Cytoplasmic Protein mRNA Interaction Mediates cGMP-modulated Translational Control of the Asialoglycoprotein Receptor*

Expression of the asialoglycoprotein receptor by the human hepatocellular carcinoma cell line HuH-7 in response to intracellular cGMP concentrations was previ- ously shown to be regulated at the translational level. In a cell-free system, initiation of asialoglycoprotein recep- tor mRNA translation was dependent on the presence of the 7-methylguanylate cap site and was independent of 8-bromo-cGMP levels in which the cells were grown prior to RNA isolation. Stable transfection of COS-7 cells with deletion constructs of the asialoglycoprotein receptor H2b subunit localized the cGMP-responsive cis- acting element to the mRNA 5 (cid:42) -untranslated region (UTR). Addition of biotin (an activator of guanylate cy- clase) induced the expression of (cid:98) -galactosidase present as a chimeric plasmid containing the H2b 187-nucleotide 5 (cid:42) -UTR. An RNA gel retardation assay identified a 37-nucleotide cognate sequence within this 187-nucleotide region. Titration of the 5 (cid:42) -UTR with a cytosolic fraction isolated from HuH-7 grown in the presence or absence of 8-bromo-cGMP or biotin provided direct evidence for an RNA-binding protein responsive to intracellular levels of cGMP. Based on these findings, it seems reasonable to propose that reduction of intracellular levels of cGMP by biotin deprivation results in a negative trans-acting factor associating with the 5 (cid:42) -UTR of asialoglycoprotein receptor mRNAs, thereby inhibiting translation. Regulated expression of cell-surface lectins has been implicated adapted by at 135 at °C for buffer excess or without 187-nucleotide incubation continued proteinase K completely abolished this protein-de-pendent assay. mixture resolved on a 4% low cross-linked PAGE and prepared for autoradiography.

Regulated expression of cell-surface lectins has been implicated in such diverse processes as endocytosis, bacterial and viral infection, regulation of cell proliferation, homing of lymphocytes, and metastasis of cancer cells (1). The asialoglycoprotein receptor (ASGR) 1 is the hepatocellular prototype of a cell-surface lectin responsive to the differentiated state of the liver cell (for review, see Ref. 2). In addition to being a model of receptor-mediated endocytosis (3), the presence of ASGR on hepatocytes provides a membrane-bound active site for cell-tocell interactions (4,5), has made possible the selective targeting of chemotherapeutic agents (6) and foreign genes (7), and has also been implicated as a site that mediates hepatitis B virus uptake (8).
A human hepatoma cell line (HepG2) has provided a convenient model to investigate ASGR biosynthesis. When HepG2 cells were grown to confluence in a minimal essential medium or in a chemically defined medium containing a variety of hormones and growth factors supplemented with dialyzed fetal bovine serum, expression of ASGR was reduced by 60 -70% (9). The low molecular weight factor required for the restoration of ASGR expression was isolated, purified, and identified as biotin (10). Similar results were obtained with a second hepatocellular carcinoma cell line, HuH-7 (11), indicating that the effect was not cell line-specific. Though usually not considered as part of an induction pathway, the effects of biotin upon the steady state expression of ASGR could be mimicked by the addition of the second messenger 8-bromo-cGMP , and these additions were not additive (11,12). This suggested that the effect of biotin may have been mediated through changes in the cGMP level via biotin activation of the membrane-associated guanylate cyclase (13).
Estimates of the steady state level of ASGR mRNA suggested that cGMP-regulated expression of ASGR was at the posttranscriptional level (11,12). Polysome analysis of ASGR subunits H1 and H2 mRNAs indicated that the addition of 8-Br-cGMP caused a shift of ASGR mRNA from the ribonucleoprotein fraction into a translationally active membrane-associated polysomal pool. Although the biochemical mechanisms have not been determined, cGMP has been suggested to regulate the expression of other proteins at a translational level (13,14) and has been shown to increase total protein synthesis in isolated hepatocytes (15).
