Parathyroid Hormone-related Protein Interacts with RNA*

Parathyroid hormone-related protein (PTHrP) is a secreted protein that acts as an autocrine and paracrine mediator of cell proliferation and differentiation. In addition to its biological activity that is mediated through signal transduction cascades, there is evidence for an intracellular role for PTHrP in cell cycle progression and apoptosis. These effects are mediated through a mid-region nuclear targeting sequence (NTS) that localizes PTHrP to the region of the nucleolus where ribonucleoprotein complexes form in vivo. In this work, we show that endogenous, transfected, andin vitro translated PTHrP proteins bind homopolymeric and total cellular RNAs at salt concentrations up to 1 m. A peptide representing the PTHrP NTS was effective in competing with the wild-type protein for RNA binding, whereas a similar peptide representing the nucleolin NTS was not. Site-directed mutagenesis revealed that the binding of PTHrP to RNA was direct and was dependent on preservation of a core GXKKXXK motif, embedded in the PTHrP NTS, which is shared with other RNA-binding proteins. The current observations are the first to document RNA binding by a secreted cellular protein and predict a role for PTHrP in regulating RNA metabolism that may be related to its localization in the nucleolus of cells in vivo.

Parathyroid hormone-related protein (PTHrP) 1 is a secretory protein that is both structurally and functionally related to PTH, the major regulator of calcium homeostasis (1,2). As a consequence of sequence homology in amino acids 1-34, PTH and PTHrP bind to a common G-protein-coupled receptor in bone and kidney through which they elicit a spectrum of biological activity related to calcium and phosphate homeostasis (3). In addition to these effects elicited through signal transduction cascades, we (4,5) and others (6,7) have shown that some of the biological activity of PTHrP is mediated through amino acids 87-107, which constitute a nuclear/nucleolar targeting sequence (NTS). The PTHrP NTS bears some similarity to both the lysine-rich bipartite sequences seen in proteins such as nucleolin (8) and the arginine-rich sequence in the retroviral protein Tat (9) that mediates binding of the protein to RNA as well as nucleolar targeting. The PTHrP NTS also contains, at its amino terminus, a lysine-rich motif similar to the consensus core that has been identified in numerous double-stranded RNA-binding proteins (10).
Nucleoli are the sites within the nucleus that are involved primarily in the transcription and processing of ribosomal RNA and its assembly into ribonucleoprotein complexes prior to export into the cytoplasm (11). As such, they are prominent in interphase cells undergoing rapid protein synthesis and disappear as distinct entities in mitotic or metabolically inactive cells. Morphological analysis at electron microscopic resolution defines the nucleolus as a membrane-free organelle containing regions of varying electron density (12). In previous work, we localized PTHrP, by immunoelectron microscopy, to the dense fibrillar component of nucleoli in tissue sections from fetal rat bone (4). This region represents complexes of newly transcribed 45 S rRNA and protein that are subsequently processed and assembled into ribosomes. The complex series of events involved in ribosome biogenesis depends on the presence of numerous nucleolar proteins that have known RNA binding properties.
In a manner similar to protein/protein interactions and protein/DNA interactions, the specific and stable association of RNA and protein is most often mediated through recognition motifs (13). In many cases, a larger module containing several repeats of shorter motifs is required for efficient interaction, whereas in other instances, a self-contained motif appears to be sufficient. Perhaps the best characterized and most widely recognized RNA-binding proteins are those that contain multiple copies of either ribonucleoprotein (RNP) (14) or K homology (KH) domains (15). These modules stretch over 80 -100 amino acids and are often found in conjunction with RGG boxes (13). Other proteins that have been loosely categorized on the basis of structural homology and potential biological function are those such as Tat that contain an arginine-rich motif. Yet another large (Ͼ300) group of loosely categorized proteins, to which TRBP (TAR-binding protein) (16), GCN2 (17), and PKR (18) belong, are those that bind to double-stranded RNA. These proteins contain one or more copies of a double-stranded RNA (dsRNA)-binding motif with a lysine-rich core.
