Control of insulin mRNA stability in rat pancreatic islets. Regulatory role of a 3'-untranslated region pyrimidine-rich sequence.

Stabilization of insulin mRNA in response to glucose is a significant component of insulin production, but the mechanisms governing this process are unknown. We presently observe that insulin mRNA is a highly abundant messenger and that the content of this mRNA is mainly controlled by changes in messenger stability. We also demonstrate specific binding of the polypyrimidine tract-binding protein to a pyrimidine-rich sequence located in the 3'-untranslated region (3'-UTR) of insulin mRNA. This binding was increased in vitro by dithiothreitol and in vivo by glucose. Inhibition of polypyrimidine tract-binding protein binding to the pyrimidine-rich sequence by mutation of the core binding site resulted in a destabilization of a reporter gene mRNA. Thus, glucose-induced binding of polypyrimidine tract-binding protein to the 3'-UTR of insulin mRNA could be a necessary event in the control of insulin mRNA levels.

Glucose is the main regulator of insulin biosynthesis (1). Besides controlling insulin biosynthesis by modulating protein synthesis initiation and elongation rates (2), glucose also stimulates the production of insulin by increasing insulin mRNA levels (3). This effect is achieved by a selective stimulation of insulin gene transcription (4) as well as by an increase in insulin mRNA stability (5). However, the relative contribution of transcriptional versus post-transcriptional mechanisms to the control of insulin mRNA levels is under debate. According to the original view, insulin mRNA is a highly abundant messenger (6), which is not affected by short term glucose challenges (7), and therefore controlled to a large extent by changes in messenger stability (5). The half-life of insulin mRNA was assessed to be 29 h at a low glucose concentration and 77 h at a high glucose concentration (5). Recent observations, however, challenge this view by suggesting that insulin mRNA contents of insulin-producing cells are rapidly and dramatically increased by glucose and that this effect is mediated by stimulation of insulin gene transcription (8 -10). In addition, it has also been reported that glucose destabilizes, rather than stabilizes, recently formed insulin mRNA (9).
Very little is known on the mechanisms by which glucose, according to the original view, promotes an increased stability of insulin mRNA. A recent study has identified the 3Ј-untrans-lated (3Ј-UTR) 1 region as a critical region for glucose-mediated control of rat insulin II mRNA stability (11). We have observed that the 3Ј-UTR of rat insulin mRNA contains a pyrimidinerich segment directly downstream of the coding sequence (Fig.  1A). Interestingly, similar pyrimidine-rich segments are located to the insulin mRNA 3Ј-UTR of several mammalian and non-mammalian species including humans (Fig. 1B). Pyrimidine-rich segments are also present in 3Ј-UTRs of other longlived messengers, such as ␣-globin, (I)-collagen, 15-lipoxygenase, and tyrosine hydroxylase (TH) (Ref. 12; Fig. 1A). It appears that the 37-39-kDa poly(C)-binding protein (PCBP), also known as ␣CP or hnRNP-E, binds to the pyrimidine-rich motifs of these mRNA, thereby assembling a ribonucleoprotein complex, the ␣-complex, which results in stabilization of the mRNA (13). Another interesting pyrimidine-rich sequence-binding protein is the 56-kDa polypyrimidine tract-binding protein (PTB) (14). As other hnRNP proteins, PTB assists in processing, transport, and translation of mRNAs. However, in insulinproducing MIN6 cells PTB mRNA contents are increased 5-fold by glucose (15). This may indicate that the expression of PTB is under cell-specific control and that PTB contributes to the ␤-cell phenotype as a glucose-responsive secretor of insulin.
The aim of the present study was first to quantify insulin mRNA in rat pancreatic islets. A high insulin mRNA content would strongly support the original view that insulin mRNA contents are mainly controlled by post-transcriptional mechanisms. Second, we aimed at characterizing the molecular event by which glucose increases insulin RNA stability. We have therefore studied the putative role of the pyrimidine-rich insulin mRNA 3Ј-UTR in the regulation of insulin mRNA stability.

