Nuclear Targeting by the Growth Factor Midkine

Ligand-receptor internalization has been traditionally regardedas part of the cellular desensitization system. Low-densitylipoprotein receptor-related protein (LRP) is a large endocytosisreceptor with a diverse array of ligands. We recently showedthat LRP binds heparin-binding growth factor midkine. Here wedemonstrate that LRP mediates nuclear targeting by midkine andthat the nuclear targeting is biologically important. Exogenousmidkine reached the nucleus, where intact midkine was detected,within 20 min. Midkine was not internalized in LRP-deficientcells, whereas transfection of an LRP expression vector restoredmidkine internalization and subsequent nuclear translocation.Internalized midkine in the cytoplasm bound to nucleolin, anucleocytoplasmic shuttle protein. The midkine-binding siteswere mapped to acidic stretches in the N-terminal domain ofnucleolin. When the nuclear localization signal located nextto the acidic stretches was deleted, we found that the mutantnucleolin not only accumulated in the cytoplasm but also suppressedthe nuclear translocation of midkine. By using cells that overexpressedthe mutant nucleolin, we further demonstrated that the nucleartargeting was necessary for the full activity of midkine inthe promotion of cell survival. This study therefore revealsa novel role of LRP in intracellular signaling by its ligandand the importance of nucleolin in this process.

INTRODUCTION

Top

Introduction

Extracellular signaling molecules such as growth factors andcytokines bind to plasma membrane receptors that activate theirown kinases and/or recruit adapter proteins. These events onthe plasma membrane have been regarded as the onset of signaling.Subsequent to receptor binding of ligands, the ligand-receptorcomplexes are internalized and delivered to specific vesicularcompartments (e.g., early and late endosomes and lysosomes),leading to desensitization. While the ligands are often degraded,the receptors themselves are either degraded or recycled backto the cell surface.

Mounting evidence indicates that nuclear targeting by extracellularsignaling molecules plays an indispensable role in their biologicalactivities. For example, acidic fibroblast growth factor (aFGF)and Schwannoma-derived growth factor need nuclear localizationfor their mitogenic activity (28, 33, 76). For basic FGF (bFGF),increases in ribosomal gene transcription and cell proliferationare tightly correlated to the nuclear translocation of bFGF(1, 5). Thus, signals from the cell surface receptor and translocationof the ligand to the nucleus cooperate and play roles in thebiological activities of many extracellular signaling molecules.The translocation of ligands across the plasma membrane is dependenton their own plasma membrane receptors. Nuclear localizationsignals (NLSs) of ligands themselves have been implicated innuclear translocation for many ligands, such as platelet-derivedgrowth factor A (PDGF A) (11), PDGF B (41), aFGF (28, 76), gammainterferon (82), interleukin 1 ${alpha}$ (75), interleukin 1ß(24), and interleukin 5 (29). However, the precise mechanismof nuclear targeting by extracellular signaling molecules ispoorly understood.

Midkine was first identified as the product of a retinoic acid-responsivegene that is up-regulated in the differentiation system of embryonalcarcinoma cells (32, 70). Its important roles have been implicatedin various aspects of biology, such as neuronal survival anddifferentiation (48, 73, 79), carcinogenesis (10, 30, 51, 72),and tissue remodeling (31, 53, 80). At the cellular level, midkinepromotes cell growth (47, 48, 68), cell survival (54, 58, 73),cell migration (26, 43, 59, 66), and plasminogen activator activity(38). Although midkine does not have an apparent NLS, it islocalized in the nucleus in hemangioma cells (67) and in cellsin a variety of tumor tissues (data not shown). Recently, weidentified low-density lipoprotein (LDL) receptor-related protein(LRP) as a midkine-binding protein (49). Because the LRP antagonistreceptor-associated protein (RAP) suppresses midkine-mediatedneuronal cell survival, it has been suggested that LRP is acomponent of the functional midkine receptor (49).

