Functional Domains of Recombinant BactericidaVPemneability Increasing Protein (rBPIZ3)*

The 23-kDa recombinant amino-terminal bactericidal/ permeability increasing protein fragment (rBPIzs) has all of the antibacterial and antiendotoxin properties of the holoprotein. In the current studies, we have identified multiple active domains within rBP12s with chemi- cal and proteolytic cleavage fragments and with syn- thetic overlapping peptides. We also demonstrate a novel, high affinity heparin binding property for rBP12s, in addition to its established bactericidal and lipopoly- saccharide binding properties. Cleavage fragments and synthetic, overlapping peptides of rBPIzs were analyzed for inhibition of the lipo- polysaccharide-induced Limulus amebocyte lysate reaction, for bactericidal activity, and for heparin binding. Three separate, active domains were identified in amino acid regions 17-45,66-99, and 142-169. Asingle synthetic peptide (86-99) was bactericidal. These results indicate that rBP12s is comprised of three separate functional domains which contribute to the high affinity interaction of rBPIz3 with Gram-negative bacteria. The individual activity of each domain and the cooperative in- teraction among domains provide the basis for developing analogues with increased biologic efficacy.

tericidal and LPS binding assays (5,8). The carboxyl-terminal half of the protein may be involved in localizing the molecule within the neutrophil granule, but its precise function is unknown (6,9).
Functional domains may exist within the 25-kDa amino-terminal fragment of BPI. Many bactericidal proteins share the common structural motif of an amphipathic a helix characterized by a cationic region which is opposed by a hydrophobic region. Bactericidal sites can be formed by a continuous amino acid sequence or by discontinuous sequences that are brought together by the three-dimensional structure of the molecule (10). For example, a continuous bactericidal site is found at the amino terminus of the cecropins, a family of antibacterial peptides found in the hemolymph of lepidopteran insects (11). Alternatively, a discontinuous bactericidal site can be found in the bactenecins from bovine neutrophils (12) and in the defensins (13). To date, the bactericidal region(s) within BPI have not been described.
Many mammalian proteins isolated from blood bind to soluble heparin or cell surface heparan with a heparin binding domain characterized by a cluster of basic and hydrophobic amino acids (14). This similarity of the heparin binding motifs to the bactericidal/LPS binding motifs prompted us to analyze the potential heparin binding by rBPIz3. Heparin is a sulfated polysaccharide with a net negative charge and a molecular mass range of 6 to 25 kDa (15). In general, heparin binding proteins can be categorized into two classes. The first class consists of proteins which require heparin as a cofactor to elicit a specific biologic response. Examples of this first type of heparin binding protein include fibroblast growth factors (16, 171, endothelial cell growth factors (18), and antithrombin 111 (19). The second class of heparin binding proteins consists of proteins which neutralize the heparin-dependent stimuli of the first class of heparin binding proteins. Examples of this second type include platelet factor IV (20) and thrombospondin (21). Other proteins such as interleukin 3 (221, interleukin 8 (231, complement C l q (24), granulocyte/macrophage colony stimulating factor (22), and heparin binding protein (CAP37, azurocidin) (25) also bind to heparin, but the function of this binding has not been determined. Heparin may protect these proteins from proteolytic degradation and/or serve as a target for localization of proteins to heparan-rich cells ( i e . endothelial cells) or to the heparan-rich extracellular matrix.
We report herein that in addition to its bactericidal and high affinity LPS binding properties, human rBPIz3 binds specifically to heparin. We also describe the fragmentation of rBPIZ3 by cyanogen bromide (CNBr) or by endoproteinase Asp-N digestion and subsequent analysis of the fragments for LPS interactions, bactericidal activity, and heparin binding. Overlapping peptides of the rBPIz3 sequence were also synthesized and analyzed for activity to further define the functional domains of the recombinant molecule. We have identified three separate, active domains within the rBPIz3 molecule that contribute t o the overall biologic activity.

