Regulation of α1(I) Collagen Messenger RNA Decay by Interactions with αCP at the 3′-Untranslated Region

Liver fibrosis is characterized by an increased deposition of extracellular matrix proteins, including collagen type I, by activated hepatic stellate cells (HSCs). Previous studies have shown that this increase is mediated primarily by a post-transcriptional mechanism. In particular, the RNA-binding protein αCP binds to the α1(I) collagen 3′-untranslated region (UTR) and stabilizes this RNA in activated, but not quiescent, HSCs. This study examines the role of αCP in the decay of transcripts containing the collagen 3′-UTR in extracts obtained from NIH fibroblasts and quiescent and activated HSCs. Using an in vitro decay system, αCP binding activity was competed out with the addition of wild type oligonucleotides, but not with mutant oligonucleotides. Competition of αCP binding activity increased the rate of decay of wild type transcripts containing the αCP 3′-UTR binding site, but not of transcripts containing a mutated binding site. Quiescent HSC extracts contain no αCP binding activity and have no difference in the rate of decay of transcripts with wild type and mutant binding sites for αCP. The addition of recombinant αCP was sufficient to increase the half-life of the wild type transcript, whereas that of the mutant transcript was minimally changed. In vitro decay assays performed with activated HSC extracts that contain αCP binding activity demonstrate a markedly reduced decay rate of wild type compared with mutant transcripts. In vivo small interfering RNA experiments targeting αCP showed a reduction of the binding activity of αCP and a concomitant reduction in intracellular levels of α1(I) collagen messenger RNA. In conclusion, this study demonstrates the direct role of αCP in the stabilization of α1(I) collagen messenger RNA by blocking RNA degradation in activated HSCs.

most chronic stimuli, HSCs are transformed from a quiescent to that of an activated phenotype (4,5). Isolation of HSCs from normal livers and plating them on plastic results in a spontaneous activation of the HSCs that closely mimics the activation observed in the liver in vivo (6). This culture-induced activation includes an increase in ECM production, including ␣1(I) collagen, loss of vitamin A stores, increased proliferation rate, and increased expression of smooth muscle ␣-actin (7,8). The increase in ␣1(I) collagen expression is primarily because of an increase in the half-life of the ␣1(I) collagen mRNA molecule, from 1.5 h in quiescent HSCs to greater than 24 h in activated HSCs (9,10). The increase in ␣1(I) collagen mRNA stability coincides with an increase in the binding activity of ␣CP (hnRNP E2, PCBP2) to a C-rich region in the ␣1(I) collagen 3Ј-untranslated region (UTR) (10). Mutant mRNAs lacking this binding region do not demonstrate increased stability in activated HSCs. ␣CP binds to the 3Ј-UTRs of several other mRNAs, including ␣-globin, tyrosine hydroxylase, and 15-lipoxygenase, suggesting that it may play a regulatory role for multiple mRNAs (11)(12)(13)(14)(15).
The hnRNP family of RNA-binding proteins consists of many members, most of which are alternatively spliced and have several isoforms resulting in a very large family of proteins. They are involved in many different aspects of RNA metabolism, from pre-mRNA splicing and mRNA shuttling to mRNA stability, transcriptional regulation, and translation (16 -22). Some hnRNP proteins have RNA binding activity regulated directly by specific kinase phosphorylation, suggesting kinases may play important roles in regulating gene transcription or translation in response to certain stimuli (21,23). The hnRNP family member ␣CP contains three RNA-binding K domain homology domains and is alternatively spliced to produce several discrete isoforms (24,25). Yeast two-hybrid studies have demonstrated that ␣CP may form homodimers and have identified other nucleic acid-binding proteins with which it may interact, such as poly(A)-binding protein (PABP) (13,26).
