RNA-protein interactions within the 3' untranslated region of vimentin mRNA
RNA-protein interactions within the 3 ' untranslated region of vimentin mRNAZendra E. Zehner*, Rebecca K. Shepherd, Joanna Gabryszuk1, Tzu-Fun Fu, May Al-Ali and W. Michael Holmes1
Department of Biochemistry and Molecular Biophysics, Box 980614 and 1Department of Microbiology and Immunology and The Massey Cancer Center, Medical College of Virginia, Virginia Commonwealth University, Richmond, VA 23298-0614, USA
Received January 24, 1997; Revised and Accepted June 20, 1997
ABSTRACT
Several functions have been attributed to protein binding within the 3' untranslated region (3'UTR) of mRNA, including mRNA localization, stability, and translational repression. Vimentin is an intermediate filament protein whose 3' untranslated sequence is highly conserved between species. In order to identify sequences that might play a role in vimentin mRNA function, we synthesized 32P-labeled RNA from different regions of vimentin's 3'UTR and assayed for protein binding with HeLa extracts using band shift assays. Sequences required for binding are contained within a region 61-114 nucleotides downstream of the stop codon, a region which is highly conserved from Xenopus to man. As judged by competition assays, binding is specific. Solution probing studies of 32P-labeled RNA with various nucleases and lead support a complex stem and loop structure for this region. Finally, UV cross-linking of the RNA-protein complex identifies an RNA binding protein of 46 kDa. Fractionation of a HeLa extract on a sizing column suggests that in addition to the 46 kDa protein, larger complexes containing additional protein(s) can be identified. Vimentin mRNA has been shown to be localized to the perinuclear region of the cytoplasm, possibly at sites of intermediate filament assembly. To date, all sequences required for localization of various mRNAs have been confined to the 3'UTR. Therefore, we hypothesize that this region and associated protein(s) might be important for vimentin mRNA function such as in localization.
INTRODUCTION
The importance of the mRNA 3' untranslated region (3'UTR) as a repository for signals determining mRNA processing, polyadenylation, export, stability, localization, translation and cell cycle regulation has become apparent (1 -5 ). A summary of this work indicates that critical RNA-protein interactions are involved and a multitude of RNA elements and binding proteins are being identified as additional contributors to gene regulation. A thorough analysis of these complex elements and proteins will be required in order to fully understand cellular regulation.
Although poly(A) RNA is homogeneously distributed throughout the cytoplasm of cells, various cytoskeletal and other mRNAs have been shown to be localized to specific cellular compartments (6 -11 ). For example, [beta]-actin mRNA is localized to the lamellipodia of chick fibroblasts, whereas [alpha]-actin and vimentin mRNA are found in a perinuclear configuration (9 ). In addition, vimentin mRNA is localized to the costameres in myotubes allowed to differentiate in tissue culture (12 -14 ). Because vimentin transcription is down-regulated during myogenesis in vivo, it is difficult to assess the physiological importance of this observation, but it does illustrate the possibility that the same mRNA may localize differently in the cytoplasm of different cell types.
To date, all signals required for mRNA localization, whether in Drosophila or higher eukaryotes, have been localized to the 3'UTR (6 ,15 ). In some cases, these signals can be multiple and quite complex (16 -18 ). When 3'UTR sequences required for [beta]-actin mRNA localization were attached to the [beta]-galactosidase ([beta]-gal) gene, all [beta]-gal synthesis was directed to the lamellipodia (19 ,20 ). Furthermore, protein synthesis was not required for mRNA localization (21 ). Disruption of microtubules had no effect on localization, but microfilaments were required for the correct sorting of actin mRNA (22 ). Singer coined the term `zip-code' to identify those sequences which direct mRNAs to particular cellular locations (11 ,23 ).
In the case of intermediate filament proteins (IFPs) like vimentin, it is reasonable that mRNAs are localized. IFPs are coiled-coil polymers postulated to contain as many as 32 chains assembled into the 10 nm filament (24 -28 ). Pulse-chase experiments have indicated that the pool of soluble vimentin in the cell is actually the coiled-coil tetramer, an important intermediate in filament assembly (29 ). In embryonic muscle cells and fibroblasts, over one-half of the newly-synthesized vimentin is found immediately associated with the cytoskeleton (14 ). These results suggest the possibility that translation and filament polymerization may be linked. In this regard, co-translational assembly may ensure the close proximity of nascent chains for ease of filament polymerization and suggest the importance of vimentin mRNA localization as a prerequisite for optimal filament formation.
