Recognition of LPS by TLR4 initiates inflammatory responses inducing potent antimicrobial immunity. However, uncontrolled inflammatory responses can be detrimental. To prevent the development of septic shock during an infection with Gram-negative bacteria, the immune system has developed mechanisms to neutralize LPS by specialized proteins. In this study, we report the recombinant expression and functional characterization of the mouse homolog of human bactericidal/permeability-increasing protein (BPI). Purified recombinant mouse BPI was able to neutralize LPS-mediated activation of macrophages and to block LPS-dependent maturation of dendritic cells. Recombinant mouse BPI neutralized the capacity of Gram-negative bacteria to activate immune cells, but did not influence the stimulatory properties of Gram-positive bacteria. Unlike human BPI, mouse BPI failed to kill or inhibit the growth of Pseudomonas aeruginosa. Together, these data demonstrate that murine BPI is a potent LPS-neutralizing protein that may limit innate immune responses during Gram-negative infections.
During an infection, the innate immune system must rapidly and specifically detect the presence of foreign pathogens to trigger appropriate protective responses. Nonetheless, the duration and intensity of the host inflammatory reaction must be tightly regulated to avoid the fatal consequences of septic shock. Host recognition of bacterial pathogens is mediated mainly by the TLRs that are activated by unique and invariant microbial structures (1). Microbial recognition by the innate arm of the immune system rapidly induces a series of proinflammatory cascades and stimulates a definitive adaptive immune response (2). Numerous studies have demonstrated that the activation of TLR4 by LPS, a complex and abundant glycolipid of the outer membrane of Gram-negative bacteria, is essential to robust host defense (3).
LPS is one of the most potent stimulators of the innate immune system (4). Mammals possess an array of LPS-binding proteins that are referred to as the bactericidal/permeability-increasing (BPI)3 protein family and are characterized by the expression of a homologous protein domain that is shared by the family members (BPI protein domain). In addition to BPI itself, members of this protein family include the LPS-binding protein (LBP), phospholipid transfer protein, cholesterol ester transfer protein, and the palate lung and nasal epithelium clone (PLUNC) proteins (5, 6). These proteins have been shown to initiate LPS recognition by immune cells or to sequester and neutralize LPS. Specifically, LBP is an acute phase protein produced mainly in the liver and present in the serum that transfers LPS monomers from micelles to the primary LPS-binding receptor CD14 at basal expression levels. At high concentrations, LBP is able to inhibit LPS by delivering LPS to plasma lipoproteins (7). LPS-CD14 interactions lead to the recruitment of transmembrane TLR4 in conjunction with an extracellular soluble protein termed MD-2, thereby leading to a conformational change in MD-2 which is sensed by TLR4 (8). This sequence of the binding modus of MD-2 after interaction with its ligand to TLR4 could be recently confirmed by a structure model of TLR4 and MD-2 along with the LPS antagonist Eritoran (9). Minute amounts of LPS trigger the formation of this protein complex and activate a powerful cellular response in the host (10). In contrast, the high-affinity binding property of human BPI to the lipid A moiety of LPS at the nanomolar range prevents the transfer of LPS to cellular receptors and facilitates its uptake by macrophages through a phagocytic pathway. This mechanism leads to the elimination of LPS and to the down-regulation of inflammatory cell activation (11). BPI is highly conserved in evolution with orthologs of this protein being expressed in teleost fish (12), rabbit (13), rat (14), mice (15, 16), humans (17), and invertebrates (18). The initial incorporation of the PLUNC proteins into the BPI family was based solely on structural similarities, although an interaction of PLUNC proteins with LPS has been demonstrated (19). The exact function of the PLUNC proteins is difficult to predict because the BPI family consists of proteins that bind LPS yet mediate opposing biological effects, such as LBP and BPI (20). Short PLUNC1 protein has been shown to limit the growth of Mycoplasma hominis inside lung epithelial cells but the precise mechanism for this action remains unsolved (21).
Recently, a mouse homolog of human BPI was identified and the expression and regulation of the expression of this protein was analyzed (15, 16). In this report, we have analyzed the potential of purified recombinant mouse (rm) BPI protein to interfere with cell activation in response to LPS as well as intact bacteria. We demonstrate that rmBPI specifically neutralizes LPS-mediated activation of APCs.