LRP belongs to the LDL receptor family. There are five prototypemembers of the family: LDL receptor, ApoE receptor 2, very low-densitylipoprotein (VLDL) receptor, LRP, and LRP2/Megalin. The majorfunctions of these receptors are to endocytose and deliver theirligands to lysosomes for degradation or catabolism (27, 39,65). There are over 30 identified ligands of these receptors,including ApoE lipoproteins, ${alpha}$ ₂-macroglobulin, plasminogen activator,and plasminogen activator inhibitor-1 complexes, lipoproteinlipase, and thrombospondin-1 (21). Among them, ApoE lipoproteinsare common ligands for all members, whereas ${alpha}$ ₂-macroglobulinis a specific ligand for LRP (22). In addition, it was recentlyreported that some members of the LDL receptor family functionas signaling membrane receptors. ApoE receptor 2 and VLDL receptorare reelin receptors, which play a crucial role in neuronal-cellmigration during embryogenesis and which utilize adapter proteinDisabled-1 for intracellular signaling (12, 25, 71). Recentlyidentified members of the family LRP5 and -6 function, togetherwith Frizzled, as Wnt receptors, which are important for bodyaxis determination, neuronal differentiation, and carcinogenesis(45, 56, 69, 74). These findings suggest that LRP may also functionas a signaling membrane receptor for midkine. However, takinginto account that LRP exhibits the strongest endocytosis activityamong the LDL receptor family members (42), it is also an intriguinghypothesis that LRP mediates midkine internalization and isinvolved in nuclear targeting by midkine. In this paper we reportthe involvement of LRP and nucleolin, a nucleocytoplasmic shuttleprotein, in the nuclear targeting of midkine. We found that(i) LRP delivers exogenous midkine not only to the endocytosis/degradationsystem but also to the nuclear transport system, (ii) midkineendocytosis via LRP is followed by cytoplasmic nucleolin-midkineassociation, which leads to the nuclear translocation of midkine,(iii) NLS-deficient nucleolin traps midkine in the cytoplasmand inhibits its nuclear localization, and (iv) nuclear targetingis needed for the full antiapoptotic activity of midkine.

MATERIALS AND METHODS

Top

Materials and Methods

Reagents. Yeast-produced human midkine protein (26) was a generous giftfrom S. Sakuma (Meiji Milk Co. Ltd., Odawara, Japan). N-terminallybiotinylated human midkine was prepared as described previously(61). Rabbit antibasigin antibodies were raised against theextracellular domain of mouse basigin (15). ¹²⁵I-Na, ECL, proteinA-Sepharose CL-4B, and protein G-Sepharose 4 fast flow werepurchased from Amersham Pharmacia Biotech. DNase I and the pGEM-TEasy Vector System I were from Promega. pcDNA3.1+ was from Invitrogen.Lipofectamine Plus was from Life Technologies. Fluorescein isothiocyanate(FITC)-conjugated streptavidin, the anti-FLAG M2 monoclonalantibody, the tetramethyl rhodamine isocyanate-conjugated anti-mouseimmunoglobulin G (IgG) antibody, the peroxidase-conjugated monoclonalantibiotin antibody (BN-34), and propidium iodide were fromSigma. The antinucleolin mouse monoclonal antibody (MS-3) wasfrom Santa Cruz Biotechnology. The antihemagglutinin (HA) ratmonoclonal antibody (3F10) was from Roche. N-Nitrosocysteinewas from Dojin. The Lab-Tek chamber slides were from Nunc. Restrictionenzymes and a ligation kit were from Takara. The horseradishperoxidase-conjugated anti-mouse IgG antibody was from Seikagaku.Cisplatin was from Bristol-Myers Squibb.

Cells and culture. Mouse L cells were maintained in Dulbecco's modified Eagle'smedium (DMEM) supplemented with 10% fetal bovine serum and wereseeded onto dishes, chamber slides, or coverslips a day beforeeach experiment. Mouse embryonic fibroblasts (MEF) (PEA13 [LRP^-/-,LRP deficient, homozygous, CRL-2216], PEA10 [LRP^+/-, LRP deficient,heterozygous, CRL-2215], and MEF-1 [LRP^+/+, simian virus 40-transfected,CRL-2214]) (77) were obtained from the American Type CultureCollection and cultured as described above.

Radioiodination of midkine. The purified midkine protein was radioiodinated by the chloramine-Tmethod. Ten micrograms of midkine in 90 µl of 50 mM sodiumphosphate buffer, pH 7.2, was added to 1 mCi of ¹²⁵I-Na and50 µl of a freshly prepared solution of chloramine-T (1mg/ml in 50 mM sodium phosphate buffer, pH 7.2). After 1 minat 20°C, the reaction was stopped by adding 50 µlof 5 mM sodium bisulfite and 100 µl of 150 mM potassiumiodide. Free ¹²⁵I-Na was separated from ¹²⁵I-midkine by chromatographyon Sephadex G-25 (PD10; Amersham Pharmacia Biotech) equilibratedwith 50 mM sodium phosphate buffer, pH 7.2, containing 0.25%bovine serum albumin. Analysis of ¹²⁵I-midkine by sodium dodecylsulfate-15% polyacrylamide gel electrophoresis (SDS-15% PAGE)followed by autoradiography gave a single band of 17 kDa. Thespecific activity of ¹²⁵I-midkine was 5 x 10⁷ cpm/µg.

