Edouard Bertrand*, Micheline Fromont-racine, Raymond Pictet, and Thierry Grange

We have adapted to RNA molecules the ligation-mediated polymerase chain reaction (LMPCR) procedure of genomic sequencing [Mueller, P. R. & Wold, B. (1989) Science 246, 780-786]. This new procedure, the reverse ligation-mediated PCR (RLPCR), is sufficiently sensitive to allow "in vivo" footprinting of minor RNA species. It is based on the ligation of an RNA linker of known sequence to every 5' end resulting from the cleavage of total cellular RNA. Target RNA molecules are specifically reverse-transcribed and the resulting products are amplified by PCR. The localization of the initial 5' ends is ultimately determined on a sequencing gel. To demonstrate the validity of this strategy, we have used RNase Ti treatment of permeabilized cells and RLPCR and have detected in vivo iron-depletion-dependent footprints on two iron-responsive elements of the transferrin receptor mRNA. RNA-protein interactions play a critical role in many biological processes (for a review, see ref. 1). Dedicated RNAbinding proteins recognize specific RNA sequences and/or structures that participate in the regulation of gene expression at levels as diverse as transcription initiation (2), premRNA splicing and polyadenylylation (3), nucleocytoplasmic transport of mRNA (4), specific mRNA localization in the cytoplasm (5), mRNA stability (6, 7), and translation (8, 9). A well-studied example is the regulation mediated by the iron-responsive element-binding protein (IRE-BP; ref. 10; for a review, see ref. 11). This protein interacts specifically with sequences involved in the iron-dependent regulation of ferritin mRNA translation (8, 9) and transferrin receptor (TfR) mRNA stability (6, 7). The affinity of the IRE-BP for the iron-responsive element (IRE) is dependent upon the intracellular iron level; it is roughly 200-fold higher in the presence of an iron chelator than in the presence of an iron source (12). Iron is believed to modulate this affinity through direct interaction with the protein (13). The interaction of the IRE-BP with the IRE has different functional consequences depending on the location of the IRE inside the mRNA and the overall arrangement of the regulatory sequences. In the ferritin mRNA, the IRE is localized within the 5' untranslated region (UTR) and the interaction of the IRE-BP is believed to repress translation of the mRNA (9). In the TfR mRNA, five IREs are localized in the 3' UTR, and the interaction of the IRE-BP is believed to prevent mRNA degradation (6, 14). The so-called "footprinting" method is one of the most informative techniques to analyze the interaction of proteins with nucleic acids. It relies on the protein-induced changes of the reactivity of nucleic acids toward a modifying agent. This technique has been instrumental in the characterization ofthe large number of DNA-binding proteins currently known in eukaryotic cells. Further insight into the participation of these proteins in the regulation ofgene expression was gained when it became possible to analyze their interaction with DNA inside the cell by "in vivo" footprinting (15). The cornerstone of the in vivo footprinting approach, especially when one works with higher eukaryotes, is the ability to detect events involving rare molecules diluted in a complex population. For example, in mammals, a given base in a given regulatory sequence will represent less than -10-9 of the DNA initially present inside the cell. The most sensitive method for the visualization of in vivo footprints is the ligation-mediated PCR (LMPCR) procedure of Mueller and Wold (16), which allows exponential amplification of the signal. Our current knowledge of RNA-binding proteins is much less advanced than that of DNA-binding proteins in part because of technical limitations. In particular, no amplification method suitable for in vivo footprinting of RNA molecules is available. Since the average amount ofRNA existing in a mammalian cell corresponds to 5 x 1010 nucleotides (17), the in vivo analysis of the interaction of proteins with RNA present at most in a few hundred copies per cell requires methods with a sensitivity similar to those used for the in vivo footprinting of a single-copy gene. We present a procedure derived from the LMPCR method (16) that allows mapping at single-base resolution of cleavage points introduced into a specific RNA species diluted in a complex RNA population. We illustrate the power of this technique by the detection in vivo of iron-depletion-dependent