Purification of a RNA-binding Protein from Rat Liver

In cultured rat hepatocytes the degradation of phosphoenolpyruvate carboxykinase mRNA might be regulated by protein(s), which by binding to the mRNA alter its stability. The 3′-untranslated region of phosphoenolpyruvate carboxykinase mRNA as a potential target was used to select RNA-binding protein(s) from rat liver by the use of gel retardation assays. A cytosolic protein was isolated, which bound to the phosphoenolpyruvate carboxykinase mRNA 3′-untranslated region and other in vitro synthesized RNAs. The protein was purified to homogeneity; it had an apparent molecular mass of 400 kDa and consisted of identical subunits with an apparent size of 24.5 kDa. Sequence analysis of a tryptic peptide from the 24.5-kDa protein revealed its identity with rat ferritin light chain. Binding of ferritin to RNA was abolished after phosphorylation with cAMP-dependent protein kinase and was augmented after dephosphorylation with alkaline phosphatase. Weak binding was observed in extracts from okadaic acid-treated cultured hepatocytes compared with untreated cells. Preincubation of ferritin with an anti-phosphoserine or an anti-phosphothreonine antibody attenuated binding to RNA, while an anti-phosphotyrosine antibody generated a supershift indicating that phosphoserine and phosphothreonine but not phosphotyrosine residues were in close proximity to the RNA-binding region. Ferritin is the iron storage protein in the liver. Binding of ferritin to RNA was diminished in the presence of increasing iron concentrations, whereas the iron chelator desferal was without effect. It is concluded that ferritin might function as RNA-binding protein and that it may have important functions in the general regulation of cellular RNA metabolism.

In cultured rat hepatocytes the degradation of phosphoenolpyruvate carboxykinase mRNA might be regulated by protein(s), which by binding to the mRNA alter its stability. The 3-untranslated region of phosphoenolpyruvate carboxykinase mRNA as a potential target was used to select RNA-binding protein(s) from rat liver by the use of gel retardation assays. A cytosolic protein was isolated, which bound to the phosphoenolpyruvate carboxykinase mRNA 3-untranslated region and other in vitro synthesized RNAs. The protein was purified to homogeneity; it had an apparent molecular mass of 400 kDa and consisted of identical subunits with an apparent size of 24.5 kDa. Sequence analysis of a tryptic peptide from the 24.5-kDa protein revealed its identity with rat ferritin light chain. Binding of ferritin to RNA was abolished after phosphorylation with cAMP-dependent protein kinase and was augmented after dephosphorylation with alkaline phosphatase. Weak binding was observed in extracts from okadaic acid-treated cultured hepatocytes compared with untreated cells. Preincubation of ferritin with an anti-phosphoserine or an antiphosphothreonine antibody attenuated binding to RNA, while an anti-phosphotyrosine antibody generated a supershift indicating that phosphoserine and phosphothreonine but not phosphotyrosine residues were in close proximity to the RNA-binding region. Ferritin is the iron storage protein in the liver. Binding of ferritin to RNA was diminished in the presence of increasing iron concentrations, whereas the iron chelator desferal was without effect. It is concluded that ferritin might function as RNA-binding protein and that it may have important functions in the general regulation of cellular RNA metabolism.
In the regulation of RNA metabolism protein-RNA interactions play a central role. This is well documented for the splicing process, which eliminates intervening sequences from the primary transcript in the ribonucleoprotein particle of the spliceosome (1), for the transport of RNA from the nucleus to the cytoplasm (2), for translational initiation, which is mediated by protein factors recognizing the initiation region on the mRNA and positioning the ribosome to the translational start site (3), for the storage of translationally inactive mRNA (4) or for spatial localization of mRNA in the cell (5). Even if the mechanism and regulation of mRNA degradation are far from being understood, it is clear that RNA-protein interactions have a key role (6,7). Besides its function in RNA transport and translation the poly(A) tail has been shown to protect poly(A) carrying mRNAs from rapid degradation by interaction with the poly(A)-binding protein PABP (8,9). Many mammalian mRNAs contain AU-rich portions in their 3Ј-untranslated region, which after being occupied by AU-binding proteins render the RNA instable (10 -12).
In cultured rat hepatocytes and in rat hepatoma cells the expression of the key control enzyme in gluconeogenesis, phosphoenolpyruvate carboxykinase (PCK), 1 was stimulated at the transcriptional level by glucagon (13) or cAMP (14,15). Insulin, the glucagon antagonist at large, inhibited the glucagon-(cAMP-)stimulated gene transcription (13,16). Yet, besides transcriptional regulation the expression of PCK is also regulated at the level of mRNA degradation. In cultured rat hepatocytes insulin and the proinflammatory cytokine interleukin 6 accelerated the degradation of glucagon-induced PCK mRNA (13,17). The acceleration of PCK mRNA degradation was probably due at least in part to the insulin-enhanced expression of a ribonuclease. 2 In FTO-2B rat hepatoma cells cyclic AMP stabilized PCK mRNA (18,19). The stability of PCK mRNA seems to be regulated by proteins binding to the 3Ј-untranslated region of the mRNA. In cultured rat hepatocytes binding of a protein was enhanced after glucagon treatment, which correlated with PCK mRNA accumulation. Insulin prevented the increase in protein binding, which correlated with the decrease in PCK mRNA levels (20,21). In FTO-2B cells cAMP caused reduced binding of a protein to the 3Ј-UTR of PCK mRNA (22).
