MicroRNAs (miRNAs) constitute a novel, phylogenetically extensive family of small RNAs (∼22 nucleotides) with potential roles in gene regulation. Apart from the finding that miRNAs are produced by Dicer from the precursors of ∼70 nucleotides (pre‐miRNAs), little is known about miRNA biogenesis. Some miRNA genes have been found in close conjunction, suggesting that they are expressed as single transcriptional units. Here, we present in vivo and in vitro evidence that these clustered miRNAs are expressed polycistronically and are processed through at least two sequential steps: (i) generation of the ∼70 nucleotide pre‐miRNAs from the longer transcripts (termed pri‐miRNAs); and (ii) processing of pre‐miRNAs into mature miRNAs. Subcellular localization studies showed that the first and second steps are compartmentalized into the nucleus and cytoplasm, respectively, and that the pre‐miRNA serves as the substrate for nuclear export. Our study suggests that the regulation of miRNA expression may occur at multiple levels, including the two processing steps and the nuclear export step. These data will provide a framework for further studies on miRNA biogenesis.
Recent discoveries of an abundant class of ∼22 nucleotide (nt) RNAs, microRNAs (miRNAs), have widened the scope of eukaryotic gene regulation (Lagos‐Quintana et al., 2001; Lau et al., 2001; Lee and Ambros, 2001; Mourelatos et al., 2002). Over a hundred miRNA genes have been discovered so far from biochemical and bioinformatic studies of Caenorhabditis elegans, Drosophila melanogaster and Homo sapiens. The number is expected to increase sharply as the survey for miRNAs continues. Only nine out of 40 miRNAs identified by Dreyfuss et al. (Mourelatos et al., 2002) overlapped with those charted by Tuschl et al. (Lagos‐Quintana et al., 2001), although both groups used the same cell line (HeLa cells) for the cloning. Moreover, many miRNAs appeared in the cloning only once, implying that there remain rare unidentified miRNAs.
The paradigm for the function and biogenesis of miRNAs has been provided by lin‐4 and let‐7, which were originally identified by genetic analysis of C.elegans developmental timing (Lee et al., 1993; Reinhart et al., 2000). lin‐4 and let‐7 act as post‐transcriptional repressors of their target genes when bound to their specific sites in the 3′ untranslated region of the target mRNA (Lee et al., 1993; Wightman et al., 1993; Moss et al., 1997; Olsen and Ambros, 1999; Slack et al., 2000). The functions of other miRNAs are currently unknown. However, given the diversity in sequence and expression pattern, miRNAs are expected to play various roles in a wide range of regulatory pathways (Lai, 2002).
Little is known about the biogenesis of miRNAs. RNAs of ∼70 nt as well as ∼22 nt have been detected from northern blot analyses for most miRNAs, suggesting that these ∼70 nt RNAs are the precursors of ∼22 nt miRNAs (Lee et al., 1993; Pasquinelli et al., 2000; Reinhart et al., 2000; Grishok et al., 2001). An interesting, common feature of miRNAs is that in silico secondary structure prediction indicates that the ∼70 nt putative precursor forms a stem–loop structure with some bulges (Lagos‐Quintana et al., 2001; Lau et al., 2001; Lee and Ambros, 2001; Mourelatos et al., 2002). Because Dicer, the multi‐domain RNase III‐type protein, was found to function in generating small interfering RNAs (siRNAs) that are similar in size (21–25 nt) (Hammond et al., 2000; Bernstein et al., 2001), a prediction was made that Dicer also functions in the processing of ∼70 nt RNAs into mature miRNAs. Indeed, it was later shown that immunoprecipitated Dicer generates ∼22 nt miRNAs from in vitro synthesized ∼70 nt RNAs (Ketting et al., 2001; Knight and Bass, 2001). Moreover, when the human (Helicase‐MOI) and C.elegans (dcr‐1) homologs of Dicer were knocked down or knocked out, the ∼70 nt RNAs accumulated while ∼22 nt mature miRNAs disappeared (Grishok et al., 2001; Hutvagner et al., 2001; Ketting et al., 2001; Knight and Bass, 2001). Therefore, the ∼70 nt RNAs are the precursors of mature miRNAs and are accordingly referred to as ‘pre‐miRNAs’. Apart from Dicer, Alg‐1 and Alg‐2 in C.elegans appear to be involved in miRNA maturation, although their precise roles in processing are not clear (Grishok et al., 2001). More recently, a number of human miRNAs have been found in association with eIF2C2, Gemin3 and Gemin4 to form an ∼15S RNP complex (Mourelatos et al., 2002). The function of this complex and each component protein remains to be determined.
