Endonucleolysis in the turnover of insulin-like growth factor II mRNA.

The overlapping transcription units constituting the rat insulin-like growth factor II (IGF-II) locus generate multiple mRNAs by using different promoters. Three promoters have been identified, giving rise to 4.6-, 3.8-, and 3.6-kilobase mRNAs. The latter, originating from promoter P3, is the most abundant IGF-II mRNA in the rat liver cell-line BRL-3A. Moreover, a non-polyadenylated 1.2-kilobase (kb) transcript and a 1.8-kb tail fragment are prominent transcripts at steady-state. In this study, we show that the 1.8-kb tail fragment is uncapped and sediments as a 30 S ribonucleoprotein particle, and is thus not actively engaged in protein synthesis. In contrast, both the 3.6-kb mRNA and the 1.2-kb transcript cosediment with polysomes. In the presence of cytoplasmic extracts, the full-length 3.6-kb mRNA is cleaved into the 1.8-kb tail fragment and a similar-sized upstream fragment. The cleavage occurs between a putative hairpin and a phylogenetically conserved guanosine-rich region which forms a stable higher order RNA structure in the presence of K+. We suggest that endonucleolysis is the initial step in IGF-II mRNA decay and that this event may participate in the post-transcriptional regulation of IGF-II production.

* This work was supported by the Carlsberg and Novo Foundations and the Danish Natural Science and Medical Research Councils. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed University Institute of Biological Chemistry B, Sdlvgade 83, DK-1307 Copenhagen K,  conferred by an A + U-rich sequence in the 3'-untranslated region of the mRNAs. In addition, a cytosolic protein that binds to four AUUU repeats has been identified in lymphocytes (Malter, 1989). In the case of the transferrin receptor mRNA, multiple hairpins in the 3"untranslated region of the RNA have been recognized as iron-responsive elements determining the stability of the mRNA (Mulner and Kuhn, 1988), and a binding protein with an apparent molecular mass of 90 kDa has been purified (Walden et al., 1989;Rouault et al., 1990). This implies that primary sequence, higher order RNA structure, and attached proteins may be involved in destabilizing the mRNA. However, other features may be operational, as illustrated for @-tubulin mRNA where instability of the mRNA depends on recognition of the first four amino acids of @-tubulin as they emerge from the ribosome (Yen et al., 1988).
Insulin-like growth factor I1 (IGF-11)' is a mitogenic polypeptide consisting of 67 amino acids which plays an important role in fetal growth and development (reviewed by Humbel, 1989). The single copy gene for IGF-I1 comprises six and nine exons in rat and human, respectively. In the rat, exons 4, 5, and 234 nucleotides of exon 6 provide the coding region for the prepropeptide (Fig. 1). By initiating transcription at three different promoters (Pl, P2, P3), the gene generates multiple mature transcripts of 4.6 kb (exons 2 and 4-6), 3.8 kb (exons 1 and 4-6), and 3.6 kb (exons 3 and 4-6) (Frunzio et al., 1986;Soares et al., 1986;Gray et al., 1987;Ueno et al., 1987;de Pagter-Holthuizen et al., 1988;Holthuizen et al., 1990). Additional IGF-I1 transcripts between 1.2 and 3.0 kb have been detected in fetal rat tissues and transformed cells, but their origins are uncertain. Recent studies have suggested that some of these smaller transcripts occur through the use of alternative polyadenylation signals in the primary transcript (Chiarotti, 1988;Ueno et al., 1989). Moreover, a transcript corresponding to the distal 1.8 kb of exon 6 has also been detected, and it was inferred that it originated from a new transcription unit within exon 6 of the rat IGF-I1 gene (Matsuguchi et al., 1989).
An alternative explanation for the presence of the 1.8-kb transcript, and other transcripts that do not exhibit 5' termini corresponding to the three promoters, may be that they are intermediate mRNA fragments following endonucleolytic cleavage of the abundant 3.6-kb mRNA. In this study, we have investigated the latter possibility and further characterized the translational status of endogenous rat IGF-I1 transcripts. The results demonstrate that the major 3.6-kb mRNA is translated in vivo and can be processed into two 1.8-kb transcripts by the inclusion of a cytoplasmic extract in vitro. The cleavage site is located in the trailer of the mRNA at a The abbreviations used are: IGF-11, insulin-like growth factor 11; kb, kilobase(s); SDS, sodium dodecyl sulfate; bp, base pair(s), RNP, ribonucleoprotein. FIG. 1. Linear map of the rat IGF-I1 gene. Exons are indicated by boxes and numbered according to Holthuizen et al. (1990). Filled boxes mark the coding region for the prepropeptide, and the sites of transcription initiation are indicated (PI, Pz, and Pj). Exons 1-3 encode alternative 5'-untranslated regions, and exon 6 encodes the 3'-untranslated region. Moreover, the three exons of the upstream insulin gene are indicated (Ins). phylogenetically conserved guanosine-rich region that is able to adopt a stable higher order RNA structure.