EXPERIMENTAL PROCEDURES
Reduction and Alkylation of rBPIzrThe production and purification of rBPIZ3 will be described in detail elsewhere. The rBPIZ3 was reduced and alkylated prior to proteolysis by cyanogen bromide or endoproteinase Asp-N. The protein was desalted by cold (4 "C) acetone precipitation (1:l v/v) overnight and pelleted by centrifugation (5000 x g) for 10 min. The rBPI,, pellet was washed twice with cold acetone and dried under a stream of nitrogen. The rBPI,, was then reconstituted to 1 mdml in 8 M uredO.1 M Tris, pH 8.1, and reduced by addition of 3.4 nm dithiothreitol (Calbiochem) for 90 min at 37 "C. Alkylation was performed by the addition of iodoacetamide (Sigma) to a final concentration of 5.3 nm for 30 min in the dark at room temperature. The reduced and alkylated protein was acetone-precipitated and washed as described above, and the pellet was redissolved for either CNBr or Asp-N digestion.
Cyanogen Bromide Cleavage-Reduced and alkylated rBPI,, was dissolved in 70% trifluoroacetic acid (TFA) (protein sequencing grade, Sigma) to a final protein concentration of 5 mg/ml. Cyanogen bromide (Baker Analyzed Reagent, VWR Scientific, San Fransisco, CA) dissolved in 70% TFA was added to give a final 2:l ratio of CNBr to protein (w/w). This is approximately a 75-fold molar excess of CNBr over methionine residues in rBP12,. The reaction was purged with nitrogen and allowed to proceed for 24 h in the dark at room temperature. The reaction was terminated by adding 9 volumes of distilled water with subsequent freezing (-70 "C) and lyophilization.
Endoproteinase Asp-N Digestion of rBPIz,The reduced and alkylated rBPIZ3 was solubilized at 5.0 mg/ml in 8 M uredO.1 M Tris, pH 8.1. An equal volume of 0.1 M Tris, pH 8.1, was added so that the final conditions were 2.5 mg/ml protein in 4 M uredO.1 M Tris, pH 8.1. Endoproteinase Asp-N from Pseudomonas fiagi (Boehringer Mannheim) was added at a 1:lOOO (w/w) enzymexubstrate ratio, and the reaction was allowed to proceed for 6 h at 37 "C. The reaction was terminated by addition of TFA to a final concentration of 0.1%, and the samples were fractionated by reverse phase HPLC.
Reverse Phase HPLC of Fragment Mixtures-The CNBr and Asp-N fragment mixtures were purified on a Zorbax Protein Plus C3 column (4.6 x 250 mm, 300-Apore size, MACMOD Analytical Inc., Chadds Ford, PA). The column was eluted with a gradient from 5% acetonitrile in 0.1% TFA to 80% acetonitrile in 0.1% TFA over 2 h at 1.0 mWmin. Fragment elution was monitored at 220 nm using a Beckman System Gold HPLC. The column heating compartment was maintained at 35 "C, and fractions were collected manually, frozen at -70 "C, and dried in a Speed Vac concentrator. Fragments were then solubilized in 20 nm sodium acetate, pH 4.0/0.5 M NaCl prior to use. Amino acid sequence and molar concentrations of the cleavage fragments were determined by Edman degradation using an Applied Biosystems Model 477A protein Sequencer and Model 120A analyzer.
Mass Spectrometry-Electrospray ionization mass spectrometry (ESI-MS) was performed on a VG Bio-Q mass spectrometer. Molecular masses were obtained by mathematical transformation of the data (26).
Peptide Synthesis-Peptides were synthesized in a 96-well plate format using solid phase pin technology as recommended by the manufacturer (Cambridge Research Biochemicals Ltd., Wilmington, DE). Briefly, the sequence of rBPI,,  was divided into 47 different 15-mer peptides that progressed along the sequence of rBPIZ3 by initiating a subsequent new peptide every fifth amino acid. Peptides were synthesized in duplicate to minimize possible synthesis errors. Control pins were subjected to the same synthesis steps as the other pins without the addition of activated FMOC (N-(9-florenyl)methoxycarbonyl)amino acids. Peptides were cleaved from the solid-phase pin support using an aqueous basic buffer (0.1 M sodium carbonate, pH 8.3). A quantitative diketopiperazine cyclization occurs under these conditions, resulting in cleaved peptides with a cyclo(lysylproly1) moiety on the carboxyl terminus (27).
