INTRODUCTION

Bacterial endotoxin or lipopolysaccharide (LPS),¹ a major component of the outer membrane of Gram-negative bacteria, is a potent mediator of the inflammatory response. Because Gram-negative sepsis remains one of the primary causes of severe systemic inflammation in hospitalized and immunocompromised patients, there is great interest in characterizing proteins involved in the biological response to LPS. In this paper, we focus on two LPS-binding proteins, lipopolysaccharide-binding protein (LBP) and bactericidal/permeability-increasing protein (BPI). LBP and BPI are members of a family of lipid transfer/lipopolysaccharide-binding proteins that also includes cholesteryl ester transfer protein (CETP) and phospholipid transfer protein (PLTP). These proteins share significant sequence homology and all bind lipophilic substrates (1).

Involved in a complex array of responses to LPS, LBP is a 60-kDA serum glycoprotein that binds the lipid A portion of the LPS molecule to form a high affinity LBP·LPS complex (2). This complex potentiates the cellular response to LPS via interaction with the monocytic differentiation antigen CD14 (3, 4). LPS can be transferred from LBP to CD14 (3, 4), present as either a membrane-bound protein on myeloid cells or a soluble serum protein that interacts with endothelial and some epithelial cell lines to elicit an inflammatory response. Recent evidence suggests that LBP may additionally be involved in the neutralization of LPS via interaction with serum lipoproteins (5, 6) or through the internalization of a LBP·LPS·CD14 complex by neutrophils (7).

BPI is a 55-kDa protein found in granules of mature neutrophils and, like LBP, interacts with LPS to form a high affinity complex. BPI, however, binds LPS with higher affinity than does LBP (8-10), and BPI·LPS complexes do not stimulate monocytes or endothelial cells (11-13). The binding of LPS by BPI also prevents binding to LBP, neutralizing the inflammatory activity of LPS (14-16). BPI and its recombinant N-terminal fragments have been demonstrated to provide protection against challenge with bacteria or purified bacterial endotoxin in several animal models (17-22) and, more recently, in human clinical trials (23). Additionally, Rogy et al. (22) have tested a protein chimera of BPI and LBP, NCY103, in an endotoxin challenge model in baboons.

Both BPI and LBP contain 456 amino acids and show an approximate 45% homology at the amino acid level which is distributed over the entire protein sequence. Interestingly, the genes for BPI and LBP lie adjacent to each other in the human genome, suggesting that they might have arisen from a gene duplication event (24). LPS binding is a property of the N-terminal half of both LBP and BPI (9, 16, 25-27), and a proteolytic N-terminal fragment of BPI (26) served as the basis for generating the recombinant forms of BPI (9) and LBP (27). The C-terminal region of LBP is required for CD14 interaction (27-29), whereas the function of the C terminus of BPI is less well characterized. There is evidence that the C-terminal region of BPI possesses some LPS neutralization activity, especially in regard to smooth LPS (possessing long chain polysaccharides) (16). In addition to binding LPS, the N-terminal fragment of BPI also binds heparin (30).

To examine how the structure of these closely related proteins affects their opposing functions, we have created fusion proteins with shuffled N-terminal and C-terminal domains from LBP and BPI and have tested the fusion in both in vitro and in vivo model systems. Our results demonstrate that each fusion protein retains the LPS binding characteristics of the N-terminal domain of the parent protein and that the transduction of the LPS-induced inflammatory signal is dependent on the presence of the LBP C-terminal domain. Additionally, our data suggest that the heparin binding characteristics of BPI, LBP, and the fusion proteins play a role in their pharmacokinetics.

MATERIALS AND METHODS

Cloning of the cDNAs encoding the N-terminal BPI (rBPI₂₁) protein (31) and the N-terminal and full-length LBP (rLBP₂₅ and rLBP) proteins (9) have been described previously. Cloning of full-length rBPI was done in a similar manner as described for full-length rLBP (27). The cloned cDNAs were then inserted into the mammalian expression vectors described below.