EXPERIMENTAL PROCEDURES
Preparation of Total RNA and Poly(A)+ RNA-Total cellular RNA was isolated according to Chirgwin et al. (1979), with minor modifications. Briefly, the BRL-3A cells (ATCC, CRL 1442) were dissolved in 5 M guanidinium thiocyanate, and total cellular RNA was pelleted through a cushion of 5.7 M CsC1. The RNA pellet was resuspended in 20 mM Tris-HC1, 2 mM EDTA, 0.2% SDS containing 200 pg/ml autodigested proteinase K and incubated for 1 h at 37 "C. After phenol/chloroform extraction the RNA was ethanol precipitated and redissolved in water. Poly(A)+ RNA was obtained by oligo(dT)cellulose chromatography (Aviv and Leder, 1972).
Northern Analysis-The RNA was fractionated in 1.0% agaroseformaldehyde gels (Lehrach et al., 19771, transferred to Hybond-NTM nylon membrane, UV cross-linked, and hybridized with a 32P-labeled cDNA probe generated by random priming (Feinberg and Vogelstein, 1983) or with 5'-[32P]labeled oligodeoxynucleotide probes. Hybridization with cDNA probes was performed as previously described (Nielsen et al., 1990). When using oligodeoxynucleotide probes, hybridization was performed at 58 "C for 3 h in 4 X SSPE, 5 X Denhardt's solution, 0.5% SDS, 150 pg/ml denatured salmon sperm DNA, and filters were washed for 20 min in 4 X SSPE, 0.1% SDS at 48 "C and for 20 min in 4 X SSPE, 0.1% SDS at 58 "C. Experiments using Northern analysis were carried out three times, independently.
cDNA and Oligodeoxynucleotide Probes-A cDNA probe specific for the rat preproIGF-I1 coding region and oligodeoxynucleotide probes specific for the exons 1,3, and 6 were used. The 545-bp cDNA probe extends from one nucleotide downstream of the initiation codon to three nucleotides downstream of the termination codon as previously described (Stylianopoulou et al., 1988). The oligodeoxynucleotide probes were complementary to positions 214-239 (probe sequence: GCAACGCCCAGTCCGTTGGAAGACCC) and 63-85 (probe sequence: CTGAAGTTGGATAAGGAGGCCGC) in rat leader exons 1 and 3, respectively, and to positions 2689-2712 (probe sequence: GGCAGGTAATTTAGGGTGCCTCG) in the distal part of exon 6 (Sussenbach, 1989). Moreover, a ribonucleotide probe specific for (3-actin was used as previously described (Geijer et al., 1990).
Immunoprecipitation of m'G-capped mRNAs-0.5 pg of poly(A)+ RNA in 100 p1 of NET buffer (50 mM Tris-HC1, pH 7.6, 150 mM NaC1, 1 mM EDTA, 0.1% Nonidet P-40) containing 1000 units/ml RNasin and 500 pg/ml Escherichia coli tRNA was mixed with 20 pl of monoclonal antibody HT-20 (1 mg/ml) (Bochnig et al., 1987) and agitated for 1 h at 4 "C. Then 20 pl of protein A-Sepharose (Pharmacia), precoated with affinity purified rabbit immunoglobulins to mouse immunoglobulins (Dakopatts, Denmark), were added and incubation was continued for an additional 1 h at 4 "C. Finally, the complex was precipitated by centrifugation at 14,000 X g for 1 min. The precipitate was washed three times with ice-cold NET buffer before it was resuspended in 100 pl of NET buffer. The precipitate and the supernatant were phenol/chloroform extracted, and the RNA was ethanol precipitated and subjected to Northern analysis.