To extend our previous observations toward defining the role played by PTHrP in the region of the nucleolus engaged in rRNA synthesis and processing, we investigated the possibility that PTHrP binds RNA. We now demonstrate that both endogenous and transfected PTHrP proteins bind poly(G) homopolymeric RNA, GC-rich double-stranded RNA, and total cellular RNA. The interaction is of high relative affinity and is depend-ent on the presence of a core lysine motif shared with other RNA-binding proteins.

MATERIALS AND METHODS
Plasmids and Modifications-The previously described plasmids pPTHrP/3 (rat PTHrP) and p⌬87-107PTHrP/3 (4) were modified to include an in-frame Myc tag. Oligonucleotides myctag 3 (5Ј-cggaattcgcgtctatggaacaaaagctgattagcgaagagg-3Ј) and myctag 4 (5Ј-ccgctcgagtcagttgttcaggtcctcttcgctaatcagc-3Ј) were used to generate a polymerase chain reaction product encoding the Myc epitope flanked by a 5Ј-EcoRI site and a 3Ј-XhoI site. After digestion with EcoRI and XhoI, the fragment was ligated into the pcDNA3 expression vector (Invitrogen) to generate pcDNA3myc. Oligonucleotides myctag 1 (5Ј-cgggatcccacgatgctgcggaggctg-3Ј) and myctag 2 (5Ј-ggaattcatgcgtccttgagctggg-3Ј) were used to generate PTHrP DNA fragments with a 5Ј-BamHI site and a 3Ј-EcoRI site using pPTHrP/3 and p⌬87-107PTHrP/3 as templates. The fragments were digested with BamHI and EcoRI and ligated into the BamHI/EcoRI site of pcDNA3myc. The resulting plasmids, PTHrPmyc and PTHrPmyc-NTS, were verified by DNA sequencing. Mutations were introduced into the target plasmid pPTHrP/3 using the Chameleon mutagenesis kit (Stratagene). The following series of mutant cDNAs was generated with selectively altered residues within one of three distinct basic domains in the PTHrP NTS: M1, 87 GKKKK 91 to 87 GEEKI 91 , M2, 96 KRREQ 100 to 96 KGTEL 100 ; and M3, 102 KKKRR 106 to 102 IIERG 106 . The resulting cDNAs, M1PTHrP, M2PTHrP, and M3PTHrP, were modified to include an in-frame Myc tag as described above. Human FMR1 was used as a positive control to demonstrate binding to ribosomal RNA proteins (19).
Cell Culture and Transient Transfections-All cell culture reagents were purchased from Life Technologies, Inc., and all plasticware (Falcon) was from Becton Dickinson Labware (Lincoln Park, NJ). The PTr cell line (a kind gift of G. N. Hendy, Calcium Research Laboratory, Montreal, Canada) was maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated calf serum and antibiotic/antimycotic. COS-1 cells were maintained and transfected as described previously using a mock transfection for control (4). Briefly, 48 h after transfection, cells were lysed with buffer for binding assays and immunoblot analysis. PTr cells plated at a density of 2 ϫ 10 6 /150 mm 2 were grown to 80% confluence, washed twice with phosphate-buffered saline, and harvested by scraping into lysis buffer (150 mM NaCl, 20 mM Tris (pH 7.4), 20 mM NaF, 0.1 mM sodium vanadate, 1% Triton X-100, 500 M phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, and 15 g/ml aprotinin). Cell lysates were cleared of debris by centrifugation prior to use in binding assays.
In Vitro Transcription and Translation-Transcription and translation were performed in vitro using pPTHrP/3 or p⌬87-107PTHrP/3 as a DNA template with the TNT T7 coupled rabbit reticulocyte lysate system (Promega) according to the manufacturer's instructions. 1 g of plasmid DNA was added to the rabbit reticulocyte lysate containing T7 RNA polymerase and 500 Ci/ml [ 3 H]leucine (ICN). 2 l of the translation product was removed to represent the total cell lysate prior to dividing the remaining product equally for RNA binding experiments.