Overexpression of nucleolin and nucleolin mutants. Mouse nucleolin cDNA (nucleotides [nt] 1 to 2124) (6) taggedwith the coding sequence for a FLAG epitope sequence was obtainedfrom the RNA of mouse fibroblasts by means of PCR with reversetranscription. Several nucleolin mutants, in which the NLS (encodedby nt 841 to 897), N terminus (encoded by nt 1 to 831), or Cterminus (encoded by nt 907 to 2124) was deleted were designedas shown in Fig. 6. For mutants with acidic stretches, acidicstretches were tentatively defined as follows: AS1, encodedby nt 1 to 135; AS2, encoded by nt 136 to 510; AS3, encodedby nt 511 to 654; AS4, encoded by nt 655 to 831. Subcloningof PCR products was performed with pGEM-T Easy Vector SystemI. After DNA sequencing, nucleolin expression vectors were constructedin pcDNA3.1(+). Each vector was transiently transfected intocells with Lipofectamine Plus a day before each experiment.Cells were lysed with a lysis buffer (20 mM Tris-HCl [pH 7.5],300 mM sucrose, 60 mM KCl, 15 mM NaCl, 0.5 mM EDTA, 0.5% [vol/vol]Triton X-100, 1 mM phenylmethylsulfonyl fluoride [PMSF], 2 µgof aprotinin/ml, 10 µg of leupeptin/ml, and 1 mM sodiumvanadate) and then centrifuged at 10,000 x g at 4°C for10 min. A supernatant (cell lysate) and its immunoprecipitatewith the anti-FLAG M2 monoclonal antibody were prepared andanalyzed by Western blotting with the anti-FLAG M2 monoclonalantibody, the horseradish peroxidase-conjugated anti-mouse IgGantibody, and the ECL system.

View larger version (62K):
[in this window]
[in a new window]
FIG. 6. Mapping of midkine-binding sites in nucleolin. (A) Structures of nucleolin and its derivatives. AS, acidic stretch; GAR, glycine- and arginine-rich domain; MK, midkine; RBD, RNA-binding domain. At the right, the ability to bind to midkine (+ or -) is shown. (B) Various FLAG-tagged nucleolin expression vectors were transfected into LRP^+/+ cells. After incubation with ¹²⁵I-midkine at 4°C, cells were allowed to internalize ¹²⁵I-midkine by incubation at 37°C for 20 min. The binding between ¹²⁵I-midkine and overexpressed FLAG-tagged nucleolin was assessed by precipitation with anti-FLAG antibody as described in Materials and Methods.

Kinetic analysis of endocytosis and cell fractionation. The binding of ¹²⁵I-midkine was carried out for 2 h at 4°Cwith 20 ng of ¹²⁵I-midkine/ml in culture medium. Unbound ligandwas removed by washing with ice-cold culture medium, and midkine-boundcells were incubated for various times at 37°C in prewarmedmedium. At each time point, the plates were quickly placed onice and the medium was replaced with ice-cold phosphate-bufferedsaline (PBS). Cells were harvested and fractionated into detergent-solubleand nuclear fractions by a method previously described (1, 17,81). Briefly, the cells were harvested by trypsinization (0.125%trypsin, 0.025% EDTA in PBS at 37°C for 10 min) and thenlysed in 1 ml of 20 mM Tris-HCl, pH 7.5, containing 300 mM sucrose,60 mM KCl, 15 mM NaCl, 0.5 mM EDTA, 0.5% (vol/vol) Triton X-100,1 mM PMSF, 2 µg of aprotinin/ml, 10 µg of leupeptin/ml,and 1 mM sodium vanadate (nuclear preparation buffer). The detergent-solublefraction was obtained by removal of nuclei with low-speed centrifugation(800 x g for 5 min at 4°C) (57). The pellet was furtherwashed three times with the same buffer to obtain nuclei. Microscopicanalysis demonstrated the absence of cytoplasmic membranes andorganelles in the nuclear fraction. The nuclei were solubilizedin 0.25 ml of nuclear preparation buffer comprising 0.4% (wt/vol)SDS, 1 mM MgCl₂, and 10 U of DNase I for 3 min at 4°C, followedby the addition of 0.75 ml of nuclear preparation buffer. Thenuclear extract was incubated at 37°C for 5 min and at roomtemperature for an additional 10 min and then centrifuged at14,000 x g for 10 min (57). Both fractions were analyzed bySDS-15% PAGE, followed by autoradiography.

For some assays, cells were fractionated into membrane and cytosolicfractions. After harvest by trypsinization, cells were washedwith ice-cold culture medium and then with the nuclear preparationbuffer without Triton X-100 and homogenized in the nuclear preparationbuffer without Triton X-100. After removal of the nuclear pelletby low-speed centrifugation (800 x g for 5 min at 4°C),high-speed centrifugation (100,000 x g for 1 h at 4°C) wasperformed to separate membranes and vesicles from soluble proteins.The soluble fraction was regarded as the cytosolic fraction.The pellet was resolved in nuclear preparation buffer containing0.5% Triton X-100 (membrane fraction).

