Detection and isolation of RNA-binding proteins by RNA ... - CiteSeerX

3816–3822 Nucleic Acids Research, 1997, Vol. 25, No. 19

 1997 Oxford University Press

Detection and isolation of RNA-binding proteins by RNA-ligand screening of a cDNA expression library Rudolf Sägesser1,*, Emilio Martinez1, Mina Tsagris1,2 and Martin Tabler1 1Institute

of Molecular Biology and Biotechnology, Foundation for Research and Technology–Hellas and of Crete, PO Box 1527, GR-71110 Heraklion, Crete, Greece

2University

Received July 7, 1997; Revised and Accepted August 15, 1997

ABSTRACT A screening assay for the detection of RNA-binding proteins was developed. It allows the rapid isolation of cDNA clones coding for proteins with sequence-specific binding affinity to a target RNA. For developing the screening protocol, constituents of the human U1 snRNP were utilized as model system. The RNA partner consisted of the U1-RNA stem–loop II and the corresponding protein consisted of the 102 amino acid N-terminal recognition motif of the U1A protein, which was fused to β-galactosidase and expressed by the recombinant lambda phage LU1A. Following binding of the fusion protein to nitrocellulose membranes, hybridization with a 32P-labeled U1-RNA ligand was carried out to detect specific RNA–protein interaction. Parameters influencing the specificity and the detection limit of binding were systematically investigated with the aid of the model system. Processing the nitrocellulose membranes in the presence of transition metals greatly increased the signal:background ratio. A simple screening protocol involving a single-buffer system was developed. Specific RNA–protein interaction could be detected in the presence of a large excess of recombinant phages from a cDNA library. Only moderate binding affinities (Kd = 10–8 M) were required. The suitability of the RNA-ligand screening protocol was demonstrated by the identification of new viroid-RNA binding proteins from tomato. INTRODUCTION RNA is unique amongst all naturally occurring macromolecules, since it is a carrier of genetic information and may also have catalytic properties. However, RNA molecules normally associate with specific proteins in order to exert their cellular function as a ribonucleoprotein. Therefore, RNA–protein interactions play a key role in fundamental processes of all living organisms, such as transcription, nuclear transport, translation and infection by RNA viruses. For example, RNA maturation and translation rely on RNA–protein interactions that direct the assembly of the catalytic machinery in order to form spliceosomes and ribosomes. Telomerase contains the genetic information to maintain the

* To

telomere sequence due to its bound RNA template. RNase P consists of a catalytic RNA and an accessory protein which is essential for the full activity in vivo. Biochemical approaches for the isolation of RNA-binding proteins are often difficult and laborious. Despite the central importance of RNA–protein interactions, only few screening approaches are available for their detection and rapid analysis. For example, a combinatorial phage display library (1,2) allowed the selection of artificial peptides which were bound to a specific RNA. Other in vitro procedures used a labeled RNA-ligand for the detection of RNA–protein interaction (3–5). The former procedures rely on a modified protocol for the isolation of DNA-binding proteins (6,7). However, detection of cDNA clones encoding new and biologically relevant RNA-binding proteins has not yet been demonstrated by RNA ligand screening. In this report, we describe a screening assay in which RNA-binding proteins are efficiently and rapidly detected in a large cDNA expression library. Our approach combines the advantages of described DNA-ligand and RNA-ligand screening techniques in vitro. The simple screening procedure presented here may be of general usefulness for the isolation of hitherto undiscovered RNA-binding proteins.

MATERIAL AND METHODS Clones for the U1A/U1-RNA model system A plasmid containing cDNA that encoded the human U1A wild-type protein, as well as the plasmid that contained the sequence of the U1-RNA stem–loop II (8) were kind gifts from Dr Nagai, Cambridge, UK. Digestion with NdeI and HindIII released a fragment of ∼300 bp corresponding to the N-terminal residues 1–102 of wild type U1A protein. After filling in the NdeI 5′ overhang with Klenow polymerase, the fragment was subcloned in frame with the lacZ gene into the EcoRV–HindIII sites of pBluescript SK (Stratagene, La Jolla, USA). Subsequently, the entire insert was cut out by EcoRI–XhoI digestion and ligated into the same sites of lambda ZAP II (Stratagene, La Jolla, USA). In vitro packaging of recombinant phage was carried out using home-made packaging extract (9), yielding lambda clone LU1A.

whom correspondence should be addressed. Tel: +30 81 394364; Fax: +30 81 394408; Email: [email protected]

