Hepatoma-derived Growth Factor SIGNIFICANCE OF AMINO ACID RESIDUES 81–100 IN CELL SURFACE INTERACTION AND PROLIFERATIVE ACTIVITY*

Hepatoma-derived growth factor (HDGF) has proliferative, angiogenic, and neurotrophic activity. It plays a putative role in the development and progression of cancer. When expressed in cells, the mitogenic activity of HDGF depends on its nuclear localization, but it also stimulates proliferation when added to the cell culture medium. A cell surface receptor for HDGF has not been identified so far. We investigated the interaction of various purified recombinant HDGF fusion proteins with the cell surface of NIH 3T3 fibroblasts. We showed that binding of a HDGF- (cid:1) -galactosidase fusion protein to the cell surface of NIH 3T3 fibroblasts was saturable, occurred with high affinity ( K D (cid:2) 14 n M ), and had a prolif- erative effect. We identified a peptide comprising amino acid residues 81–100 within the amino-terminal part of HDGF that bound to the cell surface of NIH 3T3 cells with saturation and affinity values similar to those of HDGF. When added to primary human fibroblasts, this peptide stimulated proliferation. Substitution of a single amino acid (K96A) within this peptide was sufficient to abolish its binding to the cell surface and its proliferative activity. In contrast, when expressed transiently in NIH 3T3 cells, a HDGF- (cid:1) -galactosidase fusion protein in which amino acid residues 81–100

Hepatoma-derived growth factor (HDGF) has proliferative, angiogenic, and neurotrophic activity. It plays a putative role in the development and progression of cancer. When expressed in cells, the mitogenic activity of HDGF depends on its nuclear localization, but it also stimulates proliferation when added to the cell culture medium. A cell surface receptor for HDGF has not been identified so far. We investigated the interaction of various purified recombinant HDGF fusion proteins with the cell surface of NIH 3T3 fibroblasts. We showed that binding of a HDGF-␤-galactosidase fusion protein to the cell surface of NIH 3T3 fibroblasts was saturable, occurred with high affinity (K D ‫؍‬ 14 nM), and had a proliferative effect. We identified a peptide comprising amino acid residues 81-100 within the amino-terminal part of HDGF that bound to the cell surface of NIH 3T3 cells with saturation and affinity values similar to those of HDGF. When added to primary human fibroblasts, this peptide stimulated proliferation. Substitution of a single amino acid (K96A) within this peptide was sufficient to abolish its binding to the cell surface and its proliferative activity. In contrast, when expressed transiently in NIH 3T3 cells, a HDGF-␤-galactosidase fusion protein in which amino acid residues 81-100 were deleted still had proliferative activity, whereas a fusion protein containing only the 81-100 peptide did not. Our results suggest the existence of a plasma membrane-located HDGF receptor for which signaling depends on amino acid residues 81-100 of HDGF. This region differs from the one that has been recently identified to be essential for mitogenic activity depending on the nuclear localization of HDGF. Thus, HDGF exerts its proliferative activity via two different pathways.
It has been shown to have growth factor activity for hepatoma cells, fibroblasts, smooth muscle cells, and endothelial cells (1,2,4,5). In addition, it has recently been shown to be neurotrophic (6). In recent years, five homologous proteins have been identified by different approaches (4,(7)(8)(9). Four of these proteins have been termed HRP-1 to HRP-4, and the fifth protein has been named p52/75 or LEDGF. Because of the high similarity between the different family members, the amino-terminal region (amino acids residues 1-100) of these proteins has been termed the HATH region (9). In contrast, the length and amino acid sequences of carboxyl-terminal regions of HRPs vary considerably, suggesting a modular structure of these proteins, which is supported by NMR data analysis of human HDGF (10). Expression of HDGF mRNA as well as protein is developmentally regulated, and various results point to an important function of HDGF during tissue development, protection, and repair (6,(11)(12)(13)(14)(15)(16)(17)(18). In addition, studies on different types of cancers suggest a possible role of this growth factor in cancer development and neovascularization (19 -24). However, compared with other growth factor families, our current knowledge regarding HDGF and HRPs is still limited.