It was the goal of this investigation to identify and characterize protein(s), which might contribute to the regulation of PCK mRNA degradation. By gel retardation assays a cytosolic protein binding to PCK mRNA was identified, which was purified and biochemically characterized. Peptide sequence analysis showed identity of the protein with rat ferritin light chain. The binding of ferritin to RNA was regulated by phosphorylation/dephosphorylation and by iron. Competition experiments showed that ferritin did not bind preferentially to PCK mRNA but to a variety of different RNAs. It is suggested that ferritin may play a regulatory role in PCK mRNA degradation as well as in the general cytosolic mRNA turnover. The identification of ferritin as RNA-binding protein is discussed in view of an old hypothesis and recent publications, which proposed ferritin to play a role in the general metabolism of cytosolic RNA (23,24).

Primary Cultures of Rat Hepatocytes and Induction Experiments-
Liver cells were isolated from male fed Wistar rats (200 -250 g) by the collagenase perfusion technique. Cells were seeded on 6-cm diameter plastic culture dishes (Greiner, Nü rtingen) and maintained in culture in the presence of 100 nM dexamethasone (Sigma, Deisenhofen) and 0.5 nM insulin (Sigma, Deisenhofen) to improve cell viability as described (13). After 48 h of culture, cells were washed twice with medium containing only 100 nM dexamethasone. Fresh medium containing in addition 0.1 nM glucagon (Serva, Heidelberg) was added and the culture was continued for another 2 h. Then insulin at 10 nM and/or the phosphatase inhibitor okadaic acid (ICN, Meckenheim) at 0.5 M and/or the transcriptional inhibitor actinomycin D at 10 g/ml (Sigma, Deisenhofen) were added for another 3 h. Finally the cells were harvested for preparation of protein extracts.
Preparation of Cytosolic Extracts from Cultured Rat Hepatocytes-Cells from one culture dish were washed twice with ice-cold phosphatebuffered saline, scraped off in 800 l of homogenization buffer containing 10 mM Tris/HCl, pH 7.4, 250 mM sucrose, 0.5 mM CaCl 2 , 0.5 mM dithioerythreitol (Serva, Heidelberg), 0.5 mM phenylmethylsulfonyl fluoride (Serva, Heidelberg), and homogenized in a glass homogenizer with a motor-driven Teflon pestle for 6 min at maximal speed. The homogenate was centrifuged at 1,500 ϫ g for 20 min, the resulting supernatant at 8,500 ϫ g for 15 min, and the then resulting supernatant at 120,000 ϫ g for 1 h to obtain a post-polysomal, cytosolic extract (S120 extract). After heating to 50°C for 15 min the crude extract was centrifuged at 14,000 ϫ g for 15 min (Fig. 3). The resulting supernatant was brought to 35% saturation with ammonium sulfate and stirred on ice for 30 min. After centrifugation of the slurry for 20 min at 14,000 ϫ g the supernatant was brought to 65% saturation with ammonium sulfate and stirred again on ice for 15 min. The slurry was centrifuged at 14,000 ϫ g for 20 min, the pellet was suspended in dialysis buffer, which contained 10 mM Tris/HCl, pH 7.4, 10% glycerol, 0.5 mM each of dithioerythreitol and phenylmethylsulfonyl fluoride and the extract was dialyzed (cut off M r 3,500, Spectrum Medical Industries) overnight at 4°C (cytosolic extract CE) (Fig. 3). Protein concentration was determined by a commercial kit (Bio-Rad, Mü nchen). The extract was stored at Ϫ70°C. The same procedure was performed for the preparation of cytosolic extracts from whole rat liver or other organs as indicated. In this case tissue was homogenized in 10 volumes of homogenization buffer.
Purification of a RNA-binding Protein from Rat Liver-3.5 g of swollen DEAE Sephadex A-25 (Pharmacia LKB Biotechnologies Inc., Freiburg) pre-equilibrated with buffer A (10 mM Tris/HCl, pH 7.4, and 0.5 mM EDTA) was incubated in a batch procedure for 30 min at 4°C with 10 ml of cytosolic extract (CE) containing about 100 mg of total protein. Buffer A containing unbound proteins was collected and combined with buffer A resulting from two subsequent washes of the gel. Bound protein was eluted by washing with buffer B (10 mM Tris/HCl, pH 7.4, 0.5 mM EDTA, and 1 M NaCl). Eluted protein and the unbound material were concentrated separately by ultrafiltration (Centriplus 100, Amicon Inc., Beverly, MA). The unbound protein fraction called FT1 (for historical reasons because this procedure was processed in a column at the beginning of the study) was further processed with fast protein liquid chromatography on the anion exchange matrix Source 30 Q (Pharmacia, Freiburg) in an HR 5/5 column (Pharmacia, Freiburg) (Fig. 3). The column was equilibrated with three successive washes with buffer A. The sample was loaded onto the column and the resin washed with buffer A until protein was no longer detectable in the flow-through (FT2). Bound protein was eluted in a linear gradient ranging from 0 to 1 M NaCl at a flow rate of 0.5 ml/min. Fractions (1 ml) were assayed for RNA binding activity in the gel retardation assay as described below. RNA binding activity containing fractions were pooled (range of fractions 6 -10) (Source extract SO), concentrated, and equilibrated with buffer A by ultrafiltration on Centricon 100 and loaded onto a Superdex 200-pg column (Column volume, V c ϭ 135 ml) equilibrated with buffer A (Fig. 3). Elution was performed in buffer A at a flow rate of 0.25 ml/min and 0.5-or 0.6-ml fractions were collected and assayed for RNA binding activity (range of fractions 36 -38) (Superdex extract). For molecular mass determination the Superdex column was calibrated with standard proteins in a gel filtration calibration kit (Sigma, Deisenhofen). SDS-polyacrylamide electrophoresis under reducing conditions revealed that the Superdex extract contained one prominent protein band and only minor contaminations. Therefore it was used to raise an antiserum against the purified RNA-binding protein in rabbits.
