Identification and characterization of heparan sulfate-binding proteins from human lung carcinoma cells.

The heparan sulfate proteoglycan/heparin-binding proteins of the human lung carcinoma cell line LX-1 have been identified, partially purified, and characterized. Analysis of the binding of [3H]heparin to membranes isolated from LX-1 cells indicated the presence of two classes of binding sites, with Kd values of approximately 2 x 10(-10) and 4 x 10(-8) M and corresponding Bmax values of 1 x 10(5) and 2 x 10(7) binding sites/cell. Binding was also observed with isolated heparan sulfate chains and with intact heparan sulfate proteoglycan isolated from two different cell types. With each ligand, binding was inhibited by addition of unlabeled heparin. The binding proteins were extracted from LX-1 cell membranes in detergent solution, and two size classes of binding proteins were identified by overlaying transblots of electrophoretically separated proteins with radioactive ligands. These two classes of binding proteins were shown to contain doublets with estimated molecular masses of approximately 16 kDa (HSBP1A and HSBP1B) and approximately 32 kDa (HSBP2A and HSBP2B). The proteins were partially purified by heparin-Sepharose chromatography and shown to bind heparin and heparan sulfate proteoglycan. By amino acid composition, N-terminal amino acid sequence, and reactivity with antibody, HSBP1A was shown to be very similar to histone 2B; HSBP1B may also be related to histone 2A. HSBP2A and HSBP2B, however, did not react with antibodies to the major histones and had compositions different from one another and from HSBP1.

Heparan sulfate proteoglycans are components of plasma membranes (1,2), basement membranes (3,4), and interstitial matrices (5,6). Heparin is a related molecule that has many properties similar to those of the HS' side chains of HSPGs and is often used as an analogue of HS; however, in vivo, heparin usually occurs as free glycosaminoglycan chains within mast cell granules. The wide distribution of HSPGs has led to a number of studies on the biological functions of these molecules. These studies have indicated potential roles for HSPGs in regulation of several cellular phenomena, e.g. smooth muscle cell proliferation (7), angiogenesis attachment (9, lo), and neuromuscular junction formation (11). However, the mechanism of action of HSPGs in these phenomena is not clearly understood. Studies from several laboratories (12)(13)(14)(15)(16) suggest that one possible mechanism of action of extracellular matrix molecules is via interaction with binding proteins or "receptors" on the cell surface. Binding of heparin or HS to a number of cell types, including smooth muscle cells (17), hepatocytes (la), endothelial cells (19), and melanoma cells (20), has been described. However, in most cases, the binding proteins involved have not been characterized.
Lankes et al. (21) have isolated and characterized a 7%kDa heparin-binding protein from bovine uteri that may be involved in the inhibition of smooth muscle cell proliferation, and Winer and Ax (22) have isolated three 14-16-kDa heparin-binding proteins from plasma membranes of granulosa cells (22). We have identified a HS/heparin-binding protein from mouse B-16 melanoma cell membranes with a molecular mass of -14 kDa (20). Interaction of extracellular matrix HSPG with these melanoma cells appears to be involved in the regulation of the production of a collagenase stimulatory factor (23). In this study, we have identified, partially purified, and characterized two size classes of HSPG/heparin-binding proteins (HSBPl and HSBPS) with estimated molecular masses of -16 and -32 kDa, respectively, from membranes of the human lung carcinoma cell line LX-l. Each of the two classes of proteins can be resolved into two components whose amino acid compositions are similar in some respects, but are not identical.
The -16-kDa proteins (HSBPlA and HSBPlB) are closely related or identical to histones, whereas the -32-kDa proteins (HSBPBA and HSBPBB) appear to be unique HSPG/heparin-binding proteins. Assays-The binding assays for whole cells, membranes, and membrane extracts have been described before (20). Briefly, the cells or membranes prepared as described above were mixed with radioactive ligand, incubated, centrifuged, and washed, and the radioactivity associated with the cell or membrane pellet was measured as described (20). A solid-phase assay was used to measure binding of radioactive ligands to extracts and other soluble fractions.