Binding to Homopolymeric RNA and DNA-Ribonucleotide and deoxyribonucleotide binding assays were performed essentially as described previously (20,21) using the following synthetic homopolymers: poly(A), poly(C), poly(G), poly(U), or double-stranded poly(G)⅐poly(C) and poly(dG)⅐poly(dC) coupled to Sepharose beads (Sigma). The beads were washed and reconstituted to 50% (v/v) in phosphate-buffered saline. Binding reactions were carried out in 0.2 ml of lysis buffer containing 2 l of in vitro translated product, 100 g of total protein from transfected COS-1 cells, or 300 g of total protein from PTr cells. Binding was performed at the indicated NaCl concentrations for 1 h at 4°C on a rocking platform. The beads were pelleted by brief centrifugation and washed three times in lysis buffer prior to resuspension in 20 l of Laemmli sample buffer. Bound proteins were eluted from the nucleic acid by heating at 95°C for 5 min, resolved on 15% SDSpolyacrylamide gels, and analyzed either by autoradiography or by immunoblot analysis.
Immunoblot Analysis-The polyclonal antiserum raised against rat PTHrP-(67-86) has been previously characterized (22). The monoclonal antibody 9E10, specific for the human c-Myc epitope (23), was harvested as ascites fluid from mice 7 days after implantation of the hybridoma (American Type Culture Collection). Horseradish peroxidase-conjugated secondary antisera were purchased from Sigma (goat anti-mouse IgG) and Pierce (goat anti-rabbit IgG). Proteins denatured in Laemmli sample buffer were separated by size on 15% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. Mem-branes were blocked overnight with 5% nonfat milk protein in Trisbuffered saline with 0.05% Tween 20 and incubated overnight with either the anti-Myc or anti-PTHrP-(67-86) primary antibody. The protein bands were detected using an appropriate horseradish peroxidaseconjugated secondary antibody and the enhanced chemiluminescence system (Amersham Pharmacia Biotech).
Binding to Total Cellular RNA-Total cellular RNA was labeled by incubating HeLa cells, grown to 50% confluence, with 500 Ci of [ 32 P]orthophosphate (ICN) overnight (21) and isolated using TRIzol (Life Technologies, Inc.) according to the manufacturer's instructions. Proteins were immunoprecipitated, with either control IgG or anti-Myc antibody, from COS-1 cells transiently transfected with the various PTHrP constructs. 3 ϫ 10 6 cpm of labeled RNA was incubated with the equivalent of one-third of a cell lysate for 30 min at 4°C. The beads were washed three times in lysis buffer and counted in a Beckman scintillation counter. Immunoblot analysis of the counted samples was performed as described above using anti-Myc antibody to verify the expression level of the immunoprecipitated proteins. To examine the profile of RNAs recovered in the binding assays, bound RNAs, eluted from the precipitated proteins by denaturing in formamide buffer at 70°C for 5 min, were subjected to formaldehyde-agarose electrophoresis and transferred to a nitrocellulose membrane that was exposed to x-ray film overnight. An aliquot of unlabeled cellular RNA was run in the presence of ethidium bromide to identify 28 S and 18 S ribosomal bands on the membrane.
Peptide Synthesis and Competition Binding Assay-Peptides corresponding to amino acids 87-107 of PTHrP (PTHrP NTS) or amino acids 281-301 of nucleolin (nucleolin NTS) were synthesized with a carboxylterminal biotin tag at the Sheldon Institute for Biotechnology (McGill University, Montreal). Competition assays were performed by incubating 2 l of in vitro translated PTHrP with poly(G)-Sepharose in the presence of the peptides at concentrations of 10 Ϫ7 to 10 Ϫ3 M. The bound proteins were separated on 15% SDS-polyacrylamide gels and analyzed by immunoblotting with anti-Myc antibody. For the peptide binding studies, 10 Ϫ7 M biotinylated peptides were incubated either with poly(G)-Sepharose in lysis buffer containing up to 1000 mM NaCl or with poly(A)-, poly(C)-, or poly(U)-Sepharose in the presence of 250 mM NaCl. The beads were washed, and the bound peptides were eluted by heating in Laemmli sample buffer prior to separating on 20% SDSpolyacrylamide gels. The peptides were detected by immunoblot analysis with horseradish peroxidase-conjugated streptavidin (Stratagene).