Preparative Isoelectric Focusing-200 -250 mg of protein in the S120 extracts was incubated in a final volume of 50 ml in focusing puffer (10 mM Tris/HCl, pH 7.4, 2 M urea, 2% Ampholyte pH 3-10) and filled into the precooled (4°C) Rotofor ® cell (Bio-Rad, Mü nchen). Isoelectric focusing was performed at constant power of 12 watts for 5-6 h. During this time voltage increased from 290 to 900 V and the current decreased from 41 to 13 mA. At the end of the run samples were fractionated as described in the manufacturer's manual and pH was measured in each sample. To separate ampholytes from protein, the samples were incubated in 1 M NaCl for 30 min at 4°C. Then they were dialyzed against 10 mM Tris/HCl, pH 7.4, 0.5 mM EDTA, and 10% glycerol overnight. Each fraction was assayed for RNA binding activity as described below.
Five different focusing buffers were tested containing CHAPS, urea, glycerol, digitonin, and/or Tris/HCl at different concentrations and combinations to minimize denaturation of protein during the focusing procedure. The Tris/HCl buffer described here yielded the best results.
Preparation by PCR of Double-stranded DNA Templates for in Vitro Transcription-Plasmid pBS-PCK722 containing a 722-base pair PstI-EcoRI PCK cDNA fragment, which represented 120 translated and 602 untranslated bases of the 3Ј-portion of PCK mRNA (bases 1890 -2611, Ref. 14), was used as the template for the production of PCR products. To obtain various DNA templates for the generation of PCK transcripts of different lengths and regions from the PCK mRNA 3Ј-UTR the following primers were used for PCR: sense primer 1 and 2 contained at the 5Ј-end the T7-RNA polymerase promotor sequence (5Ј-TAATAC-GACTCACTATAG-3Ј) followed by genuine PCK mRNA sequences from bases 2010 to 2030 and 2265 to 2291 (14). Antisense primers 1 and 2 contained genuine PCK mRNA sequences from bases 2235 to 2259 and 2583 to 2603 (14). The combinatorial use of the primer pairs sense primer 1 and antisense primer 2, sense primer 1 and antisense primer 1, sense primer 2 and antisense primer 2 yielded PCR products PCK1, PCK2, and PCK3, which contained either the total 3Ј-UTR (bases 2010 -2603), the 5Ј-portion (bases 2010 -2259), or the 3Ј-portion (bases 2265-2603) of the PCK mRNA 3Ј-UTR (14). PCR was performed with 1 ng of pBS-PCK722 and 80 pmol of each primer in 10 mM Tris/HCl buffer, pH 8.3, containing 50 mM KCl, 1.5 mM MgCl 2 , 1 mM of each dCTP, dTTP, dATP, and dGTP and finally 1 unit of Taq DNA polymerase. PCR was performed in 35 cycles (45 s at 95°C, 45 s at 55°C, 60 s at 72°C). The resulting PCR products were taken as templates for a second round of PCR under the same conditions. These PCR products were purified by a commercial kit (Qiagen, Dü sseldorf) and the DNA content was measured spectrophotometrically.
In Vitro Transcription-The PCR products PCK1, PCK2, and PCK3 were used as templates to generate PCK1, PCK2, and PCK3 transcripts. Transcription reactions were carried out with 10 nM PCR products (see above) in a final volume of 40 l in transcription buffer (Stratagene, Heidelberg), containing 0.5 mM adenosine-, cytidine-, and guanosine 5Ј-triphosphate each, 0.025 mM uridine 5Ј-triphosphate, 55 pmol of uridine 5Ј-[␣-32 P]triphosphate (600 Ci/mmol) (ICN, Meckenheim), 10 mM dithioerythreitol, and 25 units of human placental ribonuclease inhibitor (Amersham, Braunschweig). The reaction was started by addition of 25 units of T7 RNA polymerase (Stratagene, Heidelberg). After incubation for 45 min at 37°C another 25 units of T7 RNA polymerase were added and the reaction was continued for 45 min at 37°C. The reaction was terminated by adding 10 g of tRNA (Boehringer, Mannheim) and 25 units of human placental ribonuclease inhibitor. Subsequently template DNA (PCK1, 2, or 3) was digested with 1 unit of RNase-free DNase I (Boehringer) for 15 min at 37°C. The mixture was brought to 300 mM sodium acetate and 8 mM dithioerythreitol; the volume was adjusted to 70 l with diethyl pyrocarbonatetreated water. Transcripts were purified by extraction of proteins with 70 l of buffered phenol (Amresco, Solon), pH 4.3. Phases were separated by centrifugation for 5 min in a microcentrifuge and the aqueous phase was applied to a Nick column (Pharmacia, Freiburg) to remove unincorporated nucleotides. Human histone H1°and mouse arylsulfatase A transcripts were generated by the same procedure using plasmid DNA kindly provided by Drs. D. Doenecke (Göttingen) and V. Gieselmann (Kiel).