3817 Nucleic Acids Acids Research, Research,1994, 1997,Vol. Vol.22, 25,No. No.119 Nucleic

3817

Preparation of blocking RNA

Adsorbing recombinant proteins to nitrocellulose membranes (plaque lift)

Yeast RNA (SIGMA R-6625, RNA type VI, torula yeast; SIGMA-Aldrich, Germany) (1 g) was dissolved in 10 ml TE buffer, adjusted to pH 7.5 with 5 N NaOH, and incubated with 5 µg proteinase K (Promega, Madison, USA) for 2 h at 37
C. After extraction of the RNA with RNase-free phenol–chloroform [1:1 (v/v), adjusted to pH 8.0], the RNA was precipitated in 2.5 vol ice-cold EtOH (–20
C) for 5 min, centrifuged at 1800 g for 10 min at 4
C and washed in 70% EtOH. The wet pellet was resuspended with DEPC-treated H2O (9) to give a final concentration of 50 mg/ml yeast blocking RNA (assuming 1 OD260 = 40 µg/ml RNA), and 1 ml aliquots were distributed to Eppendorf tubes and stored at –20
C. Total RNA from Escherichia coli XL-1 was prepared using a slightly modified protocol according to Oelmüller et al. (10). An aliquot of 100 ml bacterial cells from the late exponential phase was centrifuged, the pellet resuspended in 10 ml ice-cold AE-buffer (20 mM NaAc, pH 5.5, 1 mM EDTA) and extracted with 1:1 vol hot phenol (pre-warmed at 65
C). The DNA–RNA mixture was subjected to two additional phenol extractions in the presence of 1/10 vol 3 M NaAc, pH 5.5, precipitated in 2.5 vol 100% EtOH and washed with 75% EtOH. The pellet was resuspended in 1 ml TE buffer; 5 U RNase-free DNase I (Boehringer Mannheim, Germany), 40 U RNasin (Promega, Madison, USA) were added and incubated for 30 min at 37
C. Then, 100 µg proteinase K (Promega, Madison, USA) were added, and incubation continued for another hour. The total RNA was phenol extracted, precipitated in 2.5 vol 100% EtOH, the pellet resuspended in 0.5 ml TE buffer and the concentration determined using the OD260/OD280 ratio (9). The yield was ∼5–10 mg total RNA/100 ml bacterial cells.

Escherichia coli strain XL-1 (Stratagene, La Jolla, USA) was used as host strain throughout the studies for the lytic growth of lambda ZAP II (Stratagene, La Jolla, USA) and recombinant derivatives. Bacteriophage infection and plating of E.coli XL-1 cells were performed after infection according to Sambrook et al. (9) with slight modifications. A detailed description of the procedure for screening cDNA expression libraries is given below to ensure optimal conditions. A single colony of E.coli XL-1 was grown overnight at 30
C in 50 ml LB medium (9), supplemented with 0.2% maltose, 10 mM MgSO4 and 10 µg/ml tetracycline. Cells were collected by centrifugation at 700 g for 10 min at 4
C, carefully resuspended in ice-cold sterile 10 mM MgSO4, adjusted to OD600 = 1.5 and stored for maximally 2 days at 4
C. Bottom agar (1.2%) was freshly prepared by boiling 250 ml NZY medium (9), supplemented with 10 µg/ml tetracycline, and poured into Petri dishes (ø 85 mm). The plates were dried upside down with open lids for 30 min at 37
C and then kept at 37
C before pouring the top agar. Bacterial cells were infected by mixing 100 µl of E.coli XL-1 (OD600 = 1.5) with 10 µl bacteriophages [∼1–1.5 × 104 plaque forming units (p.f.u.)/plate] in sterile glass tubes and incubation for 30 min at 37
C. For each plate, 2.5 ml top agar (sterile NZY medium containing 0.7% electrophoresis grade agarose, pre-warmed at 50
C) was added, shortly vortexed and immediately poured onto bottom agar. The plates were incubated for the first 30 min upside down with open lids at 42
C, and incubation was continued for 3–4 h with closed lids until pinpoint plaques appeared. Nitrocellulose membranes [ø 82 mm, Immobilon-NC (HAHY) membrane, Millipore, Bedford, USA] were impregnated for 30 min in 20 mM IPTG, overlaid on the plates and incubated overnight at 37
C. Hybridization with an RNA-ligand