Because vimentin and [alpha]-actin mRNAs are directed to a different cellular compartment than [beta]-actin, different localization signals and components should be required. Ultimately, we want to search for cellular factors that direct mRNA to the periphery of the nucleus rather than the lamellipodia. Because of the documented importance of the 3'UTR in such cellular functions, we began by examining protein binding within this region of vimentin mRNA. A comparison of vimentin mRNA sequences from Xenopus to man exhibits striking sequence homology even within the non-coding 3'UTR (30 ). To determine protein binding regions, we synthesized 32P-labeled RNA from the entire 3'UTR, as well as from distinct sub-domains. By band shift assays we have defined a protein binding domain and shown that binding is specific. RNA solution probing studies suggest that this region has a well-defined secondary structure. Due to the sequence conservation of this domain across vertebrates, we hypothesize that it may be important for vimentin mRNA function.
MATERIALS AND METHODS
Subcloning of the 3'UTR of human vimentin mRNA
The entire human vimentin 3'UTR and various sub-regions thereof wereobtained by PCR using a human vimentin cDNA as the template (a kind gift of D.Bloch and P.Leder, Harvard University, Cambridge, MA). Each of the 5'-primers contained the 2 nucleotides (nt), gc, (to make the EcoRI restriction site internal and guarantee maximum enzymatic digestion), an EcoRI restriction site (italics), and the T7 promoter sequence (bold) followed by specific primers (capital letters) to the designated regions of the vimentin cDNA. The sequence position designations refer to the number of bases after the stop codon. For example, P#1 (5'-gcgaattctaatacgactcactatagggCACTCAGTGCAGCAATA-3') begins at base 11 and ends at base 27 downstream of the stop codon. Other 5'-primers contain the same first 28 nt plus vimentin sequence as follows: P#2 (5'-GTTCTTAACAACCGACA-3') bases 139-156; P#6 (5'-CAAGAATAAAAAAGAAATCC-3') bases 37-56; and P#7 (5'-CTTAAAGAAACAGCTTTCA-3') bases 61-79. The 3'-primers contain 2 nt (to ensure complete digestion), plus a BamHI restriction site (italics), followed by the desired 3' vimentin sequence (in capital letters) as follows: P#3'L (5'-atggatccGTTTTTCCAAAGATTTATT-3') bases 302-320; P#3'S (5'-atggatccAAAGTATTCTAGCACAAGA-3') bases 208-226; P#8 (5'-cgggatccGTTAAGAACTAGAGCT-3') bases 132-147; and P#10 (5'-ctggatccTATCTTGCGCTCCTG-3') bases 100-114. The indicated primers were used to construct the following templates: P#1 and P#3'L for transcript 11/320, P#1 and P#3'S for transcript 11/226, P#6 and P#8 for transcript 37/147, P#7 and P#8 for transcript 61/147, P#6 and P#10 for transcript 37/114, and P#2 and P#3'S for transcript 139/226.
PCR reactions used 50 ng of EcoRI-linearized template and 1 [mu]g of each primer for 30 cycles. Following PCR, fragments were digested with the restriction enzymes EcoRI and BamHI and cloned into similarly digested pUC18. The content of each subclone was verified by DNA sequencing.
Template 81/108 was synthesized by annealing the oligonucleotides P#14 (5'-attctaatacgactcactatagggCACGGAAAGACGTCAAAAAGTCCTCGCG-3') bases 81-108, and P#15 (5'-gatcCGCGAGGACTTTTTGACGTCTTTCCGTGccctatgtgagtcgtattag-3') bases 108-81, followed by direct cloning into EcoRI and BamHI digested pUC18.
Preparation of 32P-labeled RNA transcripts
Plasmids to be transcribed into RNA were first linearized by BamHI digestion. Transcripts 11/137 and 11/73 were made by first digesting the starting template (clone 11/226) with MaeI or AluI, respectively, thereby generating run-off transcripts which terminated at the specific internal restriction site. Preparative amounts of T7 polymerase transcripts were prepared from 25-50 [mu]g of template in the presence of transcription buffer (350 mM HEPES pH 7.5, 30 mM MgCl2, 2 mM spermidine and 40 mM dithiothreitol), 40 U of RNasin, 7 mM each of the four NTPs, and 60 U of T7 polymerase (1.5 U/[mu]g) purified as described (31 ). Following incubation for 2 h at 40oC, the RNA was purified by HPLC on a TSK2000 molecular sieve column. When required for solution probing studies, RNA was end-labeled with [[gamma]-32P]ATP using T4 polynucleotide kinase following 5'-phosphate removal with either bacterial (32 ,33 ) or shrimp alkaline phosphatase per manufacturer's directions, and gel purified as described below.