Plasmids encoding the LBP_(1-197)BPI_(200-456) fusion protein P4160_(L-B) or the BPI_(1-199)LBP_(198-456) fusion protein P4161_(B-L) were constructed by combining appropriate portions of these molecules at ClaI restriction sites engineered by overlap polymerase chain reaction mutagenesis within homologous regions of the BPI or LBP cDNA. The introduced ClaI sites did not modify the encoded amino acid sequences within these regions of the LBP and BPI proteins.

The mammalian expression vectors for all proteins in this study were constructed by cloning DNA encoding the protein of interest into vectors containing the human cytomegalovirus promoter, the mouse light chain 3-untranslated region, and the Escherichia coli guanine-xanthine phosphoribosyltransferase gene for selection of transfectants in mammalian cells (32).

Generation of transfectant cell lines and protein purification of rLBP₂₅ (27), rBPI, rBPI₂₁ (31), and rLBP (27) were as described previously. The P4160_(L-B) fusion protein was purified according to the procedure outlined in Theofan et al. (27), whereas the protocol described in Horwitz et al. (31) was followed for the purification of P4161_(B-L). All proteins were purified to greater than 98% purity from either CHO-K1 transfectants (rBPI₂₁,P4160_(L-B), P4161_(B-L)) or CHO-DG44 (a DHFR mutant of CHO Toronto obtained from Dr. Lawrence Chasin, Columbia University, New York) transfectants (rBPI, rLBP₂₅, rLBP). Concentrations of the purified proteins were determined by absorbance at 280 nm using the following set of values for E(1 ng/ml/cm): rBPI₂₁, 1.02; rBPI, 0.58; rLBP₂₅, 0.71; rLBP, 0.63; P4160_(L-B), 0.54; P4161_(B-L), 0.62.

One hundred µg of rBPI₂₁, rBPI, or rLBP were iodinated using 100 µl of lactoperoxidase glucose oxidase-immobilized beads (Enzymobeads, Bio-Rad) in 100 µl of phosphate-buffered saline, 1.0 mCi of iodine-125 (Amersham Corp., IMS30), and 50 µl of 55 mM D-glucose for 45 min at room temperature. The reaction was quenched by centrifugation for 1 min. ¹²⁵I-ligand was purified by gel filtration with Sephadex G25 (Pharmacia Biotech Inc.) equilibrated with 5 mM sodium citrate, pH 7.2, 150 mM NaCl, 0.2% pluronic F68 (BASF, Mount Olive, NJ) and 0.002% Tween 80. ¹²⁵I-Ligand recovery and specific activity were determined by trichloroacetic acid precipitation. For binding analysis, 96-well plates were coated with lipid A (LIST Biological Laboratories Inc., Campbell, CA) by treating with 50 µl of a 2.5 µg/ml solution of lipid A in phosphate-buffered saline at 4 °C overnight. Plates were then washed three times with Dulbecco's modified Eagle's medium (Life Technologies, Inc.) + 1.0% bovine serum albumin in 10 mM HEPES, pH 7.4 (DBH). 50 µl of 0.5 nM ¹²⁵I-rBPI₂₁ or ¹²⁵I-rLBP in DBH were mixed with 50 µl of 20 pM to 10 µM unlabeled ligand, added to the wells, and incubated for 5 h at 4 °C. Binding was terminated by washing the wells three times with DBH and quantitated by extracting the wells with 1 N NaOH and counting the resultant extract in a Packard Gamma counter. Data from direct and competitive binding experiments were analyzed by the weighted nonlinear least squares curve fitting program Ligand adapted for Macintosh (MacLigand (33)). Objective statistical criteria (F, test, extra sum squares principle) were used to evaluate goodness of fit and for discriminating between models. Nonspecific binding was treated as a parameter subject to error and was fitted simultaneously with other parameters.