Inhibition of the Limulus Amebocyte Lysate Response to LPS-rBPI,, fragments and synthetic peptides were tested for inhibition of the Limulus amebocyte lysate (LAL) response to LPS. r B P L fragmentdpeptides were added to Eppendorf tubes together with a constant amount (4 ng/ml) of Escherichia coli 0113 LPS (Associates of Cape Cod, Woods Hole, MA). ARer incubation at 37 "C for 3 h, 360 pl of phosphate-buffered saline was added to each tube to yield a final LPS concentration of 200 pg/ml. Fifty pl of diluted samples were then transferred to the wells of Immulon I1 strips (Dynatech, Chantilly, VA) with an equal volume of Limulus amebocyte lysate (Quantitative Chromogenic LAL Kit, Whittaker Bioproducts, Inc., Walkersville, MD). Following incubation at room temperature for 25 min, 100 pl of chromogenic substrate (prepared according to manufacturer's specifications) was added to each well. After incubation for 20 min, the reaction was stopped by the addition of 100 pl of 25% acetic acid. Absorbance was quantitated at 405 nm using a V , , microplate reader (Molecular Devices, Menlo Park, CA).
Radial Difision Bactericidal Assay-The rBPIZ3 synthetic peptides, CNBr cleavage fragments, and Asp-N digestion fragments were analyzed for bactericidal activity using the rough mutant E. coli 55 in a radial diffiion assay (28). Specifically, an overnight culture ofE. coli 55 was diluted 150 into tryptic soy broth (Difco) and incubated for 3 h to midlog phase. The bacteria were centrifuged at 1500 x g for 5 min in a Sorvall RT6000GB, resuspended, and washed with 10 nm sodium phosphate buffer, pH 7.4, and finally resuspended in phosphate buffer (5 ml). Bacterial concentration was determined by measuring absorbance at 590 nm (1.25 x lo9 colony forming unitdml yield Asgo = 1.00). The bacteria were diluted to 4 x lo6 colony forming unitdml in 10 ml of molten underlayer agarose. The underlayer agarose was composed of 1% agarose M (Pharmacia, Uppsala, Sweden), 3% tryptic soy broth, and 0.02% Tween 20 (Sigma) dissolved in phosphate buffer. The mixture was poured into a square, 100 x 100 x 15-mm Petri dish (Baxter Scientific Products, McGaw Park, IL), and allowed to solidi& A series of wells were punched into the agarose with a sterile 3-mm punch attached to a vacuum source. A volume of 10 pl containing rBPIZ3, fragments, or peptides (0.1-10 pg) was added to individual wells. The plates were incubated at 37 "C for 3 h, and 10 ml of molten overlayer agarose was poured on top. Overlayer agarose was composed of 1% agarose, 6% tryptic soy broth in the phosphate buffer. The overlayer was allowed to harden, and the plates were incubated overnight at 37 "C. Bactericidal zones were visualized by staining the plates for 24 h with a solution of 0.002% Coomassie Brilliant Blue R-250 (Bio-Rad), 27% methanol (Sigma), and 15% formaldehyde (Sigma) in distilled water. The bactericidal zones were measured with a micrometer (e0.05 mm), and data are expressed as area in nun2.
Direct Heparin Binding A s s a p A direct [3H]heparin binding assay (29) was modified by using a polyvinylidene difluoride membrane (Immobilon-P, Millipore) to bind proteindpeptides with a higher capacity than nitrocellulose (30). The hydrophobic membrane was hydrated by successively aspirating 100 pl each of ethanol and then water through each well of a 96-well Multiscreen IP plate as recommended by the manufacturer.
rBPI,, or BPI fragmentdpeptides (0.5-5 pg in 100 pl of PBS) were adsorbed to the membrane by aspiration. The membrane was then blocked with 100 pl of 0.1% bovine serum albumin in phosphate-buffered saline, pH 7.4 (blocking buffer). Dilutions of PHIheparin (0.03-20 pCi/ml, average M , = 15,000, DuPont NEN) were made in blocking buffer, and 100 pl were incubated in all wells for 1 h at 4 "C. The unbound heparin was aspirated through the membrane and the wells were washed three times with blocking buffer, dried, and punched from the plates for liquid scintillation quantitation in an LKB Model 1217 scintillation counter. Results are expressed as the total number of countdmin bound minus the counts/min bound by control wells which received blocking buffer only.