Gel Retardation Assay-Standard binding reaction was carried out in a final volume of 40 l with 40 g or as indicated total protein and about 5 ng of 32 P-radiolabeled RNA in binding buffer containing 10 mM Tris-HCl, pH 7.4, 100 mM KCl, 3 mM MgCl 2 , 3 mM EDTA, and 40 g of RNase A (Boehringer, Mannheim). Samples were incubated for 90 min at 37°C. After addition of 5 l of electrophoresis buffer containing 10% glycerol and 0.01% bromphenol blue, reaction mixtures were separated on a 5% native polyacrylamide gel (34 ϫ 16 cm) for 2-3 h at 20 mA (about 6 volt/cm) at room temperature. After electrophoresis gels were dried and exposed to Hyperfilm MP (Amersham, Braunschweig) in general overnight at Ϫ70°C or analyzed with a PhosphorImager (Molecular Dynamics, Krefeld). Competition experiments were carried out by addition of unlabeled competitor RNAs to the binding reaction 3 min prior to the addition of the labeled transcript.
In Vitro Phosphorylation and Dephosphorylation-In the phosphorylation reaction 40 g of protein in the CE was treated with 80 units of the catalytic subunit of the cAMP-dependent protein kinase (Sigma) for 30 min at 37°C in 40 l of phosphorylation buffer containing 10 mM Tris/HCl, pH 7.4, 4.8 mM MgCl 2 , and 0.85 mM ATP. In the dephosphorylation reaction 40 g of protein in the CE was treated with 7 units of calf intestinal alkaline phosphatase (U. S. Biochemical Corp., Cleveland, OH) for 30 min at 37°C in 40 l of dephosphorylation buffer containing 10 mM Tris/HCl, pH 7.4, and 0.01 mM ZnCl 2 . Controls were performed without addition of kinase or phosphatase.
Detection of Phosphorylated Amino Acids by Anti-phosphoamino Acid Antibodies-1 g of protein in the Superdex extract ( Fig. 4) was incubated with 0.5 or 1 g of anti-phosphoserine, anti-phosphothreonine (Biomol, Hamburg), or anti-phosphotyrosine antibody (Amersham, Braunschweig) or with an unrelated antibody in 10 mM Tris/HCl, pH 7.4, and 0.05% Triton X-100 in a final volume of 10 l at 4°C overnight. After the incubation the mixture was submitted to the standard RNA binding assay.
Immunoprecipitation of Ferritin with an Anti-ferritin Antibody-40 mg of protein A coupled to Sepharose CL-4B beads (Pharmacia, Freiburg) was swollen in 1.5 ml of TS buffer (15 mM Tris/HCl, pH 7.4, 150 mM NaCl) for 4 h at room temperature. The swollen gel was collected by centrifugation and washed 3 times in TS. The final pellet (200 l) was resuspended in 200 l of TS and combined with 20 l (90 g) of anti-ferritin antiserum. The mixture was incubated overnight at 4°C and subsequently for 4 h at room temperature. 60 l from this mixture were combined with 40 g of protein in the CE and incubated overnight at 4°C. Precipitates were pelleted by centrifugation at 14,000 ϫ g. The supernatant was taken off (supernatant 1) and the pellet was washed 2 times with 30 l of TS. The resulting supernatants were combined with supernatant 1. 25 g of protein in this ferritin-free CE was applied to the gel retardation assay.
Miscellaneous-SDS-polyacrylamide gel electrophoresis was performed according to Laemmli (25). Mass spectroscopy and sequencing of tryptic peptides of the purified protein were conducted by Drs. B. Schmidt and T. Selmer (Göttingen, Germany). Antiserum against the purified protein from the Superdex extract was generated in rabbits by Eurogentec (Seraing, Belgium).

Characteristics of Cytosolic Protein
Binding to PCK mRNA-To identify and characterize proteins, which might be involved in the regulation of PCK mRNA degradation, a gel retardation assay was established using cytosolic protein extracts from rat liver and a 32 P-labeled transcript, which was produced by in vitro transcription from a PCR-generated DNA fragment containing the entire 3Ј-untranslated region of PCK mRNA (PCK1 transcript, 594 bases long; position 2010 to 2603 in PCK mRNA; Ref. 14).
When the binding reaction was performed in the absence of RNase A two dominant RNA-protein complexes (complex 1 and 2) formed with PCK1 transcript and cytosolic protein(s) from rat liver. While formation of complex 2 did not require the presence of Mg 2ϩ complex 1 formed only in the presence of both Mg 2ϩ and EDTA (Fig. 1A, lanes 3-5 versus lanes 4 -6). Addition of RNase A yielded the specific formation of only complex 1 alone, essentially only in the presence of equimolar concentrations of a divalent cation (Mg 2ϩ or Mn 2ϩ or Ca 2ϩ or Zn 2ϩ ) and EDTA as shown representatively for Mg 2ϩ /EDTA (Fig. 1A, lane 10 versus lanes [7][8][9]. This complex was only obtained with protein from rat liver cytosol and not with bovine serum albumin or with nuclear protein extracts (Fig. 1A, lanes 11 and 12).
Complex formation required incubation at 37°C for at least 15 min (Fig. 1B, lanes 4 -6) and did not occur at 4°C (Fig. 1B,  lanes 1-3), indicating that a temperature-sensitive step was involved in RNA-protein complex formation. Heating the cyto-solic extract to 70°C abolished complex formation (not shown). Pretreatment of the extracts with proteinase K did not destroy the RNA binding activity of the protein (Fig. 1B, lane 11). Pretreatment of the extracts with a protease mixture containing proteinase K, trypsin, and Pronase E prevented complex formation (not shown). These data show that complex 1 was due to an interaction of protein(s) with PCK1 transcript.
RNA-protein interactions can be expected to be dependent on salt concentration and pH. Binding of the rat liver cytosolic protein to PCK1 transcript was still observed at 0.4 M KCl but was lower compared with 0.1 M KCl (Fig. 1B, lane 6 versus 7). Protein binding to PCK1 transcript was stronger at pH 4.0 compared with pH 9.0 (Fig. 1B, lanes 8 -10). Preparative isoelectric focusing (IEF, "Experimental Procedures") enriched the RNA binding activity in fractions between pH 4.0 and pH 5.2, which showed that the RNA-binding protein was acidic.