In mammalian cells, translation of most mRNA species appears to occur by the association of the preinitiation complex at or near the 7-methylguanylate cap structure at the 5Ј-untranslated region (UTR) of mRNA and scanning downstream to the site of protein synthesis initiation (16). The 60 S ribosomal subunit is subsequently recruited to the complex, and translation begins. Recovery of the ASGR mRNA in the ribonucleoprotein fraction during biotin deprivation suggested that intracellular levels of cGMP may play a significant role in modulating the initiation phase of ASGR mRNA translation. The bimodal polysomal distribution of ASGR mRNA was characteristic of a class of mRNAs that were inefficiently translated (17). Current evidence suggests that mRNAs in these functionally distinct fractions differ structurally or through the proteins they interact with (18). Within this group of mRNAs, most interactions between RNA and cytosolic proteins were defined by motifs localized to the 5Ј-UTR (19).
In the present study, the potential role of the 5Ј-UTR as the cis-acting element governing the cGMP-modulated expression of ASGR was established. In vitro transcription coupled with an RNA gel retardation assay defined the cis-acting element within a 37-nucleotide region. In addition, the effective concen-tration of a cytoplasmic protein trans-acting fraction was shown to be responsive to cGMP and biotin deprivation.

EXPERIMENTAL PROCEDURES
DNA Constructs-The 5Ј and 3Ј-UTR regions of the H2b cDNA of ASGR were deleted using polymerase chain reaction to introduce unique restriction sites (XbaI, 6 nucleotides upstream of ATG translation start site or BamHI, 9 nucleotides downstream of the translation stop site). The resulting constructs were subcloned into either pcDNA3 for selection of stable transfectants in COS-7 or pGEM-4Z for in vitro transcription. The 5Ј-UTR of the ASGR H2b was prepared by polymerase chain reaction with the addition of HindIII ends and cloned into a ␤-galactosidase reporter vector (pSV-␤-galactosidase, Promega) 296 nucleotides upstream of the lacZ coding region start site. The nucleotide sequences of the polymerase chain reaction-generated 5Ј and 3Ј inserts were confirmed by the dideoxy chain termination method with Sequenase (DNA Sequence Facility, Albert Einstein College of Medicine).
Cell Culture, Transfection, and Analysis of Cell Extracts-COS-7 and HuH-7 cells were cultured in Eagle's minimum essential medium (MEM) supplemented with 10% fetal bovine serum (FBS) or dialyzed FBS with or without 8-Br-cGMP or biotin. Stable transfectants of COS-7 cells (20) resistant to 400 g/ml G418 were subcloned, and those with the highest level of ASGR expression in MEM ϩ 10% FBS as determined by Western blot (21) were utilized. Transient transfection of HuH-7 with the chimeric plasmid was mediated by LipofectAMINE (Life Technologies, Inc.) following the manufacturer's instructions. 5 hrs after transfection, cells were harvested by trypsinization and replated at a 1:3 ratio in MEM supplemented with 10% dFBS. 24 hrs later, the medium was changed to MEM ϩ 10% dFBS with or without 10 Ϫ7 M biotin. Three days after transfection, cells were harvested, and ␤-galactosidase activity was measured following the manufacturer's instructions (Invitrogen). Near-confluent HuH-7 cultures (1 ϫ 10 6 cells) were labeled with 200 g/ml [ 35 S]Met/Cys (Pro-mix, Amersham Corp.) for 1 h, followed by a 2-h chase. ASGR was immunoprecipitated, resolved on 10% SDS-PAGE, and the gel was processed for fluorography. Western blot and immunoprecipitation protocols as well as antibodies used in this study have been previously described (11).
Northern Blot Analysis-Total cytoplasmic RNA was isolated by extraction with guanidine thiocycanate (22) from approximately 5 ϫ 10 7 cells (HuH-7 or transfected COS-7). For Northern blot analysis, RNA samples were resolved on 1% agarose-formaldehyde gels and transferred to Nytran membrane. The blots were hybridized at high stringency with random prime-labeled ([␣ -32 P]dCTP) H1-and H2-specific probes (12,23). Equal loading and transfer of RNA was verified by staining the membrane with methylene blue.