Analytical amounts of RNA were synthesized from 5-10 [mu]g of linearized template and internally labeled by incorporation of either [[alpha]-32P]ATP, [[alpha]-32P]CTP or [[alpha]-32P]UTP in the presence of transcription buffer (40 mM Tris-HCl pH 8.0, 20 mM MgCl2, 1 mM spermidine, 0.1% Triton X-100 and 5 mM dithiothreitol), 60 U of RNasin, 0.25 [mu]M of the labeled nucleotide plus 250 [mu]M of the remaining nucleotides and 24 U of T7 polymerase (1.5 U/[mu]g) for 2-4 h at 40oC. In later RNA synthesis reactions, the transcription buffer was changed to the HEPES-containing buffer noted above for preparative RNA synthesis. For purification, 32P-labeled RNA was heated at 60oC for 4 min prior to loading on either a 6 or 8% polyacrylamide gel (electrophoresis at 1200 V, 40 mA for ~4 h). The band was excised and the RNA removed either by electrophoresis or soaking overnight in Maxam-Gilbert buffer (0.5 M ammonium acetate, 10 mM magnesium acetate, 0.1 M EDTA, and 0.1% SDS). Following ethanol precipitation, the RNA was washed twice with 70% ethanol prior to use.
Preparation of HeLa whole cell and nuclear extracts
HeLa whole cell extract (a postribosomal supernatant) was prepared from HeLa cells as described (34 ,35 ). Protein was concentrated by ammonium sulfate precipitation and dialyzed against 50 mM Tris-HCl pH 7.9, 100 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol, 1% Nonidet P-40 and 20% glycerol (v/v), to remove the ammonium sulfate, and stored in aliquots at -70oC. Nuclear extract was prepared as described (35 ). Protein concentration was determined by the Bradford assay and was 20 [mu]g/[mu]l for the whole cell extract and 35 [mu]g/[mu]l for the nuclear extract.
RNA band shift assays
32P-labeled RNA (4-6 fmol, 50-100 000 d.p.m.) was incubated with 25-40 mM Tris-HCl pH 7.5, 5-20 mM MgCl2, 150 mM KCl and 4-168 [mu]g HeLa whole cell extract or 5 [mu]g nuclear extract in a final volume of 20 [mu]l for 15 min on ice, after which 50 [mu]g of heparin was added for an additional 5 min. Following the addition of 5 [mu]l of 50% (v/v) glycerol (and tracking dye to only the free RNA samples), complexes were resolved on a 5% polyacrylamide gel containing 5% glycerol (ratio of 40:1 acrylamide:bis) in 0.5* Tris-borate buffer (45 mM Tris-HCl pH 8.3, 45 mM borate, 2.5 mM EDTA) at 15-25 mA for 2-4 h at 4oC. Gels were dried and exposed to film with an intensifying screen at -70oC overnight.
Competition band shift assays were done with 4 fmol of 32P-labeled RNA (base 11/226) and 4 [mu]g of whole cell extract (40 000 d.p.m.). This amount of extract was chosen because it yielded 60% of maximal probe binding activity. Either specific, unlabeled RNA (11/226) or non-specific, brome mosaic virus RNA (BMV) (36 ), were added at 25-200-fold excess at the beginning of incubation, prior to extract addition. The counts contained in the shifted or free RNA bands were quantified with a Fuji BioImage Analyzer BAS2000 and the percent shifted calculated as the amount of shifted material divided by the sum of free and shifted counts per lane.