Regulated RNA decay can be analyzed in vitro by following a radiolabeled RNA containing the sequence of interest over time in cytoplasmic extracts (26 -29). Polyadenylated reporter RNAs incubated with cell extracts of interest have been utilized to show the dependence of ␣-globin mRNA stability to a region in the 3Ј-UTR of the message (26,30). To elucidate the role of the 3Ј-UTR in regulating ␣1(I) collagen mRNA stability, we measured relative stabilities of RNA transcripts containing wild type (WT) and mutated (MUT) ␣1(I) collagen 3Ј-UTRs using both activated and quiescent HSCs extracts.
The use of small interfering double-stranded RNAs (siRNAs) is a newly described method for assessing gene function in cell culture models (31,32). siRNAs cause a sequence-specific degradation of messages targeted by the siRNA, effectively down-regulating endogenous gene expression (33)(34)(35). To confirm that the in vitro decay assay had in vivo relevance, we used siRNAs to selectively inhibit ␣CP levels in activated HSCs and observed what effects this had on collagen ␣1(I) mRNA.
This study utilized lysates from quiescent and activated HSCs to demonstrate that the stability of the ␣1(I) collagen mRNA is directly mediated by binding of ␣CP to its cognate binding site in the 3Ј-UTR. Furthermore, selectively inhibiting expression of ␣CP in rat HSCs resulted in decreased ␣1(I) collagen mRNA steady state levels, demonstrating a direct role for ␣CP in the stabilization of ␣1(I) collagen message in activated HSCs.
Lysates-Confluent dishes of cells, either NIH 3T3 or HSCs, were washed one time in 5 ml of phosphate-buffered saline and scraped off the dish. The cells were pelleted at 10,000 ϫ g and incubated in a hypotonic buffer (10 mM Tris, pH 7.6, 1 mM potassium acetate, 1.5 mM magnesium acetate, 2 mM dithiothreitol, 0.5 mM Pefabloc, 2 g/ml aprotinin, 0.5 g/ml leupeptin, 50 M Na 3 VO 4 ) on ice and cells lysed with 5-10 strokes using a 1.5-ml pestle (Nalgene 749521-1590). Cellular debris was removed by centrifugation at 16,000 ϫ g and the supernatant collected as cytoplasmic lysates. S100 lysates were generated from the cytoplasmic lysates by centrifugation in a TLS-55 rotor at 100,000 ϫ g for one hour in a Beckman Ti20 tabletop ultracentrifuge. Glycerol was added to a final concentration of 10% and the samples stored at Ϫ80°C.
Plasmids and Sequences-The wild type collagen ␣1(I) 3Ј-UTR (WT 3Ј-UTR) (GenBank TM accession number Z78279) was cloned into pBluescript using HindIII and SalI to allow in vitro transcription with the T3 promoter. The mutant 3Ј-UTR (10) that lacks ␣CP binding capability was cloned into the same site to generate the MUT3Ј-UTR transcript. The addition of 50 adenosines to the end of the 3Ј-UTR was performed essentially as previously described (27). The oligonucleotide containing the poly()A sequence was ligated into the digested pBluescript-3Ј-UTR vector at the SalI and a blunted XhoI site and the plasmid DNA transformed into competent DH5␣ Escherichia coli.
Transcripts and Probes-To generate capped and radiolabeled probes from WT 3Ј-UTR and MUT 3Ј-UTR, mCAP analog (Stratagene) was added to an in vitro transcription reaction containing 1 g of NSI linearized plasmid, 50 units of RNasin (Promega), 250 mM dithiothreitol, 1 mM ATP, 1 mM CTP, 0.3 mM GTP, 0.25 mM UTP with 0.06 Ci of [ 32 P]UTP and 20 units of T3 RNA polymerase. Probes were purified in a 4% acrylamide gel containing 7 M urea and radiolabeled bands excised from the gel and extracted in stop solution (400 mM NaCl, 2.5 mM Tris, pH 7.6, 0.1% SDS). Oligonucleotides were radiolabeled using 3 l of [␥-32 P]ATP (10 mCi/ml) and 10 units of T4 polynucleotide kinase with standard protocols (37).