RESULTS
In previous studies, we identified sequence homology between amino acids 87-107 of the cellular protein PTHrP and amino acids 48 -57 of the retroviral regulatory protein Tat (4). In the case of Tat, the same amino acids represent an argininerich RNA-binding motif. Taken together with our previous observation that PTHrP localizes to the dense fibrillar component of the nucleolus, we sought to determine if PTHrP is an RNA-binding protein and if the interaction is mediated through its NTS.
Binding of Transiently Expressed PTHrP to Cellular RNA and Homopolymeric RNA-We first examined the possibility that transiently expressed PTHrP could associate with total cellular RNA. COS-1 cells were transfected with an expression vector encoding full-length rat PTHrP with the Myc epitope tag at the carboxyl terminus (PTHrPmyc) to facilitate identification of the expressed protein. Lysates of PTHrPmyc-transfected or mock-transfected COS-1 cells were immunoprecipitated with control (IgG) or anti-Myc antibody. The immunoprecipitated proteins were then incubated with 32 P-labeled total cellular RNA and washed, and the bound RNA was quantitated by scintillation counting. A significant proportion of the radiolabeled RNA was observed in anti-Myc immunoprecipitates of PTHrPmyc-transfected cells, but not in mock-transfected cells or IgG immunoprecipitates (Fig. 1A). Immunoblot analysis of the counted samples was then performed, using anti-Myc antibody, to verify the equality of expression levels of the immunoprecipitated proteins (data not shown). To determine whether the PTHrP NTS was required for this interaction, we transfected COS-1 cells with an expression plasmid encoding PTHrPmyc devoid of the NTS (PTHrPmyc-NTS) and analyzed the lysates for RNA binding. De-spite an equivalent level of PTHrP expression, Myc immunoprecipitates of PTHrPmycϪNTS-expressing cells showed no significant binding to radiolabeled RNA (Fig. 1A), suggesting that the NTS is required for the association of PTHrP with RNA.
To investigate whether PTHrP, like many other RNA-binding proteins (13), associated with homopolymeric RNA, COS-1 cells expressing full-length PTHrPmyc were lysed, and equal amounts of protein were incubated with poly(A), poly(C), poly(G), or poly(U) homopolymeric RNA covalently coupled to Sepharose beads. Immunoblot analysis of PTHrPmyc-expressing COS-1 cell lysates reproducibly demonstrated two to three distinct protein species that bound poly(G) homopolymeric RNA, but not poly(A), poly(C), poly(U), or Sepharose beads alone (Fig. 1B). The proteins most probably represent prepro-PTHrP and the mature protein. PTHrP protein devoid of the NTS (PTHrPmycϪNTS) was unable to associate with homopolymeric RNA (Fig. 1B). To assess the relative affinity of the PTHrPmyc/RNA interaction, we performed binding studies in the presence of a high salt concentration, which has been shown in the past to disrupt nonspecific protein/RNA interactions (24). The association between PTHrPmyc and poly(G)-Sepharose was stable and persisted at a salt concentration of 1 M (Fig. 1C). These data suggest that PTHrP is a bona fide RNA-binding protein and that the association requires all or part of the NTS.
Profile of RNAs Recovered in Binding Studies-To identify the species of RNA that bound to PTHrP immobilized on the Sepharose beads, the binding studies were repeated using total RNA harvested from 32 P-labeled HeLa cells and protein expressed in COS-1 cells from cDNAs encoding wild-type PTHr-Pmyc, PTHrPmycϪNTS, and human FMR1myc. RNAs eluted from the protein-bound beads were then subjected to formaldehyde-agarose gel electrophoresis and autoradiography. As shown in Fig. 2A, ribosomal RNA species bound to both wildtype PTHrPmyc and human FMR1myc, but not to PTHrPmycϪNTS. Fig. 2B demonstrates immunoblot analysis of the proteins harvested from transfected COS-1 cells that were used for RNA binding.