Chemicals
In this assay, the wells of a microtiter plate were first coated with the material to be assayed (20). After washing the wells with PBS containing 0.1% bovine serum albumin, they were incubated with radioactive ligand at 25 "C for 30 min and washed again. The radioactivity bound to each well was determined by dissolving the bound material in 2% SDS as described previously (20

Binding Parameters
for Heparin, HS, and HSPG-The binding of radioactive GAGS and HSPG to the LX-l membrane preparation was investigated using optimal conditions for binding that were established in separate experiments. Fig. 1 shows the binding of [3H]heparin to membranes derived from LX-l cells on mixing a constant amount of the membranes with increasing concentrations of ligand. Saturation was obtained, and [3H]heparin binding was inhibited completely in the presence of excess unlabeled heparin. Binding was obtained with intact LX-l cells as well as with isolated membranes (Fig. 1). However, the amount of heparin bound to the cells was lower than that to membranes and varied between experiments. This may have been due in part to the tendency of the intact cells to form aggregates in suspension or to the presence of internal as well as cell-surface binding sites. Binding of ["Hlheparin to intact cells and to isolated membranes was also measured after washing the cells or membranes with 0.5 M NaCl. This treatment did not significantly alter the amount of binding in either case, implying that the binding sites are tightly associated with the cells and membranes.
Analysis of the [3H]heparin binding data (Fig. 2), using the LIGAND computer program of Munson (29), supported a two-site model, with apparent Kd values of 2 x 10-i' and 4 X lo-' M. The maximum number of binding sites/cell (B,,,) was 1 x lo5 for the higher affinity sites and 2 X lo7 for the lower affinity sites.
Binding to LX-1 membranes was also obtained using radiolabeled HS, and this binding was also completely inhibited with excess unlabeled heparin. However, binding was considerably less than that for heparin; for this reason and since the specific activity of HS was relatively low, binding could  LX-l cells.
The data from Fig. 1 were analyzed by the LIGAND program of Munson (29). Kd values of 2 x 10-l" and 4 X lo-" M and B,,, values of 1 X lo" and 1.7 X 10' binding sites/cell were obtained. Since HSPG is the natural component of extracellular matrices, we investigated the binding of radioactive HSPG, obtained from two different sources, to LX-l membranes. Fig.  3 shows the binding of HSPG from bovine endothelial cells and the inhibition of binding by heparin. A similar pattern of binding was obtained with radiolabeled HSPG isolated from human colon carcinoma cells. Inhibition of HSPG binding by heparin was 85-90% at concentrations up to 100 ng, but was only -70% at 400 ng of added HSPG. The latter result may have been due to the reduced ratio of unlabeled heparin to labeled HSPG at this concentration of HSPG or to participation of the protein core as well as the HS chains of HSPG in binding. Scatchard analysis of these data again gave curvilinear plots, but these were difficult to interpret since HSPG is likely to be a multivalent ligand and since the data suggest the presence of at least two binding sites. Assuming two sites, approximate calculations from the Scatchard plot gave Kd values of 4 X 10-l' and 6 X 10-l' M.  (20) and by others with hepatocytes (18). This can probably be explained by differences in the degree of sulfation in different heparan sulfate preparations such as has been observed in binding studies with hepatocytes (18). No significant inhibition of binding of either ligand was obtained with hyaluronic acid, and low levels of inhibition were observed with chondroitin sulfate. Dermatan sulfate showed 46 and 69% inhibition of radiolabeled heparin and proteoglycan binding, respectively. Identification of Binding Proteins-To investigate the approximate molecular size of the binding proteins, we performed an overlay binding assay where the proteins were first separated on SDS-PAGE and transblotted onto nitrocellulose membrane and then were incubated with the radioactive ligand. Fig. 4 shows an autoradiograph of such a nitrocellulose transblot of membrane proteins after incubation with radiolabeled heparin (he 3) or HSPG (lane 1). With each ligand, binding was seen to occur mainly to two protein doublets, one with an estimated molecular mass of -16 kDa (HSBPl) and the other with an estimated molecular mass of -32 kDa (HSBP2).