The main cellular localization of HDGF is nuclear, although in some cells, HDGF can be found in the cytosol (3,11,12,25). HDGF has two nuclear localization signals, one in the conserved HATH region, and the other in the carboxyl-terminal area specific for the different family members. The nuclear localization has been shown to be a prerequisite for the mitogenic activity of intracellularly expressed HDGF (5,26). Several groups have shown independently that HDGF is mitogenic not only when expressed in cells but also when applied exogenously (2,3). Despite its purification from cell culture medium and its mitogenic properties as an exogenous ligand, HDGF lacks a classical signal sequence and is only poorly secreted by transfected cell lines. In this respect, it resembles FGF-1 and FGF-2. Current data for these FGF family members suggest that these growth factors may be released from cells via a mechanism independent of the secretory endoplasmic reticulum-Golgi-plasma membrane pathway (27)(28)(29)(30). For exogenously applied HDGF and its related protein, LEDGF, it has also been shown that they can be internalized, and it has been proposed that the mitogenic action of exogenous HDGF depends on internalization and subsequent nuclear translocation (26,31). Internalization would require a cell surface receptor, and in this respect, it is important to note that HDGF strongly binds heparin-like glycosaminoglycans (HLGAGs) (4,10). However, the mechanism of internalization has not yet been revealed. Alternatively, exogenous HDGF may also bind to a hitherto unknown plasma membrane receptor coupled to a proliferative signal transduction pathway. In accordance with this, rapid phosphorylation of Erk1/2 upon addition of HDGF to pulmonary epithelial cells has recently been demonstrated (44). With the exception of HLGAGs, however, no specific cellular receptor for HDGF or the related proteins has been identified so far.
In this study, we examined the mode of action of HDGF on NIH 3T3 fibroblasts. By using various HDGF-␤-galactosidase fusion proteins, we searched for the HDGF region responsible for binding to the cell surface. We report the identification of a peptide region in the amino terminus of HDGF that was sufficient to interact with the cell surface and activate proliferation of fibroblasts. Deletion of this peptide region demonstrated that HDGF signals from the cell surface of fibroblasts, rather than after internalization and translocation to the nucleus of these cells.

MATERIALS AND METHODS
Unless otherwise stated, all chemicals were from Sigma. Construction of ␤-Galactosidase and GFP Fusion Proteins-For bacterial expression of ␤-galactosidase fusion proteins, the coding region of the enzyme containing six histidine residues at the carboxyl terminus was amplified by PCR using the plasmid pCMV␤ from Clontech, which contains the complete ␤-galactosidase coding region. The cDNA coding for the ␤-galactosidase-histidine fusion protein (␤GalHis) was cloned into TRC99a vector (Amersham Biosciences) via a KpnI and HindIII restriction site inserted in the primers used for amplification (for sequences of primers and shape of the recombinant protein products, see Table I and Fig. 2). All HDGF fragments were amplified by PCR using primers containing EcoRI (sense) and KpnI (antisense) restriction sites (for primers, see Table I). DNA fragments coding for fewer than 20 amino acids were not generated by PCR but by hybridization of the respective sense and antisense oligonucleotides (for sequences, see Table I). After digestion with EcoRI and KpnI, the fragments were inserted 5Ј of the ␤GalHis coding region in Trc99a vector. In all cases, the ATG within the NcoI restriction site of the vector served as the starting point of translation.
Histidine-tagged GFP fusion proteins were prepared by amplifying the coding region of GFP using plasmid eGFP-C3 (Clontech) as template for PCR (for primer sequences, see Table I). The PCR product was purified, digested with HindIII and KpnI, and exchanged with the ␤GalHis sequence in the Trc99a vector described above.
For eucaryotic expression, the coding sequences of the fusion proteins were cut from Trc99a vector by EcoRI and HindIII and cloned into pCDNA3/Zeo vector (Clontech) with an altered multicloning site containing an ATG in front of the EcoRI site.