Binding of Endogenous PTHrP to Homopolymeric RNA-To further explore the apparent specificity of the interaction between PTHrP and poly(G) homopolymeric RNA, the PTr cell line was used as a source of endogenous PTHrP. Lysates of PTr cells were incubated with different homopolymeric RNAs conjugated to Sepharose beads, and the bound proteins were separated by SDS-polyacrylamide gel electrophoresis and analyzed by immunoblotting with PTHrP-(67-86) antiserum. As was true of transfected PTHrP, endogenous PTHrP bound specifically to poly(G), but not to poly(C)-, poly(A)-, or poly(U)-Sepharose or Sepharose beads alone (Fig. 3A). The bound protein migrated with an apparent molecular mass of 27 kDa, and the association was capable of withstanding salt concentrations up to 1 M, indicating that the interaction was specific and of apparent high affinity (Fig. 3B).
Effects of Amino Acid Substitutions within the PTHrP NTS on RNA Binding-To define the minimum sequence within the NTS required for RNA binding activity, we tested several PTHrP constructs that contain mutations in this region (Fig.  4A). We have shown previously that mutation of 87 GKKKK 91 to 87 GEEKI 91 (M1PTHrPmyc) effectively blocked the nuclear/nucleolar targeting function of the PTHrP NTS. Substitution of 102 KKKRR 106 for 102 IIERG 106 (M3PTHrPmyc) was only partially effective in this respect, and mutation of 96 KRREQ 100 to 96 KGTEL 100 (M2PTHrPmyc) had little or no effect on targeting the protein to the nuclear compartment (33). To determine whether the proteins expressed from these cDNAs bound RNA, COS-1 cells were transfected with the expression vectors, and the lysates were used for homopolymeric binding studies. All of the proteins bound poly(G)-Sepharose at 250 mM salt, but considerable differences were observed at higher salt concentrations (Fig. 4B). The binding of both M1PTHrPmyc and M3PTHrPmyc to poly(G)-Sepharose was absent or severely impaired at 750 mM salt, whereas the binding of M2PTHrPmyc was unaffected at that salt concentration (Fig. 4B). These data suggest that the overall composition of the basic residues in the NTS affects its interaction with poly(G) homopolymeric RNA. We also examined the ability of the mutant proteins to associate with total cellular RNA (Fig. 4C). M2PTHrPmyc and wildtype PTHrPmyc bound radiolabeled cellular RNA with similar apparent affinities. However, the ability of M1PTHrPmyc to bind RNA was severely impaired, and that of M3PTHrPmyc was partially impaired (Fig. 4C). These observations were consistent with the data demonstrating compromised binding of the mutant PTHrPmyc proteins to poly(G)-Sepharose.
PTHrP NTS Peptide Competes with in Vitro Translated PTHrP for RNA Binding-We examined the possibility that the interaction between PTHrP and RNA was direct by incubating in vitro translated PTHrP with poly(G) RNA. Radiolabeled PTHrP was generated by incubating the full-length rat PTHrPmyc cDNA in the presence of [ 3 H]leucine in a rabbit reticulocyte lysate. In vitro translated PTHrP bound poly(G)-Sepharose, and the binding persisted at a 1 M salt concentration (Fig. 5A), suggesting that the association between PTHrP and RNA was direct. The specificity of the association was then tested by performing competition binding studies in the presence of peptides corresponding to either the PTHrP NTS or the nucleolin NTS. Amino acids 87-107 of PTHrP and amino acids 281-301 of nucleolin are similar lysine-rich bipartite motifs that target the respective proteins to the nuclear compartment (4,8). When in vitro translated PTHrPmyc was incubated with poly(G)-Sepharose in the presence of the PTHrP NTS peptide at a concentration of 10 Ϫ4 M, the interaction between PTHrPmyc and RNA was effectively blocked (Fig. 5B). The nucleolin NTS peptide, on the other hand, was ineffective in this capacity even at 10 Ϫ3 M (Fig. 5B). The high concentration of peptide required to dislodge in vitro translated PTHrPmyc from poly(G) homopolymeric RNA was a further indication that the association was specific and of apparent high affinity. These competition data suggested that the PTHrP NTS peptide might bind RNA directly. This possibility was tested by incubating biotinylated PTHrP NTS peptide with the different homopolymeric RNAs. The bound peptides were separated by SDS-polyacrylamide gel electrophoresis and detected by horseradish peroxidase-conjugated streptavidin. As was true of the full-length protein, the PTHrP NTS peptide bound specifically to poly(G)-Sepharose beads (Fig. 5C), and the binding withstood salt concentrations up to 1000 mM. The nucleolin NTS, on the other hand, was incapable of binding to any of the synthetic homopolymers. These experiments confirmed that the interaction between the PTHrP NTS and poly(G)-Sepharose was specific and direct and was not merely a function of its ionic strength.