Several miRNA genes exist as clusters of 2–7 genes with intervals as short as a few nucleotides long (Lagos‐Quintana et al., 2001; Lau et al., 2001; Mourelatos et al., 2002). The expression profiles of clustered genes are highly similar, raising the possibility that transcription of such miRNAs is controlled by common regulatory sequences. One possibility is that a single promoter drives transcription of the clustered miRNA genes. If so, the nascent transcripts must be polycistronic and therefore would require an additional unknown processing step to produce the ∼70 nt pre‐miRNAs before Dicer can participate and generate mature miRNAs. In other words, maturation of miRNAs may occur through at least two steps: (i) processing of the nascent transcript to the ∼70 nt precursor; and (ii) processing of the ∼70 nt to the ∼22 nt mature miRNA. Alternatively, it is also plausible that the ∼70 nt precursor itself is the nascent transcript. In this case, each miRNA in the cluster is transcribed from its own promoter, but the same enhancer controls transcription for coordinated expression.
Another open question regarding miRNA biogenesis is about the subcellular localization and transport of miRNAs. Nucleocytoplasmic transport (especially export) is critical in the expression and functions of RNAs (Wolin and Matera, 1999; Warner, 2001; Will and Luhrmann, 2001; Dreyfuss et al., 2002; Reed and Hurt, 2002). For instance, pre‐mRNAs are retained in the nucleus until splicing is successfully carried out, so that only correctly processed mRNAs can pass the ‘quality control’ and become available for cytoplasmic translation (Chang and Sharp, 1989; Legrain and Rosbash, 1989; Luo and Reed, 1999; Maquat and Carmichael, 2001). Biogenesis of the UsnRNAs is also closely linked to the nucleocytoplasmic transport. The UsnRNAs, with the exception of U6 and U6atac, must first be exported to the cytoplasm where snRNP assembly is initiated (Hamm and Mattaj, 1990; Izaurralde et al., 1995; Fornerod et al., 1997). Following modifications and Sm core assembly, the UsnRNAs are re‐imported to the nucleus to complete snRNP assembly and to participate in pre‐mRNA splicing (Palacios et al., 1997; Huber et al., 1998; Yu et al., 1998). Studies of the localization and transport of miRNAs are likely to reveal important aspects of miRNA expression and function.
In this study we explore the mode of miRNA expression. By employing various in vivo and in vitro analytical methods, we demonstrate the polycistronic nature of clustered miRNAs and their stepwise processing in different subcellular locations.
Primary precursors for miRNAs are longer than the ∼70 nt pre‐miRNAs
Our initial approach to address this issue was to search for transcripts that are longer than the proposed ∼70 nt precursors. These long transcripts are likely to be the nascent miRNA transcripts that serve as the primary precursors for pre‐miRNAs. Total RNA from HeLa cells was subject to RT–PCR. The primers were chosen to bind outside the boundary of the predicted ∼70 nt stem–loop clusters so that we could detect transcripts covering the whole cluster. Two miRNA clusters were examined in this study: mir‐23∼27∼24‐2 containing three miRNA genes and mir‐17∼18∼19a∼20∼19b‐1 containing five (Lagos‐Quintana et al., 2001). Both the clusters reside in intergenic sequences, making it unlikely that the transcripts are part of or read‐through of other gene transcripts. PCR products of the expected size (403 and 791 bp) were detected (Figure 1A and B, respectively), suggesting that the nascent transcripts are polycistronic and that these clusters may be transcribed as single transcriptional units. To show that the products of the RT–PCR originate from RNA but not from contaminated genomic DNA, reverse transcriptase was omitted during the first‐strand synthesis (lanes 2). As an additional control, total RNA was pre‐treated with RNase A before first‐strand synthesis (lanes 3). The same approach was applied to mir‐30a, which is a single gene and not part of a cluster. A PCR product of the expected size (171 bp) was detected (Figure 1C, lane 1). Different sets of primers were also used for RT–PCR to cover a broader region around miR‐30a, resulting in PCR products of 301 nt, indicating that the long transcript is over 301 nt (data not shown). In agreement with our observation, several expressed sequence tag (EST) clones containing some miRNAs, such as miR‐23∼27∼24‐2, were found in public databases, as noted previously by Tuschl et al. (Lagos‐Quintana et al., 2001). Therefore, miRNAs in general may exist as transcripts that are longer than the ∼70 nt pre‐miRNAs.