Cleavage of IGF-II Transcripts-0.5 pg of poly(A)+ RNA in 100 p1 of incubation buffer (20 mM Tris-HC1, pH 8.0, 1.5 mM MgClZ, 140 mM KCl, 0.5 mM dithiothreitol, 500 pg/ml E. coli tRNA) was heated at 56 "C for 5 min followed by slow cooling to room temperature, before 1000 units/ml RNasin was added at 0 "C. Then 3 ~1 (10 fig of protein) of a post-mitochondrial supernatant from rat liver BRL-3A cells or human IN 157 rhabdomyosarcoma cells, lysed in the presence of Nonidet P-40 (see above), were added, and incubation continued for 3 and 6 h at 0 "C. At the end of the incubation samples were phenol/chloroform extracted and ethanol precipitated, and RNA was subjected to Northern analysis.
In experiments concerning the subcellular localization of the cleavage activity, polysomes were isolated from a 20-47% sucrose gradient as above. The top of the gradient was used as a source of soluble proteins. Polysomal fractions were pooled and the KC1 concentration was adjusted to 500 mM, followed by a centrifugation at 100,000 X g for 3 h at 4 "C. The supernatant was used as a source of polysomal salt wash, while the pellet was resuspended in 100 pl of lysis buffer. The endonucleolytic activity of the fractions was assayed on 0.5 pg of renatured poly(A)+ RNA from BRL-3A cells for 4 h at 0 "C as described above, but with the omission of RNasin.
Primer Extension-0.5 pmol of 5'-[32P]dTTGACCTGTGCCT-CCTCCTGACT (complementary to positions 1572-1550 in rat exon 6) and 1 pg of poly(A)+ from BRL-3A cells were heated at 95 'C for 1 min and annealed at 56 'C for 2 h in 12 pl of 10 mM Tris-HC1 (pH 6.9), 40 mM KC1,0.5 mM EDTA. The annealing buffer was changed from 40 mM in KC1 to 40 mM in NaCl when the effect of the alkali metal ion was examined. Extension was carried out for 30 min at 45 "C by adding 8 pl of a mixture containing 125 mM Tris-HC1, pH 8.4, 25 mM MgCI2, 5 mM dithiothreitol, 500 pM dNTPs, 3 units of avian myeloblastosis virus reverse transcriptase (Life Sciences, FL). Primer extension analysis of in vitro transcripts was carried out as described above, except that 1 pmol of 5' end-labeled primer was annealed with 1 pmol of RNA template in 6 pl for 20 min, and the ratios between dideoxy-and deoxynucleoside triphosphates were 0.3, 0.4, 0.4, and 0.1 for G, A, T, and C, respectively.
In Vitro Transcription-A 1156-bp BglII-XbaI fragment from the human exon 9 (positions 2312-3467) was inserted into the polylinker of an in vitro transcription vector that exploited the T7 RNA polymerase system (Studier and Moffatt, 1986). The recombinant was digested with AluI (position 2671 in the inserted human exon 9), and an in vitro transcript of 388 nucleotides was generated by the addition of T7 RNA polymerase (Milligan and Uhlenbeck, 1989). The transcript corresponds to positions 2312-2671 in exon 9 and exhibits 28 extraneous nucleotides (GGCCUGCAGGUCGACUCUAGAACU-AGUG) at the 5' terminus originating from the polylinker, and it was purified by electrophoresis in a 8% polyacrylamide, 7 M urea gel. The full-length transcript was visualized by UV shadowing and eluted with 250 mM sodium acetate, pH 6.0, 1 mM EDTA. After extraction with phenol/chloroform and precipitation with ethanol, the fulllength transcript was dissolved in water.
A 5' end-labeled transcript of 208 nucleotides was prepared in a similar way, but the BglII-XbaI fragment was inserted closer to the T7 RNA polymerase promoter so 14 extraneous nucleotides (GGCU-AGAACUAGUG) were present in the transcript. Linearization was carried out by digestion with SmaI (position 2503 in the human exon 91, and [Y-~*P]GTP was included in the in vitro transcription mixture (Milligan and Uhlenbeck, 1989). The 5' end-labeled transcript of 208 nucleotides was purified by denaturing polyacrylamide gel electrophoresis as described above.