Materials and Methods
Mice and bacterial strains
C57BL/6 mice were purchased from Charles River.
For in vitro stimulation, Salmonella enterica serovar Typhimurium (S. typhimurium) NCTC 12023, Escherichia coli ATCC 25922, Pseudomonas aeruginosa ATCC 27853, Neisseria meningitidis (clinical isolate), Staphylococcus aureus ATCC25923, group B Streptococcus (GBS, clinical isolate) as well as Listeria monocytogenes (clinical isolate) were used.
293T HEK cells, RAW274.6 macrophages (American Type Culture Collection), and RAW macrophages stably transfected with a NK-κB luciferase reporter construct (RAW-κB) (22) were cultured with RPMI 1640 supplemented with 5% FCS, 10 mM HEPES, 10 mM sodium pyruvate, 50 μM 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin (complete medium).
Cloning of murine BPI into an expression plasmid, transient and stable transfection, and immunoblotting
The mouse BPI coding sequence was cloned into the pCMV-FLAG 1 (Sigma-Aldrich) expression vector. A full-length PCR product was obtained using primer pairs to amplify the open reading frame of mouse BPI without the putative leader sequence. A preparative PCR was conducted with the Triple Master Mix (Eppendorf) and separated on an agarose gel. The 1.4-kb DNA fragment of the PCR was purified with a Gel Extraction Kit (Promega) and directly ligated into the plasmid using compatible restriction sites. The cloned cDNAs were subjected to automated sequencing using the Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems). To identify recombinant protein expression, the plasmid was transiently transfected into 293T HEK cells using polyethyleneimine (Sigma-Aldrich) (23) and the cell lysates as well as supernatants were collected 48 h later. In case of stable expression, the BPI plasmid construct was transfected along with pcDNA3.1 hygromycin (Invitrogen) to provide a selection marker. After selection with hygromycin B, single clones were isolated and the supernatant and cell lysates were collected. Recombinant mouse BPI-FLAG protein was purified from the supernatant by M2 Agarose beads (anti-FLAG M2 Affinity Gel Freezer-safe; Sigma-Aldrich) and dissolved in SDS-PAGE sample buffer. Supernatant as well as cell lysates were subjected to SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Thereafter, the blots were incubated with an anti-FLAG M2 Ab (mouse anti-FLAG mAb; Sigma-Aldrich) and detected with a donkey anti-mouse IgG-HRP (Dianova) using the ECL Plus reagent (GE).
Generation of an anti-mouse BPI antiserum
An antiserum was generated from two peptides derived from the coding sequence of mouse BPI (peptide 1, CRGHRDPCPIHPPVMRFVPN (257–276 aa) and peptide 2, CQLLSNNGRFQLNTDFLR (309–326 aa)). The synthesis of the peptides, coupling to keyhole limpet hemocyanin, immunization of the rabbit, and collection of the serum was performed by SeqLab. The antiserum was purified by affinity chromatography with the peptides used for immunization to obtain a highly purified anti-mouse-specific polyclonal antiserum.
Purification of rmBPI
Supernatants were collected from three independently generated stable clones expressing rmBPI and purified by affinity chromatography using a fast protein liquid chromatographer (ÁKTA; GE). An Anti-FLAG M2 Affinity Gel was prepacked into an empty chromatography column and the supernatant was applied to the column. The bound protein was eluted with 0.1 M glycine (pH 3.5) and the eluted fractions were immediately neutralized with 0.5 M Tris (pH 7.4). Fractions that contained the recombinant protein were subjected to dialysis in PBS (pH 7.4) and transferred to a siliconized tube to avoid the loss of the protein during short-term storage. The protein content were determined by a protein assay based on bicinchoninic acid (Uptima). The purity of the protein fraction was analyzed by SDS-PAGE followed by Coomassie blue staining and by immunoblot using an Ab that detects either the FLAG epitope tag or the murine BPI protein, respectively.