Histidine-tagged HDGF was produced as described elsewhere (4). Sequences of all recombinant cDNA constructs were verified by cycle sequencing on an ABI 310 automatic sequencer (Applied Biosystems, Foster City, CA) according to the protocol supplied by the manufacturer.
Production and Purification of Recombinant Proteins-All recombinant proteins were produced and purified by means of their carboxylterminal His 6 tag as described elsewhere (4). Purity and correct molecular weight of the recombinant proteins were examined by Coomassie Blue staining of SDS-PAGE gels and Western blotting using antibodies against the carboxyl-terminal histidine residues. Protein concentrations were determined by comparison with known concentrations of bovine serum albumin on a Coomassie Blue-stained SDS-PAGE gel and by the DC protein assay (Bio-Rad).
Cell Culture-All media and chemicals for cell culture were purchased from Invitrogen, and all plastic material was purchased from BD Biosciences. NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM glutamine, and 10% fetal bovine serum. Subconfluent cells were passaged every 3 days by treatment with 0.05% trypsin/ EDTA for up to 2 months. Primary human skin fibroblasts were cultured in the same medium as described for NIH 3T3 cells and passaged once a week.
HDGF Binding Assay-Cells were seeded into wells of a 96-well microtiter plate at a density of 12,000 cells/well. After 24 h, cells were fixed in 4% paraformaldehyde for 10 min at 4°C and washed with ice-cold phosphate-buffered saline. Free protein binding sites were blocked with 200 l of blocking buffer (2% (w/v) bovine serum albumin in phosphate-buffered saline) for 2 h at 4°C. Binding of the different ␤GalHis fusion proteins was performed in binding buffer (50 mM sodium phosphate, 1 mM MgCl 2 , 300 mM NaCl, and 2% (w/v) bovine serum albumin, pH 7.4) for 2.5 h at 4°C. After washing three times with washing buffer (50 mM sodium phosphate, 300 mM NaCl, and 0.05% (v/v) Tween 20, pH 7.4) and one time with washing buffer containing no Tween 20, bound ␤-galactosidase activity was determined by adding 100 l of 1 mM chlorophenol red-␤-D-galactopyranoside (CRPG substrate; Roche Applied Science) in binding buffer to each well. After 1 h at 37°C, color development was measured at 595 nm in a microplate reader. Alternatively, cell surface-bound ␤-galactosidase fusion proteins were visualized by incubation with 250 l of ␤-galactosidase substrate (40 mg/ml X-gal in dimethylformamide, 200 mM K 4 Fe(CN) 6 / K 3 Fe(CN) 6 , and 2 mM MgCl 2 ) at 37°C for 3 h.
Proliferation Assay-Mitogenic activity was determined by measuring the incorporation of tritium-labeled thymidine into DNA of proliferating cells (32). Primary human skin fibroblasts were seeded into 24-well plates at a density of 6,000 cells/cm 2 and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. After 6 days, medium was replaced by Dulbecco's modified Eagle's medium containing no fetal bovine serum but containing different concentrations of the respective histidine-tagged recombinant proteins or synthetically produced peptides (EMC, Tuebingen, Germany). In addition, each well received 8.9 kBq [ 3 H]thymidine (specific activity, 365 MBq/mmol; Amersham Biosciences), and cells were incubated for an additional 20 h. Cells were harvested and processed as described previously (32), and incorporation of radioactivity was measured in a 1900CA liquid scintillation analyzer (Packard).
For the determination of cell numbers after transfection, the MTS assay was performed according to the protocol supplied by the manufacturer (Promega).
Fluorescence Correlation Spectroscopy-The setup of the microscope was as described elsewhere (33). For measurements on living cells by fluorescence correlation spectroscopy, all proteins where produced as GFP fusion proteins in HEK 293 cells and purified by means of their histidine tags as described. All cells were tested for binding of the fluorescent ligands at the concentrations given under "Results." In addition, for all cells and ligands, laser scans were recorded as described by Hegener et al. (34,35).