Structural Probing with RNase TI-The purified 5' end-labeled transcript of 208 nucleotides in 20 pl of 30 mM Tris-HC1, pH 7.8, 5 mM MgC12, 100 mM KCl, containing 4 pg of E. coli tRNA, was renatured by heating at 95 "C for 1 min, followed by a 56 "C treatment for 10 min, and slowly cooling to room temperature. The renatured transcript was digested with 0.01 unit of RNase TI for 30 min at 0 "C. RNase TI was inactivated by phenol and chloroform extractions, and digests were ethanol-precipitated before gel electrophoretic analysis. The purified 5' end-labeled transcript was also digested with 0.005 unit of RNase TI (G-specific) or 0.005 unit of RNase Up (A-specific) in the presence of 4 pg of E. coli tRNA carrier essentially as described by Donis-Keller (1980).

RESULTS
Steady-state Transcripts-IGF-I1 transcripts expressed in rat liver BRL-3A cells a t steady-state were characterized by Northern analysis (Fig. 2). When a coding region probe (CR) was used the major transcripts in total RNA were 3.6 and 1.2 kb. The previously described 4.6-kb mRNA (Soares et al., 1986), originating from promoter Ps, was only apparent after very long exposures (data not shown). Hybridization with a probe specific for exon 3 (Ex3) and exon 1 showed that all the IGF-I1 transcripts, except the 4.6-kb mRNA, contained exon 3 and appeared in the same relative amounts as determined by the coding region probe. No mRNAs were identified by the exon 1 probe in BRL-3A cells. A probe complementary to the distal part of exon 6 (E&) hybridized with the 3.6-kb mRNA and a 1.8-kb transcript. Finally, a minor band of 2.4 kb was detected, which also contained the exon 3 leader and the preproIGF-I1 coding region. The extent of polyadenylation of the various transcripts in BRL-3A cells was determined by oligo(dT)-cellulose chromatography. The 1.2-kb mRNA was completely recovered in the flow-through (A)-. In contrast, the 3.6-kb mRNA and the 1.8-kb transcript from exon 6 were exclusively recovered in the poly(A)+ fraction (A)+. In brief, BRL-3A cells contain three abundant transcripts originating from the rat IGF-I1 gene: a mature polyadenylated 3.6-kb mRNA derived from promoter PB; a non-polyadenylated 1.2kb transcript that also contains both the exon 3 leader and the preproIGF-I1 coding region but lacks about 2.4 kb of exon 6; and a polyadenylated 1.8-kb tail fragment which corresponds to the distal part of exon 6. n 7 G Capping of IGF-II Transcripts-The polyadenylated 1.8-kb tail fragment could be an independent transcript or it could arise from an endonucleolytic event in the cytoplasm.  ( E x 6 ) . Autoradiography was for 8-24 h at -80 "C with intensifying screens. (Bochnig et ul., 1987). The precipitate and the supernatant were subjected to Northern analysis with a probe to the distal part of exon 6 (Fig. 3). The 3.6-kb mRNA, which acts as an internal standard, was found in both the precipitate (track 2) and the supernatant (truck 3), whereas the 1.8-kb tail fragment is exclusively found in the supernatant (truck 3). Although the binding of the HT-20 antibody could not reach equilibrium during the l -h incubation at 4 "C, the 1.8-kb tail fragment was not precipitated and is, by inference, uncapped. Therefore, it is unlikely that the latter fragment is an independent transcript.
Subcellular Localization of IGF-II Transcripts-The.translational status of the 1.8-kb tail fragment, together with that of the other IGF-I1 transcripts, was examined by a combination of sucrose gradient and Northern analyses. BRL-SA cells were solubilized by Nonidet P-40 treatment in an isotonic buffer, and the lysate was centrifuged through a 20-47% sucrose gradient in the presence of 5 mM MgC12 or 10 mM EDTA. The upper part of Fig. 4 shows the sedimentation profiles, and the lower part provides the results from the Northern analysis with the coding region probe (CR) or the exon 6 probe ( E x 6 ) . Both the 3.6-kb mRNA and the 1.2-kb transcript cosedimented with polysomes but were also found in fraction 1 which sedimented a t about 30 S. The polysomal transcripts were released by EDTA, which strongly suggests that the two transcripts are located in polysomes (Perry and Kelley, 1968;Henshaw, 1968). In contrast, the 1.8-kb tail fragment was only identified at the top of the gradient in fraction 1, which implies that this transcript is not translated.