In Vitro Decay Assay-In vitro decay assays were performed essentially as previously described (27). Briefly, the decay reaction was performed in a 40-l reaction volume containing 1 g/l S100 lysate and 40,000 cpm radiolabeled probe and 20,000 cpm of a radiolabeled 87-mer oligonucleotide as a control for loading and pipetting. Time points were taken at the indicated times and immediately added to a 300-l stop solution (see above), and 5 g of yeast tRNA was added. The reaction mixture was layered on top of 300 l of water-equilibrated phenol and the entire mixture immediately vortexed. After 5 min of centrifugation (16,000 ϫ g), the aqueous phase was collected and the RNA precipitated with 2 volumes of 95% ethanol. RNA was suspended in 7 l of formamide running dye and run on a 4% acrylamide gel containing 7 M urea.
Recombinant Proteins-The recombinant GST-␣CP 2 was generated as described (38,39). Briefly, GST-␣CP 2 was expressed from the pGEX-3X plasmid in DH5␣ E. coli by induction with 0.1 M isopropyl-1thio-␤-D-galactopyranoside for 3 h prior to purification. Recombinant protein was purified after cell lysis with the addition of 500 l of a 50% slurry of glutathione-Sepharose beads. The beads were washed three times with ice-cold phosphate-buffered saline and recombinant protein eluted using 50 mM reduced glutathione in 100 mM Tris, pH 7.6, 100 mM NaCl 2 .
Quantitation of Results-The amount of radioactive probe at the different time points was quantitated using the PhosphorImager software, ImageQuant 5.2. The zero time point was defined as 100% of the probe, and lanes were standardized for loading and/or precipitation variation by normalizing to the amount of the radiolabeled 87-mer oligonucleotide in the reaction. All experiments were repeated at least three times.
RNA Isolation-Total cellular RNA from HSCs was isolated as previously described (36).
siRNA-siRNAs (Table I) were synthesized using the Ambion Silencer siRNA kit according to the manufacturer's recommendations. Cy3-Luc siRNA was purchased from Dharmacon (Lafayette, CO). HSCs were transfected with 100 pmol siRNA/well in 6-well dishes with 2 l of LipofectAMINE in a total volume of 1 ml following the manufacturer's recommended protocol. Cells were analyzed 72 h later after harvesting RNA or cellular extracts as described above.
Electrophoretic Mobility Shift Assay-Electrophoretic mobility shift assays were performed as previously described (39). Briefly, 10 g of cytoplasmic lysates were incubated in a binding buffer (12.5 mM HEPES, pH 7.9, 15 mM KCl, 0.25 mM EDTA, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.2 mg/ml tRNA, 50 units/ml RNasin, 10% glycerol) with 10,000 cpm of probe for 30 min on ice. Binding reactions were electrophoresed in 6% native gels at 250 V, gels dried, and signals quantitated by PhosphoImager analysis using ImageQuant 5.2 software (Amersham Biosciences). Mixing EMSAs included 10 g of activated HSC extracts incubated with either 10 g of quiescent extracts or 10 g bovine serum albumin as a negative control.
Western Blotting-Cell lysates were prepared as described above. Lysates were supplemented with SDS loading buffer, separated by 10% SDS-PAGE, and electrotransferred onto nitrocellulose. Blots were incubated with 5% BLOTTO (TBST ϩ 5% nonfat dry milk) and incubated with anti-␣CP antibodies. ␣CP antibodies were a generous gift from M. Kiledjian.