Preparation of the RNA probes A plasmid (8) was used as template for the synthesis of the U1-RNA. The RNA transcript consisted of the 46mer U1-RNA stem–loop II containing nucleotides 50–92, with three additional guanines at the 5′ end. Radioactively labeled U1-RNA stem–loop II (designated U1-RNA in this study) probe was generated from 1 µg of EcoRI-linearized plasmid DNA. The DNA was incubated in a final reaction volume of 20 µl, containing 1× T3 RNA polymerase reaction buffer (Boehringer Mannheim, Germany), 500 µM rATP, rGTP and rCTP, 5 µM UTP, 20 U RNasin (Promega, Madison, USA), 1 µl [α-32P]UTP (800 Ci/mmol; Amersham, Buckinghamshire, UK) and 1 U T3 DNA-dependent RNA polymerase (Boehringer Mannheim, Germany) and incubated for 1.5–2 h at 37
C. After addition of 1 U RNase-free DNase I (Boehringer Mannheim, Germany) and further incubation for 5 min at 37
C, the transcript was extracted 1× with phenol–chloroform [1:1 (v/v), adjusted to pH 8.0], fractionated on a 2 ml Bio-Gel A-0.5m (100–200 mesh) gelfiltration column (Bio-Rad Laboratories) equilibrated with TE buffer. RNA transcripts were stored at –20
C. Typical labeling reactions yielded U1-RNA probes with an activity between 20 and 25 × 106 c.p.m. The integrity of the RNA-probe was routinely checked on a 5% denaturating polyacrylamide gel (9) before screening was carried out.

Unless otherwise noted, the whole procedure was performed with a single screening buffer (SB buffer) containing 15 mM HEPES, pH 7.9, 50 mM KCl, 0.1% (w/v) Ficoll 400-DL (SIGMAAldrich, Germany), 0.1% (w/v) Polyvinyl-Pyrrolidon PVP-40 (SIGMA-Aldrich, Germany), 0.01% (v/v) Nonidet P-40 (SIGMA-Aldrich, Germany), 0.1 mM MnCl2, 0.1 mM ZnCl2, 0.1 mM EDTA and 0.5 mM DTT, dissolved in 0.1% (v/v) DEPC-treated nanopure H2O. For pre-hybridization and hybridization, 100 µl of 50 mg/ml yeast blocking RNA was added to 50 ml SB buffer. A crystallizing dish (ø 120 mm, 60 mm high) was extensively washed with soap before use, wrapped in aluminum foil and kept overnight in a 150
C oven to destroy any remaining RNase activity. After overnight incubation and prior to removal, membranes were labeled with black ink to mark their orientation relative to the Petri dish. They were dried for 2–5 min on Whatman 3MM paper and transferred into the crystallizing dish for the first wash in 100 ml SB buffer (16 membranes were maximally screened per dish). All subsequent steps were carried out on a gyratory shaker at 150 r.p.m. (Bioblock Scientific, Rotatest 74579) at 4
C. Four prewashes with 100 ml SB buffer each were carried out (each for 5 min) to remove bacterial debris and material loosely bound to the membranes. After quantitatively decanting the SB buffer, the membranes were prehybridized for 30 min in 50 ml SB buffer, supplemented with 0.1 mg/ml of yeast blocking RNA. After the addition of the 32P-labeled RNA