EXPERIMENTAL PROCEDURES
Materials-The RNA oligonucleotides were from Scandinavian Gene Synthesis AB (Köping, Sweden). The rabbit anti-PCBP serum was a generous gift from the laboratory of Ellie Ehrenfeld. David Helfman's and Carol Bromstycks' groups kindly supplied the mouse PTB and rabbit anti-hnRNP K antibodies.
Isolation and Culture of Pancreatic Islets-Adult Sprague-Dawley rats, from a local colony, were used. Islets were isolated by collagenase digestion (17). The islets were cultured in RPMI 1640 supplemented with 10%, Fetal Clone II serum (Hyclone Europe Ltd., Cramlington, UK). The islets were cultured free-floating 5-10 days with the medium changed every second day (16).
Human islets were obtained from the Central Unit of the ␤-Cell Transplant, Brussels, where the islets were isolated and maintained in culture as described previously (17).
Quantitative Real-time PCR-Total RNA from 20 -30 islets was isolated using the Ultraspec TM Total RNA Isolation System (Biotech Laboratories, Houston, TX). First strand cDNA was synthesized using random nonamers and avian myeloblastosis virus reverse transcriptase (Promega, Madison, WI). PCR reactions were carried out with the Lightcycler real-time PCR instrument, using the Faststart DNA Master CYBR Green I kit (Roche Molecular Biochemical, Mannheim, Germany). The following primers were used: insulin forward, 5Ј-ACAGC-ACCTTTGTGGTCC; insulin reverse, 5Ј-GGACTCAGTTGCAGTAGT-TC; ␤-actin forward, 5Ј-GCCCTGGCTCCTAGCACC; ␤-actin reverse, 5Ј-CCACCAATCCACACAGAGTACTTG. Unknown samples were run in duplicates against an external standard curve with known quantities of linearized pRI7 plasmid, which contains the rat insulin cDNA (18), or linearized ␤-actin cDNA plasmid. The amplification efficiency was similar for the reverse-transcribed islet RNA samples and pRI7 ( Fig. 2A).
␤-Cell number per islet was determined using flow cytometry. Briefly, islets were trypsinized for 5 min at 37°C to generate free islet cells. The cells were then fixated for 5 min in 4% paraformaldehyde, permeabilized with 1% saponin, and incubated for 20 min at room temperature with a guinea pig anti-insulin antibody. A fluorescein isothiocyanate-labeled anti-guinea pig antibody was added, and the cells were analyzed in a FACSCalibur flow cytometer (Becton-Dickinson Instruments). Cells with increased FL 1 -fluorescence were gated and counted as ␤-cells.
Rat islets were homogenized in 100 l of homogenization buffer (19), by using a Pellet Pestle® Motor (Kontes, Scientific Glassware/Instruments, Vineland, NJ). The homogenates were centrifuged at 13,000 ϫ g for 10 min. The supernatant was used as the cytosolic fraction and the pellet as the nuclear fraction. The pellet was resuspended in an equal volume homogenization buffer supplemented with 1% Triton X-100.
The RNA-protein binding reaction was performed essentially as de-scribed previously (19). To half the reactions 11 mM dithiothreitol (DTT) was added, before addition of the probe. In some cases, the reaction mixtures were divided into two aliquots, one which was cross-linked by exposed to uv radiation (5 milliwatts/cm 2 ) for 5 min and then analyzed by reducing SDS-PAGE, the other was used directly for non-denaturing gel electrophoresis. In the latter case, the samples were applied on a 7% polyacrylamide gel and electrophoresed in 0.5 ϫ TBE (45 mM Tris borate, 1 mM EDTA). The gel was fixated, dried, and exposed to a film overnight at Ϫ70°C. Elution and Analysis of Gel Slices Containing the RNA Oligonucleotide-Protein Complex-RNA oligonucleotide-protein binding reactions were run on a non-denaturing gel, and the position of the retarded and radioactive RNA oligonucleotide-protein complex was visualized by exposing an x-ray film to the gel. The lanes were then cut in three parts: above, at, and below the RNA-protein complex position. Proteins were eluted from the gel slices in 5 mM Tris acetate, pH 8.0, 0.1% SDS, and 0.1 mM EDTA under agitation overnight. The eluates were then concentrated on Centricon microconcentrators (Amicon, Beverly, MA), and the proteins were separated on a 12% SDS-PAGE and electroblotted to a nitrocellulose filter. The filters were hybridized with the monoclonal anti-PTB 3 antibody (20). Horseradish peroxidase-conjugated antimouse antibody (1:1000) was used as secondary antibody, which was detected by the Amersham ECL system (Amersham Biosciences, Inc.).