Validation of RNA in Vitro Decay Assay (IVDA) in NIH
Cells-To study the factors involved in ␣1(I) collagen mRNA degradation, we utilized radiolabeled 7mG-capped and polyadenylated transcripts as RNA substrate and cytosolic S100 protein extracts in reaction mixtures (40,41). Our initial studies to investigate a regulated decay of the collagen 3Ј-UTR (WT 3Ј-UTR) utilized extracts from NIH 3T3 cells that express type I collagen and exhibit the same binding activity seen in activated HSCs (Fig. 1, lane 5). Extracts from NIH 3T3 cells (Fig.  1, lane 2) and activated HSCs (lanes 4 and 5) HSC, but not quiescent HSCs (lane 3), have binding activity for the collagen 3Ј-UTR. This binding activity has previously been identified as ␣CP using supershift assays with ␣CP-specific antibodies (39). To examine whether RNA decay occurred via a regulated, poly(A)-dependent process in NIH 3T3 extracts, an in vitro decay assay containing increasing amounts of unlabeled poly(A) homopolymers was performed. The addition of excess cold poly(A) homopolymers to the reactions will increase the decay rate if decay occurs by a poly(A)-dependent process (26,30). The addition of increasing amounts of poly(A) homopolymers increased the rate of decay of the WT probe ( Fig. 2A, lanes  3-5), whereas addition of other poly(U) had little effect on the rate of decay ( Fig. 2A, lane 6), thus confirming a poly(A)-dependent process.
Having established that the decay reactions occurred via a poly(A)-dependent process, we investigated whether the ability of the RNA probe to be bound by ␣CP had any effect on the rate of decay in the NIH extracts. The decay rates of the WT probe, which ␣CP binds, and the MUT probe, which ␣CP fails to bind when using NIH 3T3 extracts, demonstrated an 11-fold increase in stability at 60 min of the WT probe compared with the MUT probe (Fig. 2B). Thus, ␣CP, either alone or in a complex, is capable of stabilizing the transcript in NIH extracts. Our data of the ␣1(I) collagen 3Ј-UTR is consistent with that described for the ␣-globin mRNA and provides the first evidence that ␣CP can directly stabilize the ␣1(I) collagen 3Ј-UTR.
Competition of ␣CP Binding Decreases WT Transcript Stability-To confirm that ␣CP binding to the ␣1(I) collagen 3Ј-UTR RNA probe inhibits the decay of the mRNA, competition studies were performed using an unlabeled consensus binding sequence oligonucleotide (WT-oligo) for ␣CP. By conducting the IVDA under conditions where ␣CP binding is abolished by the addition of unlabeled WT-oligo, we observed an increase in the decay rate of the WT 3Ј-UTR transcript to a rate very similar to the MUT 3Ј-UTR transcript (Fig. 2, C and D). The addition of a mutant oligonucleotide (MUT-oligo) of the same size incapable of binding ␣CP had no effect on the decay rate of the WT 3Ј-UTR (Fig. 2, C and D). Interestingly, the control reaction using the MUT 3Ј-UTR probe lacking the ␣CP binding site also showed a slight increase in the rate of decay with the unlabeled WT-oligo competitor, but not the unlabeled MUT-FIG. 2. WT transcripts are more stable than MUT transcripts in NIH extracts. A, an in vitro decay assay using NIH 3T3 S100 extracts with WT 3Ј-UTR transcripts to determine the components of the decay reaction using different homopolymers to compete off RNA-binding proteins. The addition of increasing amounts of poly(A) homopolymer increased the decay rate (5, 10, and 20 ng, respectively, lanes [3][4][5], whereas addition of poly(U) (25 ng, lane 6) did not affect the decay rate compared with control reaction (lane 2). Time 0 probe shown in lane 1. Percentage of probe compared with time 0 is shown below each lane. B, decay rates of WT 3Ј-UTR transcripts (0, 5, 10, 20, 40, and 60 min, lanes 1-6, respectively) is slower than the decay rate of MUT 3Ј-UTR (0, 5, 10, 20, 40, and 60 min, lanes 7-12, respectively) containing transcripts in NIH 3T3 S100 extracts. C, addition of 10 ng of WT oligo will compete off ␣CP and increase the rate of decay of the WT 3Ј-UTR transcript (0, 1, 2.5, and 5 min, lanes 1-4), but the MUT oligo has no effect (0, 1, 2.5, and 5 min, lanes [5][6][7][8]. The MUT 3Ј-UTR transcript displays a modest increase in decay with WT-oligo competition (lanes 9 -12), whereas the MUT-oligo has no effect (lanes 13-16). D, graphical analysis of the IVDA decay rates using WT or MUT probes with the addition of WT or MUT competing oligos. Assays were performed in triplicate and a representative assay shown. oligo competitor (Fig. 2C). This may reflect ␣CP sequestering PABP from the transcript because of a direct interaction of PABP with ␣CP (26,30). Together these data confirm that ␣CP binding to the ␣1(I) collagen 3Ј-UTR transcript confers stability in a sequence-specific manner.