We next performed northern blot analysis for miR‐30a using denaturing agarose gels in an attempt to characterize the full‐length transcript (Figure 2). A transcript of ∼600 nt was detected from HeLa cells when using the probe of 301 nt surrounding miR‐30a. This transcript does not seem to be abundant, since northern blotting never gave a strong signal. A long transcript for miR‐23∼27∼24‐2 has not been detected under similar conditions. Of note, we also tried 5′‐ and 3′‐RACE (rapid amplification of cDNA ends) for miR‐30a and miR‐23∼27∼24‐2, which has not been successful so far, supporting the notion that these transcripts are rare at steady‐state level, possibly due to rapid processing.
Maturation of miRNAs occurs through at least two steps
In order to verify that these long transcripts are authentic precursors of mature miRNAs, we developed an in vitro processing system that is similar to the in vitro splicing system (Kim et al., 2001). Cell‐free assay systems have proven invaluable for dissecting the molecular basis of physiological events, as elegantly exemplified by the in vitro splicing system originally reported by Krainer et al. (1984) and the cell‐free system for RNA interference established by Tuschl et al. (1999).
In brief, the PCR products were inserted downstream of the T7 promoter and used for in vitro transcription. Uniformly radiolabeled RNA was then mixed with total cell extract prepared from HEK293T cells. Following incubation at 37°C for 90 min, the resulting RNA from the reaction mixture was extracted and analyzed on a denaturing polyacrylamide gel. The input RNA that contains miR‐30a as well as its surrounding sequences resulted in four major fragments in the presence of cell extract (Figure 3A, lane 2). The sizes of two fragments were ∼65 and ∼23 nt, strikingly reminiscent of the pre‐miRNAs and mature miRNAs, respectively, which have been observed previously by northern blot analyses of endogenous RNAs (Lagos‐Quintana et al., 2001; Lau et al., 2001; Lee and Ambros, 2001). The antisense transcript did not produce any discrete fragments (Figure 3A, lane 5), indicating that the cleavage is not due to non‐specific nuclease activity. To examine whether the ∼65 nt fragment is indeed the precursor of ∼23 nt fragment, as would be expected for miRNAs, the fragments from in vitro processing reaction were gel purified and used for an additional round of in vitro processing (Figure 3B). The ∼65 nt fragment, but not the others, was processed to generate the ∼23 nt fragments in the presence of cell extract (Figure 3B, lane 5), demonstrating that the ∼65 nt fragment is the precursor of the ∼23 nt fragments. Therefore, the input RNA was processed first into ∼65 nt and then into ∼23 nt fragments. The identities of the ∼195 and ∼110 nt fragments are currently under investigation. They may be the side products or the intermediates from the first‐step processing.
In order to confirm the identity of the ∼23 nt fragment, we carried out primer extention analysis (Figure 3C). The fragments of ∼23 nt were detected from the in vitro processed products (Figure 3C, lane 3) as well as from HeLa total RNA (Figure 3C, lane 1), verifying that the ∼23 nt fragment is the authentic mature miR‐30a‐s. When a similar experiment was carried out to detect miR‐30a‐as that is released from the opposite strand of pre‐miR‐30a, only a marginally detectable level of miR‐30a‐as was observed from both HeLa total RNA and in vitro processing reaction (data not shown). Two recent reports showed that miR‐30a‐as (originally named miR‐30) is less abundant than miR‐30a‐s (previously termed miR‐97) (Lagos‐Quintana et al., 2002; Zeng et al., 2002). Thus, there seems to be a certain degree of asymmetry in processing of this miRNA.
Given that miRNA processing occurred correctly without the original 5′‐ and 3′‐end sequences, it appears that the 5′‐ and 3′‐end sequences are not required for processing. Furthermore, a shorter RNA substrate (171 nt) covering miR‐30a was processed in vitro as efficiently as the 371‐nt‐long RNA substrate (data not shown), suggesting that the cis‐acting element(s) for processing is located in close proximity to, and/or inside, pre‐miRNA sequences.