3818 Nucleic Acids Research, 1997, Vol. 25, No. 19 probe (to give a final activity of ∼0.25–0.5 × 106 c.p.m./ml), the hybridization continued for 2 h. Unspecifically bound radioactivity was removed by washing the membranes four times for exactly 5 min in 100 ml SB buffer. The membranes were completely dried onto Whatman 3MM paper, wrapped in Saran foil and exposed to Kodak X-Omat AR film for 24 h at –70
C with the support of a Dupont Lightning-Plus intensifying screen. Putative positive clones were picked and purified to homogeneity in a secondary and tertiary screening round according to Sambrook et al. (9). Northwestern blot analysis Recombinant E.coli XL-1 harboring plasmid pMS-X1 was grown at 30
C in LB medium containing 100 µg/ml ampicillin. Expression of recombinant protein was induced at OD600 = 0.8 by the addition of 1 mM IPTG to the cells and cultivation continued for 3 h at 30
C. Non-induced and induced cells (1 ml) were spinned and the supernatant discarded. Bacterial pellets were resuspended in 100 µl extraction buffer (100 mM Tris–HCl, pH 8.0, 200 mM NaCl, 1 mM EDTA, 3 mM MgCl2, 1 mM DTT, 0.5 mM PMSF, 10 µM Leupeptin, 10 µM Aprotinin, 10 µM Pepstatin and 10% glycerol). An aliquot of 10 µl of each cell extract was mixed with 10 µl loading buffer (100 mM Tris–HCl, pH 6.8, 200 mM DTT, 4% SDS, 0.1% bromophenol blue and 20% glycerol) and heated for 5 min at 95
C. Protein samples were resolved using 10% SDS–polyacrylamide gel electrophoresis and blotting was carried out according to Zaidi and Malter (11) onto nitrocellulose membrane (Protan Nitrocellulose, Schleicher & Schuell) at 30 V (constant voltage) for 16 h in Tris–glycine–methanol buffer (25 mM Tris, pH 8.3, 200 mM glycine and 20% methanol) at 4
C with TransBlot apparatus (BioRad, Hercules, USA). Unlike in the original protocol (11), the transferred proteins were renatured overnight in SB buffer and the temperature was reduced to 4
C. Prehybridization and hybridization was carried out as described in the previous section. Tomato expression library and viroid probes A cDNA expression library was prepared from leaves of Lycopersicum esculentum (cultivar Rentita) using a ZAP-cDNA GigapackII Gold cloning kit (Stratagene, La Jolla, USA). Labeled (+) RNA transcripts of potato spindle tuber viroid (PSTVd) were prepared by in vitro transcription using plasmids pHa106 for a slightly larger than monomeric unit (12) or pSP-Av5.8(–) for a multimeric unit (13). RESULTS Expression of recombinant U1A protein fragments The purpose of this study was to develop a screening technique for the identification of RNA-binding proteins. For establishing the proper experimental conditions we took advantage of the human U1A protein which specifically interacts with human U1-RNA (8,14). Both molecules represent a well characterized model system for RNA–protein interaction. The peptide fragment A102wt, containing the first 102 amino acids of the human U1A protein, had shown full binding affinity to U1-RNA (8,14). It was shown that this peptide also bound tightly to a sub-fragment of the U1-RNA, the 46mer stem–loop II

Figure 1. Test of a described hybridization protocol for the U1A/U1-RNA model system. A plaque-lift contained a mixture (1:5) from U1A expressing lambda phage LU1A and from non-recombinant lambda ZAP II. Hybridization with 32P-labeled U1-RNA (0.25 × 106 c.p.m./ml) followed exactly the protocol according to Qian et Wilusz (3). Already at low plaque densities it became difficult to discriminate plaques caused by recombinant and non-recombinant phages.

of U1-RNA (8), which was also used here. The truncated A102wt peptide was expressed in a stable manner and folded properly in E.coli cells (1,2,8). We inserted the corresponding cDNA into bacteriophage lambda ZAP II and expressed it as fusion protein with β-galactosidase from the recombinant lambda LU1A. Testing the U1A/U1-RNA model system Several published protocols (3–5) for the in vitro detection of an RNA–protein interaction were tested for their ability to detect also a specific RNA–protein binding caused by the U1A/U1-RNA model system. The protocols are based on an earlier description for the screening of an expression cDNA library for the identification of DNA-binding proteins (6,7). When we tested them, only the protocol described by Qian and Wilusz (3) showed a positive signal (Fig. 1). However, the positive signal was no longer detectable at plaque densities >2 × 103 p.f.u./plate (ø 85 mm) even after the specific radioactivity was raised to 107 c.p.m./ml (not shown). If the strength of a RNA–protein interaction is unknown prior to screening, one would need to apply very low plaque densities. This protocol, therefore, did not seem completely satisfactory for screening of an entire library, especially if rarely expressed RNA-binding proteins or moderate binding affinities are to be identified. Improving the screening sensitivity and quality We inspected various parameters which could improve the signal:background ratio of the screening procedure systematically. First, the effect of NaCl in the removal of unspecific RNA-binding during the final wash was tested. Plaque lifts from non-recombinant lambda ZAP II were exposed to 32P-labeled U1-RNA probes and then subjected to the washing buffer supplemented with NaCl at concentrations between 50 and 500 mM. High salt washing in the presence of NaCl or KCl showed only negligible reduction of the background (Fig. 2A). Since a high salt concentration might also remove specifically bound RNA, we decided to carry out further final washings in the presence of only 50 mM KCl. Second, we used a single buffer (SB buffer) for pre-hybridization and hybridization, supplemented with 0.1 mg/ml yeast blocking RNA (Fig. 2B). This greatly improved the reproducibility and resulted also in an increased signal:background ratio. In the previous protocol, Qian et Wilusz (3) pre-hybridized the mem-