Cell surface biotinylation. Cell surface biotinylation was performed with EZ-link (Pierce)in accordance with the manufacturer's protocol. After biotinylation,cells were lysed with ice-cold radioimmunoprecipitation assaybuffer (10 mM sodium phosphate [pH 7.0], 150 mM NaCl, 0.1% SDS,1% NP-40, 1% sodium deoxycholate, 2 mM EDTA, 1 mM PMSF). Proteinswere subjected to SDS-PAGE and Western blotted with the mousemonoclonal antibiotin antibody.

Indirect immunofluorescence staining. To visualize the subcellular localization of midkine, nucleolin,and Cy3-labeled RAP, indirect immunofluorescence staining wasperformed. For these experiments, cells were grown on glasscoverslips or chamber slides and then treated with 200 ng ofbiotinylated midkine/ml and/or 1 µg of Cy3-labeled RAP/ml.The cells were washed three times with ice-cold PBS, fixed with4% paraformaldehyde in PBS for 10 min at 4°C, permeabilizedwith 1 or 4% Triton X-100 in PBS at 4°C for 20 min, andthen blocked with 1% bovine serum albumin in PBS at 4°Cfor 30 min. The treated cells were incubated at 20°C for1 h with FITC-conjugated streptavidin. Finally, they were extensivelywashed with PBS and then examined with a confocal microscopicsystem (MRC1024 system). For counterstaining, detection of thenuclei was performed with 0.5 µg of propidium iodide/ml,which detected double-stranded nucleic acids. For the immunodetectionof endogenous nucleolin or overexpressed nucleolin and a mutantnucleolin tagged with FLAG, cells were stained first with theantinucleolin mouse monoclonal antibody or the anti-FLAG M2monoclonal antibody and then with the tetramethyl rhodamineisocyanate-conjugated anti-mouse IgG antibody.

Cell survival assay. Cells transfected with the vector for NLS-deficient nucleolin( ${Delta}$ NLS) or the control vector were seeded on chamber slides inDMEM supplemented with 10% fetal bovine serum. After overnightincubation, culture medium was replaced with fresh DMEM withoutserum. After a further 48 h of incubation, cells were pretreatedwith or without midkine (100 ng/ml) for 3 h and then treatedwith 100 µM cisplatin (an anticancer drug) or 200 µMS-nitrosocysteine (a donor of NO) for 24 h. Percentages of apoptoticcells were estimated with the In Situ apoptosis detection kit(Takara) based on the terminal deoxynucleotidyltransferase-mediateddUTP-biotin nick end labeling (TUNEL) method (18). Cells werecounterstained with 0.2 µg of propidium iodide/ml. Apoptoticand healthy cells were counted in five fields at 200x magnification(total cell number per field was approximately 500). Statisticalanalysis was performed with a paired t test. P values <0.05were regarded as significant.

RESULTS

Top

Results

Midkine is internalized and translocated to the nucleus. The kinetics of exogenous midkine internalization were studiedby exposing cells to ¹²⁵I-labeled midkine at 20 ng/ml. We firstcompared the efficiencies of two available methods (low pH andtrypsinization) for removing cell surface-bound ¹²⁵I-midkine.¹²⁵I-midkine was allowed to bind to the cell surface at 4°Cfor 2 h. Unbound midkine was removed by washing the cells withice-cold medium. The cell surface-bound fraction representedapproximately 50% of the applied ¹²⁵I-midkine. To remove thecell surface-bound fraction, cells were incubated with eithera low pH buffer (150 mM NaCl, 50 mM acetic acid) or 0.125% trypsinand 0.025% EDTA in PBS. Most of the cell surface-bound ligandswere removed by either method, but approximately 1% of the applied¹²⁵I-midkine remained with the cells, which probably reflectedthe not-removed cell surface ¹²⁵I-midkine fraction (see theband at 0 min in Fig. 1A). We thus used the trypsin method inthe subsequent experiments.

View larger version (29K):
[in this window]
[in a new window]
FIG. 1. Time course of midkine internalization. L cells (a mouse fibroblast cell line) were incubated with ¹²⁵I-midkine at 4°C for 2 h. Cells were washed with ice-cold medium to remove unbound ¹²⁵I-midkine and then incubated at 37°C to initiate internalization of exogenous ligands. At the indicated time points, cells were treated with trypsin to remove cell surface-bound proteins, and detergent-soluble and nuclear fractions were isolated. (A and B) Autoradiography after SDS-15% PAGE of detergent-soluble (A) and nuclear (B) fractions. (C and D) Relative densitometrical densities of A and B, respectively, with the maximum value set at 100%.

By a shift of the temperature to 37°C after the cells werewashed with ice-cold medium, the cells were induced to internalizecell surface-bound ¹²⁵I-midkine. The time course of ¹²⁵I-midkineinternalization is shown in Fig. 1. The amount of internalizedmidkine in the detergent-soluble fraction reached a peak ataround 10 min (Fig. 1A and C), whereas the nuclear fractiondid not peak until 20 min (Fig. 1B and D). Note that the recoveredmidkine appeared to be intact because it was indistinguishablein size from the intact form (Fig. 1A and B). The amount ofinternalized midkine decreased after these points, reachingthe level at 0 min at around 60 min. Consistent with continuousintracellular midkine degradation, ¹²⁵I counts released intothe medium increased until 60 min (data not shown).