The human monocytic cell line THP-1 was obtained from the American Type Culture Collection (Rockville, MD). The cells were cultured in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% heat-inactivated (56 °C for 30 min) fetal bovine serum (HyClone Laboratories, Logan, UT), 1 mM glutamine, 1 mM pyruvate, 0.2 µM -mercaptoethanol. For induction of surface CD14, THP-1 cells (3 × 10⁵ cells/ml) were incubated with 80 nM 1,25-dihydroxyvitamin D₃ (Biomol Research Laboratories, Plymouth Meeting, PA) for 3 days as described (27, 36). CD14 expressing THP-1 cells were harvested and washed twice with Dulbecco's phosphate-buffered saline without Ca²⁺ or Mg²⁺ (phosphate-buffered saline; Life Technologies, Inc.) and resuspended in RPMI medium. They were then diluted to 2 × 10⁶ cells/ml in serum-free medium (SFM) made from RPMI medium plus 1% HB101 (Irvine Scientific, Irvine, CA), 100 units/ml penicillin, 100 µg/ml streptomycin, and 2 mM glutamine. For neutralization studies (Fig. 2), LPS (E. coli 0128:B12 LPS (Sigma)) was used at 1 ng/ml in SFM containing 1 µg/ml rLBP. For TNF induction studies (Fig. 3), 1 ng/ml 0128 LPS was used in SFM. Cells (2 × 10⁵/well) were plated, and after protein addition were incubated at 37 °C for 2 h with 5% CO₂. TNF levels in the culture supernatants were measured using a TNF ELISA kit (T Cell Sciences, Cambridge, MA) according to the directions specified by the manufacturer. Each sample was assayed in duplicate.

These experiments were done as described previously (30), and each sample was assayed in duplicate. Apparent affinity constants and binding capacities were calculated using the GraFit software (Erithacus Products Ltd.)

Data shown in Figs. 2 and 4 was analyzed by 2-way between groups analysis of variance with Fischer's protected least significant difference post hoc tests using SuperANOVA software (ABACUS Concepts Inc. Berkeley, CA).

All experimental protocols followed the guidelines according to Guide for the Care and Use of Laboratory Animals published by the U.S. Department of Health and Human Services and National Institutes of Health. Male CD® rats (Charles River, Wilmington, MA) weighing 250-300 g were used in all experiments. Animals were received healthy and housed in conventional cages and received standard laboratory chow and water ad libitum in an environmentally controlled animal room with 12-h light-dark cycles.

Plasma samples were assayed for rBPI₂₁ or rBPI using affinity purified rabbit anti-rBPI₂₃ as the capture antibody and biotin conjugated rabbit anti-rBPI₂₃ as the detection antibody as described (37). Plasma standards were prepared by adding known amounts of rBPI₂₁ or rBPI to rat plasma for the determination of percent recovery. The plasma concentrations in the samples were then corrected based on the recovery values.

Similarly, LBP and fusion proteins were also assayed by ELISA. For rLBP₂₅ and rLBP, affinity purified polyclonal rabbit anti-rLBP antibody was used as the capture antibody, and biotin-labeled, affinity-purified rabbit anti-rLBP antibody was used as the secondary antibody. For P4161_(B-L), rabbit anti-rBPI₂₃ antibody was used for the capture antibody, and biotin-labeled, affinity purified rabbit anti-rLBP antibody was the detection antibody. For P4160_(L-B), mouse monoclonal anti-C-terminal rBPI was used as the capture antibody (6C2), and biotin-labeled, affinity purified rabbit anti-LBP₂₅ was used for the secondary antibody.

Each protein was administered via the tail vein at a dose of 1 mg/kg to three male CD^® rats for a given experiment. Blood samples were collected from the retro-orbital sinus into tubes containing sodium citrate (Sigma). The plasma was extracted and stored at 70 °C until assayed. For rBPI₂₁, blood samples were collected at selected times from 0.5 min to 2 h after administration of rBPI₂₁. For rLBP₂₅, rLBP, and P4160_(L-B), blood samples were collected at selected times from 0.5 min to 72 h after administration of dose. For P4161_(B-L), blood samples were collected at selected times from 0.5 min to 8 h after administration of dose.