3819 Nucleic Acids Acids Research, Research,1994, 1997,Vol. Vol.22, 25,No. No.119 Nucleic

3819

Figure 3. Influence of the RNA integrity. Lambda LU1A was mixed with non-recombinant lambda ZAP II. Hybridization was done in SB buffer, but U1-RNA probes were used from three independent transcript preparations. (A) Non-degraded U1-RNA probe showed negligible background. (B) Slight degradation of U1-RNA gave an increased background. (C) Extensive degradation resulted in a very strong background and ‘white’ plaques.

Figure 2. Parameters influencing the background signal. Uniform plaque lifts from non-recombinant lambda ZAP II were used for the hybridization with U1-RNA. (A) Influence of NaCl. U1-RNA remained tightly bound to the membrane, even after a final four-times wash for 5 min each in the presence of 0.5 M NaCl. (B) The addition of total yeast RNA (0.1 mg/ml) as blocking agent in the pre- and hybridization buffer reduced the background drastically.

branes with an excess of 5 mg/ml yeast RNA, but omitted the blocking substrate during RNA probing in the hybridization buffer. Third, we tested different buffer systems. A change of HEPES buffer to Tris buffer, phosphate buffer or citrate buffer gave similar results, but the use of 15 mM HEPES buffer, pH 7.9, gave the best signal:background ratio. However, pH adjustment was critical and resulted in a considerable background increase below pH 7.5. Fourth, we found that some detergents showed a positive effect on the signal:background ratio. For example, neutral detergents like Tween-20 or Nonidet P-40 (both at a concentration of 0.01%) could decrease the background significantly without affecting the positive signal. Inclusion of 0.01% SDS in the buffer reduced both the positive and the background signal. The RNA-ligand was completely washed off at a concentration of 0.5% SDS. In contrast, the presence of cationic detergents like CTAB or DTAB resulted in a strong increase of the background signal. Fifth, the influence of several membrane types on the signal/background was tested. HAHY or HATF type membrane (Immobilon-NC membranes, Millipore, Bedford, USA) gave excellent results with respect to the sharpness of the positive signal and the signal:background ratio. Hybond C membrane (Amersham, Buckinghamshire, UK) gave similar results, however this membrane was relatively rigid and therefore difficult to handle. The size of the membranes was crucial. Large membranes produced a non-homogenous overall background which often prevented the detection of positive signals. Small membranes (ø 82 mm) were beneficial in two ways. First, they produced a homogenous background allowing the detection of positive signals even at the membrane periphery. Second, the appearance of false positive signals was negligible, thus facilitating the screening procedure. Finally, we found that even minor degradation of the RNA probe severely deteriorated the detection limit of the screening protocol (Fig. 3).

Transition metals greatly increase the signal:background ratio By a plaque lift procedure, not only recombinant proteins become transferred to the nitrocellulose membrane, but also proteins originating from lysed E.coli cells. Most likely, these are proteins that bind RNA in a sequence-unspecific fashion, creating a general background signal. As mentioned above, the inclusion of carrier RNA reduced the background. In addition to that, we tested a wide range of divalent metal ions for their ability to reduce background originating from non-recombinant lambda phages. Neither Mg2+ nor Ca2+ affected the background. Instead, the specific hybridization signal obtained by binding of U1-RNA to U1A protein was gradually washed off when concentrations of Mg2+ or Ca2+ approached 25 mM. Quite different results were obtained in the presence of 0.1 mM Mn2+ or Zn2+ in the screening buffer. Here, the hybridization signal from non-recombinant lambda phages was completely suppressed (Fig. 4). A similar reduction of the background was observed by the use of 0.1 mM Fe2+, Co2+ or Ni2+. However, we found the best signal:background ratio by a combination of Mn2+ and Zn2+, each at a concentration of 0.1 mM (Fig. 4B and C). The intensity of the positive signal was still high, but decreased at concentrations above 1 mM Zn2+ or Mn2+, respectively. No degradation of the RNA probe could be observed during hybridization unless concentrations of 10 mM Zn2+ or more were applied (Fig. 4D). Assessing the detection limit of RNA–protein interaction We examined whether the improved protocol, including the background-suppressing transition metals Zn2+ and Mn2+ (0.1 mM each), could be used to detect a single phage clone which expressed the recombinant U1A protein amongst an increasing number of unrelated phages. Such a high plaque density mimicked the screening of an expression library. For this purpose, lambda phage LU1A was randomly mixed with increasing numbers of recombinant lambda phages from a cDNA expression library. As Figure 5 shows, a positive signal generated by recombinant U1A was unambiguously detectable, even at the highest plaque density (25 000 p.f.u./plate) tested. It is noteworthy in this context that confluent plaques created by high phage densities did not significantly affect the hybridization signal.