The binding activity containing IEF fractions were used for an estimate of the binding affinity of the protein to RNA. Increasing the amount of protein from 10 to 100 g at a constant amount of the PCK1 transcript (16 fmol) increased complex formation in a sigmoidal manner. Half-maximal protein binding was reached at 38 g, maximal binding at 80 g of protein (not shown). The binding protein from rat liver (which was identified as rat ferritin light chain, see below) was en- Specificity of Cytosolic Protein Binding to RNA-Binding of cytosolic protein to PCK1 transcript should be hindered in the gel retardation assay by RNA molecules, which share sequence and/or structure similarities with PCK1 transcript. To study the binding preference of the protein for different RNAs, the gel retardation assay with PCK1 transcript was performed in the presence of single-stranded RNA homopolyribonucleotides (poly(A), poly(C), poly(G), and poly(U)) ( Fig. 2A, upper panel).
In the presence of poly(U) (500 and 1000-fold excess by weight) binding of protein to PCK1 transcript was abolished and with tRNA (2000 and 4000-fold molar excess) binding was effectively attenuated. Poly(A) competed for protein binding to a lesser extent than poly(U), whereas poly(C) exhibited weak and poly(G) no competition for binding. Binding to labeled PCK1 transcript was also investigated in the presence of unlabeled transcripts of PCK1, PCK2, and PCK3, which contained either the entire 3Ј-UTR of PCK mRNA (PCK1 bases 2010 -2603) or the 5Ј (PCK2 bases 2010 -2259) or the 3Ј (PCK3 bases 2265-2603) (14) part of it in a 150 -1200-fold molar excess (Fig. 2,  bottom part). With all three unlabeled transcripts binding of cytosolic protein to PCK1 transcript was effectively attenuated. However, when the PCR DNA fragments, which were used as templates for the generation of the transcripts, were included in a 500-fold excess by weight in the binding reaction, no interference with protein binding was observed.
Protein binding to PCK1 transcript was compared with binding to the 3Ј-UTR of the human histone 1°(HIS) mRNA, which encodes a nuclear protein, and a portion of the 3Ј-UTR of the mouse arylsulfatase A mRNA, which encodes a lysosomal protein. With cytosolic protein from rat liver HIS and arylsulfatase A transcripts formed a complex, which had the same electrophoretic mobility as the complex with PCK1 transcript (Fig.  2B).
The tissue distribution of the cytosolic RNA-binding protein was studied. Cytosolic protein extracts from rat liver (LI), kidney (KI), spleen (SP), and lung (LU) were applied to the gel retardation assay with the PCK1 transcript. Whereas rat liver extracts exhibited strong protein binding to the PCK1 transcript, extracts from kidney, spleen, and lung showed weak but significant binding forming a complex with slightly higher mobility in the case of spleen and lung (Fig. 2C).
These data show that the cytosolic RNA-binding protein was prevalent in rat liver and that the protein bound specifically to all transcripts investigated but not to DNA. The fact that poly(U) and to a lesser extent poly(A) competed for protein binding might indicate that single-stranded A-and/or U-rich sequences were involved in the contact of the protein to RNA, which are present in multiple regions in the 3Ј-untranslated part of PCK mRNA (14).
Purification of the RNA-binding Protein and Identification as Ferritin-The RNA-binding protein was purified by sequential chromatography of cytosolic protein of rat liver on the weak anion exchange resin DEAE 25, the strong anion exchange material Source Q, and the final gel filtration over Superdex 200-pg column (Fig. 3). The enrichment of the RNA-binding protein from the CE was assessed in the gel retardation assay with PCK1 transcripts by counting the radioactivity in the The Superdex-purified native protein was processed for sequencing by carboxymethylation under reducing conditions and was subsequently digested with 1% trypsin. After fractionation of the tryptic digest by high performance liquid chromatography a suitable peptide was chosen for sequencing. Sequencing was terminated after completion of 9 amino acids. The resulting peptide sequence ALFQDVQKP was used for a homology search in the Swiss protein sequence data bank. The search revealed sequence identity with amino acids 76 -84 in the rat ferritin light chain (Swiss-Prot AC: P02793).
Gel retardation assays were performed with commercial ferritin (Sigma; 99% pure) and PCK1 transcript and compared with gel retardation assays with the protein purified here from rat liver. Both protein preparations formed a protein-RNA complex with PCK1 transcript, which had the same electrophoretic mobility. Complex formation with the commercial ferritin required the same divalent cation, EDTA, and pH conditions as the purified ferritin from rat liver (data not shown). The purified protein was used to generate a polyclonal antibody. In Western blots this antibody detected both the commercial ferritin and the ferritin purified in this study.
The antibody was incubated overnight at 4°C with the CE or the purified protein preparation (Fig. 4A, lanes 1-4). In gel retardation assays with the preincubated protein preparations and with PCK1 transcript, complex 1 did not form, indicating that the ferritin antibody prevented protein binding to PCK1 transcript. To show that this was not due to an interaction of the antibody with PCK1 transcript, which could unspecifically prohibit complex formation, ferritin was immunoprecipitated with Sepharose CL-4B protein A-coupled antibody from the CE. Depletion of ferritin from CE by immunoprecipitation with protein A abolished complex formation with the PCK1 transcript (Fig. 4B, lanes 1 and 2) demonstrating an interaction of ferritin with PCK1 transcript.