A two- or three-exponential disposition function was used to describe the change in concentration with time after protein was administered. The data were fitted by weighted nonlinear least squares analysis using the software program PCNONLIN (Statistical Consultants, Inc., Lexington, KY). The pharmacokinetic parameters clearance rate (CL, ml/min/kg), steady state volume of distribution (V_ss, ml/kg), volume of distribution of the central compartment (V_c, ml/kg), and total body mean residence time (bMRT, minutes) were calculated from curve fit parameters using standard equations (38). In addition, the plasma mean residence time (pMRT, minutes) was calculated as V_c/CL.

RESULTS

To examine the specific structure/function properties and intramolecular domain interactions between the N- and C-terminal domains in BPI and LBP, we prepared the following six recombinant protein species: full-length forms of BPI (rBPI) and LBP (rLBP (27)), truncated N-terminal forms of BPI and LBP (rBPI₂₁ (31) and rLBP₂₅ (27)), and two fusion proteins. The fusion protein P4160_(L-B) was made by joining the N-terminal amino acids 1-197 of LBP to the C-terminal amino acids 200-456 of BPI. The converse construct, termed P4161_(B-L), was made by fusing the N-terminal amino acids 1-199 of BPI to the C-terminal amino acids 198-456 of LBP (see "Materials and Methods" and Fig. 1). The properties of these recombinant proteins were then compared in both in vitro and in vivo assays as described below.

Previously we have shown that BPI binds lipid A with a much higher affinity than LBP (apparent K_d values of 2-5 nM for BPI versus 60-80 nM for LBP (9, 27)). To evaluate the apparent K_d values of each recombinant protein species, we measured the ability of rBPI₂₁, rBPI, rLBP, P4160_(L-B), or P4161_(B-L) to compete with the binding of ¹²⁵I-rBPI₂₁ to immobilized lipid A. The data shown in Fig. 2 are from a representative experiment and indicate that rBPI₂₁ competes best with radiolabeled rBPI₂₁ for lipid A binding, whereas P4160_(L-B) and rLBP show significantly lower activity (p < 0.05 for rBPI₂₁ versus P4160_(L-B) and p < 0.05 for rBPI₂₁ versus rLBP). Since this is an experimental system utilizing immobilized substrate, it is possible that smaller molecules such as rBPI₂₁ have an increased ability to compete for radiolabeled rBPI₂₁ binding, due to less steric hindrance, than a larger molecule such as rBPI. However, rBPI, rLBP, P4160_(L-B), and P4161_(B-L) are all of approximately equal mass, and so we can readily compare the relative apparent K_d values of these proteins. We have also shown previously that rLBP₂₅ and rLBP have nearly identical apparent K_d values for lipid A binding (9). Thus, proteins containing the BPI N-terminal domain (rBPI₂₁, rBPI, and P4161_(B-L)) are most active in competing with radiolabeled rBPI₂₁ for lipid A binding, whereas those with the LBP N-terminal domain (rLBP and P4160_(L-B)) show significantly lower activity (p < 0.05 for rBPI₂₁ versus rLBP or P4160_(L-B)). When the various proteins were examined for their ability to compete with radiolabeled rLBP for binding to lipid A, similar trends were found. Unlabeled rLBP, rLBP₂₅, and P4160_(L-B) all had equivalent activities in inhibiting the binding of radiolabeled rLBP to lipid A, whereas rBPI₂₁, rBPI, and P4161_(B-L) were more active (data not shown).

We next examined the ability of the recombinant proteins to induce LPS-dependent TNF expression in THP-1 cells. In the presence of LPS, both rLBP and P4161_(B-L) were able to induce TNF production (Fig. 3). The C-terminal domain of rLBP is necessary for interaction with CD14 (39), and the data in Fig. 3 demonstrate that it is sufficient to confer this property on the fusion protein P4161_(B-L) possessing the LPS binding domain of rBPI.