Northwestern Hybridization-Northwestern was performed essentially as described previously by others (21). Islet proteins were separated by 12% SDS-PAGE and transferred to a membrane and renatured by incubating the membrane in renaturing buffer (21) for 2 h at room temperature. The membrane was then incubated in hybridization buffer (renaturing buffer ϩ 20 mg/ml tRNA, 1 mg/ml heparin, and 0.5 nmol of 32 P-labeled RNA oligonucleotide) for another 2 h. The membrane was further washed 3 ϫ 10 min in renaturing buffer, dried, and finally exposed to an x-ray film. The position of the radioactive band was compared with that of the PTB, which was visualized by immunoblotting, analyzed in parallel.
Construction of Reporter Gene Vectors with 3Ј-UTR Ins-PRS-Double-stranded DNA oligonucleotides with the sequences of wild-type ins-PRS (5Ј-UCCACCACUCCCCGCCCACCCCUCU), mutant 1 ins-PRS (5Ј-UCCACCACUCCCCGCCCACCACUCU), or mutant 2 ins-PRS (5Ј-UCCACCACUCCCCUCCCACCCCUCU) were synthesized with XbaI restriction sites at both ends. The oligonucleotides were then cloned into the XbaI site of the pCR TM 3-CAT vector (Invitrogen, San Diego, CA), downstream the coding sequence of the chloramphenicol acetyltransferase (CAT) reporter gene, and upstream the bovine growth hormone polyadenylation signal. All vectors were control-sequenced using dideoxynucleotide termination sequencing and the Sequenace kit (United States Biochemicals, Cleveland, OH).
Semiquantitative Reverse Transcriptase-PCR-Lipofection of dispersed islet cells was performed essentially as described previously (22). The cells were transfected with 1 g of either empty pCR3-CAT vector, pCR3-CAT ϩ wild-type ins-PRS, pCR3-CAT ϩ ins-PRS mutant 1, or pCR3-CAT ϩ ins-PRS mutant 2 and maintained in culture for 48 h. The cells from each of the four dishes were then exposed for 24 h to 2.8 or 28 mM glucose, with our without 5 g/ml actinomycin D.
The cells were collected from the previous step, and the RNA was isolated using the Ultraspec TM Total RNA Isolation reagent. The You-Prime First-strand Beads kit (Amersham Biosciences, Inc.) was used for cDNA synthesis. The sequences of the PCR primers were CAT forward, 5-GAATGCTCATCCGGAACT; CAT reverse, 5Ј-CCAGGGTC-AAGGAAGGCACGG; glyceraldehyde-3-phosphate dehydrogenase forward, 5Ј-GACCCCTTCATTGACCTCA; and glyceraldehyde-3phosphate dehydrogenase reverse, 5Ј-CCTTCTCCATGGTGGTGAA. When analyzing samples treated with actinomycin D, 30 cycles were required to see PCR products in the exponential phase. However, for samples not treated with actinomycin D, 25 cycles were sufficient. The PCR products were run in an 1.5% agarose gel and stained with ethidium bromide for quantification of 28 S rRNA.

RESULTS
Insulin mRNA Is Highly Abundant in Rat ␤-Cells-Using real-time PCR, insulin mRNA contents of in vitro cultured rat pancreatic islets were quantified. We first assessed basal and non-stimulated insulin mRNA levels by preculturing the islets for 24 h at a substimulatory glucose concentration and in the absence of serum. At these basal conditions, we observed that a ␤-cell contains 40,000 -50,000 insulin mRNA molecules and that insulin mRNA is 60 -80 times more frequent than the messenger for ␤-actin (Fig. 2B). In line with this, a short term (1 h) stimulation with a high glucose concentration did not increase the contents of insulin mRNA (Fig. 2B). Insulin has recently been reported to potently stimulate insulin gene transcription (8 -10). However, addition of 5 g/ml insulin did not affect insulin mRNA contents (Fig. 2B). These insulin mRNA estimates are based upon the assumption that the yield and efficiency of the RNA extraction and cDNA synthesis procedures are 100%. Our results might therefore be underestimates.