The subcellular localization of midkine was determined by meansof indirect immunofluorescence using N-terminally biotinylatedmidkine and FITC-conjugated streptavidin. Cytoplasmic midkinewas detected within 5 min of incubation at 37°C, whereasthe intensity of the signal in the nucleus reached a maximumat 20 min (Fig. 2), which is consistent with the results inFig. 1. The internalized midkine was detected not only homogeneously,reflecting cytoplasmic localization, but also in vesicular compartments.

View larger version (20K):
[in this window]
[in a new window]
FIG. 2. Spatial and temporal profiles of midkine internalization. L cells were exposed to midkine, which was labeled with biotin at its N terminus, as in Fig. 1. The localization of midkine (green; top and bottom panels) was determined by indirect immunofluorescence staining. The nucleus was stained with propidium iodide (red; top panels). The number in each panel indicates minutes of 37°C incubation. Bar, 10 µm.

Midkine internalization is dependent on LRP. Since midkine binds to LRP (49), we examined if midkine internalizationdepends on LRP. Wild-type MEF (LRP^+/+) internalized midkine,but LRP-deficient MEF (LRP^-/-) did not (Fig. 3). LRP^+/- cells,which are heterozygous for the LRP locus, gave the same resultsas LRP^+/+ cells (data not shown).

View larger version (19K):
[in this window]
[in a new window]
FIG. 3. Suppression of midkine internalization in LRP-deficient cells. Wild-type (LRP^+/+) and LRP-deficient (LRP^-/-) embryonic fibroblasts were used to assess the effect of LRP on midkine internalization. The procedure was the same as in Fig. 2. The number in each panel indicates minutes of 37°C incubation. Bar, 10 µm.

Differential intracellular trafficking of midkine and RAP. The profile of midkine internalization was compared with thatof RAP, since both of the molecules are ligands for LRP. After5 min of incubation at 37°C, midkine and RAP were mostlycolocalized inside the cells, probably in early endosomes (Fig.4). However this colocalization disappeared afterward. Whilemidkine was translocated to the nucleus, RAP remained in thevesicular compartments in the cytoplasm (Fig. 4). Thus, LRP-mediatedendocytosis does not always induce nuclear targeting by itsligands. The data also suggest that there is a mechanism topromote the nuclear translocation of midkine.

View larger version (21K):
[in this window]
[in a new window]
FIG. 4. Differential internalization of midkine and RAP. Both midkine and RAP are ligands for LRP. Their internalization profiles were compared. The procedure was as described in Fig. 2. Bar, 10 µm.

Cytoplasmic midkine binds to nucleolin. We previously observed that midkine binds to nucleolin in aligand blot assay (67). Nucleolin has been implicated in variousaspects of ribosome biogenesis (rDNA transcription, rRNA maturation,and ribosome assembly) and nucleocytoplasmic transport (20).Surprisingly, nucleolin was also reported to be located on thecell surface (8, 13, 35, 40). We therefore localized nucleolinin cells used in this study.

As a monoclonal antibody against nucleolin that can be usedfor immunocytochemistry was not good for immunoprecipitation,we generated a full-length nucleolin construct tagged with FLAGat its C terminus. Figure 5A shows proteins extracted by cellfractionation. LRP and basigin (a glycoprotein belonging tothe Ig superfamily) are cell surface membrane proteins. Calnexinis a membrane protein of the endoplasmic reticulum. These threewere mainly detected in the membrane fraction (Fig. 5A). Nucleolinwas detected not only in the nuclear fraction but also in thecytosolic and membrane fractions (Fig. 5A). When the cell surfacewas labeled by biotinylation, cell surface membrane proteinbasigin was biotinylated but endoplasmic reticulum membraneprotein calnexin was not (Fig. 5B). Nucleolin was itself biotinylated(Fig. 5B); thus a part of the nucleolin was located on the cellsurface.