Specificity of Binding of GAGS and HSPG-
A pattern of binding was obtained with [""S]HS which was similar to that seen with heparin (data not shown). Also, the pattern of binding was not altered by washing the LX-l cells or the isolated membranes with 0.5 M NaCl prior to processing for the overlay assay. Two other protein bands (22 and 12 kDa) were also observed, but the intensity of these Heparan Sulfate-binding Proteins two bands varied with different membrane preparations, and some times they were not present. One possibility is that these bands represent breakdown products of the binding proteins, but this remains to be investigated.
With HSPG, weaker binding was also observed in regions above and below the HSBP2 doublet (lane I), indicating the possible presence of other HSPG-binding proteins in the LX-l membranes. The binding of radioactive HSPG (lane Z), HS, and heparin (data not shown) to all of these protein bands was almost totally inhibited in the presence of excess unlabeled heparin. The inhibition of [%]HSPG binding to these bands with unlabeled heparin makes it unlikely that this binding is due to reaction with the core protein of the proteoglycans.
Thus, all three ligands bound mainly to two protein doublets, which we have designated as HSBPlA and HSBPlB for for -16 kDa doublet and HSBPBA and HSBP2B for the -32-kDa doublet (A signifying the upper band and B the lower band in each case).
The data in Fig. 4  Binding was performed as described for cholate extract of the LX-l membrane preparation was applied to a column of heparin-Sepharose, and the column was eluted with a linear gradient of O-2 M NaCl in PBS. As shown in Fig. 6, the activity was eluted between 1.0 and 1.4 M NaCl, and the peak of activity is present in fractions 10-13. SDS-PAGE of the column fractions and silver staining of the gel indicated that fractions 10 and 11 were enriched with two doublets of estimated molecular masses of -16 and -32 kDa (Fig. 7A), the same sizes as the bands seen by the overlay assay of the crude membrane extract (Fig. 4). Fractions 12-14 mostly showed the lower doublet with a trace amount of the upper doublet.
To further confirm the association of binding activity with the two doublets observed in the fractionated material, we pooled the active fractions and performed an overlay assay with [""SIHSPG.
As shown in Fig. 7B, radioactive bands were associated with two doublets identical in size to those observed by silver staining in Fig. 7A and by autoradiography of overlays of the crude membrane extracts in Fig. 4. These data strongly indicate that the polysaccharide binding activities mostly reside in the two protein doublets (HSBPlA/B and HSBPPAIB) with molecular masses of -16 and -32 kDa, respectively.
In addition to identifying the HSBP bands by overlay assay, we measured binding of ["Hlheparin to the pooled active fractions from heparin-Sepharose using a solid-phase binding assay (Fig. 8 The column fractions were analyzed for conductivity (---) and heparin binding activity (M).  after separation by SDS-PAGE, transblotting to Immobilon, and excision of the four bands from the blot. Since the bands on Immobilon blots are very sharp and clearly delineated from one another, they can be excised cleanly without cross-contamination. Table II shows the total amino acid composition of the HSBPs. The compositions are similar for many of the amino acids, e.g. all four proteins contain a high concentration of alanine and lysine residues. However, significant differences are also apparent between the four proteins. For example, the serine, glycine, and leucine contents of HSBPlA and HSBPlB are quite different, as are the aspartate, glutamate, alanine, and lysine contents of HSBPSA and HSBP2B. Likewise, there are major differences between several of the amino acids of either HSBPlA or HSBPlB and either HSBPBA or HSBP2B.