rBPI₂₃ is known to block LPS-induced gene expression in isolated adherent human monocytes (40). Thus, we characterized the ability of the recombinant proteins to block LPS-induced TNF expression in THP-1 cells. rBPI and rBPI₂₁ showed the greatest ability to inhibit TNF production (Fig. 4), indicating neutralization of LPS inflammatory activity. The fusion protein P4160_(L-B) also inhibited TNF induction but was approximately 10-fold less active then rBPI (p < 0.05 for difference of rBPI or rBPI₂₁ versus P4160_(L-B)). The C-terminal portion of BPI has been previously reported to have LPS neutralization activity (16), and in our studies, rBPI consistently showed significantly greater TNF inhibition (p < 0.05 for rBPI versus rBPI₂₁). These data confirm that the C-terminal domain of BPI contains some LPS neutralization activity, a conclusion also supported by comparison of the data from P4160_(L-B) and rLBP₂₅. Addition of the BPI C-terminal domain to rLBP₂₅ (creating the P4160_(L-B) fusion protein) results in a protein with a significantly enhanced ability to inhibit TNF production (p < 0.05 for P4160_(L-B) versus rLBP₂₅). Although the P4161_(B-L) fusion protein potentiates LPS-induced TNF expression at low concentrations (<10 nM, Figs. 3 and 4), at higher concentrations (>10 nM, Fig. 4) inhibition of TNF expression is observed. Additionally, rLBP showed LPS neutralization activity at high concentrations (>50 nM) in this and other assay systems.²

Because heparin binding also localizes to the N-terminal LPS binding domain in rBPI₂₃ (30), we examined the heparin binding characteristics of the recombinant proteins (Fig. 5). Analysis of the data revealed that although the apparent K_d values measured for LPS are quite different for N-terminal LBP versus N-terminal BPI (60-80 and 2-5 nM, respectively), the apparent heparin affinity constants for the parent, N-terminal, and fusion proteins were all very similar (between 100 to 200 nM, Table I).

Table I. Heparin binding characteristics of test proteins Protein Kda Capacityb nM ng rBPI21 129.2 51.3 rBPI 255.5 40.2 P4161(B-L) 201.1 26.3 P4160(L-B) 99.9 5.5 rLBP 121.1 5.0 rLBP25 119.5 4.6 a  Affinity constants are expressed as apparent affinity constants. b  Capacity is expressed as ng of heparin bound per µg of test protein.

The N-terminal domains of LBP and BPI did, however, show consistent differences in binding capacities for heparin (determined by analysis of the plateau in Fig. 5). The differences in measured capacities indicated the following trend: those proteins with a LBP N terminus, rLBP₂₅, rLBP, and P4160_(L-B) had very low heparin capacity, whereas rBPI₂₁ and rBPI had nearly 10-fold higher calculated binding capacities. The P4161_(B-L) protein, with the BPI N-terminal and LBP C-terminal domains, showed a 5-fold higher heparin binding capacity than rLBP (see Table I).

We next examined the pharmacokinetic characteristics of the recombinant proteins. Plasma levels of the various proteins were followed by ELISA over time after administration of 1 mg/kg in rats. rBPI₂₁ and rBPI were cleared quickly (Fig. 6 and Table II), whereas the rLBP₂₅ and P4160_(L-B) proteins had a slower rate of clearance. The rapidly cleared proteins (rBPI₂₁ and rBPI) have high heparin binding capacities, whereas the proteins with lower capacities (rLBP₂₅, rLBP, and P4160_(L-B)) were cleared more slowly. The protein P4161_(B-L), whose heparin binding capacity was found to be intermediary (Table I), had a pharmacokinetic clearance profile that was also intermediary between the BPI proteins and the LBP proteins. Although the heparin binding capacity of rLBP was similar to rLBP₂₅ and P4160_(L-B), the pharmacokinetic behavior of rLBP showed an increased mean residence time.