footprints on two IREs of the human TfR mRNA. MATERIALS AND METHODS Nuclease Treatment in Vivo and Preparation of RNA. Exponentially multiplying human hepatoma cells (Hep G2; ref. 18) were treated for 20 hr with either desferrioxamine (100 ,uM) or hemin (100 ,uM). Cells from one 100-mm plate (107 cells) were recovered in 10 ml of serum-containing medium after trypsinization, rinsed successively with 10 ml of phosphate-buffered saline and 1 ml of physiological buffer (11 mM KH2PO4/K2HPO4, pH 7.4/108 mM KCl/22 mM NaCl/1 mM MgCl2/1 mM dithiothreitol/1 mM ATP; ref. 19), and suspended in 600 ,l of physiological buffer. Cells were permeabilized and treated with RNase Ti after dilution of 100 ,ul of the cell suspension in 100 ,ul of ice-cold physiological buffer supplemented with 0.2% Nonidet P-40 and various amounts of RNase Ti (Boehringer Mannheim). After a 3-min incubation on ice, the nuclei were pelleted by a 15-sec centrifugation at 5500 x g and the supernatant was transferred to a tube containing 200 ,ul of 10mM Tris HCl, pH 7.5/150mM NaCl/5 mM EDTA/1% SDS and 400 ,ul of phenol/isoamyl alcohol, Abbreviations: LMPCR, ligation-mediated PCR; RLPCR, reverse ligation-mediated PCR; IRE, iron-responsive element; IRE-BP, IRE-binding protein; TfR, transferrin receptor; UTR, untranslated region. *Present address: Beckman Research Institute of the City of Hope, 1450 East Duarte Road, Duarte, CA 91010-0269. 3496 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Proc. Natl. Acad. Sci. USA 90 (1993) 3497 30:1. The contents of the tube were immediately mixed thoroughly. The aqueous phase was extracted two more times with phenol/isoamyl alcohol and then with chloroform, and the RNA was precipitated with ethanol. The RNA was suspended in 100 ,ul of water and the concentration of the solution was estimated on a gel by comparison with an RNA sample of known concentration. The RNA was precipitated again and suspended at 1 mg/ml in water. Untreated RNA used for chemical sequencing was prepared identically except that no RNase Ti was added. The reaction conditions for in vivo footprinting must be optimized to obtain a reliable footprint. The modification of the target RNA should be as close as possible to "single-hit" conditions in order to avoid destabilization ofthe RNA-RNA or RNA-protein complexes. However, the extent of the digestion should be sufficient to provide a good signal-tonoise ratio and a number of modified molecules large enough to be statistically significant (see discussion related to this point in ref. 20). We have observed high background at RNase Ti concentrations that were too low relative to endogenous RNase activity (most reacting bases were not guanines) and disappearance of the footprints at high RNase concentrations (data not shown). Therefore, the use of a footprinting reagent with a defined base specificity simplifies optimization of the footprinting conditions because the specific and expected pattern is easily discerned from background. Reverse Ligation-Mediated PCR (RLPCR). Oligodeoxyribonucleotides were synthesized by phosphoramidite chemistry on a Gene Assembler (Pharmacia) and purified by electrophoresis in a denaturing polyacrylamide gel followed by reverse-phase chromatography on Sep-Pack C18 cartridges (Waters). The "melting" temperature (Tm) of the oligonucleotide primers was estimated by using the formula Tm (C) = 4(G + C) + 2(A + T) (21). The hybridization temperature used during the PCR was equal to the Tm of either primer 2 or 3 minus 5°C (21). The nucleotide sequences (5' to 3') of the primers were as follows: DNA linker primer, GGGCAUAGGCUGACCCUCGCUGAAA; Pit-i primer 1, TACCACAGGCAAGTCT; Pit-1 primer 2, TATCTGCACTCAAGATGCTCCTT; TfR primer 1, CTAAATCTTAGCTTCAAC; TfR primer 2, AACTTTATTCAATTACATTTGGCTG; TfR primer 3, ATTCAATTACATTTGGCTGACGGCTG. Synthesis of the RNA linker. The RNA linker was synthesized by in vitro transcription of a synthetic oligonucleotide template with 17 RNA polymerase (22). The sequences ofthe oligonucleotides were 5'-TAATACGACTCACTATAG-3' and 5'-TTTCAGCGAGGGTCAGCCTATGCCCTATAGTGAGTCGTATTA-3'. The resulting RNA linker was purified by electrophoresis in a 10% polyacrylamide sequencing gel followed by reverse-phase chromatography. Phosphorylation ofRNA 5' ends. RNase Ti-treated RNA (7 ,ug) was incubated for 30 min at 37°C with 10 units of T4 polynucleotide kinase (Amersham) in 10 /l ofPNK buffer (50 mM Tris HCl, pH 7.6/10 mM MgCl2/5 mM dithiothreitol/0.1 mM spermine/0.1 mM EDTA) supplemented with 1 mM ATP. The samples were then kept at -70°C. RNA linker