Binding of PTHrP to Double-stranded Nucleotide Sequences-TRBP is one of a number of cellular proteins that associate with GC-rich double-stranded RNA, including the stem-loop structure formed by the Tat transactivating or TAR region of FIG. 2. Profile of cellular RNAs bound to transiently expressed PTHrP. Total cellular RNA harvested from 32 P-labeled HeLa cells was incubated with proteins precipitated from COS-1 cells transfected with cDNAs encoding PTHrPmyc, PTHrPmycϪNTS (-NTSPTHrPmyc), or human FMR1myc (hFMRmyc). A, bound RNA eluted from precipitated proteins was resolved on formaldehyde-containing 1% agarose gels and transferred to nitrocellulose membranes that were exposed to x-ray film overnight. Positions of 28 S and 18 S ribosomal bands were verified by ethidium bromide staining. Labeled RNA was eluted from anti-Myc immunoprecipitates (IP) from lysates of cells expressing PTHr-Pmyc and human FMR1myc, but not from PTHrPmycϪNTS-expressing cells or when IgG was used in place of anti-Myc antibody. B, shown are the results from immunoblot analysis of aliquots of Sepharose beads recovered by immunoprecipitation. nascent viral RNA (16). In view of the similarities between Tat and PTHrP, we investigated the possibility that PTHrP might also bind to GC polynucleotide sequences. In the presence of 250 mM salt, wild-type and mutant PTHrP proteins bound poly(G)⅐poly(C) RNA, but did not demonstrate significant binding to the corresponding deoxynucleotide sequences (Fig. 6A). The association between poly(G)⅐poly(C) and PTHrP was maintained in the presence of 1 M salt, indicating an apparent high affinity interaction (Fig. 6B). The same pattern of compromised binding of M1PTHrPmyc and M3PTHrPmyc was observed with poly(G)⅐poly(C) compared with poly(G) alone, whereas M2PTHrPmyc had little effect. These data suggest that PTHrP is capable of binding poly(G)-rich sequences in the context of both single-and double-stranded RNAs, but does not bind DNA. DISCUSSION We have shown that endogenous and transfected PTHrP proteins bind to single-and double-stranded RNAs at salt concentrations up to 1 M. The binding is direct and is highly dependent on the presence of a core GXKKXXK motif within PTHrP that is conserved among other RNA-binding proteins. A peptide corresponding to the PTHrP NTS can bind alone to poly(G)-Sepharose and can compete with wild-type PTHrP for binding to the homopolymer. Site-directed mutagenesis revealed that the RNA binding activity of PTHrP was direct and dependent on preservation of the core motif.