Next, we quantified insulin mRNA contents in response to a long term (24 h) glucose and serum stimulation. The insulin mRNA contents of control islets dropped to ϳ20,000/␤-cell (Fig.  2C). This is not surprising, because prolonged culture without glucose and serum (24 h ϩ 24 h) is known to result in ␤-cell apoptosis (23). The 24-h high glucose stimulation increased insulin mRNA contents to ϳ100,000/␤-cell (Fig. 2C). Also the ratio insulin mRNA to ␤-actin mRNA was markedly increased (Fig. 2C).
Having established that insulin mRNA is highly abundant and only slowly increased in response to glucose, we next addressed the question whether transcriptional or post-transcriptional mechanisms mediate the regulatory effect of glucose. For this purpose, islets were precultured in the presence of serum and high glucose to bring the insulin mRNA contents up to high levels. Next, some of the islets were transferred to a low glucose-containing medium with or without actinomycin D. As expected, a 16-h incubation at a low glucose concentration promoted a small (35%) decrease in insulin mRNA (Fig. 2D). Interestingly, inhibition of RNA synthesis did not decrease insulin mRNA contents and the drop induced by low glucose was of the same magnitude in the presence as in the absence of actinomycin D (Fig. 2D). This correlates well with previous findings showing that actinomycin D does not markedly reduce insulin mRNA levels (5). This is possibly explained by inhibition of synthesis of insulin mRNA degrading factors, which would mask the inhibiting effect on insulin gene transcription (5). On the other hand, ␤-actin mRNA did not decrease in response to low glucose and the actinomycin D-induced decrease was more pronounced than that of insulin mRNA (Fig.  2D). Consequently, the low glucose-induced decrease in the insulin mRNA/␤-actin mRNA ratio was of the same relative magnitude both with and without actinomycin D (Fig. 2D). These findings are in line with the previously established insulin mRNA half-life of 29 h at a low glucose concentration, 77 h at a high glucose concentration, and a ␤-actin mRNA half-life of 9 h (5, 24).
A 55-60-kDa Protein Binds to the Ins-PRS-The findings that glucose-induced changes in insulin mRNA stability control insulin mRNA contents prompted us to next investigate the mechanisms that regulate insulin mRNA stability. Probing for specific binding to the pyrimidine-rich insulin mRNA 3Ј-UTR sequence (ins-PRS), we observed binding of a 25-bp ins-PRS RNA oligonucleotide to a cytosolic protein, both in liver and islet cells. This binding resulted in a cross-linked complex with the combined molecular mass of 65-70 kDa (Fig. 3, A-C). If it were assumed that the RNA oligonucleotide increases the molecular mass of the complex with 10 kDa, the molecular mass of the RNA-binding protein would be 55-60 kDa. This binding FIG. 2. Real-time PCR quantification of insulin mRNA contents. A, amplification efficiency of linearized pRI7 plasmid insulin cDNA and rat islet insulin cDNA samples in real-time PCR reactions. Series of 10-fold dilutions were analyzed in duplicates. B, precultured rat pancreatic islets were maintained for 24 h at 1.7 mM glucose and in the absence of serum. Islets, in groups of 20 -30, were then stimulated for 60 min with 17 mM glucose or 5 g/ml insulin. Insulin mRNA, ␤-actin mRNA, and ␤-cell number were quantified as described under "Experimental Procedures." Results are means Ϯ S.E. for six independent observations. C, precultured rat islets were maintained for 24 h at 1.7 mM glucose and in the absence of serum. The islets were then stimulated for 24 h with 17 mM glucose or with 10% Fetal Clone II.