Decay of WT and MUT Transcripts Are Identical in Quiescent HSCs Extracts-Having determined that the in vitro decay reactions recapitulate in vivo RNA poly(A)-dependent decay, the stability of WT and MUT transcripts was investigated in lysates from quiescent HSC extracts. These extracts lack detectable ␣CP binding activity (Fig. 1) despite expression of ␣CP protein (10). If ␣CP binding activity is required to stabilize the collagen mRNA via the 3Ј-UTR, then the decay rates of the WT and MUT 3Ј-UTR transcripts should be similar in these extracts. As predicted, the WT 3Ј-UTR (Fig. 3A, lanes 1-6) and the MUT 3Ј-UTR (lanes 7-12) displayed almost identical decay rates using these extracts. This demonstrates the requirement that ␣CP bind to the WT 3Ј-UTR to stabilize the transcript, as observed in NIH extracts (Fig. 2B).
WT Transcripts Are Stabilized in Activated HSCs-To examine whether a difference exists in decay rates between the WT and MUT 3Ј-UTR transcripts in activated HSCs, IVDAs were performed using extracts obtained from day 14 culture-activated HSCs. A significant increase in the stability of the WT transcript was observed compared with the MUT transcript (Fig. 3B, lanes 1-6 and 7-12).
Recombinant ␣CP Increases Stability of WT Transcripts-To determine whether ␣CP specifically increases the stability of the WT 3Ј-UTR transcripts, recombinant ␣CP was added to the IVDA reactions using quiescent HSC extracts that do not contain an endogenous ␣CP binding activity (Fig. 1). Recombinant ␣CP stabilized the WT 3Ј-UTR transcript but failed to affect the stability of the WT 3Ј-UTR transcript (Fig. 4). The substantial increase in the WT but not MUT transcript indicates a direct role for the binding of ␣CP in stabilizing the ␣1(I) collagen 3Ј-UTR.
Inhibition of Endogenous ␣CP Expression Decreases ␣1(I) Collagen mRNA Levels-To investigate the role of ␣CP in vivo, FIG. 3. Increased WT transcript stability in activated, but not quiescent, HSCs. IVDA using 40 g of quiescent or activated HSC extracts. A, the rate of decay of WT and MUT 3Ј-UTR transcripts using quiescent HSC extracts lacking ␣CP binding activity. B, activated extracts possessing ␣CP binding activity demonstrate a slower rate of decay of the WT transcript than that of the MUT transcript (p Ͻ0.1 by Student's t test). siRNAs targeting the rat ␣CP mRNA (Table I) were transiently transfected into Rat1 fibroblasts, cells that express ␣1(I) collagen and possess ␣CP binding activity. Binding of the ␣CP protein, as assessed by EMSA, was inhibited ϳ30% by addition of siRNA against ␣CP compared with a control siRNA (Fig. 5A). Because the siRNA decreases the binding activity of ␣CP, we wanted to assess the effects of inhibiting ␣CP binding activity on the steady state levels of the ␣1(I) collagen mRNA. Steady state levels of ␣1(I) collagen mRNA were reduced ϳ60% in cells transiently transfected with ␣CP siRNA compared with control siRNA as determined by RNase protection assay (Fig. 5B). There was no change in glyceraldehyde-3-phosphate dehydrogenase mRNA levels. These data corroborate the role of ␣CP in stabilizing the ␣1(I) collagen mRNA. To examine the effects of inhibiting ␣CP expression in HSCs, we transfected siRNAs against ␣CP into culture-activated HSCs and assessed ␣CP binding by EMSA. A decrease of 30 -40% in ␣CP binding activity was observed in HSCs transfected with siRNAs targeting ␣CP compared with control siRNAs (Fig. 5C). We again examined the steady state levels of ␣1(I) collagen mRNA in HSCs transfected with siRNAs against ␣CP and observed a 50 -60% decrease in the steady state ␣1(I) collagen mRNA levels (Fig.  5D). Together, these experiments demonstrate that ␣CP interaction with the ␣1(I) collagen mRNA 3Ј-UTR increases stability of the mRNA molecule in activated HSCs.