The endocrine, paracrine, and autocrine mechanisms of PTHrP action, which are mediated through transduction cascades following the interaction of secreted PTHrP with the common PTH/PTHrP cell-surface receptor, are well documented. On the other hand, little is known regarding the documented intracellular role played by PTHrP. The PTHrP NTS is situated in a highly conserved region of the protein that has also been identified as a site of endoproteolytic cleavage (25). However, the presence of an intact NTS has been shown to have biological relevance not only in protecting serum-deprived chondrocytes from apoptosis (4), but also as a requirement for PTHrP-induced mitogenesis in vascular smooth muscle cells (6). Others have shown that PTHrP is targeted to the nucleolus of cells in G 1 and suggest that its subcellular distribution is cell-cycle dependent (7). As a first step toward defining the molecular basis for these intracellular actions of PTHrP, we now show that PTHrP is a sequence-specific RNA-binding protein and that the association with RNA is dependent on the presence of a core motif found in double-stranded RNA-binding proteins.
In the absence of highly specific PTHrP antibodies, we modified all of our expression constructs to include a Myc epitope tag. Using the anti-Myc antibody, two major species of PTHrP were commonly detected in immunoblot analysis of lysates prepared from PTHrPmyc-transfected COS-1 cells. They migrated with apparent molecular sizes slightly larger (due to the 11-amino acid Myc tag) than those previously detected with an antiserum raised against the amino terminus of PTHrP (4). An antiserum that was raised against synthetic PTHrP-(67-86) recognized a major species of endogenous PTHrP that bound to poly(G)-Sepharose and migrated with an apparent size similar  4). C, a similar pattern of affinity for radiolabeled cellular RNA was evident in that binding of M1PTHrPmyc was reduced to near background levels and that of M3PTHrPmyc was reduced to ϳ50% of that seen for the intact protein or M2PTHrPmyc.
to that of prepro-PTHrP. Thus, it appeared that more than one species of PTHrP could associate with homopolymeric RNA.
The strength of binding of a protein to a specific nucleotide sequence, whether it be DNA or RNA, can be assessed by performing the binding studies at high salt concentrations. If the binding is weak or requires additional sequences outside of those being tested, then the association will not withstand an elevation in salt concentration (13). In this context, binding of the KH domain proteins hnRNP K and FMR1 to RNA is mediated through a combination of KH domains and an RGG box (24,26). Binding of in vitro translated, wild-type FMR1 to poly(G)-Sepharose was severely compromised at a salt concentration of 500 mM. Binding of the prototypical hnRNP K protein to RNA, on the other hand, was resistant to 1 M salt, as was demonstrated for PTHrP in our own work. Homopolymeric RNA binding in the presence of high salt concentrations was abrogated by deletion of the RGG box or the carboxyl-terminal KH domain in FMR1 (24). Variable alterations in binding to RNA at high salt concentrations resulted from point mutations of critical Ile residues in the three KH domains in hnRNP K (26). In a similar manner, we have shown that deletion of the PTHrP NTS results in abrogation of binding to poly(G)-Sepharose and that conservative mutations within the NTS lead to variable alterations in binding to RNA at high salt concentrations. It is interesting to note that, despite their lack of structural similarities, PTHrP and FMR1 both bind 28 S and 18 S species of ribosomal RNA with an apparent equal affinity at low salt concentrations. As is true for many other RNA-binding proteins, the physiological significance of the different structural motifs and their RNA binding affinities and specificities awaits further investigation.
In the absence of the NTS, there was no specific binding to total cellular RNA or to homopolymeric RNA. However, deletion of a stretch of 20 amino acids could influence the overall structure of PTHrP, thus preventing it from making contact with its RNA target. We therefore sought to map the point of contact between PTHrP and RNA using a set of NTS mutants that had previously been generated for immunofluorescent localization of the protein. In those studies, it was shown that substitution of the amino-terminal lysines was most effective in preventing nuclear/nucleolar targeting of PTHrP (33). More recently, others have demonstrated that proteins in which either the amino-or carboxyl-terminal lysine tract was deleted failed to stimulate proliferation when expressed in vascular smooth muscle cells, as did intact PTHrP (6). We now show that substitution of these same basic amino acids, particularly the amino-terminal stretch, compromises binding of the protein to both synthetic and total cellular RNAs. This is in keeping with the observation that PTHrP 87 GXKKXXK 93 constitutes a core motif that is also found in the functionally heterogeneous group of proteins containing dsRNA-binding domains. The general structure of the dsRNA-binding domain is ␣␤␤␤␣, and it is believed that the junction between the third ␤-sheet and the ␣-helix plays a critical role in RNA binding (10). PTHrP does not conform to these structural requirements in that it has helix-breaking prolines at positions 86 and 94 that flank the core motif and does not possess the extended dsRNA-binding domains. Despite this lack of overall homology, mutation of specific lysines within the core motif resulted in severely compromised binding of PTHrP to RNA in our own studies as well as in those of others examining binding of the dsRNA-binding domain proteins PKR (27) and GCN2 (17) to RNA. In the case of PTHrP, substitution of other lysine or arginine residues in the NTS had only a modest or no effect on binding.