An Additional Factor Besides Ferritin Was Necessary for Efficient RNA-Protein Complex Formation-Ferritin formed a RNA-protein complex with PCK1 transcript. This complex was stronger with binding activity containing fractions from the anion exchange chromatography compared with the purified Superdex 200-pg fraction, indicating the loss of a putative factor, which promoted complex formation between ferritin and the PCK1 transcript (Fig. 3). 15 g of protein of cytosolic protein extract were subjected to the gel retardation assay with PCK1 transcript and increasing amounts of Superdex extract fractions 36 -38 (Fig. 5A). The combination of 15 g of protein CE with 4.8 g of protein Superdex extract stimulated complex formation 10-fold compared with Superdex extract alone. This implies that the CE contained a component, which was lost during gel filtration. Therefore the gel retardation assay using PCK1 transcript and 1 g of protein Superdex extract was completed with fractions from the Superdex chromatography following the Superdex extract fraction 37 (Fig. 5B). Addition of an aliquot derived from the seventh fraction after the ferritincontaining fraction (i.e. fraction 44 of total fractions) increased complex formation 3-fold. This fraction corresponded to a molecular mass range of about 180,000 Da. It obviously contained a factor capable of stimulating complex formation between ferritin and PCK1 transcript. In the absence of ferritin, protein from this fraction alone did not form any complex with PCK1 transcript. The nature of the factor remains at present unknown because it had not been further purified and characterized. It is, however, clear that the factor does not act on ferritin binding to RNA by increasing the amount of binding protein but by enhancing its affinity or its activity. This can be concluded because protein binding of the combined 4.8 g of Superdex extract and of 15 g of CE was not additive to protein binding of each of the fractions alone (Fig. 5A). Moreover, in the experiment described in Fig. 5B the amount of protein in Superdex extract was kept constant (1 g) and only the addition of fraction 44 increased protein binding.
In Vitro Phosphorylation/Dephosphorylation Regulated Complex Formation between Ferritin and RNA-If ferritin should play a role in the regulation of mRNA turnover, it is reasonable to assume that its binding to RNA should be regu-  4). In the control, extracts were incubated in the absence of antibody (lanes 1 and 3). B, in a second experimental approach the antiferritin antibody was coupled to Sepharose CL-4B immobilized protein A as described under "Experimental Procedures" and used for depletion of ferritin from 25 g of protein of rat liver CE prior to the gel retardation assay with PCK1 transcript (lane 2). In the control, the extract was only incubated with protein A (lane 1). 5. Enhancement of RNA-protein complex formation by an  unknown factor. A, increasing amounts (0.3-4.8 g of protein) of Superdex extracts (SD) were added to the gel retardation assay with PCK1 transcript and with 15 g of protein of CE. B, fractions from the Superdex chromatography were tested in the gel retardation assay with PCK1 transcript and the ferritin containing Superdex fractions 36 -38 for the activity, which stimulated RNA-protein complex formation. A representative experiment is shown with Superdex fraction 44, which contained the enhancing activity. All other tested fractions were inactive. lated by phosphorylation/dephosphorylation. To confirm this assumption, the purified ferritin was treated in vitro in the presence of ATP with the catalytic subunit of cAMP-dependent protein kinase prior to the binding reaction. This treatment abolished protein binding to PCK1 transcript (Fig. 6A, lanes 1  versus 2). Treatment with alkaline phosphatase prior to the binding reaction enhanced binding of the purified ferritin to PCK1 transcript (Fig. 6A, lanes 3 versus 4). The loss of RNA binding activity of ferritin after treatment with the catalytic subunit of cAMP-dependent protein kinase could be restored by subsequent treatment with alkaline phosphatase (not shown). The data show that the RNA binding activity of ferritin purified from rat liver was regulated by phosphorylation/ dephosphorylation.

FIG.
To specify which amino acid residues were phosphorylated, the purified ferritin was incubated overnight with specific antibodies against phosphoserine, phosphothreonine, and phosphotyrosine or with an unrelated antibody. After this incubation the gel retardation assay was performed with the PCK1 transcript. Incubation with the anti-phosphotyrosine antibody generated a "supershift" (Fig. 6B) indicating that the ferritin contained phosphorylated tyrosine residues. Binding of the antibody to phosphotyrosines did not interfere with protein binding to RNA. In contrast, preincubation with the anti-phosphoserine or anti-phosphothreonine antibodies decreased complex formation (Fig. 6B) indicating that binding of the antibodies to phosphorylated serine and threonine residues in the protein interfered with the binding to RNA. Apparently, binding of the antibody to phosphoserine or phosphothreonine masked the RNA-binding region. Unrelated antibodies did not affect protein binding (not shown).
Inhibition of Protein Phosphatases Decreased Binding of Ferritin to RNA in Vivo-In cultured rat hepatocytes glucagon increased PCK mRNA levels transiently to a maximum after 2 h. When insulin was given at the maximal increase at 2 h it accelerated PCK mRNA degradation (13). The serine/threonine phosphatase inhibitor okadaic acid given at the maximal increase at 2 h inhibited the degradation of PCK mRNA in the absence and presence of insulin, indicating the involvement of dephosphorylation events in the control of PCK mRNA degradation. 2 Therefore, to demonstrate that phosphorylation and dephosphorylation may also play a role in the regulation of RNA binding activity of ferritin in vivo, extracts were prepared from 48-h cultured rat hepatocytes treated for 2 h with glucagon and thereafter in addition for another 3 h with insulin in the absence and presence of okadaic acid. Ferritin in cytosolic extracts from okadaic acid-treated cells, irrespective whether treated simultaneously with insulin or not, possessed lower binding activity to the PCK1 transcript than protein from non-okadaic acid-treated cells (Fig. 7). This implies also that in vivo phosphorylation of the protein decreased and dephosphorylation increased binding to the PCK mRNA. Ongoing gene transcription did not play a role in the regulation of protein binding to PCK1 transcript, because addition of the transcriptional inhibitor actinomycin D to the hepatocytes did not affect formation of complex 1. The half-life time of ferritin in the liver was determined at 50 -70 h (27). Therefore, it is unlikely that under the chosen experimental conditions, i.e. preparation of extracts 3 h after application of insulin, okadaic acid, or actinomycin D, changes in binding activity were due to changes in ferritin content of the hepatocytes. It can be assumed that the decrease in protein binding to PCK1 transcript was rather due to the decrease in the affinity or activity of ferritin binding to RNA.