Results are means Ϯ S.E. for six independent observations. D, precultured rat islets were maintained for 24 h at 17 mM glucose in the presence of serum. The islets were then transferred to 1.7 mM glucose with or without the presence of 5 g/ml actinomycin D for 16 h. Results are means Ϯ S.E. for eight observations. * denotes p Ͻ 0.05 using Student's paired t test. ϩ denotes p Ͻ 0.05 using one-way analysis of variance and Bonferroni's test.
was not observed in the absence of uv cross-linking, nor when a mutated RNA oligonucleotide, in which four centrally located C were replaced by A (ins-PRS mutant 3), was used (Fig. 3A). In liver cells, there was also prominent binding to the ins-PRS RNA oligonucleotide by a 40-kDa protein (Fig. 3A). The identity of this protein is not known, but the MW corresponds well to PCBP, which has been reported to bind to the pyrimidine-rich sequence of TH mRNA (25). Binding of the 40-kDa protein to the ins-PRS RNA oligonucleotide was, however, not consistently observed in islet cells (Fig. 3, B and C).
The Binding Activity of the 55-60-kDa Protein to Ins-PRS Is Stimulated by Reducing Agents and Glucose-Using cytosolic extracts from non-stimulated rat pancreatic islets, we observed that the reducing agents DTT (Fig. 3B) or ␤-mercaptoethanol (data not shown) increased binding of the 55-60-kDa protein to ins-PRS. The enhancing effect of DTT was dose-dependent and reached maximum at 5-10 mM (data not shown). The increased binding of the 55-60-kDa protein to ins-PRS in response to DTT was also observed when binding reactions were analyzed by non-denaturing gel electrophoresis (Fig. 3D). In this case, binding was visualized by retardation of the radioactive RNA oligonucleotide, and the intensity of the retarded band corresponded well to the intensity of the 65-70-kDa protein-RNA oligonucleotide complex observed in parallel cross-linking experiments (Fig. 3, C and D). The effect of reducing agents was not as clear in islet homogenates stimulated with a high glucose concentration as compared with non-stimulated islets (Fig. 3, C and D). Instead, the ins-PRS binding activity of glucose-stimulated islets was enhanced in the absence of DTT and not further increased in the presence of DTT. These findings suggest that the ins-PRS binding activity is maximally stimulated in the presence of DTT and that differences observed in binding activity in the absence of DTT may reflect the in vivo activity of the 55-60-kDa protein. Thus, by expressing FIG. 3. Binding activity to the ins-PRS RNA oligonucleotide in liver and rat islet extracts. A, rat liver extract was incubated with 32 P-labeled wild-type or mutant 3 ins-PRS RNA oligonucleotide. 11 mM DTT was added to some of the samples. Following the binding reaction, some of the samples were uv-cross-linked and analyzed by SDS-PAGE. The arrowhead to the right indicates the position of the 65-70-kDa protein-RNA oligonucleotide complex. Below this complex can also be seen the 50-kDa protein-RNA oligonucleotide complex. B, islet cytosolic extracts were incubated with wild-type ins-PRS RNA oligonucleotide with or without 11 mM DTT and with different concentrations of the molecular cross-linker suberic acid bis(N-hydroxysuccinimide ester). The arrowhead to the right indicates the position of the 65-70-kDa protein-RNA oligonucleotide complex. C, rat islets were incubated for 60 min at 1 or 28 mM glucose. Cytosolic extracts were then prepared and allowed to bind to wild-type ins-PRS RNA oligonucleotide in the presence or absence of DTT. Each sample was subsequently divided into two aliquots, of which one was uv-cross-linked and the other was not (D). Non-cross-linked samples were analyzed by non-denaturing gel electrophoresis. E, islet cytosol extracts from rat and human islets were incubated with rat ins-PRS or human ins-PRS 32 P-labeled RNA oligonucleotides, respectively. Binding reactions were analyzed by non-denaturing gel electrophoresis.
ins-PRS binding activity as the ratio binding activity ϪDTT/ ϩDTT, we correct for differences in total amount of the 55-60-kDa protein present in islet samples. When analyzing the effect of glucose, it was found that the ins-PRS binding activity in rat islet cytosol was significantly increased by 28 mM, but not by 5 mM glucose (Fig. 4). In nuclear extracts, both 5 and 28 mM glucose increased ins-PRS binding activity as compared with no glucose (Fig. 4). Thus, glucose-induced stabilization of insulin mRNA is paralleled by increased binding of the 55-60-kDa protein to ins-PRS.