View larger version (49K):
[in this window]
[in a new window]
FIG. 5. Midkine binds to LRP and then to nucleolin. (A) Cell fractionation for detection of protein location. Cyt, cytosol; Memb, membrane; Nuc, nucleus. (B) Western blot analysis after cell surface biotinylation. Cell surface membrane protein basigin was biotinylated, but not endoplasmic reticulum membrane protein calnexin. Nucleolin was biotinylated. Antibodies used were anti-FLAG (F), anticalnexin (Cal), antibasigin (Bsg), and antibiotin (Bio) antibodies. IP, immunoprecipitation. (C) Cells were allowed to endocytose ¹²⁵I-midkine as described for Fig. 1. Cells were harvested after 10 min of 37°C incubation. ¹²⁵I-midkine (MK) was barely endocytosed by LRP^-/- cells (Lysate) even when nucleolin-FLAG was overexpressed. The association of ¹²⁵I-midkine and nucleolin-FLAG took place only in L and LRP^+/+ cells, not in LRP^-/- cells (IP: ${alpha}$ -FLAG). (D) Effect of LRP overexpression on midkine internalization in LRP^-/- cells. ¹²⁵I-midkine was internalized and associated with nucleolin in LRP^-/- cells only when LRP was overexpressed. (E) Temporal profiles of midkine internalization by cell fractionation. LRP^+/+ cells transfected with an LRP expression vector (mLRP4T100) were used. ¹²⁵I-midkine was associated with LRP in the membrane (Mem) fraction. In contrast, ¹²⁵I-midkine was associated with nucleolin in the cytosolic (Cyt), but not membrane, fraction.

The roles of two key molecules (LRP and nucleolin) in midkineinternalization were then determined. Consistent with the datashown in Fig. 3, L cells (an LRP-expressing mouse fibroblastcell line) and LRP^+/+ cells internalized midkine but LRP^-/-cells did not (Fig. 5C, lysate). When nucleolin-FLAG was overexpressed,¹²⁵I-midkine was coprecipitated with nucleolin-FLAG in L andLRP^+/+ cells but not in LRP^-/- cells (Fig. 5C). Therefore, midkineendocytosis depends on LRP and midkine binds to nucleolin onlywhen midkine is endocytosed. Figure 5D supports these data.As midkine binds to the second and fourth ligand-binding domainsof LRP (K. Kadomatsu, unpublished data), we used HA-mLRP4T100.This construct is a miniform of LRP containing the fourth ligand-bindingdomain with a HA tag at the N terminus, which is functionalin endocytosis if an appropriate ligand is applied (52). InLRP^-/- cells, ¹²⁵I-midkine was internalized only when HA-mLRP4T100was overexpressed (Fig. 5D, lysate). Cotransfection of nucleolin-FLAGand HA-mLRP4T100, but not nucleolin-FLAG alone, enabled nucleolinto bind midkine (Fig. 5D, IP: ${alpha}$ -FLAG).

These data were further supported by Fig. 5E. In this experiment,we examined the temporal profile of midkine internalizationby cell fractionation. In the membrane fraction (probably mainlyendosomes), the amount of ¹²⁵I-midkine was relatively constantfrom 5 to 20 min on incubation at 37°C. From this fraction,midkine was coimmunoprecipitated with LRP but not nucleolin.In contrast, the amount of midkine in the cytosolic fractionreached a maximum at 10 min. Midkine was coimmunoprecipitatedwith nucleolin in this fraction. Taken together, the data indicatethat midkine is endocytosed by LRP and is translocated to thecytosol, where it binds to nucleolin. As nucleolin is knownas a nucleocytoplasmic shuttle protein, it is possible thatmidkine is translocated to the nucleus as a cargo of nucleolin.

A nucleolin mutant affects midkine localization. We next determined the midkine-binding site in nucleolin. Thestructure of nucleolin can be divided into four parts (Fig.6). The N-terminal domain contains four acidic stretches (AS1,-2, -3, and -4), major phosphorylation sites, and sites fornucleolin-protein interaction. The second part is a bipartiteNLS, which is essential for the nuclear translocation of nucleolin.The third part contains the four RNA-binding domains (RBD1,-2, -3, and -4), essential for RNA binding. At the C terminusis the GAR domain, which is rich in glycine, arginine, and phenylalanineresidues and which is important for both the nucleolin-proteininteraction and the function of the RNA-binding domains (Fig.6). We generated various nucleolin mutant constructs (Fig. 6).The results in Fig. 6B indicate that the N-terminal domain ofnucleolin, particularly the acidic region, is responsible formidkine binding. One acidic stretch among the four was sufficientfor midkine binding.

The localization of midkine and endogenous nucleolin was thenexamined. Figure 7 shows pictures of LRP^+/+ cells after 20 minof 37°C incubation. Following serum stimulation of starvedcells, 90% of cells showed a predominantly nucleolar localizationfor nucleolin and nuclear localization for midkine (Fig. 7,left). The remaining 10% of cells showed only a nuclear andcytoplasmic localization for nucleolin and a cytoplasmic localizationfor midkine (Fig. 7, left). This close relation of midkine andnucleolin in intracellular localization suggests the possibleinvolvement of nucleolin in nuclear targeting by midkine.

View larger version (28K):
[in this window]
[in a new window]
FIG. 7. Effect of an NLS-deficient nucleolin on midkine subcellular localization. (Left) Localization of endogenous nucleolin and exogenously added midkine. LRP^+/+ cells were incubated under serum starvation for 2 days and then stimulated with 10% fetal bovine serum for 12 h. The subsequent procedure was the same as for Fig. 2. Pictures of cells at 20 min after 37°C exposure are shown. Arrow, nucleolar localization of nucleolin and nuclear localization of midkine; arrowhead, nuclear and cytoplasmic localization for nucleolin and cytoplasmic localization for midkine. (Right) LRP^+/+ cells were transfected with nucleolin expression vectors. Bar, 10 µm.