Statistical Analysis-All values are shown as the mean Ϯ S.D. Oneway analysis of variance was performed for comparison of data between two or more groups. A probability value of p Ͻ 0.05 was considered to be statistically significant, and a value of p Ͻ 0.01 was considered to be statistically highly significant. Binding curves were calculated using Origin V. 7.5 software (OriginLab Corp., Northampton, MA).

HDGF-␤-Galactosidase Fusion Protein Binding to NIH 3T3
Cells Is Saturable and Accompanied by a Proliferative Response-In proliferation assays, HDGF displays growth factor activity when added to the medium of various cell types (1, 2, 4, 5, 36). A prerequisite for this activity is the interaction of HDGF with the cell surface. A HDGF receptor, however, has not yet been identified. To find evidence for a putative HDGF receptor, we expressed and purified a HDGF-␤-galactosidase fusion protein. Fig. 1, A and B, shows that the recombinant HDGF-␤GalHis fusion protein (see Fig. 2 for description) bound to NIH 3T3 cells, whereas ␤GalHis used as a control did not. Cell binding studies using increasing amounts of the HDGF-␤GalHis fusion protein demonstrated that this interaction was saturable. Half-maximal binding occurred at ϳ14 nM (Fig. 1C). In addition, HDGF-␤GalHis added to the cell culture medium stimulated the incorporation of [ 3 H]thymidine into primary human fibroblasts (Fig. 5B).
Identification of Amino Acid Residues Involved in HDGF Cell Surface Binding-To delineate the regions of the growth factor responsible for cell surface binding, we constructed various histidine-tagged HDGF-␤-galactosidase fusion proteins (for designation of the fusion proteins, see Fig. 2) and investigated the ability of the purified proteins to bind to the surface of fixed and unfixed NIH 3T3 fibroblasts.
Structurally, all members of the HDGF family share a conserved amino-terminal 100-amino acid HATH region and pos-  (n ϭ 4). Absorbance at 595 nm is given as a measure of ␤-galactosidase activity. B, HDGF cell surface binding depends on amino acid residues 81-100 of the HATH region. The HATH region of HDGF was deleted in increments of 20 amino acids, and binding of each construct (14 nM) to the cell surface of NIH 3T3 fibroblasts was compared with that of full-length HDGF (designation of fusion proteins is as described in Fig. 2). Cell surface binding is lost by deleting the last 20 amino acids (amino acids 81-100) of the HATH region (HATH80). Values are given as means Ϯ S.D. (n ϭ 4). C, amino acid residues 81-100 of the HATH region are sufficient for cell surface binding. ␤GalHis fusion proteins encompassing the entire HDGF or amino acid residues 81-100 only were produced and purified as described and examined at a concentration of 14 nM for cell surface binding ability. The HATH81-100 fusion protein is able to bind to the cell surface. Values are given as means Ϯ S. D. (n ϭ 4). D, HATH81-100 cell surface binding is saturable. Purified ␤GalHis fusion protein of amino acids 81-100 (HATH81-100-␤GalHis) was added to NIH 3T3 cells at increasing concentrations. Cell-bound ␤-galactosidase activity was determined and is sess a non-conserved carboxyl-terminal non-HATH region (4,9). To determine which part of HDGF is responsible for cell surface binding, we fused the HATH region and non-HATH region to ␤GalHis and performed cell binding studies as described above. The results are shown in Fig. 3A and demonstrate that the amino-terminal HATH region harbored the cell surface binding activity of HDGF, whereas the carboxyl-terminal non-HATH region was unable to bind to the surface of NIH 3T3 fibroblasts.