To demonstrate that our observations are not restricted to a certain RNA substrate, we used the same assay conditions for miR‐15∼16 and miR‐23∼27∼24‐2 (Figure 4A and B). The ∼70 nt fragments as well as the 21–23 nt fragments were observed from both the miRNA clusters. Thus, this cell‐free system successfully recapitulates endogenous miRNA maturation and provides a general analysis tool for miRNA processing.
Our data show that miRNA maturation consists of at least two sequential steps: (i) the long transcripts are first processed into the ∼70 nt intermediate (pre‐miRNAs); and (ii) then into the final products of ∼22 nt (mature miRNAs). We suggest that the long transcript be designated ‘pri‐miRNA’ (primary precursor for miRNA).
Three different forms of miRNAs
Pri‐miRNAs could be detected by an additional method, RNase protection assay (RPA) (Figure 5). The probe was designed to cover part of the miRNA gene such that three different forms of miRNA (pri‐, pre‐ and mature miRNAs) can be detected simultaneously as three fragments of different sizes. From HeLa total RNA, three specific bands were detected representing pri‐miR‐23, pre‐miR‐23 and mature miR‐23 (Figure 5A, lanes 1 and 2). We cannot exclude the possibility that pri‐miRNA might not be a single RNA species of the nascent transcript but might also represent some unidentified intermediates of the first‐step processing. RT–PCR and RPA are extremely sensitive methods to detect the region to which the probes anneal. Therefore, if the nascent transcript is subject to rapid processing, the nascent transcript and the processing intermediates would appear as a discrete band in RT–PCR or RPA, whereas in northern blot analysis they would appear as one or more weak bands. We are currently investigating the mechanism and the factors of the first‐step processing.
Subcellular compartmentalization of miRNA processing
The RPA method enabled us to examine the subcellular localization of each miRNA form. From the fractionated nucleus and cytoplasm of HeLa cells, RNA was extracted and used for RPA (Figure 5B). Interestingly, pri‐ and pre‐miR‐23 are concentrated in the nucleus (Figure 5B, lane 2), while mature miR‐23 is mainly in the cytoplasm (Figure 5B, lane 3). Thus, the first‐step processing is likely to occur in the nucleus. The arena of the second‐step processing is less clear. It may occur in the nucleus and then the resulting mature miRNAs may be rapidly exported to the cytoplasm. Alternatively, the pre‐miRNAs are exported and subject to the second‐step processing in the cytoplasm.
We tackled this issue by taking advantage of our in vitro processing system. Following subcellular fractionation, the nucleoplasmic and cytoplasmic fractions were used for in vitro processing, separately or in combination (Figure 6A). In the nucleoplasmic fraction, only the first‐step reaction took place efficiently (Figure 6A, lanes 1–4), whereas in the cytoplasmic fraction neither of the reactions occurred effectively (Figure 6A, lane 8), showing that the processing machinery for the first step is restricted to the nuclear compartment. Only when the cytoplasmic fraction was added to the nucleoplasmic fraction was mature miRNA produced (Figure 6A, lane 7), which indicates that the second step is confined to the cytoplasm. With the cytoplasmic fraction only, the second step failed to occur because the cytoplasm lacks the enzymatic activity for the preceding reaction (Figure 6A, lane 8). Western blotting was performed to show the efficiency of the nucleocytoplasmic fractionation (Figure 6B). hnRNP C1/C2 and eIF4E are well known for their localization in the nucleus and the cytoplasm, respectively, at steady‐state levels.
Our data, together with the recent identification of Dicer as an miRNA processing factor, suggest a model for how miRNA is transcribed, processed and exported (Figure 7). Clustered miRNA genes constitute single, and possibly autonomous, transcriptional units, although this may not be an absolute rule for all the clustered miRNA genes. The polycistronic nascent transcripts are processed into ∼70 nt pre‐miRNAs while in the nucleus, by a yet‐to‐be‐identified factor(s). Single miRNA genes that are not part of a cluster are also expressed as nascent transcripts that are longer than pre‐miRNAs and need to be processed into pre‐miRNA forms in the nucleus. It is the pre‐miRNA form that is recognized and exported by an export factor. The second‐step processing occurs in the cytoplasm where pre‐miRNAs are met by Dicer, and possibly by other factors, and processed into the final products of ∼22 nt.