Band shift assays were also performed in the presence of specific antibodies to human vimentin (a kind gift from P.Traub, Max Planck Institute, Heidelberg, Germany), desmin (D8281, a rabbit antibody to the chicken gizzard protein purchased from Sigma), glial fibrillary acidic protein (GFAP: G-A-5, a monoclonal antibody purchased from Sigma), actin (two different antibodies; C4, a monoclonal antibody to vertebrate actin from Boehringer Mannheim Biochemicals, and A-2066, a rabbit anti-actin affinity-isolated antibody from Sigma), the histone 3'-end stem-loop binding protein, SLBP (a kind gift from W.Marzluff, University of North Carolina, Chapel Hill, NC), human nucleolar protein, B23 (a kind gift from C.Hutchinson, University of Dundee, Dundee, Scotland), human La protein (a kind gift from J.Keene, Duke University Medical Center, Durham, NC) and hnRNPC (a kind gift from G.Dreyfuss, University of Pennsylvania School of Medicine, Philadelphia, PA). All antibodies purchased from manufacturers or received from the cited individuals were used as recommended. Two sets of assays were conducted. The antibody was added to the protein extract prior to the addition of RNA or it was added after RNA-protein complex formation. Following incubation, complexes were resolved by PAGE as described above. In both cases, no inhibition of band shifting activity was detected.
UV cross-linking of protein and RNA
Incubation of 15 fmol (250 000 d.p.m.) of 32P-labeled RNA (base 11/205 or 226) with 60 [mu]g HeLa whole cell extract was as described above for band shift assays. Following incubation, the samples were exposed to UV light (254 nm, 100 W) at a distance of 5 cm for 10 min on ice. RNA was removed by digestion with 1 U of RNase T1 at 37oC for 2 h. The resulting 32P-labeled proteins were resolved by 10% SDS-PAGE in the presence of molecular weight markers. The gel was dried and exposed to XAR film with an intensifying screen at -70oC for 3 days.
Northwestern blot
HeLa whole cell or nuclear extract (5.5 [mu]g) was separated by 9% SDS-PAGE. Following electrophoresis, proteins were allowed to re-fold by soaking for 30 min in TNED buffer (10 mM Tris-HCl pH 7.5, 50 mM NaCl, 0.1 mM EDTA and 1 mM dithiothreitol) and then transferred to nitrocellulose by electrophoresis at 75 V for 1 h. The nitrocellulose was pre-hybridized overnight in 5% (w/v) non-fat dry milk, 50 mM Tris-HCl pH 7.5, 50 mM NaCl, 0.1 mM EDTA and 1 mM dithiothreitol. Following a wash with TNED buffer, hybridization with 32P-end-labeled RNA (37/147, 1.33 * 106 d.p.m.) was in TNED buffer for 2.5 h at room temperature with shaking. Following hybridization, the blot was washed three times, 10 min each, with TNED buffer. After drying, autoradiography was as described above for 10 days.
Determination of RNA secondary structure in solution
The T7 transcript (37/147) was prepared and 32P-labeled at the 5'-end as described above. The single-strand nuclease (ChS) was purified from wheat chloroplasts as described (37 ) and had an activity of 0.17 U/[mu]l. The double-strand RNase, V1 (0.7 U/[mu]l), and single-strand RNases T1 (10 U/[mu]l) and U2 (10 U/[mu]l) were obtained from Pharmacia. An RNA sequence ladder was generated by limited alkaline hydrolysis of 32P-labeled RNA (1 * 105 d.p.m.) in bicarbonate buffer pH 9.2, at 90oC for 6 min in the presence of bulk yeast tRNA (2 [mu]g). To generate an A- or G-specific ladder, a sample of 32P-labeled RNA was incubated with 0.12 U of RNase U2 or T1, respectively, for 12 min at 55oC followed by the addition of 2 [mu]g of tRNA. 32P-labeled RNA was incubated separately with 0.2 and 0.4 U of ChS nuclease or 0.01, 0.011, 0.014 and 0.017 U of RNase V1 for 5 min at room temperature in TMK buffer (20 mM Tris-HCl pH 7.5, 10 mM MgCl2, 100 mM KCl), or 0.1 and 0.3 U of RNase T1 in TM buffer (50 mM Tris-HCl pH 7.2, 10 mM MgCl2). Reaction mixtures were phenol-chloroform extracted, adjusted to 0.3 M sodium acetate pH 6.0, and ethanol precipitated with 3 vol ethanol. Pellets were centrifuged, air-dried and re-suspended in loading buffer (20 mM sodium citrate pH 5.0, 7 M urea, 1 mM EDTA, 0.025% xylene cyanol and 0.025% bromophenol blue). Samples were separated by 8% PAGE (acrylamide/bisacrylamide ratio of 19:1) containing 8 M urea and 0.5* TBE. Gels were autoradiographed after drying.