To further delineate the region of HDGF responsible for cell surface binding, we deleted various parts of the HATH region (see Fig. 2) and examined the capability of the respective proteins to bind to the surface of NIH 3T3 fibroblasts (Fig. 3B). Deletion of amino acids 81-100 abolished cell surface binding completely. Thus, amino acids 81-100 of the HATH region are necessary for cell surface binding. To examine whether this region is also sufficient for binding, we generated the HATH81-100-␤GalHis fusion protein. Fig. 3C shows that the HATH81-100-␤GalHis fusion protein was indeed able to bind to NIH 3T3 fibroblasts almost as efficiently as a fusion protein comprising the entire HATH region. In addition, cell surface binding assays with different concentrations of the HATH81-100-␤GalHis fusion protein demonstrated that binding of this protein to the cell surface was saturable and occurred with about the same affinity as the HDGF-␤-galactosidase fusion protein (compare Figs. 1C and 3D).
To further examine the importance of this peptide region for HDGF-cell surface interaction, we introduced a substitution at position 96 (lysine to alanine) in the HATH81-100-␤GalHis and the HDGF-␤GalHis fusion proteins, respectively. Interestingly, this amino acid substitution abolished the cell surface interaction of the HATH81-100-␤GalHis protein completely and caused a significant reduction in cell surface binding of the fusion protein containing full-length HDGF (Fig. 3E).
Because HDGF has been shown to bind to HLGAGs, we examined the ability of the HDGF-␤GalHis and HATH81-100-␤GalHis fusion proteins to bind to heparin. In contrast to the HDGF-␤GalHis fusion protein, the HATH81-100-␤GalHis fusion protein did not bind to heparin, demonstrating that binding to the cell surface does not occur via HLGAGs (data not shown).
To exclude the possibility that the data presented above depend on the particular fusion protein and technique applied, we used fluorescence correlation spectroscopy, which allowed us to study interactions of ligands and receptors on viable cells. For this purpose, we expressed and purified proteins in which HDGF and amino acids 81-100 of the HATH region were fused to GFP.
Fluorescence correlation spectroscopy allowed us to measure mobilities of fluorescent molecules in a defined small volume (ϳ100 fl). First, we obtained autocorrelation curves for the different constructs in solution in the absence of cells (Fig. 4,  •). These curves allowed the calculation of similar single time diffusion constants ( free ) for GFP (0.153 Ϯ 0.017 ms), HDGF-GFP(0.261 Ϯ 0.021 ms), and HATH81-100-GFP (0.219 Ϯ 0.035 ms). During measurements in the presence of cells in volumes including parts of the cell surface, HDGF-GFP and HATH81-100-GFP (but not non-fused GFP) showed significant shifts in the corresponding autocorrelation curves (Fig. 4, E). These shifts indicate lower mobilities of the proteins, which occur when fluorescent molecules are no longer freely diffusible but are bound to the cell surface. From these curves, time diffusion constants ( bound ) for the GFP fusion proteins of 2.279 Ϯ 1.252 ms (HDGF-GFP) and 1.509 Ϯ 0.935 ms (HATH81-100-GFP) in addition to free were calculated, indicating that they are bound to the plasma membrane. Values for non-fused GFP were unaltered.
To control for the specificity of the binding of HATH81-100-GFP and to investigate whether the identified peptide region indeed interacts with a cell surface binding site for HDGF, cells were preincubated with 400 nM histidine-tagged HDGF before addition of HATH81-100-GFP. Fluorescence correlation measurements with 35 nM HATH81-100-GFP on the pretreated cells revealed a reduction of binding of about 70% when compared with untreated cells (data not shown).
The Peptide Comprising Amino Acid Residues 81-100 of HDGF Has Mitogenic Activity-Next we tested a synthetically produced peptide encompassing amino acids 81-100 of the HATH region for its mitogenic activity. This peptide stimulated the proliferation of primary human fibroblasts significantly at concentrations of 37 and 185 nM, respectively. In contrast, a peptide substituted at lysine 96 had no effect (Fig. 5A). Mutant fusion proteins bearing the K96A substitution (HDGFK96A-␤GalHis) or missing amino acids 81-100 (⌬HDGF-␤GalHis) were not able to increase the proliferation of fibroblasts significantly when added to the medium (Fig. 5B). Thus, amino acid residues 81-100 of HDGF are important for the mitogenic activity of this growth factor.
