me-
Table 1. D. melanogaster miRNAs. The sequences given represent the most
abundant, and typically longest, miRNA sequence identified by cloning;
miRNAs frequently vary in length by one or two nucleotides at their 3⬘
termini. From 222 short RNAs sequenced, 69 (31%) corresponded to
miRNAs, 103 (46%) to already characterized functional RNAs (rRNA, 7SL
RNA, and tRNA), 30 (14%) to transposon RNA fragments, and 20 (10%)
sequences had no database entry. The frequency for cloning a particular
miRNA as a percentage relative to all identified miRNAs is indicated. Results
of Northern blotting of total RNA isolated from staged populations of D.
melanogaster are summarized. E, embryo; L, larval stage; P, pupa; A, adult; S2,
Schneider-2 cells. The strength of the signal within each blot is represented
from strongest (⫹⫹⫹) to undetected (⫺). let-7 stRNA was probed as the
control. GenBank accession numbers and homologs of miRNAs identified by
database searching in other species are provided in (21).
miRNA Sequence (5⬘ to 3⬘)
Freq.
(%)
E
0to
3 hours
E
0to
6 hours
L1
⫹
L2
L3 P A S2
miR-1 UGGAAUGUAAAGAAGUAUGGAG 32 ⫹ ⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹⫹⫹ –
miR-2a* UAUCACAGCCAGCUUUGAUGAGC 3
miR-2b* UAUCACAGCCAGCUUUGAGGAGC 3 ⫹⫹ ⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹ ⫹⫹⫹
miR-3 UCACUGGGCAAAGUGUGUCUCA 9 ⫹⫹⫹ ⫹⫹⫹ –––––
miR-4 AUAAAGCUAGACAACCAUUGA 6 ⫹⫹⫹ ⫹⫹⫹ –––––
miR-5 AAAGGAACGAUCGUUGUGAUAUG 1 ⫹⫹⫹ ⫹⫹⫹ ⫹/– ⫹/––––
miR-6 UAUCACAGUGGCUGUUCUUUUU 13 ⫹⫹⫹ ⫹⫹⫹ ⫹/– ⫹/––––
miR-7 UGGAAGACUAGUGAUUUUGUUGU 4 ⫹⫹⫹ ⫹⫹ ⫹/– ⫹/– ⫹/– ⫹/– ⫹/–
miR-8 UAAUACUGUCAGGUAAAGAUGUC 3 ⫹/– ⫹/– ⫹⫹⫹ ⫹⫹⫹ ⫹ ⫹⫹⫹ –
miR-9 UCUUUGGUUAUCUAGCUGUAUGA 7 ⫹⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹/– –
miR-10 ACCCUGUAGAUCCGAAUUUGU 1 ⫹ ⫹ ⫹⫹ ⫹⫹⫹ ⫹/– ⫹ –
miR-11 CAUCACAGUCUGAGUUCUUGC 7 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹ –
miR-12 UGAGUAUUACAUCAGGUACUGGU 7 ⫹ ⫹ ⫹⫹⫹⫹⫹ ⫹⫹⫹⫹/–
miR-13a* UAUCACAGCCAUUUUGACGAGU 1 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹ ⫹⫹⫹ ⫹⫹⫹
miR-13b* UAUCACAGCCAUUUUGAUGAGU 0
miR-14 UCAGUCUUUUUCUCUCUCCUA 1 – – – – – – –
let-7 UGAGGUAGUAGGUUGUAUAGUU 0 – – – – ⫹⫹⫹ ⫹⫹⫹ –
*Similar miRNA sequences are difficult to distinguished by Northern blotting because of potential cross-hybridization of probes.
R EPORTS
26 OCTOBER 2001 VOL 294 SCIENCE www.sciencemag.org856
diated by members of the Argonaute (RDE-1/
AGO1/PIWI) protein family. Two members of
this family, ALG-1 and ALG-2, have recently
been shown to be critical for stRNA processing
in C. elegans (13). Members of the Argonaute
protein family are also involved in RNAi and
posttranscriptional gene silencing. In D. mela-
nogaster, these include Argonaute2, a compo-
nent of the siRNA-endonuclease complex
(RISC) (22), and its relative Aubergine, which is
important for silencing of repeat genes (23). In
other species, these include RDE-1 in C. elegans
(24); Argonaute1 in Arabidopsis thaliana (25);
and QDE-2 in Neurospora crassa (26). In ad-
dition to the RNase III Dicer (12, 13), the Ar-
gonaute family represents another evolutionary
line between RNAi and miRNA maturation.