Table II. Pharmacokinetic parameter values (mean ± S.E., n = 3) Protein Vc Vss CL pMRT bMRT ml/kg ml/kg ml/kg/min min min rLBP 55.7  ± 9.0 208  ± 77 0.45  ± 0.11 134.3  ± 29.1 437.6  ± 54.8 P4160(L-B) 94.2  ± 9.6 176  ± 19 3.39  ± 0.26 28.0  ± 3.0 52.0  ± 4.8 rLBP25 46.8  ± 0.6 269  ± 60 5.14  ± 0.45 9.2  ± 0.8 51.8  ± 8.9 P4161(B-L) 50.6  ± 9.5 94  ± 21 9.81  ± 1.70 5.1  ± 0.1 9.4  ± 0.7 rBPI21 44.8  ± 8.5 122  ± 35 30.59  ± 4.56 1.45  ± 0.1 3.8  ± 0.5 rBPI 66.6  ± 12.5 84  ± 18 31.03  ± 4.15 2.1  ± 0.1 2.7  ± 0.3

DISCUSSION

BPI and LBP are members of a lipid transfer/lipopolysaccharide binding protein family that includes CETP and PLTP (1, 41, 42). These proteins share a great deal of sequence similarity, and all bind amphipathic lipid ligands: LBP and BPI both bind LPS, and CETP and PLTP bind phospholipids. CETP binds neutral cholesteryl esters and triglycerides, and LBP, PLTP, and CETP carry their lipid ligands as monomeric units that are transferred to lipoprotein complexes in free solution (1, 5). Recent evidence suggests that PLTP, in addition to transferring lipids between lipoprotein particles, may also be capable of neutralizing LPS (44). Mutational analysis of CETP function has demonstrated that neutral lipid binding and transfer activities are properties of the C-terminal region and are distinct from phospholipid binding and transfer (43). This family of proteins may all share a characteristic of strong and independent domain-specific functionality, which seems to be confirmed in our analysis of fusion proteins of the N-terminal and C-terminal domains of LBP and BPI. Our studies demonstrate that although the peptide sequences of these two members of this family are highly similar, the unique and domain-specific functions of BPI and LBP are upheld, both in vivo and in vitro, regardless of protein context.

The N-terminal domains of both LBP and BPI contain major LPS-binding sites. Previously we have demonstrated that rBPI₂₃ and rBPI have nearly identical apparent K_d values for lipid A binding (4.3 and 4.1 nM, respectively), whereas LBP and its N-terminal fragment bind LPS much less avidly (9, 27). Our results presented in Fig. 2 confirm both the difference in LPS binding affinity between LBP and BPI and demonstrate that the C-terminal regions have little effect on the initial binding of the N-terminal domains to LPS. This conclusion is supported by the observation that the rBPI and P4161_(B-L) proteins have very similar abilities to compete with rBPI₂₁ binding to immobilized lipid A. Likewise, rLBP and P4160_(L-B) are very similar, but lower, in their competitive abilities (Fig. 2). Thus, the N-terminal domains of BPI and LBP are able to bind LPS in a manner that is independent of the C-terminal domain, and the BPI N-terminal domain retains its higher affinity for LPS regardless of whether any C-terminal domain is present and regardless of its identity.

The higher affinity of the BPI N-terminal domain for LPS binding compared with LBP is also borne out in its ability to neutralize LPS in cell-based biological assays. When we examined the ability of the proteins in this study to inhibit LPS-induced TNF production in a monocytic cell line (Fig. 4), we found that those proteins that contain a BPI N-terminal domain had the greatest activity in LPS neutralization. In fact, although both BPI and LBP are capable of binding LPS, rBPI and rBPI₂₁ have 10-50-fold more neutralization activity than the fusion protein P4160_(L-B) which contains the LBP N-terminal domain. The truncated LBP recombinant protein, rLBP₂₅, was not able to inhibit TNF production under these experimental conditions.

Our data demonstrate that the opposing functions of these two proteins are due to the properties of their C-terminal domains. It has been shown previously that the C-terminal domain of LBP is required for the transfer of LPS to CD14 (27). As shown in Fig. 3, the LBP C-terminal domain allows interaction with the CD14 protein and transfer of the LPS molecule when it is associated with either a BPI or LBP N-terminal LPS-binding domain.