The Mitogenic Activity of the HATH 81-100 Peptide Depends on Cell Surface Interactions-HDGF acts proliferatively not only when added to cells but also when expressed within cells. In the latter case, the mitogenic activity of HDGF depends on its nuclear localization. In addition, it was shown that extracellular HDGF can be internalized. This leaves the possibility that the HATH81-100-␤Gal fusion protein added to the medium may not act through a cell surface receptor but could be internalized and activate HDGF pathways intracellularly. To exclude this possibility, we deleted amino acids 81-100 from HDGF-␤GalHis fusion protein (designated ⌬HDGF-␤GalHis). We expressed cDNAs coding for HDGF-␤GalHis, HDGFK96-␤GalHis, ⌬HDGF-␤GalHis, and ␤GalHis transiently in NIH 3T3 cells and measured the growth rate of transfected cells by the MTS assay. HDGF-␤GalHis, HDGFK96A-␤GalHis, and ⌬HDGF-␤GalHis fusion constructs stimulated proliferation after transfection, whereas HATH81-100-␤GalHis showed no effect (Fig. 5C). These results clearly demonstrate that HDGF harbors two different proliferative activities: one localized inside the carboxyl-terminal non-HATH region responsible for intranuclear activity (26,37), and the other located in the HATH81-100 region exerting its mitogenic activity from the cell surface.
To further exclude that extracellular HDGF signals through its HATH region after internalization and translocation to the nucleus, we investigated whether HDGF can be internalized. Recent studies on HEK 293 cells demonstrated that a GFP-HDGF fusion protein added to the cell culture medium can be internalized and is able to translocate into the nucleus within 72 h after addition (26). We added a HDGF-GFP and HDGF-␤-galactosidase fusion protein to the cell culture medium of primary human fibroblasts and NIH 3T3 cells for time periods of up to 72 h. Cells were investigated by fluorescence microscopy or ␤-galactosidase activity determinations. We were not displayed as absorption at 595 nm. E, lysine 96 contributes to HDGF cell surface interaction. Wild-type and K96A-substituted full-length HDGFor HATH81-100-␤GalHis fusion proteins, respectively, were examined for cell surface binding activity. The K96A substitution reduces (HDGFK96A-␤GalHis) or abolishes (HATH81-100 K96A-␤GalHis) the cell surface interaction of the fusion proteins, respectively. Values are given as means Ϯ S. D. (n ϭ 4). **, p Ͻ 0.01 between wild-type and substituted protein. Absorbance at 595 nm is given as a measure of ␤-galactosidase activity.
able to detect these fusion proteins in the cytosol or in the nuclei of these cells (data not shown). To exclude the possibility that these techniques might be too insensitive to detect inter-nalization, we used the HDGF-GFP fusion proteins to perform fluorescence correlation spectroscopy. This technique is able to detect interaction and internalization of even single fluorescent molecules in living cells with very high sensitivity (35, 38 -42). In addition to HDGF-GFP, we performed fluorescence correlation spectroscopy measurements with a HATH81-100-GFP fusion protein. GFP and GFP fused to the human immunodeficiency virus-derived TAT peptide, which has been shown previously to induce internalization of the fluorescent protein into different kinds of cells (43), were used as negative and positive controls, respectively.