Endogenous Regulation of ␣CP in HSCs-To investigate the mechanism of ␣CP binding regulation during HSC activation, we examined the cellular distribution of ␣CP protein in quiescent and activated HSCs. In quiescent HSC lysates, ␣CP was located primarily in the nuclear fraction with very low ␣CP protein levels observed in the cytoplasmic fraction. However, in the activated HSC extracts the total amount of ␣CP protein expression increased and exhibited a relocalization of ␣CP to the cytoplasmic fraction (Fig. 6A). Thus, activated HSCs contain higher ␣CP protein levels, and the ␣CP protein demonstrates a markedly different distribution. For comparison, the binding activity of PABP was similar in extracts from quiescent and activated HSCs (data not shown).
To determine whether the lack of binding activity observed in quiescent HSCs is solely because of low levels of protein expression in the cytoplasm or whether there is active repression of ␣CP binding in the quiescent HSCs, we examined binding activity in mixed extracts. Incubation of activated HSC extracts with equal amounts of bovine serum albumin had no effect on the observed binding activity. However, the addition of equal amounts of quiescent HSC extracts reduced the bind-ing activity of the activated HSC extracts by 50% in the cytoplasmic extracts. (Fig. 6B.) In the nuclear fractions, the binding activity was not affected significantly. DISCUSSION Culture-induced activation of HSCs is accompanied by an ϳ50-fold increase in the steady state levels of ␣1(I) collagen mRNA (10,42). This dramatic increase results primarily from an increase in the half-life of the ␣1(I) collagen mRNA from 1.5 to Ͼ24 h. The increased half-life of the ␣1(I) collagen mRNA correlates with increased binding activity of ␣CP to the ␣1(I) collagen 3Ј-UTR following HSC activation (10). This study demonstrates that ␣CP plays a direct role in stabilizing of the ␣1(I) collagen mRNA in activated, but not quiescent, HSCs.
Utilizing IVDAs, proteins can be assessed for their role in mRNA degradation or stabilization. RNAs that only differ in their ␣CP binding sites display different decay rates with NIH 3T3 cell extracts, supporting a role for ␣CP in mRNA decay. Competition of ␣CP binding using an unlabeled WT-oligo results in a rapid decay of transcripts containing a WT ␣CP binding site, confirming a direct role of ␣CP in mRNA decay. We investigated the role of ␣CP in the differential regulation of ␣1(I) collagen mRNA expression in quiescent and activated HSCs. Quiescent HSCs express only trace amounts of type I collagen, the ␣1(I) collagen mRNA has a short half-life, and the cells contain no detectable ␣CP binding activity (2,5,36,39). The decay rates with quiescent extracts for WT and MUT 3Ј-UTR RNA are nearly identical, as expected in extract that lack ␣CP binding activity. However, when WT and MUT 3Ј-UTR RNAs were incubated with lysates from activated HSCs, which produce collagen, have a long half-life of the ␣1(I) collagen mRNA, and contain ␣CP binding activity, a significant decrease in the decay rate was observed. This suggests that ␣CP binding activity is responsible for stabilizing the ␣1(I)  collagen mRNA by the 3Ј-UTR in activated HSCs. Although different extracts cannot be directly compared, the relative rate of decay of different transcripts in the same extract provides comparison of the specificity and rate of decay because preparation variables are not a factor. Therefore, binding activity of ␣CP present in activated, but not quiescent HSCs, contributes to the dramatic increase of the collagen mRNA half-life.