Besides the abundant steady-state transcripts, fraction 1 (CR) from the Mg'+-containing gradient also contained the 2.4-kb transcript that was hardly detectable a t steady-state. During the preparation of ribonucleoprotein (RNP) particles, decay of the 3.6-kb mRNA was extensive, and the relative amounts of the transcripts were different from the situation a t steadystate. This observation was reinforced by the hybridization signal of fraction 1 to the exon 6 probe, where the 3.6-kb mRNA could not be detected a t this exposure. The breakdown of the 3.6-kb mRNA is selective as hybridization with a @actin probe showed that this mRNA was stable during the procedure (results not shown).
We infer that the 3.6-kb mRNA is translated, and that the 1.8-kb tail fragment is not directly engaged in protein synthesis. The translational status of the non-polyadenylated 1.2kb transcript cannot be assessed unequivocally from this experiment, since its presence in polysomal fractions may result from breakdown of the polysomal3.6-kb mRNA during preparation.
Cleavage of Nuked IGF-11 Trunscripts-Since experiments to trace a precursor-product relationship among the steadystate transcripts by subjecting cells to 10 pg/ml actinomycin D did not reveal substantial changes in the transcript pattern in uiuo during a 6-h period (data not shown), and since the rapid conversion taking place during preparation of endogenous RNP particles is difficult to control, an attempt was made to identify the initial endonucleolytic cleavage in naked full-length mRNA. Fig. 5 shows an autoradiograph from a Northern analysis of renatured poly(A)+ RNA that had been subjected to 10 pg of cytoplasmic proteins from BRL-SA cells (tracks 2 and 3 ) or from human IN 157 rhabdomyosarcoma cells (track 4 ) that also express IGF-I1 (Nielsen et al., 1990). The IGF-I1 RNAs in the added lysates would not interfere with the Northern analysis since they correspond to less than 0.1 pg of total RNA. Whereas the P-actin mRNA is inert to this treatment and exhibits a similar hybridization signal in each track, the full-length mRNA shows a smaller hybridization signal when lysate has been added for 3 and 6 h (tracks 2-4) than when renatured RNA was left for 6 h without lysate (track 1 ). In parallel with the decrease in the 3.6-kb mRNA, a new 1.8-kb upstream fragment appeared and an increase in the level of the 1.2-kb transcript was apparent. Moreover, the level of the 1.8-kb tail fragment also increased during the in uitro incubation. In addition to the major effects, minor changes involved formation of a 2.4-kb tail fragment. The cleavage pattern in uitro was apparent after 30 min, and extracts from the non-IGF-I1 producing human cell-lines HT29, HCT115, HCT116, and AGS (available from ATCC) could also generate the scission (results not shown). We conclude that the naked full-length mRNA is converted into two 1.8-kb fragments in uitro. T o determine whether the trans-acting factor(s) necessary for endonucleolysis was associated with polysomes, the latter were isolated and further fractionated into a high salt wash and a pellet. Fig. 6 shows the ability of the various fractions Endonucleolytic Cleav t o cleave the 3.6-kb full-length mRNA from BRL-SA cells.
No endonucleolytic activity, as determined by both the appearance of the 1.8-kb upstream fragment and an increased amount of the distal tail fragment, was found in the fractions derived from polysomes (tracks 3 and 4 ) . In contrast, the full activity exhibited by the lysates was recovered from the top of the sucrose gradient (track 5 ) . We infer that the transacting factor(s) involved in the cleavage of IGF-I1 mRNA is not associated with ribosomes engaged in translation.
Primer Extension Analysis-The 5' terminus of the 1.8-kb tail fragment has been identified previously in both the rat and the human by S1 nuclease mapping (Matsuguchi et al., 1989;de Pagter-Holthuizen et al., 1988). The former study also included a primer extension analysis. Based on these studies and the results obtained from the in vitro conversion here, we synthesized a 23-mer complementary to positions 1550-1572 in exon 6 from the rat IGF-I1 gene, and a corresponding 21-mer complementary to positions 2524-2544 in the human exon 9, so the putative cleavage site could be identified by primer extension analysis of poly(A)+ RNA isolated from total steady-state RNA. Both primers were 5' end-labeled, annealed to RNA, and extended by avian myeloblastosis virus reverse transcriptase in the presence of K' . Fig. 7A shows an autoradiograph of the primer extension analysis. Virtually complete termination is observed a t corresponding positions in human and rat RNA, and both occur about 60 nucleotides downstream from the reported 5' termini of the 1.8-kb tail fragment based on S1 mapping analysis. In 'age of IGF-II mRNA an attempt to elucidate this apparent discrepancy, we repeated the primer extension analysis of rat RNA in the presence of Na' to examine whether the termination was due to the formation of an alkali metal ion-sensitive higher order structure in the adjacent guanosine-rich region. The result of the experiment is depicted in Fig. 7B and shows that read-through upstream of position 1503 is feasible in the presence of Na' and that a termination point corresponding to the S1 mapping data can be observed at position 1447. Northern analysis was unable to detect any transcript conversion during the annealing step a t 56 "C for 2 h, regardless of the employed monovalent cation (data not shown).