To detect internalization events for the three GFP fusion proteins, scans of cells incubated with the different constructs were performed. During these scans, the volume in which fluorescence is detected is moved from the lower plasma membrane next to the coverslip through the cytoplasm, the upper plasma membrane, and, finally, into the medium. Fluorescent proteins that were internalized can be detected as fluorescent events between the two signals of the plasma membranes. As can be seen in Fig. 6 only for TAT-GFP, peaks corresponding to the fluorescent activity of GFP could be detected inside the cell (arrow in the bottom graph), whereas HDGF-GFP or HATH81-100-GFP fluorescence could not be detected in any of the cells investigated. Thus, in the experimental setting used here, we had no indication that externally added HDGF fusion proteins can be internalized. FIG. 5. HDGF sequence harbors two distinct proliferation-inducing activities. A, wild-type and K96A-substituted peptides comprising amino acids 81-100 were produced synthetically. Different concentrations of both peptides were added in serum-free medium to human skin fibroblasts, and proliferation was quantified after 20 h by the amount of incorporated [ 3 H]thymidine. At higher concentrations, a significant difference in mitogenic activity between wild-type (pep81-100) and substituted peptide (pep81-100K96A) can be observed. Medium without peptide and medium substituted with 10% fetal bovine serum were used as negative and positive controls, respectively. Values are given as means Ϯ S.D. (n ϭ 3). p values are given in the figure. B, ␤-galactosidase fusion proteins of wild-type (HDGF-␤GalHis) and mutated HDGF bearing the K96A substitution (HDGFK96A-␤GalHis) or missing amino acids 81-100 (⌬HDGF-␤GalHis) were produced and purified as described under "Materials and Methods." 4 nM of each of the proteins was added in serumfree medium to human skin fibroblasts, and proliferation was quantified after 20 h by the amount of incorporated [ 3 H]thymidine. 4 nM of non-fused ␤-galactosidase (␤GalHis) was used as a negative control. In contrast to wild-type HDGF, none of the mutated HDGF proteins were able to stimulate proliferation compared with ␤-galactosidase when added to the cell culture medium. Values are given as means Ϯ S.D. (n ϭ 3). *, p Ͻ 0.05; ns, not significant compared with ␤GalHis control. [ 3 H]Thymidine incorporation as a measure of proliferation activity is given in cpm. C, eucaryotic expression vectors coding for ␤-galactosidase fusion proteins of HDGF bearing the K96A substitution (HDGFK96A-␤GalHis) or missing amino acids 81-100 (⌬HDGF-␤GalHis) and a peptide-␤-galactosidase fusion protein (HATH81-100-␤GalHis) were constructed as described under "Materials and Methods" and transfected into NIH 3T3 cells. Vectors coding for non-fused ␤-galactosidase (␤GalHis) and wild-type HDGF fused to ␤-galactosidase (HDGF-␤GalHis) were used as negative and positive controls, respectively. 48 h later, cell numbers were determined using the MTS assay. All constructs induced proliferation, whereas peptide 81-100 from the HATH region was not able to induce proliferation when expressed as a fusion protein inside the cell. Values are given as means Ϯ S.D. (n ϭ 4). **, p Ͻ 0.01 between constructs and the ␤GalHis control.

DISCUSSION
HDGF is a member of a family of growth factors sharing an amino-terminal conserved HATH region (4,9). For three of the six members of this family, it has been shown that they exhibit growth factor activity to various cell types (1-3, 8, 44). A putative HDGF receptor, however, has not yet been identified. Our data demonstrating saturable high affinity binding of HDGF fusion proteins to the cell surface of NIH 3T3 cells are compatible with the existence of a HDGF receptor. Furthermore, we were able to identify amino acid residues 81-100 of the HATH domain as functionally important for HDGF action. We have shown that amino acids 81-100 of the HATH region are necessary and sufficient to mediate cell surface binding with similar affinity and saturation characteristics as the fusion protein covering the entire HDGF. In addition, a protein region responsible for binding to the plasma membrane should be located on the surface of a protein. Recently, Sue et al. (10) provided NMR data that allowed three-dimensional modeling of the HATH region. Within this three-dimensional model, the HATH81-100 peptide does indeed locate to the surface of HDGF (Fig. 7A). Furthermore, the peptide is well structured. It comprises one of two ␣-helices present in the HATH region of HDGF (Fig.  7B). It is located in near proximity to the region responsible for binding to heparin but is not identical to it (10). This is in accordance with our data showing that the HATH81-100 peptide does not bind to heparin. Therefore, it can be concluded that this region binds to another (non-HLGAG), most likely proteinaceous binding site on the cell surface.