3820 Nucleic Acids Research, 1997, Vol. 25, No. 19

Figure 4. Influence of different metal ions on the hybridization signal. Plaque lifts were prepared from a plate containing ∼30 p.f.u. originating from lambda LU1A and 60 p.f.u. from lambda ZAP II. Both membranes were probed with 32P-labeled U1-RNA from the same batch. (A) Hybridization with SB buffer lacking metal ions. Both types of plaques gave a hybridization signal, but lambda LU1A was clearly distinguishable from lambda ZAP II (2 days film exposure). (B) Hybridization with SB buffer containing 0.1 mM Mn2+ and 0.1 mM Zn2+. Only plaques of lambda LU1A were visible, while no hybridization signal originating from non-recombinant lambda ZAP II could be detected (2 days film exposure). The signal intensity of LU1A was not significantly reduced compared to (A). (C) The same membrane as in (B) after 14 days of film exposure. Still no hybridization signals originating from non-recombinant lambda ZAP II were visible and the general unspecific background signal from the membrane remained low. (D) Analysis of RNA probes taken after RNA–protein hybridization from SB buffer (5% native polyacrylamide gel). Hybridization was done with increasing amounts of Zn2+ up to 10 mM with the U1A/U1-RNA model system. Optimal signal:background ratio was found for Zn2+ at a concentration of 0.1 mM. Neither degradation nor structural change of the U1-RNA could be observed after hybridization including up to 1 mM Zn2+. However, the transcript was completely degraded in the presence of 10 mM Zn2+.

Extending the screening protocol to large RNA-ligands As shown, a moderate binding affinity generated from the U1A/U1-RNA pair was detectable in a large excess of unrelated phages. However, the RNA-ligand was a relatively short molecule consisting of the 46mer U1-RNA stem–loop II (8). We applied the screening protocol also to (very) large RNA-ligands in order to see whether the size of the RNA affects the background signal. For this purpose, plaque lifts from a tomato cDNA expression library were subjected to various large RNA-ligands. For example, hybridization with monomeric/oligomeric RNAligands (up to 2028 nucleotides in length) from the potato spindle tuber viroid (PSTVd; ref. 13) gave a similar low background to that produced by U1-RNA. This encouraged us to screen the cDNA expression library from tomato for the detection of PSTVd-binding proteins. In a primary screening, ∼250 000 phages were plated. Plaque lifts were subjected to a monomeric PSTVd (+) RNA-ligand according to the protocol (Materials and Methods). The primary screen (Fig. 6A) showed 18 signals on the autoradiograph. Phage clones from corresponding plaques were analyzed in further screening rounds (Fig. 6B). Fourteen clones maintained binding affinity after

Figure 5. Detection of lambda LU1A amongst an excess of recombinant phages. A mixture of ∼20 p.f.u. of lambda phage LU1A was plated with an increasing number (5000–25 000 p.f.u.) of unrelated lambda phages originating from a tomato cDNA expression library. Plaque lifts were hybridized with U1-RNA (0.5 × 106 c.p.m./ml). Plaques became confluent at densities of 5000 p.f.u./plate or higher. Remarkably, this did not severely impair the signal strength. Positive signals could be detected also at densities of 25 000 pfu/plate. However, this required that the background was evenly distributed over the entire membrane (see Results).

Figure 6. Detection of viroid-RNA-binding proteins by the use of large RNA-ligands in the developed screening protocol (Materials and Methods). In a primary screening, ∼250 000 phages from a cDNA expression library from tomato leaves were analyzed using a 32P-labeled monomeric PSTVd (+) RNA transcript of 381 nucleotides (12). (A) Autoradiograph showing a putative clone after primary screening (arrow; the three other spots are ink labels). (B) The viroid-binding properties from the putative clone were confirmed by secondary/tertiary screening (see text for details).