Also human insulin mRNA 3Ј-UTR contains a pyrimidinerich motif (Fig. 1B). To determine whether binding in human islet cytosol extracts to the human ins-PRS occurs, we analyzed binding reactions by non-denaturing gel electrophoresis. It was observed that a similar gel retardation product was present in human islet extracts as in rat islet extracts (Fig. 3E). This indicates that a similar insulin mRNA 3Ј-UTR-protein complex is formed in human islets as in rat islets.
The 55-60-kDa Ins-PRS Binding Protein Is Probably PTB-To determine the identity of the ins-PRS-binding protein, we performed supershift analysis using antibodies specific for hnRNP-K, PCBP, and PTB. The antibodies against hnRNP-K and PCBP did not generate a supershift (Fig. 5A, results not shown). However, two different PTB monoclonal antibodies specifically abolished the ins-PRS/protein complex, an effect that was not observed with control ascites fluid (Fig.  5A). This indicates that antibody binding to PTB blocks PTB binding to the ins-PRS RNA oligonucleotide.
The Northwestern technique was utilized to probe for ins-PRS binding activity of proteins immobilized to a nitrocellulose filter. We observed ins-PRS binding activity of a renatured protein, which migrated to the same position as PTB (Fig. 5B).
Further evidence in support for PTB as the ins-PRS-binding protein was obtained by analyzing the presence of PTB at the position of the ins-PRS-protein complex in a non-denaturing electromobility shift assay gel. By eluting proteins from above, at and below the position of the ins-PRS-protein complex, we observed PTB immunoreactivity at the complex position in samples with ins-PRS probe present during the binding reaction, but not in samples without (Fig. 5C). The remaining PTB immunoreactivity was located above and not below the position of the ins-PRS-protein complex. These results indicate that the fraction of PTB that binds the ins-PRS RNA oligonucleotide alters its three-dimensional conformation so that it migrates faster in a non-denaturing gel.
PTB and Ins-PRS Stability Is Not Affected by DTT or Glucose-To exclude the possibility that increased binding of PTB to ins-PRS is due to altered stability in response to DTT or glucose, islet cell homogenates were incubated for 50 min in homogenization buffer with or without 11 mM DTT. Samples were then analyzed by immunoblotting using the anti-PTB-3 antibody and counterstained with Amido Black. It was observed that the in vitro stability of PTB (Fig. 6A) and ins-PRS (Fig. 6B) was not affected by the addition of DTT. In addition, islets were incubated for 1 h in medium containing 2.8 or 28 mM glucose, and the amount of expressed PTB was examined by immunoblotting. We could not observe any differences in PTB expression between the two groups (Fig. 6C), indicating that a 1-h incubation period is not sufficient to alter PTB contents.
Inhibition of PTB Binding to Ins-PRS Results in mRNA Destabilization-Mutation of the critical pyrimidines to purines resulted in abolished binding of PTB to ins-PRS in vitro (Fig. 3A). To assess whether mRNA stability is affected by PTB

FIG. 4. Glucose increases ins-PRS binding activity in both rat islet cytosol and nuclei fraction.
A, isolated rat islets were incubated for 60 min with different glucose concentrations and then fractionated into nuclei and cytosol. The different fractions were incubated with the 32 P-labeled ins-PRS RNA oligonucleotide, with and without DTT, and then separated with non-denaturating gel electrophoresis. Ins-PRS binding activity was expressed as ratio between the shifted ins-PRS-protein complex without and with DTT. Results are mean Ϯ S.E. for six to seven observations. * denotes p Ͻ 0.05 versus indicated control using two-way analysis of variance and Bonferroni's test. B, photograph showing a typical experiment. binding to ins-PRS in vivo, we lipofected dispersed rat islet cells with pCR TM -CAT vector with or without wild-type or mutated ins-PRS. In the presence of actinomycin D, we observed that the reporter gene mRNA containing the wild-type rat insulin I ins-PRS was equally abundant as the mRNA lacking ins-PRS (Fig. 7A). However, a mutation of one of the pyrimidines to a purine in the PTB core-binding site resulted in a marked destabilization of the mRNA, whereas a mutation of a purine to a pyrimidine outside the core-binding site had no effect (Fig. 7A). Similar results were obtained in cells incubated in the absence of actinomycin D (Fig. 7B), indicating that transcription of the reporter gene was not affected by the ins-PRS mutation. In cells expressing wild-type ins-PRS, there was no increase in reporter gene mRNA levels in response to glucose.