Heparinqffinity Chromatography-A 1.0-mg rBPI,, sample in 1.0 ml of phosphate-buffered saline, pH 7.4, was loaded onto an EconoPac Heparin Cartridge (Bio-Rad) using an EconoSystem Chromatography System (Bio-Rad). A linear NaCl gradient from 0.15 M to 2.15 M NaCl over 30 min (2.0 mumin) was used to elute the rBPIz3 from the column. Fractions were collected, and the absorbance at 280 nm was monitored.  Table I. The fragments were isolated by reverse phase ((33) HPLC (Fig. MI, and their amino-terminal sequences were determined by Edman degradation. The two largest fragments (C1 and C5) were not resolved by the C3 HPLC column; further attempts to resolve them by ion exchange chromatography were unsuccessful, presumably because they are similar in length and isoelectric point. The identities of the C1, C5 fragments within the mixture were determined by ESI-MS. The predicted mass of C 1 is 6269 (Table I) (Fig. 1 B ) were sequenced, and masses were determined by ESI-MS (Table I). A short duration digest at a 1:lOOO (w/w) enzyme:substrate ratio was used to eliminate potential nonspecific cleavages, particularly at glutamic acid. It is evident that this digestion did not proceed to completion, as one fragment (1-38) was isolated where Asp residues (amino acids 15 and 35) were not cleaved. The mass spectra of the Asp-N fragments were consistent with the predicted masses for each individual fragment. Unlike the CNBr cleavage, where the carboxyl-terminal fragment was poorly resolved, the Asp-N fragment from amino acid 116 to the carboxyl terminus was well resolved from all of the other Asp-N fragments. Bactericidal Actiuity-No bactericidal activity was demonstrated for the rBPIz3 fragments generated by CNBr or by Asp-N digestion, when tested up to 25 pmol/well. This assay detected measurable bactericidal activity with as little as 0.75 pmol of rBPIz3 per well. Reduced and alkylated rBPIz3 (up to 100 pmoywell) also was not bactericidal, while alkylated rBPIZ3 retained bactericidal activity equivalent to rBPIz3 (data not shown).

rBPI,, Fragments
Inhibition of the LPS-induced LAL Reaction-BPI inhibits LPS stimulation of the Limulus amebocyte lysate proteolytic cascade (31). The results shown in Fig. 2 demonstrate that rBPIz3 fragments from both the CNBr and the Asp-N digests also sigdicantly inhibit the LPS-induced LAL reaction. The CNBr digest fraction containing amino acid fragments 1-56 and 112-170 (Cl/C5) inhibited the LPS-induced LAL reaction with an ICso of approximately 100 n~. This ICso is approximately 10-fold higher than the ICso for rBPIz3 (9 n~) in the same assay. CNBr digest fragments not shown in Fig. 2 were noninhibitory.
A slightly different result was observed with fragments generated from the Asp-N digest, where three fragments were found to be inhibitory in the LAL assay (Fig. 2). The fragment corresponding to amino acids 116-193 (A6a) exhibited LAL inhibitory activity similar to intact rBPIz3 with complete inhibition of the LPS-induced LAL reaction at 15 IIM. The fragments corresponding to amino acids 57-104 (A4) and 1-38 (All also inhibited the LAL assay, but required 10-fold higher amounts. These results, in combination with the CNBr digest results, indicate that at least three regions of the rBPIz3 molecule have the ability to neutralize LPS activation of the LAL reaction with the most potent region appearing to exist within the 116-193 (A6a) fragment.
Binding to Heparin-rBPIz3 bound to L3H1heparin in a direct binding assay with relatively high affinity ( K d = 70 IIM) and saturable binding (Fig. 3). The binding of radiolabeled heparin was completely inhibited by a 100-fold excess of unlabeled heparin (data not shown). rBPIz3 binding to heparin was also demonstrated by heparin affinity chromatography. Bound rBPIz3 was eluted during a linear NaCl gradient (0.15 to 2.0 M NaC1) with a single peak at 0.84 M NaCl (data not shown).