Generation of dendritic cells (DCs)
DCs were derived from the bone marrow of isolated mouse femur and tibia. Marrow cells were flushed from the bones, depleted of erythrocytes, and cultured in complete medium with GM-CSF. The culture medium was replaced every third day. On day 7, suspension cells were harvested as immature DCs and plated into 12-well plates.
Stimulation of cells
RAW-κB cells as well as DCs were stimulated either in the presence or absence of different amounts of LPS. To determine whether rmBPI was able to neutralize LPS, LPS (E. coli 0111:B4; Sigma-Aldrich) was preincubated with 10, 1, 0.1, 0.01, and 0.001 μg/ml rmBPI, respectively, for 30 min before adding it to the cells. In comparison, human purified BPI (Athens Research) was used in the same concentration as rmBPI for the inhibition of LPS-mediated activation of RWA-κB cells. Sixteen hours after stimulation, the supernatants were collected, RAW-κB cell lysates were prepared for determination of luciferase activity, and stimulated DCs were processed for flow cytometry analysis. RAW-κB macrophages were also stimulated in the presence of different amounts of UV-treated bacteria. The efficacy of UV treatment to block bacterial replication was verified in each case by plating the treated and untreated samples on agar plates. The optimal bacterial number to elicit a robust activation of the NF-κB luciferase reporter construct was individually determined for each isolate. Interference of rmBPI with cellular responses was measured by bacterial stimulation in the presence of rmBPI (10, 1, and 0.1 μg/ml, respectively). Finally, RAW macrophages were stimulated with an optimal multiplicity of infection (MOI) of live E. coli, as determined by the measurement of TNF secretion into the supernatant. Live E. coli stimulation of macrophages was performed in the presence or absence of rmBPI (10 μg/ml) and supernatants were collected for analysis of its inhibitory capacity.
Mouse TNF, IL-12, and human IL-8 were analyzed from the supernatant of stimulated cells by a standard sandwich ELISA technique using matched Ab pairs (BD Biosciences) according to the manufacturer’s recommendations.
Measurement of nitrite
NO2− in supernatants was determined by the Griess assay (24).
Immature bone marrow-derived DCs (BMDCs) were cultured in the presence of 100 ng/ml LPS or 2 μM CpG, respectively. To analyze the inhibitory capacity of rmBPI, the cells were cocultured with different amounts of the recombinant protein. Twenty-four hours after stimulation, the cells were harvested and stained for the surface expression of CD11c (anti-CD11c-FITC), MHC class II (anti-H-2b-PE), and CD86 (anti-CD86-PE) (all Abs were from BD Biosciences) and analyzed by FACS.
NF-κB reporter activity
Sixteen hours after stimulation, cell lysates were processed with luciferase substrate (Promega) to measure luciferase activity.
Immunoblot of the signaling molecules
To determine the influence of rmBPI on the E. coli-induced activation of the MAPK, RAW macrophages were stimulated with 0.1 MOI of E. coli in the presence or absence of 10 μg/ml rmBPI for various periods of time. Thereafter, cell lysates were prepared with modified radioimmunoprecipitated assay buffer containing 1 mM NaVO3 and 1 μg/ml each of leupeptin, pepstatin, and aprotinin (all from Sigma-Aldrich). To detect the activation of MAPK, 100 μg of total protein was separated on a SDS-PAGE. The phosphorylation of p38 and ERK1/2 was detected by a specific phospho-p38 (New England Biolabs) and specific phospho-ERK1/2 (Promega) Abs, respectively, and the ECL Plus Western Blotting reagent (GE). Equal loading of the gel was documented with Abs specific for total p38 (New England Biolabs) and ERK1/2 (Promega) Abs, respectively.
Antibacterial activity of BPI
Purified human BPI protein as well as rmBPI was analyzed in parallel for their antibacterial activity. These assays were performed in 96-well plates in a volume of 50 μl of PBS (pH 7.4). In brief, 104 P. aeruginosa/ml were incubated with different concentrations of either purified human BPI or rmBPI for 1 h at 37°C and subsequently plated on blood agar. After incubation at 37°C overnight, bacterial colonies were quantified.