We conclude that the guanosine-rich region at the 5' terminus of the tail fragment adopts a stable higher order RNA structure that reverse transcriptase cannot traverse in the presence of K' .
T o establish whether the presence of a full-length 3.6-kb mRNA is necessary for formation of the alkali metal ionsensitive structure, a 1156-bp BglII-XbaI restriction fragment from the human exon 9 was inserted behind a T7 RNA polymerase promoter in an in vitro transcription vector. Following linearization with AluI, an in vitro transcript of 388 nucleotides was generated. The transcript was purified by denaturing polyacrylamide gel electrophoresis, eluted, and renatured in the presence of K' or Na'. The transcript was then subjected to primer extension analysis with avian myeloblastosis virus reverse transcriptase, and the resulting cDNAs were coelectrophoresed with dideoxysequencing reactions. If annealing and extension were carried out in the presence of K' (Fig. 8, track 1 ) virtually complete termination was observed a t positions indicated by arrow b. In contrast, a considerable amount of full-length cDNA (arrow a ) was generated in the presence of Na' (track 2). These results reinforce the initial observation obtained from primer extension analysis of endogenous mRNA and imply that the alkali metal ionsensitive steady-state structure can be generated locally in vitro.
Structural Probing with Ribonuclease Tl-Since the region causing termination of reverse transcription is dominated by guanosines, the ability of each guanosine to adopt an unpaired syn conformation was assessed by carrying out partial digests with RNase T1 a t strong denaturing conditions and in the presence of Mg2' and K' (Heinemann and Saenger, 1982;Christiansen et al., 1990). A transcript containing 208 nucleotides was generated by in vitro transcription with T7 RNA polymerase and concomitantly 5' end-labeled by the inclusion of [y3'P]GTP. After the end-labeled transcript had been purified by electrophoresis, it was partially digested with RNases TI or Uz (A-specific) at the denaturing conditions employed in enzymatic sequencing (Donis-Keller, 1980), or it was renatured in the presence of 5 mM Mg' and 100 mM K' and probed with RNase T1. Fig. 9 is an autoradiograph from an electrophoretic analysis of the structural probing. In track 1 the guanosines between the A2443 and AZds0 markers (track 2) are, in general, not hit by RNase T1 at the strong denaturing conditions employed in enzymatic sequencing. The notable exceptions are G2461, G2470, G2471, and G2472. Moreover, the susceptibility of the region between positions 2443 and 2480 toward RNase T1 is independent of whether the digest is carried out under denaturing (track I ) or "native" (track 3 ) conditions. This implies that the guanosine-rich region is structured in 8 M urea a t 50 "C and pH 5.0, and that the structure is similar to that adopted under non-denaturing conditions. The susceptibility of G2461 and G2471-GZ473 toward RNase TI in the presence of Mg2+ and K' suggests that these Film exposure was 2 days for tracks 1 and 2 and 6 days for the sequencing tracks. The 2320-2500 region in human exon 9 is located 2083 nucleotides downstream from the translation termination codon. nucleotides are part of flexible loop regions .

DISCUSSION
At steady-state, the three abundant transcripts originating from the IGF-I1 locus in rat BRL-3A cells are the 3.6-kb mRNA, the 1.8-kb tail fragment, and the 1.2-kb upstream transcript. This study shows that the 3.6-kb mRNA is cleaved into two similar-sized transcripts after incubation with crude cytoplasmic lysates, in vitro. The trans-acting factor(s), that is non-ribosomal and present in lysates from both human and rat cells, cleaves the mRNA trailer at a guanosine-rich region. The resulting 1.8-kb tail fragment corresponds to a stable, polyadenylated steady-state transcript that is neither capped nor actively engaged in protein synthesis.