DISCUSSION
To our knowledge, this is the first study to quantify insulin mRNA in isolated pancreatic islets. At physiological conditions, i.e. in the presence of serum, the ␤-cell content of insulin mRNAs was at least 40,000 -100,000 molecules, depending upon the glucose concentration of the culture medium. This finding is in agreement with the early and indirect assessment (6), which indicated a ␤-cell insulin mRNA content of 50,000 -150,000 molecules. A typical mammalian cell contains ϳ360,000 mRNAs in its cytoplasm (26). Assuming that the same number applies to the ␤-cell, insulin mRNA would con-stitute up to 30% of all mRNA, which correlates well with the percentage of insulin protein synthesized (1).
We did not observe any increase in insulin mRNA in response to a short term (60 min) glucose stimulation. This is, however, not surprising considering that insulin mRNA is vastly abundant in ␤-cells. The transcriptional output of insulin mRNA by the two insulin genes present in rat has not been determined experimentally. However, a theoretical calculation of the maximal insulin mRNA synthesis rate tells us that no more than 2500 transcripts/h are produced. This assessment is based upon an elongation rate of 30 nucleotides per second and that 10 RNA polymerases transcribe each insulin gene simultaneously. Although this is most likely an overestimate, it is far from the net increase of 40000 molecules, which is the minimum number of molecules necessary to double the insulin mRNA content of low glucose cultured islets in 1 h. In our hands, the insulin mRNA content is only slightly decreased by a 16 h glucose withdrawal. The same decrease was observed in islets with inhibited RNA polymerase activity. This indicates that control of insulin mRNA stability contributes significantly to the regulation of insulin mRNA levels. Indeed, it is generally agreed that the contents of abundant and long lived messengers are subject to post-transcriptional, rather than transcriptional control. This view is challenged by Leibiger and coworkers (8 -10), who report that insulin mRNA contents are increased up to 5-fold in response to a 60-min glucose stimulation. The reason for the apparent incompatibilities between the present work and the work by the Leibiger group is unclear.
Although glucose is known to specifically increase insulin mRNA stability (5), little is hitherto known of the molecular mechanisms underlying this event. However, a recent report identified the 3Ј-UTR of insulin mRNA as critical for control of messenger stability (11). The UUGAA sequence, located between the polyadenylation signal and the polyadenylation site, was suggested to be important because of its conservancy (11). Alternatively, a closer analysis of the 3Ј-UTR of rat insulin mRNA (Fig. 1A) shows that the region just up-stream of the polyadenylation signal and downstream of the termination codon contains a pyrimidine-rich sequence, with similarities to the PTB (27) and PCBP (12,28) consensus binding sequences (Fig. 1A). Protein binding to 3Ј-UTR pyrimidine-rich segments of for example TH and erythropoetin mRNA has been implicated in both constitutive and regulated stability control of the messenger (25,29). In line with this notion, we have presently observed specific binding of a 55-60-kDa protein to the ins-PRS RNA oligonucleotide. Moreover, mutation of one cytidine to an adenosine in the ins-PRS, which probably prevented protein binding, resulted in marked destabilization of the CAT reporter gene mRNA. This is in line with the work by Wicksteed et al., (11) in which it was shown that the 3Ј-UTR of insulin mRNA contains stability determinants that mediated a decreased reporter messenger stability when expressed in hepatocytes. This raises the possibility that destabilizing elements of the insulin RNA messenger are masked by ␤-cellspecific protein binding to the ins-PRS sequence.
We presently observe that a high glucose concentration, which increases insulin mRNA contents by stabilizing the messenger, stimulated protein binding to ins-PRS. This indicates that regulated protein binding to the ins-PRS element could be a necessary event in glucose-mediated insulin mRNA stabilization. On the other hand, levels of the reporter gene mRNA remained unaffected after introduction of the wild-type ins-PRS as well as after culture at a high glucose concentration. Assuming that dispersed and transfected islet cells respond to glucose as normal islet cells do, it may be that the presence of ins-PRS is not sufficient for glucose-induced stabilization of messenger stability. Thus, there may be a requirement of other cis-acting mRNA elements to obtain full control of glucoseinduced changes in mRNA stability. The induced stability of TH mRNA in response to hypoxia is for instance dependent of protein interactions with both the 3Ј-UTR and the coding region of the mRNA (29).