Nucleolin expression vectors were next transfected into LRP^+/+cells. Since LRP^+/+ cells have endogenous nucleolin, midkinelocalization may be under the control of both endogenous nucleolinand overexpressed mutant nucleolin. Overexpression of the full-lengthnucleolin did not affect midkine localization (Fig. 7, right).However, the NLS-deficient mutant ( ${Delta}$ NLS) showed a striking phenomenon:midkine appeared to be trapped in the cytoplasm and was colocalizedwith ${Delta}$ NLS (Fig. 7, right). The results indicate that ${Delta}$ NLS hasa dominant-negative effect on midkine subcellular localizationand support the idea that the endogenous nucleolin functionsin the nuclear transport of midkine.

Nuclear targeting is necessary for the antiapoptotic activity of midkine. When LRP^+/+ cells were treated with cisplatin (an anticancerdrug), apoptosis was induced: 100 µM cisplatin causedthe death of approximately 60% of the cells within 24 h. Ifcells were pretreated with 100 ng of midkine/ml for 3 h andthen treated with cisplatin, cell death was significantly suppressed[Fig. 8A, LRP^+/+ (CV) cells; P < 0.01]. This suppressionwas significantly blocked in ${Delta}$ NLS-overexpressing cells (P <0.05) [compare LRP^+/+ (CV) and LRP^+/+ ( ${Delta}$ NLS) in Fig. 8A]. InLRP^-/- cells, midkine could not protect cells from cisplatin-inducedapoptosis [Fig. 8A, LRP^-/- (CV)]. However, if a miniform ofLRP was overexpressed, midkine significantly suppressed cisplatin-inducedapoptosis [Fig. 8A, LRP^-/- (mLRP4T100)]. Similar results wereobtained when NO-induced apoptosis was examined (Fig. 8B). Theseresults suggest that the nuclear targeting is essential forthe full activity of midkine in cell survival.

View larger version (27K):
[in this window]
[in a new window]
FIG. 8. Effect of overexpression of ${Delta}$ NLS and LRP on antiapoptotic activity of midkine. Cells were transfected with the indicated vectors. Control vectors (CV) used were pcDNA3.1+ for the nucleolin ${Delta}$ NLS expression vector ( ${Delta}$ NLS) and pcDNA3 for an LRP expression vector (mLRP4T100). After 48 h of serum starvation, cells were incubated with or without midkine (100 ng/ml; MK) for 3 h. Cells were then incubated with 100 µM cisplatin (an anticancer drug) (A) or 200 µM S-nitrosocysteine (SNOC; a donor of NO) (B). Apoptotic cells were counted as described in Materials and Methods. The y axis represents the relative ratio of apoptotic cell number as revealed by the terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling method, normalizing to the value of cisplatin alone. A representative result is shown here (three independent experiments showed similar results). _*, P < 0.05; _*_*, P < 0.01; _*_*_*, P < 0.001; n.s., not significant.

DISCUSSION

Top

Discussion

The present findings indicate that LRP-mediated endocytosisnot only leads to ligand degradation but also confers a pathwayfor the nuclear localization of a ligand. Midkine is internalizedin an LRP-dependent manner and is further transported to thenucleus depending on nucleolin. The nuclear targeting of midkineis indispensable for the full activity of midkine in the promotionof cell survival.

It has been reported that nucleolin is also localized on thecell surface (8, 13, 35, 40). For example, cell surface nucleolinhas been implicated in virus infection (8, 13). During humanimmunodeficiency virus (HIV) infection of T lymphocytes, theV3 loop of HIV binds to cell surface nucleolin, facilitatingHIV attachment to cells, which can be counteracted by exogenousmidkine (9). Indeed, nucleolin was detected at the cell surfacein the present study. However, the cell surface nucleolin wasnot involved in midkine endocytosis, since only cytosolic nucleolinassociated with midkine. The laboratories of Olson and Meleseoriginally proposed a model of nucleolin as a scaffold protein,that is, the N-terminal domain of nucleolin recruits proteinsimportant in ribosome biogenesis, while RNA-binding domainsinteract with rRNA (46, 62). This model was confirmed and extendedby recent data. The N-terminal domain is needed for the firststep in rRNA processing (19). Topoisomerase I binds to the first,second, and third, but not the fourth, acidic stretches in theN-terminal domain of nucleolin (14). The idea that nucleolinrecruits topoisomerase I to sites of rDNA transcription in thenucleolus (4, 14) is supported by the findings that Saccharomycescerevisiae nucleolin ortholog Nsr1p binds to yTop1p (yeast topoisomeraseI ortholog) and is important in the subcellular localizationof yTop1p (14). The present study has further extended the modeland demonstrated for the first time that nucleolin functionsin the nuclear transport of a growth factor. It is possiblethat other ligands, besides midkine, may bind to and utilizenucleolin for their nuclear targeting. The following two findingssupport this idea: the N-terminal domain of nucleolin bindsNLSs (41, 78), and many ligands known to be localized in thenucleus contain NLSs (11, 24, 28, 29, 41, 75, 76, 82).