It has been suggested that the 1.8-kb tail fragment is generated from its own promoter within exon 6 (Matsuguchi et al., 1989) and that the corresponding human fragment is a template for protein synthesis (de Pagter-Holthuizen et al., 1988). For the rat, neither of these suggestions are compatible with our results, since the endogenous, polyadenylated, 1.8kb tail fragment is uncapped and found in an RNP sedimenting as a 30 S particle. Moreover, we show that the 1.8-kb tail 2480 -

5'
FIG. 9. Structural analysis of the guanosine-rich region with ribonuclease TI. A 5' end-labeled transcript of 208 nucleotides was partially digested with RNase T, (track 1) or RNase Uz (track 2) in 8 M urea a t 50 "C, or it was renatured in the presence of 5 mM MgCI, and 100 mM KC1 and partially digested with RNase TI at 0 "C (track 3 ) . The positions of adenosines used as markers are indicated to the left. The sequence of the guanosine-rich region is shown to the right, and boxed nucleotides are reactive a t denaturing conditions. fragment is an endonucleolytic product of the full-length 3.6kb mRNA. In human, recent transfection experiments also show that the 1.8-kb tail fragment can be generated from the trailer exon (Meinsma et al., 1991). Thus, the prime event in the decay of IGF-I1 mRNA is an endonucleolytic event and not an exonucleolytic poly(A) shortening (Bernstein and Ross, 1989). The 1.8-kb upstream fragment, which is prominent in uitro, is hardly detectable a t steady-state, so an efficient 3 / 4 5 ' exonuclease degradation probably occurs in uiuo. A possible product of such an exonucleolytic degradation is the non-polyadenylated 1.2-kb transcript, but the latter might also originate directly from an alternative endonucleolytic event, since a 2.4-kb tail fragment appears following incubation with lysates.
While there is evidence that a general 3'+5' exonuclease, associated with translating ribosomes, is involved in the processive inactivation of some mRNAs (Peltz et al., 1987), the endonucleolytic activity that cleaves IGF-I1 mRNA cannot be recovered from active ribosomes. This is in agreement with reports showing that nuclease activity is associated with mRNPs (Bandyopadhyay et al., 1990) and that translationindependent cleavage occurs in oocytes (Brown and Harland, 1990). An association of the endonuclease or other transacting factors with mRNP may explain why the 3.6-kb mRNA, in contrast to the incomplete conversion of the naked RNA in uitro, is extensively converted in the lysates, where the cleavage activity attacks its native substrate. Moreover, the activity is likely to be selective since neither endogenous nor exogenous @-actin mRNA is cleaved. The primers used in primer extension analyses of rat and human RNA are shown by lines I and 2, respectively. The horizontal arrows designate an inverted repeat which may form a hairpin in the mRNA upstream from the cleavage site, and the bracketed region is encoding the G-structure downstream from the cleavage site in mRNA. Egl I1 is the 5' end of the restriction fragment used for generating i n uitro transcripts, and SmaI is the site used for linearization in the synthesis of the i n uitro transcript of 208 nucleotides. Numbering of rat exon 6 (upper sequence) and human exon 9 (lower sequence) is according to Sussenbach (1989). and 8) and structural probing with RNase TI (Fig. 9). Loop regions are indicated with Roman numerals. The rat and human sequences of t,he depicted region are identical, except that the last uridine in rat loop 111 is a cytidine in human.
The presence of the 1.2-kb transcript on polysomes in an EDTA-releasable form suggests that it is involved in the synthesis of preproIGF-11; this result correlates with that from an in uitro translation study (Graham et al., 1986). Alternatively, the polysomal localization of the 1.2-kb transcript may merely reflect the rapid decay of the 3.6-kb mRNA during RNP preparation. The latter possibility is similar to that described for the apo very low density lipoprotein I1 mRNA (Bakker et al., 1988). Nevertheless, the 3.6-kb mRNA is much more abundant at steady-state and is likely to be the predominant template for preproIGF-I1 synthesis in uiuo.