As a further approach to assess the role of ␣CP binding in stabilizing the ␣1(I) collagen mRNA, siRNA technology was utilized. ␣CP binding activity was decreased by ϳ40% in activated HSCs transfected with siRNA, whereas a control siRNA had no effect on ␣CP binding activity. This reduction of ␣CP binding activity resulted in decreased steady state mRNA levels of ␣1(I) collagen, whereas cells treated with control siRNAs did not. These data demonstrate that the increased ␣CP binding activity in activated HSCs is responsible for increasing ␣1(I) collagen mRNA stability resulting in elevated steady state mRNA levels. ␣CP protein is redistributed from primarily a nuclear localization to distribution to both the cytoplasm and the nucleus during HSC activation. This distribution allows ␣CP to bind to the collagen ␣1(I) 3Ј-UTR and protect the mRNA from degradation. Finally, we demonstrate that the quiescent HSC cytoplasm contains an inhibitory activity that affects ␣CP binding activity in activated cytoplasmic extracts. On the other hand, we were unable to detect any post-translational changes in ␣CP to explain the differences in binding activity, such as phosphorylation, methylation, or acetylation (data not shown).
Based on our previous (10,39,(43)(44)(45)(46) and current studies, our current model (Fig. 7) of ␣1(I) collagen mRNA regulation in activated and quiescent HSCs involves ␣CP as a key component in a multilayered regulation process. The role of PABP on general mRNA turnover is well established, and evidence exists that PABP interacts with ␣CP, increasing the affinity of both proteins for RNA (30). In activated HSCs, ␣CP binds and together with PABP stabilizes the 3Ј-UTR complex and prevents loading of the degradosome, a protein complex that degrades cellular RNAs, onto the ␣1(I) collagen mRNA. We have also shown that a conserved 5Ј-stem loop structure in the ␣1(I) collagen mRNA interacts with an unidentified binding activity in activated HSCs (43,44). This 5Ј-stem loop inhibits mRNA stability and translation, while its inhibition is diminished by the ␣CP binding (45,46). We speculate that the 5Ј-binding protein can interact with either ␣CP or PABP to facilitate circularizing the mRNA and increasing translation efficiency as well as protecting the mRNA from degradation machinery. In this model, ␣CP binding would stabilize the mRNA in a dual manner: ␣CP increases binding of PABP to the poly(A) tail and also interacts with the 5Ј-stem loop protein(s). Quiescent HSCs do not have ␣CP binding activity. Therefore, PABP is not stabilized, there is no interaction with the 5Ј-stem loop, and the mRNA molecule is rapidly degraded. Upon HSC activation, however, ␣CP binds to the 3Ј-UTR and interacts with PABP and/or proteins binding to the 5Ј-stem loop, resulting in an increase in mRNA half-life, thus allowing for efficient translation (Fig. 7).
There are many similarities between the regulation of ␣1(I) collagen mRNA in activated HSCs and the regulation of ␣-globin and lipoxygenase mRNAs in erythroid cells and the tyrosine hydroxylase mRNA in response to hypoxia (17,(47)(48)(49). These mRNAs form a class of molecules post-transcriptionally regulated via 3Ј-UTR interactions with protein(s). There are three predominant classes of mRNAs that are regulated by interactions at the 3Ј-UTR: transcripts containing an AU-rich element (ARE), transcripts containing iron-responsive elements, and transcripts containing polypyrimidine-rich regions in the 3Ј-UTR. These three classes of mRNAs are regulated by interactions between cis-acting regions in the UTR and trans acting proteins via different mechanisms.
Many cell cycle mRNAs contain AREs in their 3Ј-UTRs and have very short half-lives (17). ARE-containing transcripts interact with two types of proteins that regulate message stability. AUF-1 can bind to the ARE, resulting in a rapid degradation of the transcript (50,51). This prevents message accumulation and keeps the protein levels low. Alternatively, the ELAV family of proteins can bind to ARE-containing transcripts, resulting in stabilization of the transcript, effectively allowing for a rapid increase in the cellular protein levels (28,52).