In Vitro Translation-Total RNA was isolated from HuH-7 cells grown in MEM supplemented with 10% FBS, 10% dFBS, or 10% dFBS plus 1.0 mM 8-Br-cGMP. The translation reaction was performed using a rabbit reticulocyte lysate as described by the manufacturer (Promega) in the presence or absence of 2 g/assay of m 7 GpppG, an inhibitor of cap site recognition. Following a 90-min incubation at 30°C, the [ 35 S]labeled (Pro-mix, Amersham Corp.) ASGR translation products were recovered by immunoprecipitation using a polyclonal antibody to affinity-purified human receptor (11). Samples were resolved on 10% SDS-PAGE, and gels were prepared for fluorography.
Gel Retardation Assay-A nested set of RNA probes were prepared by linearization of the full-length H2b cDNA using restriction sites localized in the 5Ј-UTR as templates for in vitro transcription from the Sp6 promoter of pGEM-4Z. The full-length (187-nucleotide) fragment was labeled by the incorporation of [␣ -32 P]UTP during transcription. The gel retardation assay was adapted from that described by Leibold and Munro (24). HuH-7 cells were homogenized, the cytosol (S-100) was prepared by centrifugation at 100,000 ϫ g for 1 h, and the aliquots were stored at Ϫ135°C. S-100 was preincubated at 0°C for 10 min in assay buffer with a 100-fold molar excess or without unlabeled transcripts prior to the addition of the labeled 187-nucleotide transcript, and incubation continued for an additional 10 min. Inclusion of proteinase K (100 g/ml) in the assay mixture completely abolished this protein-dependent assay. The mixture was resolved on a 4% low cross-linked PAGE and prepared for autoradiography.

RESULTS
Short Term cGMP-regulated Expression of ASGR-Based on our previous findings that biotin was required for expression of ASGR by HepG2 and HuH-7 cell lines, we proposed that the mechanism of biotin regulation was mediated by maintaining the intracellular level of cGMP via the activation of guanylate cyclase (11). Our original experimental protocol was modified from the steady state determination of receptor concentrations via Western blot (11) to the measurement of biosynthesis rate by pulse labeling with [ 35 S]Met/Cys. This change in protocol allowed the reduction of 8-Br-cGMP and atrial natriuretic factor (ANF) concentrations used in the present experiments to the level previously employed by others in short term protocols in cultured or isolated hepatocytes (25)(26)(27)(28)(29). HuH-7 cells were grown to near-confluence in MEM supplemented with either 10% FBS or 10% dFBS to which 10 to 1000 M 8-Br-cGMP, 10 nM ANF, or 100 M sodium nitroprusside (SNP) (shown to produce nitric oxide (25), an activator of soluble guanylate cyclase (29)) were added (Fig. 1A). Within 1 h of the addition of 500 M 8-Br-cGMP and activators of both the particulate (ANF) and soluble (SNP) guanylate cyclases (25)(26)(27)(28)(29), the biosynthetic rate of ASGR was increased by 6.7-, 8.3-, and 4.2-fold, respectively, when compared with untreated cells in dFBS alone. No difference in the abundance of specific mRNAs was detected by Northern blot analysis, regardless of whether cells were main- ANF, or (5) SNP as above. Northern blot analysis was performed using 10 g of total RNA/lane, and the transferred RNA was sequentially hybridized with probes for H1 and H2 ASGR subunits. The blot was stained with methylene blue to confirm that equal amounts of RNA had been transferred.
tained in MEM supplemented with FBS or dFBS with or without 8-Br-cGMP or its inducers (Fig. 1B). These results strongly support our hypothesis that intracellular levels of cGMP regulate the expression of the ASGR at a posttranscriptional level. When taken together with our earlier studies (11,12), these results clearly demonstrated that the molecular level of control was translational.