Binding of rBPIZ3 CNBr fragments to heparin was estimated using 100 pmol of each fragment per well with a saturating concentration of L3H]heparin (20 pg/ml). The results (Table 11) indicate that CNBr fragments containing the amino acids 71-100 (C3) and 1-56 and 112-170 (Cl/C5) bound heparin to a similar extent. The CNBr fragment 171-193 had lower heparin binding than C3 or Cl/C5. Each of the above fragments bound more heparin than the control protein, thaumatin, a protein of similar molecular weight and charge to rBPIz3.
Three Asp-N rBPIz3 fragments also demonstrated heparin binding. As seen in Table 11, the 57-104 Asp-N fragment (A4) bound the highest amount of heparin, followed by the 1-38 (AlA2) and 116-193 (A6a) fragments. These data, in combination with the CNBr fragment data, indicate that there are at least three separate heparin binding regions within rBPIZ3, with the highest capacity residing within residues 71-100.  tivity. To examine the rBPIZ3 sequence in greater detail, overlapping 15-mer peptides were synthesized and tested. Samples of 10 peptides chosen randomly were analyzed by reverse phase HPLC and found to contain one major peak accounting for more than 85% of the peak areas in the sample (data not shown). Bactericidal Actiuity-Bactericidal assays on the synthetic rBPIZ3 peptides indicated that only one peptide (85-99) was bactericidal (Fig. 4A). This same sequence is present in the CNBr fragment 71-100 (C3) and in the Asp-N fragment 57-104 (A4). Inhibition of LPS-induced LAL Reaction-Three regions of overlapping synthetic rBPIZ3 peptides inhibited the LPS-induced LAL reaction (Fig. 4B 1. The most active domain resides within residues 73-99. A second active region was seen between residues 140 and 170 and a third between residues 25 and 50. Apossible fourth minor domain was noted in the region, 118-128. These domains correspond to similar regions identified with the rBPIz3 cleavage fragments. Binding to Heparin-Three major domains of rBPIz3 synthetic peptides also bound heparin (Fig. 4C). The most active heparin-binding region encompassed residues 73-99. A second active region was evident from residues 25-50 and a third  the molecule (Fig. 5). The results from our fragmentation experiments suggest that there are three rBPIz3 domains capable of interacting with LPS. The most active cleavage fragment is found in the carboxyl-terminal region of rBPIz3 (amino acids 116-193). This region is potent in the inhibition of LPS-induced LAL activity (Fig. 2). This region includes the most hydrophobic region of rBPIz3 (amino acids 151-160) as well as all three of the cysteines in rBPIz3 (amino acids 132, 135, and 175). The hydrophobic domain (amino acids 151-160) is flanked by short sequences of cationic lysines andlor histidines. It is intriguing to postulate an intramolecular disulfide bond between two of the three cysteines (amino acids 132, 135, and 175) that would create a n LPS binding loop within the rBPIz3 molecule in a fashion analogous to the discontinuous LPS binding sites of the bactenecin (12) and defensin molecules (13). The other two regions that inhibit the LAL reaction to LPS are fragments 1-38 and 57-104. The decreased inhibition of the LAL reaction to LPS by these regions may be similar to the LPS binding by the linear polymyxin analogues (38, 39).
The results from the synthetic 15-mer rBPIz3 peptides also identify three separate, continuous domains which inhibited the LPS-induced LAL reaction. Clearly, the most inhibitory domain was between residues 73 and 99. Based on the assumption that all peptides were synthesized with approximately the same yield (supported by random HPLC analysis), the most active peptide is that containing amino acids 85-99. Further studies with larger amounts of synthetic peptides from these regions will determine the relative affinities of the three regions in competition for LPS binding assays.
Bactericidal activity was observed only in the synthetic peptide containing amino acids 85-99. The lack of bactericidal activity in the CNBr and Asp-N fragments containing residues 85-99 may be due to some alteration of the secondary structure caused by amino acids beyond the 85-99 sequence. Alternatively, some alteration during reductiodalkylation or concentration effects may have caused the lack of bactericidal activity. These fragments inhibit the LAL reaction to LPS, but the bactericidal activity is greatly diminished, if not abolished. Clearly, the synthetic peptide of residues 85-99 does have bactericidal activity (albeit at higher concentrations than rBPIZ3), and it is intriguing to postulate that this domain is responsible for the bactericidal properties of rBPIz3.