Grosshans and Slack (2002) have speculated that miRNA biogenesis may occur in a highly coordinated ‘channeling’ pathway, where miRNAs are handed over from one biogenesis factor to the next. Although it is clear from our data that the two main processing steps are separable (Figures 3B and 6A), we do not exclude the possibility that functional coupling of the steps along the miRNA biogenesis pathway is important in vivo, as shown previously in the case of mRNAs (Kim and Dreyfuss, 2001) and tRNAs (Wolin and Matera, 1999). Indeed, it may be noteworthy that the final product was produced more efficiently when the first‐ and the second‐step reaction occurred simultaneously rather than when the second‐step reaction occurred separately (compare Figure 3A, lane 2 with B, lane 5), implicating a functional coupling between the first and the second steps of miRNA processing. In light of this, it would be interesting to identify the processing factor for the first step, which is likely to be one of the key players in miRNA biogenesis.
It is intriguing that the two sequential steps of miRNA maturation are compartmentalized into the nucleus and the cytoplasm. This may leave additional room for fine regulation of miRNA expression. For instance, nuclear export of miRNA may be controlled in response to a certain signal. Based on our data, the ∼70 nt pre‐miRNA is likely to be the substrate for the nuclear export factor. Export of small RNAs, such as tRNAs and UsnRNAs, is mediated by Exportin‐t and Exportin‐1/CRM1, respectively, members of the nuclear transport receptor (NTR) family (Nakielny and Dreyfuss, 1999; Conti and Izaurralde, 2001; Kuersten et al., 2001). It remains to be seen whether the unknown export factor for miRNAs also shares the common features of the NTR family, such as the requirements for the cofactor Ran.
Evidence suggests that miRNA expression may be regulated at the level of processing. Some miRNAs in D.melanogaster appear to be processed inefficiently at early stages of embryonic development (Hutvägner et al., 2001). In the sea urchin Strongylocentrotus purpuratus, let‐7 transcripts of ∼100 nt are expressed throughout embryonic development, although mature let‐7 appears only at a later stage (Pasquinelli et al., 2000). One interesting possibility is that the pri‐miRNAs or pre‐miRNAs remain in an inert state until a developmental signal triggers their processing. It is also tempting to speculate that some pre‐miRNAs at a certain stage may be retained in the nucleus waiting for a signal for nuclear export.
Since little information is currently available on miRNA transcription and processing, our results lay a framework for further studies. Important questions remain as to what are the processing factors and mechanisms. It would also be of great interest to know whether processing is indeed regulated and, if so, what is the mechanism underlying the regulation. The human in vitro processing system developed in this study may provide a powerful tool for analysis of miRNA maturation. Understanding the mechanism of miRNA biogenesis may enable the development of new methods for functional genomics and gene therapy.
Materials and methods
Total RNA was prepared from subconfluent HeLa cells with TRIzol reagent (Gibco‐BRL). RNA (5 μg) was used for the first‐strand cDNA synthesis with SUPERSCRIPT II (Gibco‐BRL). As controls, RNA was pre‐treated with 50 μg/ml RNase A at 37°C for 2 min before the first‐strand synthesis, or the reverse transcriptase was omitted during the first‐strand synthesis. Primers used for first‐strand synthesis were as follows. For miR‐30a, 5′‐TTCAGCTTTGTAAAAATGTATCAAAGAGAT‐3′ was used for first‐strand synthesis, and 5′‐ATTGCTGTTTGAATGAGGCTTCAGTACTTT‐3′ (forward) and 5′‐TTCAGCTTTGTAAAAATGTATCAAAGAGAT‐3′ (reverse) were used for PCR amplification. For miR‐23∼27∼24‐2, 5′‐GGCAGGGGCTGCAGGCTCCAAGGGGGCTTG‐3′ was used for first‐strand synthesis, and 5′‐CGCCCGGTGCCCCCCTCACCCCTGTGCCAC‐3′ (forward) and 5′‐CCCTGTTCC TGCTGAACTGAGCCAGTGTAC‐3′ (reverse) were used for PCR amplification. For miR‐17∼18∼19a∼20∼19b‐1, 5′‐GGGGTTTGAGTTTCCCTTACTTTTCTACAG‐3′ was used for first‐strand synthesis, and 5′‐TGCTGAATTTGTATGGTTTATAGTTGTTAG‐3′ (forward) and 5′‐CACTACCACAGTCAGTTTTGCATGGATTTG‐3′ (reverse) were used for PCR amplification.