In Vitro Translation of ASGR-In mammalian cells, translation of most mRNAs appears to occur by association of a preinitiation complex at a 7-methylguanylate cap site and subsequent scanning to the translation initiation site (16). To establish the cap site status of ASGR mRNA in cells grown in FBS as compared with dFBS and dFBS supplemented with 500 M cGMP, the extent of cap-dependent in vitro translation was determined. Total mRNA was isolated from HuH-7 cells (5 ϫ 10 7 ) and translated in a rabbit reticulocyte lysate in the presence or absence of m 7 GpppG, an inhibitor of cap-dependent initiation (30). The labeled ASGR translation product was recovered by immunoprecipitation using a polyclonal antibody to affinity-purified human receptor. Resolution on 10% SDS-PAGE and subsequent fluorography indicated that inclusion of m 7 GpppG reduced translation of ASGR mRNA isolated from both control and biotin-deprived cells with or without 8-Br-cGMP to an equal extent (Ͼ90%) (Fig. 2). These results indicated that initiation of ASGR mRNA translation was cap-dependent and that addition of a 7-methylguanylated cap to ASGR mRNA was independent of biotin deprivation.
Localization of the Cis-acting H2 mRNA Cognate Sequence-To localize the cis-acting element, mutated H2 cDNAs from which the entire 5Ј or 3Ј-UTR was deleted were constructed by polymerase chain reaction amplification. These constructs, along with the full-length H2 cDNA, were cloned into the eukaryotic expression vector pcDNA3 carrying a neomycin resistance gene. Stable transfectants of COS-7 cells were selected with 400 g/ml G418. As shown in Fig. 3, deletion of the 5Ј-UTR resulted in loss of the cGMP requirement for H2 expression. In contrast, deletion of the 3Ј-UTR was without effect. These results indicated that the cGMP-responsive element was located in the 5Ј-UTR of the H2 mRNA. Northern blot analysis indicated that there was no significant difference in mRNA levels to account for this differential response to biotin deprivation or supplementation with 8-Br-cGMP (Fig. 3), supporting translational regulation in the transfected COS-7 cell lines.
Transient transfection of HuH-7 with the chimeric plasmid confirmed that the putative cis-acting element was localized within the 5Ј-UTR (Table I). Addition of biotin to the culture medium resulted in a two-fold increase in ␤-galactosidase activity. Interestingly, the presence of the 5Ј-UTR in the plasmid reduced ␤-galactosidase expression when compared with the original or a chimeric plasmid in which the 5Ј-UTR was inserted in the antisense orientation by almost 35%, even when biotin was added. This finding was consistent with the reduced level of H2b translation when compared with the H1 ASGR subunit under normal physiologic conditions (2).
RNA Gel Retardation Assay-The 5Ј-UTR (187-base pair) cDNA fragment of the H2b of ASGR was directionally cloned into pGEM-4Z vector for the generation of a nested set of 5Ј-UTR mRNA fragments by in vitro transcription for an RNAprotein binding assay (Fig. 4). RNA fragments of the 5Ј-UTR were added in 100-fold molar excess prior to the addition of the full-length 187-nucleotide-labeled RNA probe and resolution on 4% PAGE. As illustrated in Fig. 5, the failure of the Sp6-FokI transcript to inhibit the band shift assay indicated that a cognate sequence lies between 70 and 110 nucleotides relative to the Sp6 promoter. Since translational regulation due to protein-protein interactions between two regions of a tran- FIG. 2. Cap-dependent translatability of ASGR H2 subunit mRNA isolated from biotin-deprived HepG2. Cells were maintained in MEM supplemented with 10% FBS or 10% dFBS with or without 500 M 8-Br-cGMP for 24 h before the isolation of mRNA. Equal amounts of mRNA (2 g) were added to the translation mixture plus or minus m 7 GpppG. ASGR H2 translation product was recovered by immunoprecipitation using a subunit-specific antibody, as described previously (49). The resulting fluorogram was quantified by densitometry.

FIG. 3. Expression of the H2 subunit of ASGR in stably transfected COS-7 cells.