Iron Decreased Cytosolic Protein Binding to RNA in Vitro-Ferritin is the iron storage protein in the liver. This might indicate that complex formation with the purified protein and RNA can be regulated by iron. Therefore the gel retardation assay with the PCK1 transcript was performed with purified rat liver ferritin (Superdex column fractions 36 -38) after preincubation for 30 min at 37°C with increasing concentrations of iron or the iron chelating agent desferal (Fig. 8). Increasing iron concentration attenuated binding of ferritin to the PCK1 transcript in a concentration-dependent fashion. Yet, desferal did not affect protein binding to the PCK1 transcript.

DISCUSSION
Ferritin as RNA-binding Protein-The expression of PCK is in part regulated at the level of mRNA degradation (13,18,19). To find regulatory proteins which might be involved in this process, the 3Ј-UTR of PCK mRNA was used as target for the selection of RNA-binding proteins from rat liver. By this procedure a RNA-binding protein was purified and identified as rat ferritin light chain. The RNA-protein complex formed not only with transcripts containing PCK mRNA 3Ј-UTR sequences but with a variety of different mRNA 3Ј-UTR species (Fig. 2). Computer alignment with these different RNAs revealed no sequence homologies, which could be the common target sequence for protein binding. This indicates that complex, so far unidentified, RNA structures might be necessary for RNA-protein complex formation. Also searching the SWISS protein data bank for consensus peptide motifs in ferritin, which could be putative contact sites for RNA, revealed no similarities with known RNA-binding peptide motifs, e.g. with the RNP motifs in heterogeneous nuclear ribonucleoproteins (28). However, it is clear that ferritin functions as a RNA-  7. Regulation of protein binding to the 3-UTR of PCK mRNA by phosphorylation/dephosphorylation in cultured rat hepatocytes. CE were prepared from cultured rat hepatocytes, which were untreated (C), treated for 3 h (see below) with 0.5 M okadaic acid alone (C/Oka), 10 nM insulin alone (Ins), or in addition, 0.5 M okadaic acid (Ins/Oka) or 10 g/ml actinomycin D (Ins/Act). Experiments were initiated by two washings of the 48-h cultured cells with medium, which contained 100 nM dexamethasone. Then fresh medium was added, which contained in addition 0.1 nM glucagon. 2 h later insulin and/or okadaic acid or actinomycin D was added to the cells and after another 3 h extracts were prepared and the gel retardation assay performed with PCK1 transcript.
binding protein with the potency to bind to a broad range of different RNA species.
RNA binding properties of ferritin and ferritin-homologs have already been described. A prosome-like particle, which was believed to play a role in protein degradation, was isolated and turned out to be composed of identical apoferritin subunits. This particle was associated with small RNA molecules, probably various tRNA species (24). These RNAs were capable of hybridizing with globin mRNA (24,29). From Artemia cysts the protein artemin was purified, which shared homology with vertebrate ferritins and probably functions as iron storage protein (30). In addition, artemin belongs to a class of RNAbinding proteins and was proposed to maintain the integrity of the dehydrated cyst (31,32).
Ferritin is the iron storage protein in liver, which protects cells against cytotoxic effects of reactive oxygen species derived from iron-catalyzed free radical reactions (for review, Ref. 33). Three different types of ferritin were described (apoferritins), the L (light) and two variants of H (heavy) chains, of which the L type is prevalent in rat liver. The apoferritins form a holoferritin complex, which consists of 24 subunits forming a very stable shell for the storage of about 4500 iron ions as FeIIIO(OH) inside the complex (34).
Early experiments using transcriptional inhibitors showed that after an iron challenge the increase in cellular ferritin concentrations were not due to an increase in gene transcription but rather to the increase in the translation of ferritin mRNA (23). Under normal, non-iron depleted conditions ferritin concentrations remain fairly constant because of the stability of ferritin, which is reflected by the long half-life time of 50 -70 h (27). Hence, the changes in binding activity of ferritin to PCK1 transcript observed in the present study both in vitro (Fig. 6) and in the experiments using cultured hepatocytes (Fig.  7) were not due to changes in protein content but either to alterations in the binding affinity of ferritin or to changes in the concentration of active RNA-binding protein.
Translation of ferritin mRNA is controlled by iron-regulated RNA-protein interactions. The iron responsive element (IRE) was identified in the 5Ј-UTR of ferritin mRNA comprising sequences capable of forming a hairpin loop, which is recognized by the iron responsive element-binding protein (IRP). Binding of IRP is highly regulated by iron. Under iron deficiency the protein occupies the IRE so that the ferritin mRNA can no longer be translated. Under iron excess IRP complexes Fe 2ϩ causing conformational changes in the IRP, which in this state no longer binds to the IRE, so that the ferritin mRNA can be translated. The Fe 2ϩ -loaded IRP functions as cytosolic aconitase (35).