In vitro transcription
In vitro transcription was carried out as described previously (Pellizzoni et al., 1998), with the following modifications. To prepare templates for in vitro transcription, the PCR products from Figure 1 were subcloned into pGEM‐T‐easy (Promega) and clones with opposite directions were selected so that sense as well as antisense transcripts can be generated. The plasmids were designated as pGEM‐T‐30(+), pGEM‐T‐30(−), pGEM‐T‐ 15‐16(+), pGEM‐T‐15‐16(−), pGEM‐T‐23(+) and pGEM‐T‐23(−). The plasmids were linearized with SpeI. In the transcription reaction, 1 mM GTP was added instead of 0.1 mM and the cap analog was omitted.
Northern blot analysis
Total RNA (20 μg) from HeLa, U937 and HL60 cells was loaded on a 1% formaldehyde–agarose gel. The resolved RNA was transferred to a Zeta‐Probe GT blotting membrane (Bio‐Rad) overnight. Probe was prepared by the random priming method using a fragment containing a 301 bp region of pri‐miR‐30a from pGEM‐T‐30(+) as a template. Pre‐hybridization and hybridization were carried out using ExpressHyb Hybridization Solution (Clontech) following the manufacturer's instruction. The most stringent wash was carried out in 0.1× SSC and 0.1% SDS at 50°C.
In vitro processing system
Preparation of cell extract and in vitro processing were performed with some modifications to our previously reported in vitro splicing method (Kim et al., 2001). Whole‐cell extract was prepared from HEK293T cells in buffer D (20 mM HEPES–KOH pH 7.9, 100 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, 0.2 mM PMSF, 5% glycerol) by sonication, followed by centrifugation. Ten microliters of processing reaction contained 5 μl of whole‐cell extract, 1 μl of solution A (32 mM MgCl2, 5 mM ATP, 200 mM creatine phosphate), 1 U/μl RNase inhibitor (Takara) and the labeled transcripts of 1 × 104 − 1 × 105 c.p.m. The reaction mixture was incubated at 37°C for 90 min.
Primer extension analysis
The cold miR‐30a precursor was synthesized in vitro in the presence of 1 mM each NTP and incubated with HEK293T cell extract for 90 min. Following phenol extraction and ethanol precipitation, the RNA was used for annealing to 5′‐end‐labeled primer at 37°C for 40 min and subsequently for extension using SUPERSCRIPT II (Gibco‐BRL) at 50°C for 30 min. HeLa total RNA (10 μg) was used as a control. Primer sequences were 5′‐GCTTCCAGTCGAGGATGT‐3′.
Nucleocytoplasmic fractionation for RNA preparation
Subcellular fractionation of HeLa cells was carried out in RSB‐100 buffer (10 mM Tris–HCl pH 7.5, 100 mM NaCl, 2.5 mM MgCl2, 35 μg/ml digitonin), as described previously (Siomi et al., 1997). RNA was extracted from each fraction using TRIzol.
Nucleocytoplasmic fractionation for in vitro processing
For in vitro processing with the fractionated extracts, the nuclear and the cytoplasmic fractions prepared in RSB‐100 were dialyzed against buffer D before they were used for in vitro processing.
RNA (50–120 μg) was used for each assay using RNase T1 at a 1:50 dilution from the RPA III kit (Ambion). An equal amount of yeast RNA was used as a control. The probes were designed to hybridize to miRNA as well as the surrounding sequences so that the precursors will give rise to distinctly sized fragments. The probes were synthesized using T7 polymerase (Promega) directly from the PCR products that contain the T7 promoter. Primers for miR‐23 were 5′‐TAATACGACTCACTATA GGTAAGTTGGCAGCATCCTCGGTGGCAG‐3′ (forward) and 5′‐AAT TGTATACGTATTATGAAATCACATTGCCAGGGATTTC‐3′ (reverse).
Western blot analysis
Anti‐hnRNP C monoclonal antibody 4F4, a kind gift from Dr Gideon Dreyfuss, and anti‐eIE4E (Sigma) were used for western blotting.
We are grateful to members of our laboratory and to Drs Kyriacos Mitrophanous, Jin Mo Park, Sara Nakielny and Zissimos Mourelatos for critical reading of this manuscript and helpful discussion, and to Hyunjung Yang for secretarial assistance. We also thank Dr Gideon Dreyfuss for kindly providing 4F4 antibody. This work was supported by the Ministry of Science and Technology (MI‐9808‐00‐0040) and the BK21 Research Fellowship from the Ministry of Education and Human Resources Development of Korea.
- Copyright © 2002 European Molecular Biology Organization