Another member of the family of dsRNA-binding proteins is TRBP, which was isolated from a HeLa cell expression library on the basis of its ability to bind with high affinity to the GC-rich double-stranded stem of TAR RNA (16,28). The prototypical TAR-binding protein, Tat, on the other hand, appears to identify single-stranded loops in the context of the doublestranded stem (29). We have demonstrated that PTHrP binds avidly to G-rich sequences in both a single-and doublestranded context. It is therefore possible that the cellular RNA target of PTHrP might be a defined structure similar to that formed by TAR. In view of evidence that suggests that some rDNA transcription takes place in the dense fibrillar core of the nucleolus where PTHrP localizes in vivo (30), it is possible that it might play some role in the processivity of RNA polymerase I, as has been suggested for Tat in the case of RNA polymerase II (31). PTHrP could inhibit rRNA synthesis and allow the cell to arrest in G 0 /G 1 in an environment that could not support proliferation. In support of this hypothesis is evidence that transfected PTHrP was found in the nucleolus of only 10 -15% of randomly cycling cells (4) and that endogenous PTHrP localized to the nucleolus only in quiescent keratinocytes (7).
An alternative role that PTHrP might play in the nucleus is in the processing of pre-rRNA. However, despite its nucleolar localization, the lack of repetitive RNP motifs or RGG boxes appears to exclude it from the RNA recognition motif family, to which the large group of small nuclear/nucleolar RNA-binding proteins involved in RNA processing and transport belongs. One member of this family is nucleolin, which, like PTHrP, has a classic bipartite nuclear/nucleolar targeting signal composed of two polylysine tracts separated by a 10-amino acid spacer. It has been suggested that the nucleolin NTS localizes the protein to the nucleolus, where it is subsequently retained by binding to rRNA through RNP and RGG motifs (8,32). In our own studies, we have shown that the PTHrP NTS is both necessary and sufficient to localize the protein to the nucleus/nucleolus (4). We now show that a peptide corresponding to the PTHrP NTS binds directly, with specificity and apparent high affinity, to poly(G)-Sepharose, whereas a similar, equally lysine-rich peptide (nucleolin NTS) does not. In view of our data that identify the PTHrP NTS as the motif that mediates binding of the full-length protein to RNA, this observation strongly suggests that PTHrP binds RNA directly, rather than through some other molecule. Additionally, the inability of a peptide representing the nucleolin NTS to compete with PTHrP for binding to poly(G)-rich RNA supports the conjecture that PTHrP does not belong to the RRM family of RNA-binding proteins. However, the possibility that PTHrP plays some as yet undefined role in rRNA processing cannot be excluded.
It is evident from the above discussion that the classification of RNA-binding proteins on the basis of structural similarities does little to elucidate their biological function. Thus, the RNA binding characteristics of PTHrP most closely resemble those of the dsRNA-binding protein TRBP, whose biological function remains undefined. Our current work characterizing the RNA binding properties of PTHrP, combined with our previous observation that PTHrP localizes to the dense fibrillar core of the nucleolus, predicts a role for the protein in the transcription or processing of rRNA. However, further studies will be required to co-localize PTHrP with the subnucleolar sites of rDNA transcription and early rRNA processing events to clarify these issues.