secondary and tertiary screening, thus showing the reliability of the protocol. Confirming the specificity of the interaction between a newly identified protein and viroid RNA Since the screening procedure for RNA-binding proteins was carried out in the presence of a yeast total RNA extract, proteins that interact with RNA in a sequence-unspecific manner should not hybridize to the radioactively labeled RNA ligand. To confirm that the identified cDNA clones encoded for proteins that specifically interact with PSTVd RNA, two control experiments were carried out with one of the clones, called MS-X1. In a first experiment, the specificity of binding of the isolated phage LMS-X1 was examined with the plaque lift assay, using different RNA transcripts as a probe. Only when PSTVd (+) RNA was used, recombinant phages could be discriminated from the

3821 Nucleic Acids Acids Research, Research,1994, 1997,Vol. Vol.22, 25,No. No.119 Nucleic

Figure 7. Binding specificity of an identified viroid-RNA-binding protein. Recombinant lambda clone LMS-X1 encoding the viroid-binding protein was mixed with non-recombinant lambda ZAP II (1:1). Plaque lift and hybridization assays with 32P-labeled RNA-ligands were performed as described in Materials and Methods. The autoradiograph shows in (A) a positive signal due to specific binding of the monomeric PSTVd (+) RNA to the viroid-binding protein while in (B) only background signals were visible after hybridization with unrelated U1-RNA.

lambda ZAP II background (Fig. 7A). U1-RNA did not result in a positive signal (Fig. 7B). Neither an RNA transcript of plum pox virus (PPV; ref. 15), nor a plant mRNA probe of ∼1.4 kb (LeMA-1; ref. 4) gave any specific signals (not shown). This verified that the screening protocol identified sequence-specific interactions. In a second experiment, we examined whether the binding of the protein encoded by clone MS-X1 could be verified by an Northwestern assay (11). For this purpose, the recombinant lambda phage LMS-X1 was in vivo excised generating the corresponding plasmid pMS-X1 that was propagated in E.coli XL-1. Crude protein extracts were prepared before and after induction with IPTG, separated on an SDS–polyacrylamide gel, blotted on nitrocellulose membrane and hybridized with a labeled PSTVd (+) RNA transcript. A strong signal became visible after induction of the cells harboring pMS-X1 (Fig. 8). Apparently two forms of the recombinant protein were produced in E.coli, which differed slightly in size. The strong hybridization signal confirmed the specific interaction of the expressed protein with PSTVd (+) RNA. It is noteworthy that under the experimental conditions used for this Northwestern assay, no background signals were visible that originated from E.coli proteins. DISCUSSION The functional screening assay described here provides a versatile method for the identification of lambda bacteriophage cDNA clones that express RNA-binding proteins. The changes we introduced to previous protocols (3–7) seem to make the RNA-ligand screening decisively reliable. The most important changes were the introduction of a single-buffer system and the addition of carrier RNA, a detergent, and especially, of transition metals which eliminated the background signals. Great care was taken to avoid any degradation of the RNA probe. Specific binding could be detected in a large excess of unrelated lambda phage clones, which is a prerequisite for successful screening of a whole cDNA expression library. The method was elaborated with the U1A/U1-RNA pair which has a specific, but moderate binding affinity (Kd = 10–8 M; ref. 8). Interactions of similar or stronger affinities should therefore be readily detectable. We do not know whether the conditions that were optimal for the

3821

Figure 8. Northwestern blot analysis of clone pMS-X1. Crude cell extracts of E.coli XL-1 containing either pMS-X1 or pBluescript SK (control) were subjected to electrophoresis on a 10% SDS–polyacrylamide gel. Proteins were blotted to nitrocellulose membrane followed by Northwestern analysis with 32P-labeled monomeric PSTVd (+) RNA as described in Materials and Methods. Cell extracts are shown from non-induced (lanes 1 and 3) and IPTG-induced (lanes 2 and 4) E.coli XL-1 containing plasmid pBluescript SK or pMS-X1, respectively.