Regulation of Complex Formation of RNA with Ferritin-Dephosphorylation enhanced and phosphorylation attenuated formation of the RNA-protein complex between ferritin and PCK1 transcript. Anti-phosphoserine or anti-phosphothreo-nine antibodies attenuated complex formation, whereas an anti-phosphotyrosine antibody did not interfere (Fig. 6). This suggested that phosphorylated serine and threonine but not tyrosine residues are located in close proximity to the RNAbinding site of ferritin. Indeed two potential phosphorylation sites for the serine/threonine kinases casein kinase II (peptides TEVE and SQDE) and for protein kinase C (PKC; SAR) are present in ferritin. However, the function of these sites in the regulation of RNA-protein complex formation remains to be established. Phosphorylation of serine residues in ferritin has already been described earlier (36).
In cultured rat hepatocytes the degradation of PCK mRNA was accelerated by insulin and abolished in the presence of the phosphatase inhibitor okadaic acid. 2 Whereas treatment of cultured rat hepatocytes with insulin did not change RNA-protein complex formation, treatment with okadaic acid diminished complex formation (Fig. 7). This appears to show that ferritin binding to PCK mRNA was not regulated in vivo by insulin but obviously by okadaic acid-influenced phosphorylation/dephosphorylation. The diminution of complex formation in the presence of okadaic acid and the inhibition of PCK mRNA decay by okadaic acid might indicate a functional correlation between ferritin binding and mRNA degradation. In this hypothetical model dephosphorylated ferritin would function as a stimulator of PCK mRNA degradation.
In FTO-2B rat hepatoma cells cAMP stabilized PCK mRNA against degradation (18,19). A cytosolic protein had been identified, which bound specifically to different 30-mers of the PCK mRNA 3Ј-UTR in vitro (22). Protein binding was diminished after phosphorylation of the protein and after treatment of FTO-2B rat hepatoma cells with cAMP. Conversely, dephosphorylation after in vitro treatment of the cytosolic proteins by alkaline phosphatase enhanced RNA-protein complex formation (22). In the present study binding of ferritin to RNA was also diminished after phosphorylation and augmented after dephosphorylation. However, the protein described in Ref. 22 is very likely different from ferritin because of various biochemical differences. The binding activity of the protein from FTO-2B hepatoma cells was increased by KCl, it bound to specific sequences in the 3Ј-UTR of PCK mRNA, partial purification lead to the loss of an inhibitory activity and finally the molecular mass was determined at 100 kDa.
Evidence increases that cellular signal transduction pathways involving phosphorylation and dephosphorylation cycles play an important role in the control of cytosolic mRNA metabolism like mRNA degradation (for review, see Ref. 37). Various different mRNAs were destabilized in T-cells (38) but granulocyte macrophage-colony stimulating factor mRNA was stabilized in thymoma cells (39) by protein kinase C activation with phorbol ester. Phosphorylation following serine/threonine phosphatase inhibition by okadaic acid diminished C heterogeneous nuclear ribonucleoprotein binding to pre-mRNA (40). IRE binding activity was increased by phosphorylation of the IRE-binding protein by protein kinase C (41).
Functional Significance of Ferritin as RNA-binding Protein-1 g of liver contains a cytoplasmic hepatocellular volume of 0.7 ml and 1.5 nmol of ferritin (see above and Ref. 27). Therefore, the ferritin level in hepatocytes is 1.5 nmol/0.7 ml ϭ 2.1 M. In the assay PCK1 transcript was half-saturated at 0.12 M ferritin (cf. "Results"), which is more than 10-fold lower than the total cellular ferritin concentrations. However, the total ferritin level is probably clearly higher than that freely available for RNA binding. Because ferritin is associated with different cellular compartments, which are not freely interconvertable, it cannot exactly be defined, which is the level of ferritin employed in iron storage or in other cellular functions FIG. 8. Inhibition of RNA-protein complex formation by iron. 0.5 g of protein of the Superdex extract (SD) was preincubated with increasing concentrations (1, 2, and 3 mM) of ferric ammonium citrate (ferricAC) or with increasing concentrations (3, 6, and 9 mM) of the iron chelator desferal prior to the gel retardation assay with PCK1 transcript.
(27). The half-saturation concentration determined may be in the range of "free" ferritin. Therefore, it is feasible that binding of ferritin to RNA is of physiological relevance.
Some evidence exists that ferritin via its RNA-binding capacity might play a role in the general metabolism of RNA. For example, addition of ferritin-containing prosome-like particles to an in vitro translation system inhibited the translation of globin mRNA, indicating that ferritin functioned as a translational repressor (24,29). This had also been suggested in earlier reports, which claimed that ferritin bound to the 5Ј-UTR of its own mRNA and stored the mRNA in a translational incompetent form (42). Due to its similar amino acid composition and immunological cross-reactivity the RNA-binding protein from Artemia cysts had been suggested to be related to the elongation factor eEF-Ts, which, however, remains to be experimentally established (31). These data suggest that ferritin, besides its function as iron storage protein, may not have a unique but rather a general function in the regulation of RNA metabolism at various levels (23,24).
The present study had the goal to identify protein(s), which might be involved in the regulation of PCK mRNA degradation. Ferritin was isolated and its RNA binding characteristics determined ( Figs. 1-4). Because no specific binding to PCK mRNA but rather a general binding to mRNAs was detected, it seems likely that ferritin is not only involved in PCK mRNA metabolism but in general processes, which, however, at present cannot be defined. It might be relevant that in the present study RNA-protein complex formation was inhibited by iron (Fig. 8). This might link the iron loading status of the cell to the regulation of general RNA degradation.