The four protein bands were then sequenced from their N termini by the automated Edman degradation method. No Nterminal residue was detected with HSBPBA, HSBP2B, or HSBPlB, indicating that the a-amino acid group of their Nterminal residues is blocked in each case. HSBPlA, however, has the following sequence for the first 14 amino acids: Pro-Glu-Pro-Ala-Lys-Ser-Ala-Pro-Ala-Pro-Lys-Lys-Gly-Ser-, with the N-terminal amino acid being proline. Comparison of this partial amino acid sequence data of HSBPlA with other proteins (Intelli-genetics software) shows it to be identical to the N terminus of histone 2B (31). Furthermore, the total amino acid composition of HSBPlA is very similar to that of histone 2B (31).
To confirm the similarity of HSBPlA to histone 2B, Western blots were performed with antibodies to the major histones, namely H2A, H2B, H3, and H4. Antibody to histone 2B reacted with HSBPlA, but not with HSBPlB, HSBPBA, or HSBPBB (data not shown). Also, antibody to histone 2A reacted with HSBPlB, but not to the other HSBPs. None of the antibodies showed any reaction with HSBPBA or HSBP2B. Association of HSBPs with Cell Surface-The fact that binding occurs to intact LX-l cells as well as to isolated membranes and extracts suggests that the HSBPs are present, at least in part, at the cell surface. To investigate this further, we incubated a suspension of LX-1 cells with trypsin before isolating the membranes.
Since the cells were intact, only external proteins would be accessible to digestion by trypsin. Fig. 9 shows that treatment of cells with 0.1% trypsin resulted in the almost complete disappearance of HSBPl and HSBP2 from the membrane preparation, whereas 0.01% trypsin had little effect. DISCUSSION Our data indicate that human lung carcinoma LX-l cells possess membrane-associated binding sites that recognize heparin, HS, and HSPG. Analysis of the binding data for heparin suggested two classes of binding sites, with Kd values of 2 X lo-'" and 4 x lo-* M and corresponding B.,,,, values of 1 X lo" and 2 x 10'. However, other explanations such as negative cooperativity are also possible. HSPG molecules contain a number of GAG chains attached to a protein backbone, e.g. the endothelial preparation used here contains four to six HS chains/proteoglycan molecule (40). Since each of the HS chains would be expected to bind to an individual binding site, each proteoglycan molecule would interact with multiple sites. Calculations of Kd and B,,, values for multivalent ligands such as HSPG are difficult to interpret; how-ever, it is clear from approximate calculations that HSPG also interacts with the LX-l cell membranes with high affinity (& = -4 X lo-" and -6 X 10-l' M). It is likely that these high affinities of interaction are at least in part due to multivalent interaction since isolated HS chains appear to interact with lower affinity. However, the radiolabeled HS preparation used here was from 3T3 cells, and the HSPG preparations were from endothelial and colon carcinoma cells. Since HS preparations from different sources vary in their composition (18), the above interpretation of our results would require use of HS chains from the same source as the HSPG. This has not yet been possible due to the difficulty of obtaining sufficient amounts of labeled HS and HSPG of high specific activity, but current experiments are directed toward this purpose. Irrespective of these concerns, however, it is clear from the data presented that the two HSPG preparations used in this study interact with LX-l membranes with high affinity.
Most of the binding of radioactive HSPG as well as of HS and heparin itself was inhibited by heparin both in the membrane binding assays and in the overlay assays. This implies strongly that most of the HSPG binding occurred via its HS side chains, as opposed to its protein core. A low proportion of the total binding, however, may have been via the core protein as described recently for hepatocytes (32). The most logical interpretation of our data is that the major binding sites recognized by heparin, HS, and HSPG are the same sites, especially since the same group of proteins (HSBPl and HSBPB) were recognized by all three ligands in transblot overlays. However, HS from bovine kidney did not efficiently inhibit the binding of either heparin or HSPG. The ineffectiveness of this preparation of HS is probably due to its relatively low sulfate content as shown previously with hepatocytes (18) and B-16 cells (20). As with B-16 cells (20) and hepatocytes (18), hyaluronic acid did not inhibit binding of heparin, chondroitin sulfate did not inhibit significantly, and dermatan sulfate showed partial inhibition. These results taken together indicate that the iduronic acid and/or sulfate contents of HS/heparin are probably important structural determinants of binding for all three cell types.