U1A/U1-RNA model system will work for all types of RNA–protein interaction. However, it could be demonstrated that proteins could be identified that interact specifically with PSTVd-RNA, a circular, single-stranded RNA molecule of 359 bases, which has a characteristic rod-like secondary structure and fulfills its entire genetic programme as a replicon and pathogen by interaction with largely unknown proteins of its plant host (16). For the identification of the viroid-binding proteins from a tomato expression library, a large RNA-ligand was used for screening indicating that the size of the RNA is not critical. It is noteworthy that the number of signals that turned out to be artifacts during secondary and tertiary screening was very small, illustrating the selectivity and reliability of the screening protocol. In vitro RNA-ligand screening is straightforward and relatively simple. It allows the identification of poorly expressed RNAbinding proteins without the requirement for laborious and expensive purification of proteins from tissues. RNA-ligand screening complements and extends other recently introduced techniques to identify RNA-binding proteins, such as the three-hybrid system (17,18) and other in vivo genetic approaches (19,20). We have tested some of the newly identified viroid RNA-binding proteins in the three-hybrid system of SenGupta et al. (17). Preliminary results show that we can confirm the interaction also with this system, however, the positive signal is weak (Maniataki et al., unpublished). In view of this, direct RNA-ligand screening described in this report seems to have a higher sensitivity. However, it will be necessary to accumulate more data to compare the versatility of the two different screening approaches. Besides a potential higher sensitivity, another advantage of RNA-ligand screening is that standard lambda cDNA expression libraries can be utilized. In addition, a simple in vitro synthesized RNA transcript can be applied without the need to fuse it to an unrelated RNA which may influence the secondary structure and binding affinity. Thus, also mutant RNA and RNA sub-fragments can be readily tested. RNA-ligand screening may therefore find wide applications for the identification of new RNA-binding proteins.

3822 Nucleic Acids Research, 1997, Vol. 25, No. 19 ACKNOWLEDGEMENTS We are grateful to Dr Nagai, MRC, Cambridge, UK, for kindly providing plasmids. We thank M.Providaki for technical assistance in part of the project. We also thank Prof. N.Panopoulos for critically reading the manuscript. R.S. acknowledges generous financial support from the Swiss National Science Foundation, grant number 5002-37716. This work was supported in part by the European Union within the network activity of the ‘Human Capital and Mobility’ programme (ERBCHRXCT94-0635) and by the ‘BIOTECH’ programme (BIO2CT93 0400-DG12SSMA). REFERENCES 1 Laird-Offringa,I.A. and Belasco,J.G. (1995) Proc. Natl. Acad. Sci. USA, 92, 11859–11863. 2 Laird-Offringa,I.A. and Belasco,J.G. (1996) Methods Enzymol., 267, 149–168. 3 Qian,Z. and Wilusz,J. (1993) Anal. Biochem., 212, 547–554. 4 Prombona,A., Tabler,M., Providaki,M. and Tsagris,M. (1995) Plant Mol. Biol., 27, 1109–1118. 5 Werner,R., Mühlbach,H.-P. and Guitton,M.-C. (1995) BioTechniques, 19, 218–221.

6 Singh,H., LeBowitz,J.H., Baldwin,A.S. and Sharp,P.A. (1988) Cell, 52, 415–423. 7 Vinson,C.R., La Marco,K.L., Johnson,P.F., Landschulz,W.H. and McKnight,S.L. (1988) Genes Dev., 2, 801–806. 8 Jessen,T.-H., Oubridge,C., Hiang Teo,C., Pritchard,C. and Nagai,K. (1991) EMBO J., 10, 3447–3456. 9 Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. 10 Oelmüller,U., Krüger,N., Steinbüchel,A. and Friedrich,C.G. (1990) J. Microbiol. Methods, 11, 73–84. 11 Zaidi,S.H.E. and Malter,J.S. (1995) J. Biol. Chem., 270, 17292–17298. 12 Tsagris,M., Tabler,M. and Sänger,H.L. (1991) Nucleic Acids Res., 19, 1605–1612. 13 Tabler,M. and Sänger,H.L. (1985) EMBO J., 4, 2191–2199. 14 Scherly,D., Boelens,W., van Venrooij,W.J., Dathan,N.A., Hamm,J. and Mattaj,I.W. (1989) EMBO J., 8, 4163–4170. 15 Tabler,M. and Tsagris,M. (1991) Gene, 108, 175–183. 16 Symons,R.H. (1991) Mol. Plant-Microbe Interact., 4, 111–121. 17 SenGupta,D.J., Zhang,B., Kraemer,B., Pochart,P. and Fields,S. (1996) Proc. Natl. Acad. Sci. USA, 93, 8496–8501. 18 Putz,U., Skehel,P. and Kuhl,D. (1996) Nucleic Acids Res., 24, 4838–4840. 19 MacWilliams,M.P., Celander,D.W. and Gardner,J.F. (1993) Nucleic Acids Res., 21, 5754–5760. 20 Harada,K., Martin,S.S. and Frankel,A.D. (1996) Nature, 380, 175–179.