Gene regulation in response to levels of metabolites allows for subtle changes at the post-transcriptional level to alter gene expression. Genes containing iron-responsive elements are bound by a family of proteins called iron-responsive proteins (IRPs) that recognize the iron-responsive element site under low intracellular iron concentrations. This interaction stabilizes the mRNA and allows for production of the transferrin receptor to transport iron into the cell. As the cellular iron levels increase, the IRPs lose their binding capability and dissociate from the transcript, allowing the transcript to be targeted by degradation machinery. In this manner the cell is able to tightly regulate the amount of gene product in response to the availability of cellular factors (51).
Another method to regulate gene expression by mRNA stability occurs when certain genes are up-regulated in a sustained fashion during cell differentiation, such as erythroid cell maturation or HSC activation. Certain genes are up-regulated by binding of a nucleic acid-binding protein to a polypyrimidine-rich region in the 3Ј-UTR. In the ␣-globin message, ␣CP binds to a CU-rich region in the 3Ј-UTR and increases the mRNA stability, allowing for expression from the mRNA for many hours after synthesis (53). Similarly, HSCs dramatically increase the stability of the ␣1(I) collagen mRNA as the cells FIG. 6. ␣CP relocalizes to cytoplasm, which loses an inhibitory effect as HSCs become activated. Cytoplasmic and nuclear fractions from quiescent and activated HSCs were analyzed for the levels of ␣CP protein present. A, Immunoblot for ␣CP protein levels in cellular fractions. Quiescent HSCs exhibit ␣CP protein localized exclusively in the nuclear fraction (lanes 1-2), whereas activated HSC contain equivalent amounts of protein in the cytoplasmic and nuclear fractions (lanes 3-4) and a greater amount of total ␣CP protein expression. B, EMSA performed with activated HSC extracts coincubated with equal amounts of either bovine serum albumin or quiescent HSC extracts. Addition of quiescent cytoplasm lysate to activated cytoplasmic lysates reduced the endogenous binding activity of activated HSC binding activity (p Ͻ0.01 by Student's t test). Similar addition utilizing nuclear lysates had no significant effect. HSCs lack ␣CP binding activity, and the ␣1(I) collagen transcript decays rapidly. As HSCs become activated, ␣CP binding activity to the 3Ј-UTR of ␣1(I) collagen increases. ␣CP bound to the 3Ј-UTR interacts with PABP, stabilizing its interaction with the poly(A) tail, and may interact with proteins located at the 5Ј-end of the transcript, including the protein interacting with the 5Ј-stem loop structure. It is believed this interaction prevents 3Ј-exonuclease activity and acts to stabilize the ␣1(I) collagen mRNA. undergo cellular activation. ␣CP binding to a CU-rich region in the 3Ј-UTR of ␣1(I) collagen is responsible for increased stability of the ␣1(I) collagen transcript.
During liver fibrosis, quiescent HSCs undergo a phenotypic and genotypic change where they lose vitamin A stores and increase production of extracellular matrix proteins, including ␣1(I) collagen. Investigating the regulation of type I collagen in this study, we have demonstrated that (i) transcripts are stabilized by the presence of an ␣CP binding site in activated, but not quiescent HSC extracts; (ii) addition of recombinant ␣CP restores stability to transcripts containing ␣CP binding sites in quiescent HSC extracts; (iii) knockdown of endogenous ␣CP results in decreased ␣CP binding activity and a corresponding loss of endogenous collagen ␣1(I) mRNA levels; (iv) the activation of HSCs induces a subcellular redistribution of ␣CP from primarily nuclear localization in quiescent HSCs to equal cytoplasmic and nuclear distribution in activated HSCs; and (v) there is a soluble, cytoplasmic activity in quiescent HSCs that inhibits the binding activity of endogenously activated ␣CP. Taken together, this work demonstrates for the first time that ␣CP binding activity is required for collagen ␣1(I) mRNA stabilization during HSC activation and that alteration of this activity is sufficient to alter the endogenous collagen ␣1(I) mRNA levels. Thus, ␣CP is a potential target for therapy of liver fibrosis.