The full-length (H 2 FL) and the 5Ј (H 2 5D) or 3Ј (H 2 3D) deleted UTR sequences of the H2b of ASGR cDNAs were subcloned into pcDNA3 vector and stable transfectants selected with 400 g/ml of G418. The cell lines were grown to near-confluence in MEM supplemented with 10% FBS or dFBS with or without 1.0 mM cGMP. Cells were metabolically labeled with [ 35 S]Met/Cys, and the extent of ASGR polypeptide synthesis was estimated as described in Fig. 1. The failure of growing cells in dFBS to inhibit ASGR expression by the H 2 5D line strongly indicates that the cGMP cognate sequence was located in the 5Ј-UTR of H2b mRNA. The film was exposed for 72 h for H 2 3D and H 2 FL and for 18 h for the H 2 5D cell line. Northern blot analysis was performed using 10 g total RNA/lane, and the transfer was hybridized with an H2-specific probe. The blot was stained with methylene blue to confirm that equal amounts of RNA had been transferred.

TABLE I 5Ј-Untranslated region mediates biotin-dependent expression of
␤-galactosidase activity in chimeric plasmid HuH-7 cells were transfected with pSV-␤-gal plasmid without (control) or with the H2b 5Ј-UTR insert and growth for 24 h in MEM supplemented with 10% dialyzed FBS. Cells were harvested and replated at a 1:3 ratio in the same medium with or without 10 Ϫ8 M biotin supplementation. When the cells reached near-confluence (72 h postplating), the levels of ␤-galactosidase activity in cell lysates were determined. script has been reported (33), a potential and equally critical role for the upstream 1-70-nucleotide (Sp6-FokI) region cannot be eliminated by the present study. Failure of the 208-nucleotide 3Ј-UTR fragment of the H2 mRNA and the glyceraldehyde-3-phosphate dehydrogenase open reading frame mRNA to inhibit the gel retardation assay further supported the specificity of this assay (Fig. 5B).
As shown in Fig. 6, incubation of the 187-nucleotide 5Ј-UTR probe with increasing amounts of the S-100 fraction isolated from HuH-7 cells grown to confluence in MEM supplemented with FBS, dFBS ϩ 1.0 mM 8-Br-cGMP (conditions required for normalASGRexpression),ordFBSaloneshowedaconcentrationdependent gel retardation. As the concentration of S-100 protein was increased to 1 g/assay, it became evident that the S-100 fraction isolated from HuH-7 cells grown in dFBS had a higher effective concentration of RNA-binding protein than cells grown in FBS and that inclusion of 8-Br-cGMP to the cell culture reduced the concentration to the control level. This quantitative (as opposed to a qualitative difference) was consistent with the bimodal distribution of ASGR mRNA between the translationally active polysomes isolated from cells grown in FBS and the shift to the repressed state when cells were grown in dFBS (11,12). DISCUSSION In mammalian cells, translation of most mRNA species appears to occur by the association of a preinitiation complex at a 7-methylguanylate cap site and subsequent scanning downstream to the site of protein synthesis initiation (16,31). Our studies showed that cap site addition to ASGR mRNA was independent of biotin deprivation (Fig. 2). However, the recovery of the ASGR message in the ribonucleoprotein fraction during biotin deprivation (11,12) suggested that intracellular levels of cGMP play a significant role in modulating the initiation phase of translation (31).
Perhaps the best-defined example of translational regulation is that of ferritin synthesis in iron-deficient cells (16,30). Analysis of the cis-acting mRNA sequences led to the definition of the iron-responsive element with a putative stem-loop structure in the 5Ј-UTR (30). Modeling of the ASGR H2b 5Ј-UTR indicated the presence of two potential regions of secondary structure. The free energy levels (Ϫ7.4 and Ϫ8.5 kcal/mol) of these two stem-looped regions was far below that usually considered necessary to prevent recognition of a cap site (Ϫ50 kcal/mol). However, they might provide the loop structure necessary for specific recognition by exposing the RNA backbone and bases to interaction with protein groups (32,33).