In this study, we have obtained evidence for two classes of binding sites for heparin, and probably also for HSPG, on LX-l cells. A previous study (20) with B-16 cells also suggested this possibility, but those data were not conclusive in this regard. In agreement with the binding data, we detected two classes of HS/heparin-binding proteins (HSBPl and HSBPZ), as defined by their approximate molecular masses, that are candidates for the two binding sites. However, the HSPG/heparin binding affinities of the separated proteins are not yet known. Also, there may be four (rather than two) different HSBPs since HSBPlA, HSBPlB, HSBPBA, and HSBPPB appear to have significantly different amino acid compositions. Recent results' indicate that similar classes of proteins are also present in smooth muscle cell membranes. The relationship of these proteins to a 7%kDa heparin-binding protein from bovine uteri that also appears to be closely associated with the smooth muscle cell surface (21) remains to be investigated.
The binding of heparin to intact cells and the removal of all four HSBPs from intact cells by trypsinization indicate that the HSBPs are localized, at least in part, on the outer cell surface. However, partial N-terminal sequencing and reactivity with antibody to histone 2B indicate a strong homology of HSBPlA with histone 2B, and antibody reactivity also suggests that HSBPlB may be related to histone 2A.
The similarity in molecular masses suggests that the 14-16-kDa heparin-binding proteins from granulosa cell membranes (22) and the 14-kDa species from B-16 melanoma cells (20) may also be related to histones. It is possible that nuclear histones have been released from these cells during culture or harvesting, have nonspecifically become associated with the cell surface, and have remained associated with the membranes during isolation, even after extensive washing with 0.5 M NaCl. On the other hand, the presence of histone-like molecules at the cell surface that bind HSPG may be of biological importance. Several groups (33,34,36) have demonstrated the association of glycosaminoglycans with the nucleus, probably bound to histones. However, recent studies by Conrad and co-workers (35,36) indicate that extracellular HSPG is endocytosed and that a specific subpopulation of free HS chains becomes associated with nuclei via an unknown nonlysosomal pathway. A strong inverse correlation between the level of nuclear HS and cellular growth has been demonstrated (37). Thus, it is tempting to speculate that HSBPl may be involved in transport of this HS from the cell surface to the nucleus. Detailed studies of the cellular localization and role of HSBPl will be necessary to clarify this issue.
Although HSBPl may be derived from the nucleus or related to nuclear function, this does not seem likely for HSBP2. Neither HSBPZA nor HSBPZB was recognized by antibodies to the major histones (H2A, H2B, H3, and H4), consistent with their larger size compared with histones. Also, in recent studies' of smooth muscle cells using the same methodological approaches as those used in this study, we have found cellular variants that contain both the HSBPl and HSBPB doublets and variants that contain only the HSBPl doublet, i.e. the presence of HSBPB was not due to the membrane preparation method.
The functional significance of the four HSBPs and their relationship to each other are not yet clear. However, it is possible that these HSBPs, especially HSBP2, belong to the growing group of cell-surface receptors for matrix molecules (12)(13)(14)(15)(16) and thus may mediate cellular functions attributed to HSPG/heparin, e.g. inhibition of smooth muscle and mesangial cell proliferation (7,38), regulation of c-myc and c-fos mRNA levels in fibroblasts (39), angiogenesis (8), cell attachment (9, lo), neuromuscular junction formation (ll), or the above-mentioned transport of HS to the nucleus (35)(36)(37). We have also obtained evidence that extracellular matrix HSPG may regulate the production of a tumor cell-derived factor that stimulates production of fibroblast collagenase (23). Possibly, tumor cell HSBPs, such as those described here, are involved in this phenomenon. Our current work is directed toward obtaining specific antibodies and cDNAs to facilitate investigation of the relationship of the various HSBPs, their localization, and their biological roles.