For lead cleavage experiments, 32P-end-labeled RNA was supplemented with 4 [mu]g of yeast total tRNA and incubated with 4 and 12 mM lead acetate for 5 min at 20oC in 20 mM HEPES pH 7.5, 5 mM magnesium acetate, 50 mM potassium acetate. Reactions were stopped by the addition of EDTA to a final concentration of 33 mM. Samples were phenol-chloroform extracted, precipitated, re-suspended in loading buffer, and loaded on a gel as described above.
Protein fractionation on a sizing column
HeLa whole cell extract (32 mg) was fractionated on a Superdex 200 column (16 mm * 70 cm) using HPLC at a flow rate of 1 ml/min, and 70 fractions of 2 ml each were collected.
RESULTS
A comparison of vimentin's 3'UTR mRNAs across species
A number of vimentin genes have now been isolated from a wide range of species. In order to find regions of high homology which might correlate with mRNA function, we compared the 3'UTR sequence of vimentin from human to Xenopus, beginning 3' to the stop codon (set at position 1 relative to the human sequence) and extending to the first functional poly(A) site (Fig. 1 ). As noted previously, considerable homology exists between vimentin genes both in coding and non-coding regions (38 ,39 ). In Figure 1 , sequences were aligned in order to maximize homology and nucleotides exhibiting at least a 5 out of 7 match across species have been colored. There are many homologous regions. One exceptionally large region (positions 68-115) extends for 47 contiguous bases and displays 68% homology from Xenopus to human. This degree of homology is as great as that exhibited within certain coding regions of the vimentin gene. For example, when comparing the chicken sequence of exon 1 (the most divergent domain) to that of hamster, the homology is only 60% (38 ). However, within exon 6, the most conserved region, the homology increases to 90%. Thus, portions of vimentin's 3'UTR are as conserved as protein coding regions, which suggests a functional role for this region. In order to ultimately investigate this role, we pursued a further analysis of this region.
Localization of a protein binding region within vimentin's 3'UTR
To localize protein binding sites, we synthesized 32P-labeled RNA representing the entire 3'UTR of vimentin, extending from position 11 to 320 nt downstream of the stop codon (11/320), and the same region minus the poly(A) site (11/226). The ability to bind protein was assessed via a band shift assay in the presence of the non-specific competitors, tRNA and heparin. The complete 3'UTR shows considerable protein binding to increasing concentrations of a HeLa whole cell extract (Fig. 3 A, lanes 2-4). Although binding of the entire 3'UTR exhibited a broad band of shifted material, removal of the poly(A) site (11/226) resulted in a sharper band, which may be due to the elimination of the poly(A) binding site in the RNA (Fig. 3 A, lanes 6-8). Binding occurred over a wide range of KCl concentrations (100-600 mM) with no difference in band shift pattern (data not shown). Also, no difference was detected in the absence of tRNA; therefore, it was removed from subsequent band shift assays.
Confirmation of secondary structure within vimentin's 3'UTR
We next set out to determine if regions of homology and protein binding display actual secondary structure in solution. As illustrated in Figure 2 , sequences which bind protein (residues 61-114) can be folded into prominent stem-loop structures. In order to determine if such structures actually exist in solution, we carried out chemical and enzymatic probing analyses. 32P-end-labeled RNA (37/147) was prepared and cleaved with either ChS, T1 or V1 nuclease in order to determine the presence of single- or double-stranded regions. Several such experiments were conducted and a representative gel is shown in Figure 5 A. Clear and distinct regions that can be cleaved with V1 nuclease (double-stranded cuts) are evident in lanes 1-4. Similarly, T1 and Chs, single-strand-specific cleavage sites, are evident in lanes 5/6 and 7/8, respectively. In addition, 32P-end-labeled RNA was cleaved with molecular lead, which reveals regions that are single-stranded (Fig. 5 B). Three regions are distinctly hypersensitive to lead cleavage, (position 62-66, 72-81 and 93-100) and correspond roughly to those delineated with ChS and T1 cleavage.