ligation. One hundred nanograms of RNA linker was ligated to 700 ng of 5' phosphorylated cellular RNA for 16 hr at 17°C in 10 ,ul of 50 mM Tris HCl, pH 7.5/10 mM MgCl2/20 mM dithiothreitol/100 ,uM ATP containing 1 ,ug ofRNase-free bovine serum albumin (Pharmacia), 20 units of human placental ribonuclease inhibitor (Amersham), and 3 units of T4 RNA ligase (Pharmacia). The ligated products were precipitated by the addition of 1l,u of 3 M sodium acetate (pH 6.0) and 30 Al of ethanol. The pellet was rinsed with 70% ethanol and resuspended in 12 Al of water. Reverse transcription. Reverse transcription was performed in a buffer suitable for the subsequent PCR (23). Primer 1 was hybridized with the RNA for 45 min prior to reverse transcription as follows: 6 ,ul of the ligation product was mixed with 1 ,ul of primer 1 (10 ng/,lI), 1 ,ul of 60 mM MgC92, and 1 ,ul of 10 x Taq buffer (650 mM Tris HCl, pH 8.8/100mM 2-mercaptoethanol/165 mM ammonium sulfate). The sample was incubated for 5 min at 95°C and then for 45 min at 42°C. One microliter of a mixture of the four dNTPs (5 mM each) and 3 units of avian myeloblastosis virus reverse transcriptase (Stratagene) were subsequently added and the incubation was continued at 42°C for 45 min. The reaction was stopped by a 5-min incubation at 95°C. PCR amplification. The DNA was then amplified by using theDNA linker and primer 2 under conditions similar to those described by Rigaud et al. (24). To the reverse transcription reaction mixture were added 1 ,l of 10 x Taq buffer, 1 ,l of a mixture of the four dNTPs (5 mM each), 2 ,l of DNase-free bovine serum albumin (2 mg/ml; Pharmacia), 2 ,ul of 4 mM MnCl2, 2 ,ul of the DNA linker (500 ng/,ul), 2 ,4 of primer 2 (50 ng/,ul), and 1 unit of Taq DNA polymerase (Cetus). After 5 min at 95°C, the reaction was cycled 25 times (30 sec at 94°C, 3 min at 59°C, 3 min at 74°C). Labeling ofprimer 3 and PCR products. Primer 3 (100 ng, 12 pmol) was labeled at the 5' end by incubation with 50 ,uCi (1.85 MBq) of [y32P]ATP (7 pmol) and 10 units of T4 polynucleotide kinase in 20 ,l of PNK buffer for 30 min at 37°C. The reaction was stopped by a 5-min incubation at 950C. The PCR products were labeled as follows: 5/l of the PCR mixture was mixed with 0.5 /ul of the primer 3-labeling reaction mixture, 5 ng of unlabeled primer 3, and 0.5 unit of Taq DNA polymerase in 5 ,ul of 65 mM Tris HCl, pH 8.8/10 mM 2-mercaptoethanol/16.5 mM ammonium sulfate/500 ,uM each dNTP/3 mM MgCl2/400 ,M MnCl2 containing DNasefree bovine serum albumin at 200 ug/ml. After a 5-min denaturation step at 95°C the reaction was cycled five times (30 sec at 94°C, 3 min at 69°C, 3 min at 740C). Either 2 IL was loaded directly onto a sequencing gel or the material was phenol-extracted and then precipitated with ethanol to allow loading of larger quantities. Direct labeling of PCR products during PCR amplification. Sometimes the use of 5'-end-labeled primer 2 gave satisfactory results. In this case the oligonucleotide was labeled as described above and the PCR amplification was performed as described except that 2 ,ul of the labeling reaction mixture and 1 ,l of unlabeled primer 2 (10 ng/4l) were added. RESULTS AND DISCUSSION Principle of the RLPCR Procedure. The conventional PCR requires that the sequences at both ends of the region to be amplified are known. The LMPCR procedure (16) allows the exponential amplification of a DNA fragment for which the sequence of only one end is known. Its basic principle is the ligation of a linker of known sequence to the unknown end, creating a substrate suitable for PCR. Since the linker has a discrete length, a complex population of DNA can be amplified with preservation of single-base resolution. We have kept the basic principle of the LMPCR procedure but modified the order of the steps and the enzymes used in order to adapt the method to RNA molecules (Fig. 1). There are four steps to the RLPCR procedure: step 1, nonspecific ligation of an RNA linker to all available 5' phosphorylated ends by T4 RNA ligase; step 2, specific synthesis of a cDNA copy of the RNA analyzed by primer extension; step 3, PCR amplification of the cDNA; and step 4, labeling of the PCR products. Fig. 2 shows that this procedure is specific and sensitive enough to read the sequence of a minor mRNA (such as one coding for a transcription factor) after a 1-day exposure (without an intensifying screen) of a sequencing gel when as Biochemistry: Bertrand et al. 3498 Biochemistry: Bertrand et al. Total cellular RNA cleaved in vivo or in vitro: 5' P 3' LU C>U G A>G