Using specific antibodies, we have presently observed that the ins-PRS-binding protein is probably PTB. It is not clear why PCBP1 or PCBP2, which both are present in islet cells (results not shown) and which may bind ins-PRS in liver cells, do not bind ins-PRS in islet cell homogenates. PCBP is thought to bind TH mRNA, either at the PCBP consensus binding sequence present in rat TH mRNA (12) or/and at the PCBPhypoxia-inducible protein-binding site present in human TH mRNA (28) (Fig. 1A). The rat ins-PRS sequence is quite similar to these PCBP-binding sites (Fig. 1A). Thus, it is reasonable to assume that PCBP, at least in liver homogenates, could bind ins-PRS. On the other hand, one or two putative PTB-binding sites are easily identified in rat ins-PRS (Fig. 1A). In addition, pyrimidine-rich sequences similar to the PTB consensus binding site, but not to the PCBP sites, seem to be present in insulin mRNA of most species with known 3Ј-UTR sequences (Fig. 1B). Taken together, this supports our finding that PTB is the main ins-PRS-binding protein and that this event is evolutionarily conserved.
Species with no ins-PRS are frog, hagfish, and chimpanzee. The frog and particularly the hagfish are evolutionary distant species that both have insulin mRNA 3Ј-UTRs that are several hundred bases long, suggesting that these mRNAs utilize other mechanisms for messenger stability control. The chimpanzee, however, has lost 48 bp of its insulin gene due to a deletion at a site just after the reading frame (30). This results in a very short 3Ј-UTR without any pyrimidine-rich motif. Had it been possible to perform experimental studies with chimpanzee islets, it would have been very interesting to determine whether insulin mRNA levels in the chimpanzee are governed by the same signals as in rodent and human islets. Type 2 diabetes mellitus has been observed in the chimpanzee and may be more frequent than expected (31).
The main known function of PTB is to inhibit mRNA splicing (32). In addition, PTB is thought to play an important role in events such as cap-independent translation (33), RNA polyadenylation, (34) and RNA localization (35). The PTB-binding site in insulin mRNA is located just upstream of the polyadenylation signal, and it is not unlikely that PTB, in addition to regulation of mRNA stability, is involved in polyadenylation, transport, and/or translation. Moreover, PTB exists in three different isoforms, with differential alternative splicing activity, and PTB homologues with cell type-specific expression patterns have recently been identified (36). Consequently, the possibility exists that PTB isoforms and the different PTB homologues have overlapping but distinct RNA binding specificities and that a specific PTB isoform or homologue is expressed in ␤-cells with the purpose to modulate insulin mRNA levels.
Although PTB mRNA expression is up-regulated about five times in the ␤-cell line MIN6, in response to a 24-h glucose stimulation (15), we could not detect any differences in protein expression after a 1-h incubation at different glucose concentrations. This suggests that acute alterations in the PTB binding activity is regulated by post-translational modifications rather than by increased PTB gene expression. Furthermore, PTB is known to interact with other members of the hnRNP family such as hnRNP-E2 (PCBP 2), -K, and -L, which are all expressed in islets (Refs. 33 and 37, present results). Thus, the interaction of PTB with additional RNA-binding proteins may be necessary events in glucose-mediated regulation of insulin mRNA stability.
In summary, we have identified an interaction between the rat insulin I 3Ј-UTR and the protein PTB, an event that may be necessary for glucose-induced stabilization of insulin mRNA. Future studies will hopefully identify additional proteins and cis-acting elements, which co-operate in mediating a high insulin mRNA stability and availability for translation. Finally, it is possible that disturbances of the insulin mRNA metabolism contribute to the development of diabetes mellitus. Interestingly, the chromosomal localization of the PTB gene has been assigned to chromosome 14-23-p24.1 (38), which is in close proximity to the IDDM11 locus in humans (39).