Despite advanced genome projects, comput-
er-assisted detection of genes encoding func-
tional RNAs remains problematic (27). Clon-
ing of expressed, short functional RNAs, sim-
ilar to EST approaches (RNomics), is a pow-
erful alternative and probably the most
efficient method for identification of such
novel gene products (28 –31). The number of
functional RNAs has been widely underesti-
mated and is expected to grow rapidly be-
cause of the development of new functional
RNA cloning methodologies.
The challenge for the future is to define the
function and the potential targets of these novel
miRNAs by using bioinformatics as well as
genetics and to establish a complete catalog of
time- and tissue-specific distribution of the al-
ready identified and yet to be uncovered
miRNAs. lin-4 and let-7 stRNAs negatively reg-
ulate the expression of proteins encoded by
mRNAs in which 3⬘ untranslated regions con-
tain sites of complementarity to the stRNA (9–
11). Because these interaction domains are only
6 to 10 base pairs long and often contain small
bulges and G/U wobbles (9 –11), the prediction
of miRNA target mRNAs represents a challeng-
ing bioinformatic and/or genetic task. A pro-
found understanding of the expression, process-
ing, and action of miRNAs may enable the
development of more general methods to direct
the regulation of specific gene targets and may
also lead to new ways of reprogramming tissues.
References and Notes
1. S. M. Elbashir et al., Nature 411, 494 (2001).
2. S. M. Elbashir, W. Lendeckel, T. Tuschl, Genes Dev. 15,
188 (2001).
3. A. J. Hamilton, D. C. Baulcombe, Science 286, 950
(1999).
4. S. M. Hammond, E. Bernstein, D. Beach, G. J. Hannon,
Nature 404, 293 (2000).
5. P. D. Zamore, T. Tuschl, P. A. Sharp, D. P. Bartel, Cell
101, 25 (2000).
6. R. C. Lee, R. L. Feinbaum, V. Ambros, Cell 75, 843
(1993).
7. B. J. Reinhart et al., Nature 403, 901 (2000).
8. A. E. Pasquinelli et al., Nature 408, 86 (2000).
9. V. Ambros, Curr. Opin. Genet. Dev. 10, 428 (2000).
10. E. G. Moss, Curr. Biol. 10, R436 (2000).
11. F. Slack, G. Ruvkun, Annu. Rev. Genet. 31, 611 (1997).
12. G. Hutva´gner, J. McLachlan, E
´
.Ba´lint, T. Tuschl, P. D.
Zamore, Science 293, 834 (2001).
13. A. Grishok et al., Cell 106, 23 (2001).
14. E. Bernstein, A. A. Caudy, S. M. Hammond, G. J.
Hannon, Nature 409, 363 (2001).
15. Cloning of 19- to 24-nt RNAs from D. melanogaster 0-
to 2-hour embryo lysate was performed as described
(2). For cloning of HeLa miRNAs, 1 mg of HeLa total
RNA was separated on a 15% denaturing polyacryl-
amide gel, and RNA of 19- to 25-nt size was recovered.
A5⬘ phosphorylated 3⬘ adapter oligonucleotide (5⬘
pUUUaaccgcgaattccagx: uppercase, RNA; lowercase,
DNA; p, phosphate; x, 4-hydroxymethylbenzyl) and a 5⬘
adapter oligonucleotide (5⬘ acggaattcctcactAAA: upper-
case, RNA; lowercase, DNA) were ligated to the short
HeLa cell RNAs. Reverse transcription PCR was per-
formed with 3⬘ primer (5⬘ GACTAGCTGGAATTCGCG-
GTTAAA) and 5⬘ primer (5⬘ CAGCCAACGGAAT TCCT-
CACTAAA), followed by concatamerization after Eco RI
digestion and T4 DNA ligation (2). After ligation of
concatamers into pCR2.1 TOPO vectors, about 100
clones were selected and subjected to sequencing.
16. N. C. Lau, L. P. Lim, E. G. Weinstein, D. P. Bartel,
Science 294, 858 (2001).
17. R. C. Lee, V. Ambros, Science 294, 862 (2001).
18. I. Schneider, J. Embryol. Exp. Morphol. 27, 353
(1972).