The C-terminal domain of BPI has previously been reported to have LPS neutralization activity (16), and experiments presented here are consistent with those findings. In the experiment presented in Fig. 4, rBPI had a greater ability to neutralize LPS than rBPI₂₁, and P4160_(L-B) had a greater neutralization activity than rLBP₂₅. Although no significant difference was observed between rBPI₂₃ and rBPI binding to immobilized lipid A (9), a solution phase experimental system (such as that used in Fig. 4) might allow for more flexibility in interactions and so allow rBPI to bind and neutralize more LPS than rBPI₂₁. rLBP₂₅ shows no LPS neutralization activity under these assay conditions; thus, activity demonstrated by the P4160_(L-B) protein is consistent with previous observations of some amount of LPS neutralization activity in the BPI C-terminal domain. Experiments designed to examine the LPS neutralization activity of the BPI C-terminal domain alone are currently in progress and will more fully address this issue. The findings discussed above are consistent with analysis done on CETP which indicated that CETP may have more than one site for lipid interaction (43). The majority of the LPS neutralization capacity is measured in the N-terminal domain of BPI, however, since rBPI₂₁ inhibits TNF production in the THP-1 cells at approximately 10-20-fold lower protein concentration than P4160_(L-B) (Fig. 4).

The N-terminal domain of BPI has high heparin binding activity (30), and we were interested to determine if this heparin binding activity was affected within the context of a fusion protein with the LBP C-terminal domain. Also, since heparin binding co-localizes with the LPS binding domain of rBPI₂₃, and LBP also specifically interacts with LPS, we wanted to test if LBP is capable of binding to heparin. Although the measured relative binding affinities for the various proteins were all similar, heparin binding capacities for rBPI₂₁ and rBPI were found to be approximately 9-10 times those of rLBP₂₅ and rLBP (51 and 40 ng of heparin bound per µg of protein versus 5 and 4.6, Table I). Furthermore, heparin capacity appears to relate to the different serum half-lives of the various proteins measured in vivo (Table II). The rapid clearance of molecules with high heparin binding capacities such as rBPI and rBPI₂₁ suggests that these proteins might be cleared through interaction with heparin/heparan molecules in vivo. This hypothesis is further supported by the heparin binding capacity of the P4161_(B-L) protein, calculated to be intermediate between the BPI and LBP proteins (Table I), and which correspondingly had a serum half-life intermediate between BPI and LBP (Table II).

The pharmacokinetics of rLBP suggest that its C-terminal domain affects clearance rates. Although the heparin binding capacity of rLBP was found to be the same as rLBP₂₅, the plasma mean residence time of rLBP is much greater than all the other proteins in this study. Since rLBP is close in size to p4160_(L-B), and both proteins share the same LBP N-terminal domain, the decreased clearance rate of rLBP is not simply a matter of size but rather must be a property of the LBP C-terminal domain. It appears that the C-terminal domain of the rLBP serves as an interaction site with other serum proteins in vivo. Recent evidence suggesting that LBP can be associated with lipoprotein complexes in serum (5) is consistent with these pharmacokinetic observations and may indicate that clearance of LBP from the plasma occurs through a combination of interactions with other serum proteins and endothelial cells. Interestingly, the clearance rate of the P4161_(B-L) fusion protein is faster than rLBP₂₅ and may indicate that the pharmacokinetic properties of the LBP C-terminal domain, which apparently slow the clearance rate of rLBP, are dominated by the BPI N-terminal domain in this fusion protein.

The data concerning BPI, LBP, and derivative molecules presented in this study, together with the data generated in earlier studies of BPI (16) and from the mutational analysis of CETP (43), suggest that unique and independent domain-specific functionality serves as a hallmark of this lipid transfer/lipopolysaccharide-binding protein family. Thus, definition of domain structure/function relationships and the nature of interactions with lipid substrates for BPI serves as a model for similar functional characteristics of other family members. Yet this type of comparative analysis also points to the individual functions that family members can have. For example, LBP, PLTP, and CETP are all serum proteins and pass their lipophilic ligands to other acceptors, whereas BPI is normally sequestered within neutrophil granules, and the fate of LPS, once bound to BPI, is not known. In conclusion, our results clarify specific domain functions for BPI and LBP and suggest common domain structure/function relationships in the lipid transfer/lipid-binding protein family.