The significance of the HATH81-100 peptide for the action of HDGF is further supported by its ability to induce a proliferative response in treated cells. Within the sequence of amino acid residues 81-100, lysine 96 seems to play an essential role. Substitution of the residue by alanine completely abolishes the cell surface binding activity of the HATH81-100-␤GalHis fusion protein and abolishes the proliferative activity of the peptide completely.
The mitogenic activity of intracellularly expressed HDGF is located in the carboxyl-terminal non-HATH region and depends on its nuclear localization (5,26), but exogenously added HDGF has mitogenic activity (2, 3). These results leave two possible mechanisms for the mitogenic activity of exogenously added HDGF. First, the growth factor is internalized and delivered to the nucleus of the cell to exhibit its proliferative activity through the same pathway as intracellularly expressed HDGF. In this respect, it is important to note that internalization and nuclear translocation of exogeneous HDGF and LEDGF have been demonstrated (26,31). Second, HDGF acts independently of its nuclear localization through a hitherto unidentified cell surface receptor. Of course, both mechanisms are not mutually exclusive.
In the experimental setting used here, exogenous HDGF seems to exert its proliferative effect via activation of a cell surface receptor without the need of internalization because we found no evidence of HDGF internalization, even when highly sensitive techniques were applied. Because previous examinations on HEK cells, in contrast, have demonstrated internalization of HDGF (26), it seems likely that the capability of cells to internalize this growth factor depends on the cell type investigated.
In addition, the HATH81-100 peptide is located in the HATH region, whereas the proliferative activity of the growth factor expressed in the cell nucleus has been assigned to the non-HATH region (25,26). In accordance with these data, transfection of cDNAs coding for HDGF fusion proteins induced proliferation even when the HATH81-100 region was deleted. In contrast, when a cDNA of a HATH81-100 peptide fusion protein was transfected, no proliferative response was noted. The peptide exhibits mitogenic activity when applied exogeneously, but not when expressed within the cell. In contrast, a HDGF construct missing this region is inactive when applied exogenously but causes cell proliferation after transfection. This clearly demonstrates the existence of two different pathways through which HDGF exerts its proliferative activity. One pathway depends on the non-HATH region and intranuclear localization of HDGF, whereas the other depends on the HATH region and binding of the growth factor to an unknown cell surface receptor.
The latter is further supported by the observation that HDGF already induces phosphorylation of intracellular mitogen-activated protein kinases 5 min after addition to the cell culture medium of pulmonary epithelial cells (37). When we attempted to activate this pathway in NIH 3T3 cells, the increase in Erk1/2 phosphorylation after addition of HDGF to the culture medium was minor and was not comparable with FGF or platelet-derived growth factor used as positive controls (data not shown). Thus, Erk1/2 activation seems to play only a minor role in HDGF stimulation of fibroblasts. In summary, the mode of action of HDGF may vary considerably between different cell types with respect to internalization and subsequent nuclear translocation as well as the extent of activation of known intracellular signal transduction pathways. Future studies must reveal the molecular basis of this functional heterogeneity.
Finally, it should be noted that features of HDGF resemble FIG. 7. The identified peptide is located on the surface of the HATH region of HDGF. Recently, Sue et al. (10) provided structural data that allowed generation of three-dimensional models of the HATH region. The peptide comprising amino acid residues 81-100 is highlighted in dark gray. A, a space filled model of the HATH region demonstrates the localization of the identified peptide on the surface of the molecule. B, a model showing the secondary structure depictions of the HATH region. The identified peptide comprises an ␣-helix present in the HATH region. Structural data were obtained from the Protein Data Bank (accession number 1RI0). The models were predicted using Protein Explorer software (www.proteinexplorer.org) (50). those of other growth factors, in particular, FGF. Thus, FGF also binds to HLGAGs and acts through a tyrosine kinase cell surface receptor, but it can also be internalized (45)(46)(47)(48). Furthermore, its translocation from the cell surface to the nucleus depends on proteoglycans and, like HDGF, it lacks signal peptides, yet it is secreted (27,49). These similarities point to common features of the mode of action employed by these growth factors.