Characterization of protein(s) which bind vimentin's 3'UTR
The nature of protein(s) binding to vimentin's 3'UTR was investigated by a number of methods. First, the size and number of protein(s) binding were addressed by UV cross-linking of RNA-protein complexes. 32P-labeled RNA (11/205 or 11/226) was cross-linked to protein by exposure to UV light and non-cross-linked RNA removed by digestion with RNase T1. The resulting 32P-labeled protein was separated by SDS-PAGE in the presence of molecular weight standards (Fig. 6 A). For both RNAs, a single band migrating between the molecular weight markers of 34.9 and 53.2 kDa is seen.
Size fractionation of HeLa whole cell extract
It is possible that the 46 kDa RNA binding activity detected in these assays might be part of a larger complex of molecules or contain binding sites for additional polypeptide(s). In order to test this hypothesis, a HeLa whole cell extract was fractionated on a Superdex 200 molecular sieve column and band shift assays with 32P-labeled RNA (37/147) were carried out on the resulting fractions (Fig. 7 ). Interestingly, activity is detected in two major (lanes 4-6 and lanes 11-14) and a third minor region (lanes 7-10). Relevant fractions were combined as pools I, III and II, respectively. By UV cross-linking the 46 kDa species was found in pool III, which coincides with the position of elution for a 45 kDa molecular weight standard (data not shown). Surprisingly, the largest complex (pool I) which elutes at the position equivalent to 150 kDa, revealed a 66 kDa protein by northwestern blot (data not shown). No binding of the 46 kDa protein could be detected in pool I. In addition, we were unable to detect protein binding by either technique in pool II, eluting at 90 kDa. Perhaps this more minor species is unstable with purification or is refractory to these methods because it requires additional components for binding.
DISCUSSION
An analysis of the 3'UTR from vimentin mRNA across several species (human to Xenopus) indicates considerable homology (58% for yellow boxes in Fig. 1 ), which eclipses that observed for some coding regions (exon 1). By using band shift assays with various 32P-labeled RNAs from the 3'UTR, we have deduced a minimal protein binding domain within the region 61-114 downstream of the stop codon. This minimal binding domain encompasses the 47 base region which exhibits the most extensive homology (68%) from human to Xenopus. We have shown that this region in the absence of protein displays a secondary structure in the shape of a `Y' as deduced from cleavage studies with different nucleases and lead. Interestingly, this Y-shaped region has considerable secondary and possibly tertiary structure, as it contains three stems, two hairpin loops plus a central loop and at least one bulge region. Further application of the Zucker program reveals that similar structures can be drawn for the corresponding region of chicken vimentin mRNA which exhibits band shift activity with HeLa extracts (data not shown). A thorough analysis of what base sequence and structures are required for optimal binding across several species is currently underway. Overall, this region has several general features in common with other known RNA binding domains (5 ,43 -46 ). Although we have narrowed down the binding site considerably (from 300 to 47 bases), we cannot a priori deduce the structure required for protein binding. Additional solution probing studies in the presence of protein(s) will be required to complete this analysis.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the advice and encouragement of Drs Chantal and Bernard Ehresmann, Eric Westhof, Alain Krol and Richard Geige at the IBMC and Dr Irwin Davidson at IGBMC, CNRS, Strasbourg, France where this work was initiated by Z.E.Z. and W.M.H. while on sabbatical leave. Also, we wish to acknowledge the computer assistance of Mr Ian Joyce. This work was supported by United States Public Health Service Grant HL-45422 and American Cancer Society Scholar Research Award (to Z.E.Z.), a United States Public Health Service Postdoctoral Training Grant, 5 T32 H107110 (to R.K.S.) and a Poste Rouge from the CNRS, France (to W.M.H.).
33 Keith,G. (1990) In Gehrke,C.W. and Kuo,K.C. (eds) Chromatography and Modifications of Nucleosides, Chromatography Library Series. Elsevier, Amsterdam, pp. A103-A141.
43 Karn,J., Gait,M.J., Churcher,M.J., Mann,D.A., Mikaelian,I. and Pritchard,C. (1994) In Nagai,K. and Mattaj,I.W. (eds) RNA-Protein Interactions. Oxford University Press, Inc., New York, pp. 192-220.
44 Pieler,T. (1994) In Nagai,K. and Mattaj,I.W. (eds) RNA-Protein Interactions. Oxford University Press, Inc., New York, pp. 178-191.
45 Draper,D.E. (1994) In Nagai,K. and Mattaj,I.W. (eds) RNA-Protein Interactions. Oxford University Press, Inc., New York, pp. 82-102.