We failed to detect a physical association between LRP and nucleolin(data not shown). Thus, there may be some distance between thesetwo molecules. We speculate that LRP-bound midkine is internalizedand transported to the early or late endosomes. Since midkineis functional even at low pH, as assessed by using a systeminvolving midkine-induced plasminogen activator activity (37),the function of midkine may remain intact in the endosomes.Pseudomonas exotoxin (PE) provides one model for the translocationof an exogenous protein into the cytosol via LRP-mediated internalization(16, 36, 52). After binding to LRP on the cell surface, PE istranslocated as a cargo of endosomes to the endoplasmic reticulum,where furin cleaves it, producing an ADP-ribosylating fragment.This fragment is translocated to the cytosol from the endoplasmicreticulum and inactivates elongation factor 2 through ADP ribosylation,shutting down protein synthesis. Although diphtheria toxin alsoemploys receptor-mediated internalization, proteolytic processing,translocation to the cytosol, and ADP-ribosylation of elongationfactor 2, this toxin is translocated to the cytosol from acidicendosomes. As full-length midkine was recovered from the nucleus,it is not likely that dynamic processing, as for PE, for midkinetranslocation to the cytosol takes place. Midkine may provideanother mechanism of cytoplasmic translocation via LRP-mediatedinternalization.

Nucleolin is known to be phosphorylated in its N-terminal domainby casein kinase II and p34^cdc2 (3, 7, 55). In a study involvingXenopus laevis egg extracts, it was shown that p34^cdc2 phosphorylationsites improve the nuclear translocation of nucleolin when theyare dephosphorylated and enhance cytoplasmic localization whenthey are phosphorylated (63). Therefore, nuclear localizationof nucleolin depends on both its NLS and phosphorylation status.Nucleolin-bound midkine may be transported to the nucleus followingthis mechanism.

With respect to nuclear targeting by midkine, we recently foundthat the 37-kDa laminin-binding protein precursor binds midkineand is cotranslocated with midkine into the nucleus (61). Itis possible that midkine uses both nucleolin and laminin-bindingprotein precursor as shuttle proteins but that, in the cellsused in the present study, the nucleolin shuttle is the predominantone.

The intracellular domain of LRP can recruit many cytosolic signalingmolecules, including DAB1, JIP-1, PSD-95, and SHC (2, 23). Thus,we cannot exclude the possibility that LRP functions as a signalingmembrane receptor for midkine. It is rather likely that signalingfrom both the cell surface and the nucleus cooperates for thefull activity of midkine, as reported for other growth factors,such as aFGF, bFGF, and Schwannoma-derived growth factor (28,33, 76). The results in Fig. 8 support this idea.

We previously reported that pleiotrophin/HB-GAM (heparin-bindinggrowth associated molecule) also binds nucleolin in a ligandblot assay (67). Midkine and pleiotrophin/HB-GAM exhibit about50% homology and form a family of heparin-binding growth factorsthat is distinct from the FGF family and other growth factorfamilies (50). These two factors share several biological activities,such as the enhancement of neurite extension and plasminogenactivator activity and the promotion of carcinogenesis. It isthus plausible that pleiotrophin/HB-GAM uses the same pathwayas midkine does for its nuclear targeting. It is possible thatLRP (49; K. Kadomatsu, unpublished data) and receptor-type proteintyrosine phosphatase ${zeta}$ (43, 44, 59) are shared by midkine andpleiotrophin/HB-GAM. Additionally, it has been proposed thatmembrane heparan sulfate proteoglycan N-syndecan (34, 60) andreceptor tyrosine kinase ALK (64) are receptors for pleiotrophin/HB-GAM.Cell surface complex formation from these molecules and itscooperation with nuclear signaling are the next subjects ofinvestigation.

ACKNOWLEDGMENTS

We thank Justin Weinberg for reading and commenting on the manuscript.

This work was supported by grants-in-aid from the Ministry ofEducation, Science, Sports, and Culture of Japan and grants-in-aidfor Center of Excellence Research and by grants from the NationalInstitutes of Health.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biochemistry, Nagoya University School of Medicine, 65 Tsurumai-cho, Showa-ku, Nagoya 466-8550, Japan. Phone: 81-52-744-2064. Fax: 81-52-744-2065. E-mail: kkadoma{at}med.nagoya-u.ac.jp.

REFERENCES

Top