The bactericidal peptides, cecropins and magainins, are characterized by a continuous, amphipathic, a-helical region which is necessary for activity (10,401. The continuous amphipathic motif is located at the amino-terminal 10 amino acids of the cecropins (41). The bactericidal rBPIz3 peptide has high homology with the most active cecropin bactericidal sequence: cecropin A, K-W-K-L-F-K-K-I-E-K (1-10); rBPIz3, K-W-K-A-Q-K-R-F-L-K (90-99). The two other functional rBPIZ3 domains may give the whole rBPIz3 molecule a higher affinity for the outer membrane LPS of Gram-negative bacteria than the 90-99 peptide and thus, more effectively, target this bactericidal sequence to the bacterium. Identification of this bactericidal sequence within rBPIZ3 allows for future rBPIz3 analogues with increased biologic efficacy.
We observed a high degree of structural similarity between the cationidhydrophobic motif of LPS bindinghactericidal molecules and the known consensus sequences of heparin binding proteins (X-B-B-X-B-X or X-B-B-B-X-X-B-X, where X is any hydrophobic amino acid and B is any basic amino acid) (14).
rBPIz3 binding to soluble heparin in the direct binding assay and to solid phase heparin in the heparin affinity chromatog-raphy, indicates that the rBP12&eparin interaction is similar to other heparin binding proteins, including: IL-3, GM-CSF, IL-8, PF4, FGF-1, FGF-2, Clq,ATIII, and thrombospondin (22,23,20,16,29,24,19,21). All of these proteins, except thrombospondin, bind to heparin with one of the above cationid hydrophobic amino acid repeating motifs (14). Thrombospondin is a heparin binding protein that requires a Trp-Ser-X-Trp consensus sequence for heparin interactions (21). Thus, two different types of amino acid motifs are found in heparin-binding proteins. All three active domains of rBP123 contain aspects of both types of motifs (Fig. 5).
An unexpected finding from our data is that an excellent correlation exists between the synthetic rBP123 peptides that bind to heparin and those which inhibit the LPS-induced LAL reaction ( r = 0.75, p = 0.0001, n = 47). These data suggest that LPS and heparin may present similar charged arrays to the proteins with which they interact. Heparin is a highly sulfated polysaccharide with a molecular mass of 6 to 25 kDa and is composed of polymers of D-glucosamine-L-iduronic acid and Dglucosamine-D-glucuronic acid. The majority of heparin polymers consist of 8-12 disaccharide units. The overall net negative charge of heparin is due to its high content of sulfate and carboxylic acid functional groups. Sulfate esters are present at position 6 of glucosamine and position 2 of iduronic acid. Sulfamides are present at position 2 of glucosamine (15). Thus, in aqueous solution, both LPS and heparin display ionic arrays of phosphates or sulfates on a saccharide backbone. As a result, it is tempting to speculate that other proteins (e.g. polymyxins, cecropins, magainins, etc.) which bind LPS avidly may also bind tightly to heparin. Conversely, proteins which bind tightly to heparin (e.g. IL-3, GM-CSF, IL-8, PF4, Clq, ATIII, etc.) may also interact with LPS. Such interactions may lead to new clinical indications for LPSheparin binding proteins. Heparin is released from mast cell granules, and heparan, a cell-associated form, is found with a nearly ubiquitous distribution on mammalian cell surfaces and in the extracellular matrix. Future studies will determine the physiological role of rBPIZ3 heparidheparan binding. Possible roles include protection from proteolytic degradation andor tethering to the extracellular matrix as for FGF-2 (29).
In conclusion, we have identified three separate, active domains within the rBP123 molecule by two different techniques. One technique utilizes CNBr and endoproteinase Asp-N-derived fragments to dissect the whole molecule. The other approach uses synthetic overlapping 15-mer peptides of the rBPIZ3 amino acid sequence to determine the active domains. Overall, the two approaches complement each other with respect to heparin binding and LAL inhibition. Only one synthetic peptide (amino acids 85-99) was bactericidal. The contribution of all three sites, in concert, may synergize to create the total bioactivity of the I -B P I~~ molecule. These studies provide the rationale for the design of future I -B P I~~ analogues with increased biologic efficacy.