In the absence of a highly ordered 5Ј-UTR stem-loop structure, translation may be regulated by a short linear sequence (16). A highly conserved CCAUCNN sequence localized within the 5Ј-UTR of both ASGR subunit mRNAs isolated from either human or rat has been identified as a conserved RNA-binding protein cognate sequence within the 5Ј-UTR of ornithine decarboxylase (34). The presence of this conserved sequence within the putative cis-acting element as indicated by gel retardation assay (Fig. 5) supports the possibility that it may serve as a recognition motif for the cGMP responsive trans-acting factor.
One plausible explanation for cGMP-regulated expression of ASGR would be modulation of a trans-acting factor phosphorylation status. Although the modulation of ASGR by cGMP was not liver-specific (Fig. 6), it should be viewed in the context of the original finding in HepG2 cells (11,12). Since there was little, if any, cGMP-dependent protein kinase detected in hepatocytes (35,36), the classic cGMP signal transduction pathways mediated by cGMP-dependent protein kinase was presumed to be absent in liver cells (36). Therefore, if a phosphorylation/ dephosphorylation signal transduction pathway was involved in translational regulation of ASGR expression, one of the cGMP-binding phosphodiesterases (PDEs) would be the most likely effector target. As opposed to cGMP-dependent protein kinase, the various cGMP PDEs are regulated allosterically by the binding of cGMP to noncatalytic binding sites (37). Modulation of cGMP levels can either inhibit or stimulate PDE hydrolytic activity, increasing or decreasing intracellular  (Fig. 4) were added in 100-fold molar excess to the S-100 incubation mixture 10 min prior to the addition of the Sp6-SmaI 32 P-labeled probe (0.5 ng). The failure of the Sp6-FokI fragment to inhibit the probe retardation indicates that a cognate sequence lies between the FokI and HinfI fragment. B, the addition of a 100-fold molar excess of full-length 5Ј-UTR was compared with a 100-fold molar excess of the 208-nucleotide 3Ј-UTR and glyceraldehyde-3-phosphate dehydrogenase open reading frame mRNAs. The failure to inhibit the gel retardation by the 3Ј-UTR and GAPD mRNA fragments confirmed the specificity of the assay.
FIG. 6. Titration of 5-UTR-binding protein by band-shift assay. Increasing amounts of the S-100 cytosolic protein were incubated with the 187-nucleotide in vitro transcription product (10,000 cpm) before resolution on a 4% native gel. Among various preparations of S-100 isolated from cells grown in MEM supplemented with FBS or dFBS ϩ cGMP when compared with dFBS alone, the two-fold difference in protein concentration required for a positive band shift was the minimum observed. At intermediate concentrations, both shifted and free probe were presented in the lane (data not shown). The gel was dried, and the bands were localized by autoradiography. cGMP itself or cAMP (38). Indeed, a number of recent studies have suggested that cGMP-stimulated PDE may play a central role in regulating the intracellular concentrations of cAMP (39,40). Based on our previous findings that increased levels of cAMP resulted in the down-regulation of ASGR (41), induction of a cGMP-stimulated PDE resulting in a protein dephosphorylation via reduction of cAMP is a reasonable mechanism for cGMP-regulated expression of ASGR.
Presently, it may be premature to speculate that changes in the phosphorylation state of any protein via PDE mediates the effects of cGMP on translation of ASGR, especially in the light of the recent discovery of new types of cyclic nucleotide receptors that include other non-catalytic sites (37). Our understanding of the potential cGMP cascade in liver is still in its infancy, and there may yet be other schemes to account for cGMP action, such as Ca ϩ2 ion flux or induction of inositol triphosphate (42,43). Whatever the biochemical mechanism of the cGMP action may be, it is reasonable to speculate that in the absence of cGMP, a negative trans-acting factor associates with the 5Ј-UTR of the ASGR mRNA, thereby inhibiting translation. Purification of the ASGR mRNA-binding protein should provide new insight into the physiologic affect and molecular target of cGMP in the hepatocyte.