19. R. Feinbaum, V. Ambros, Dev. Biol. 210, 87 (1999).
20. M. Lagos-Quintana, R. Rauhut, W. Lendeckel, T. Tus-
chl, data not shown.
Table 2. Human miRNAs. From 220 short RNAs sequenced, 100 (45%) corresponded to miRNAs, 53 (24%) to already characterized functional RNAs (rRNA,
snRNA, and tRNA), and 67 (30%) of the sequences had no database entry. Results of Northern blotting of total RNA isolated from different vertebrate species
and S2 cells are indicated. For legend, see Table 1.
miRNA Sequence (5⬘ to 3⬘)
Freq.
(%)
HeLa
cells
Mouse
kidney
Adult
fish
Frog
ovary
S2
let-7a* UGAGGUAGUAGGUUGUAUAGUU 10 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ––
let-7b* UGAGGUAGUAGGUUGUGUGGUU 13
let-7c* UGAGGUAGUAGGUUGUAUGGUU 3
let-7d* AGAGGUAGUAGGUUGCAUAGU 2 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ––
let-7e* UGAGGUAGGAGGUUGUAUAGU 2 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ––
let-7f* UGAGGUAGUAGAUUGUAUAGUU 1
miR-15 UAGCAGCACAUAAUGGUUUGUG 3 ⫹⫹⫹ ⫹⫹ ⫹ ⫹/– –
miR-16 UAGCAGCACGUAAAUAUUGGCG 10 ⫹⫹⫹ ⫹ ⫹/– ⫹/– –
miR-17 ACUGCAGUGAAGGCACUUGU 1 ⫹⫹⫹ ––––
miR-18 UAAGGUGCAUCUAGUGCAGAUA 2 ⫹⫹⫹ ––––
miR-19a* UGUGCAAAUCUAUGCAAAACUGA 1 ⫹⫹⫹ – ⫹/– – –
miR-19b* UGUGCAAAUCCAUGCAAAACUGA 3
miR-20 UAAAGUGCUUAUAGUGCAGGUA 4 ⫹⫹⫹ – ⫹ ––
miR-21 UAGCUUAUCAGACUGAUGUUGA 10 ⫹⫹⫹ ⫹ ⫹⫹ ––
miR-22 AAGCUGCCAGUUGAAGAACUGU 10 ⫹⫹⫹ ⫹⫹⫹ ⫹ ⫹/– –
miR-23 AUCACAUUGCCAGGGAUUUCC 2 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹ –
miR-24 UGGCUCAGUUCAGCAGGAACAG 4 ⫹⫹ ⫹⫹⫹ ⫹⫹ ––
miR-25 CAUUGCACUUGUCUCGGUCUGA 3 ⫹⫹⫹ ⫹ ⫹⫹ ––
miR-26a* UUCAAGUAAUCCAGGAUAGGCU 2 ⫹ ⫹⫹ ⫹⫹⫹ ––
miR-26b* UUCAAGUAAUUCAGGAUAGGUU 1 –
miR-27 UUCACAGUGGCUAAGUUCCGCU 2 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹ ––
miR-28 AAGGAGCUCACAGUCUAUUGAG 2 ⫹⫹⫹ ⫹⫹⫹ –––
miR-29 CUAGCACCAUCUGAAAUCGGUU 2 ⫹ ⫹⫹⫹ ⫹/– – –
miR-30 CUUUCAGUCGGAUGUUUGCAGC 2 ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ––
miR-31 GGCAAGAUGCUGGCAUAGCUG 2 ⫹⫹⫹ ––––
miR-32 UAUUGCACAUUACUAAGUUGC 1 – – – – –
miR-33 GUGCAUUGUAGUUGCAUUG 1 – – – – –
miR-1 UGGAAUGUAAAGAAGUAUGGAG 0 – – ⫹ ––
miR-7 UGGAAGACUAGUGAUUUUGUUGU 0 ⫹ – ⫹/– – ⫹/–
miR-9 UCUUUGGUUAUCUAGCUGUAUGA 0 – – – – –
miR-10 ACCCUGUAGAUCCGAAUUUGU 0 – ⫹ –––
*Similar miRNA sequences are difficult to distinguish by Northern blotting because of potential cross-hybridization of probes.
R EPORTS
www.sciencemag.org SCIENCE VOL 294 26 OCTOBER 2001 857