The structure and function of the trans-acting factors involved in the endonucleolysis of the full-length IGF-I1 mRNA remain to be characterized. In bacteria, the processing of RNAs is governed by endonucleases, and their specificities appear to be dictated by a combination of primary sequence and higher order structure in their substrates (reviewed by King et al., 1986;Deutscher, 1988). So far, no eukaryotic endonucleases involved in mRNA turnover have been isolated and characterized. Moreover, the activity of the putative endonucleases is likely to be modulated by the presence of additional trans-acting factors which may regulate the binding of the endonucleases or fold the RNA into a structure which favor endonucleolysis. We have examined the target region in the 3.6-kb IGF-I1 mRNA for sequential and structural elements that are phylogenetically conserved and may participate in RNA-protein interactions. Fig. 10 shows an alignment of 300 nucleotides from rat exon 6 with the corresponding human exon 9. According to S1 mapping analyses and primer extension experiments (de Pagter- Holthuizen et al., 1988;Matsuguchi et al., 1989; and this study), the endonucleolytic cut occurs between a putative hairpin (positions 1389-1427) and a guanosine-rich sequence (positions 1470-1506). The 16bp hairpin in the 1389-1427 region in the rat (horizontal arrows in Fig. 10) is not phylogenetically conserved in its entirety, since it is only feasible to form an 11-bp stem in the human counterpart. Based on computer predictions, an additional hairpin has also been proposed to occur in the 1310-1397 region of rat exon 6, which may participate in the cleavage of the trailer (Meinsma et al., 1991).
A striking feature of the region adjacent to the putative cleavage site is the high content of guanosines (75% in the region 1467-1506). Within the entire IGF-I1 mRNA, the 1442-1506 region exhibits the highest sequence homology between rat and human, and it is therefore likely that this region is of functional significance. The fact that reverse transcriptase cannot traverse this region in K+ indicates the presence of a stable higher order RNA structure. The termination is not caused by long range interactions in the fulllength mRNA as an identical termination of reverse transcriptase is observed with an in uitro transcript of 388 nucleotides from this region. The enzyme is able to penetrate the guanosine-rich region in Na+, showing that the formation, or the stability, of the guanosine interactions is ion-sensitive. The RNase TI probing data are consistent with two different secondary structures that both include base pairing between the guanosine residues. One is a G-hairpin (Fig. 11A) with a nonreactive ascending strand in the anti conformation and a reactive descending strand in the syn conformation. The other is an intramolecular guanosine quadruplex (Fig. 11B). Whereas the ion effect is difficult to reconcile with a hairpin structure, the stability of the structure in the presence of K' supports the suggestion of a quadruplex (Williamson et al., 1989;Sundquist and Klug, 1989;Panyutin et al., 1990;Sen and Gilbert, 1990;Zimmerman et al., 1975). The low reactivity of loop I toward RNase TI does not exclude the presence of a turn in the quadruplex, since certain tetraloops are very stable (Cheong et al., 1990). It has recently been demonstrated that a guanosine-rich fragment of 5 S RNA from E. coli is able to form intermolecular quadruplexes (Kim et al., 1991), but no naturally occurring intramolecular, and therefore antiparallel, guanosine quadruplex in RNA has been reported. Here, we propose that this structural feature occurs in the IGF-I1 mRNA (Fig. 11B).
The endonucleolytic event described in this study exhibits some similarities to the cleavage of a maternal homeo box mRNA in Xenopus laevis oocytes (Brown and Harland, 1990). It is the first detectable step in the decay process; the cleavage activity is non-ribosomal; and the cleavage region is probably single-stranded. Whereas imperfectly repeated ACCU units are sufficient to render the homeobox mRNA susceptible to cleavage in a sequence-specific manner, the elements necessary for efficient cleavage of IGF-I1 mRNA appear to include highly structured flanking regions (Meinsma et al., 1991;this study). However, the function of the flanking hairpins and guanosine-rich region may be to ensure that the cleavage region is accessible and not sequestered in local RNA structures, so that sequence-specific recognition of the purine-rich target is feasible. This resembles the RNase E-dependent activity involved in E. coli RNA metabolism (Babitzke and Kushner, 1991), which appears to recognize a sequence element between two structural domains (Christiansen, 1988;Mackie, 1992).
Expression of rat IGF-I1 is initiated from a minimal promoter (Evans et al., 1988) and finished by constitutive secretion from the cell. Therefore, regulatory events governing IGF-I1 expression are likely to occur between these levels. In a previous study, we showed that the human IGF-I1 mRNAs are subject to translational discrimination (Nielsen et al., 1990). Here, we propose that an endonucleolytic cleavage of IGF-I1 mRNA provides an additional mode of modulating IGF-I1 production post-transcriptionally.