Statistical significance was calculated with the Prism 3.0 GraphPad software using the Student t test. Values of p < 0.05 were considered as significant. For the calculation of the IC50 values of BPI, the Prism 3.0 GraphPad software was used.
Expression and purification of rmBPI
For the expression of rmBPI protein, the coding sequence of mouse BPI was cloned into the expression plasmid pCMV1-FLAG. Recombinant BPI-FLAG expression by transiently transfected 293T cells was confirmed by immunoblotting without and with prior immunoprecipitation using a FLAG-specific Ab (data not shown). Having established the expression of rmBPI in this system, we then generated three independent 293T cell clones stably expressing rmBPI. Recombinant BPI was purified by fast protein liquid chromatography affinity using Abs specific for the FLAG epitope tag and consecutive dialysis of the purified protein against PBS (pH 7.4). The protein content and the purity of rmBPI were evaluated by SDS-PAGE and Coomassie blue staining (Fig. 1,A), revealing a single band of ∼60 kDa. The specificity of the 60-kDa band was further corroborated by immunoblot using FLAG Abs (Fig. 1, B and C). Thus, we were able to obtain highly purified rmBPI protein.
rmBPI inhibits the LPS-induced activation of macrophages
To determine the biological function of rmBPI, we first analyzed its effect on the activation of RAW macrophages by LPS. As expected, LPS was able to potently activate RAW-κB cells as determined by the activation of the NF-κB luciferase reporter as well as by the production and secretion of TNF (Fig. 2). In contrast, stimulation of the RAW-κB cells in the presence of rmBPI led to a dose-dependent inhibition of macrophage activation (Fig. 2). At the highest doses, rmBPI entirely antagonized the LPS-mediated activation of macrophages. This rmBPI-mediated inhibition of macrophage function was specific for LPS, because rmBPI did not inhibit the activation of RAW-κB cells by CpG DNA motifs or poly(I:C) (Fig. 2 A). Essentially identical results were obtained with rmBPI purified from three individual cell clones (data not shown).
To compare the LPS-neutralizing activity of mouse and human BPI, we performed the experiments described above for rmBPI with purified human BPI applying the same assay conditions. This approach revealed that the LPS-neutralizing activity of rmBPI is comparable to that of human BPI as determined by two different readout systems (NF-κB reporter assay: IC50 (rmBPI) = 4.9 ± 3.36 μg/ml vs IC50 (human BPI) = 1.5 ± 0.8 μg/ml; TNF suppression assay: IC50 (rmBPI) = 2.6 ± 1.6 μg/ml vs IC50 (human BPI) = 2.75 ± 2.01 μg/ml). From these experiments, we conclude that rmBPI is able to effectively neutralize the LPS-mediated activation of macrophages.
The LPS-mediated maturation of DCs is influenced by rmBPI
In an attempt to analyze whether rmBPI is able to block the LPS-induced maturation of mouse DCs, we generated BMDCs and matured them in the presence of LPS or CpG as determined by the production of TNF (Fig. 3,A), IL-12 (Fig. 3, E and F), and NO (Fig. 3, C and D) as well as the surface expression of CD86 (Fig. 3,B). The addition of rmBPI to the LPS-stimulated DCs led to a block in the maturation of DCs which was dependent on the amount of BPI used (Fig. 3, A–C and E), whereas the addition of rmBPI did not alter the activation of DCs mediated by CpG (Fig. 3, A, D, and F). These results along with the results obtained with macrophages demonstrate that rmBPI is able to robustly neutralize endotoxin and thereby prevent the activation of professional APCs (Fig. 3).
rmBPI inhibits the immunostimulatory capacity of Gram-negative bacteria
Having established the neutralizing activity of rmBPI on purified LPS derived from E. coli, we investigated whether rmBPI was also able to interfere with the immunostimulatory capacity of complex Gram-negative bacteria simultaneously engaging multiple TLRs. To this end, we analyzed macrophage function after stimulation with inactivated UV-treated Gram-negative bacteria (S. typhimurium, E. coli, P. aeruginosa, and N. meningitidis). As expected, all Gram-negative bacteria tested stimulated RAW-κB cells at low MOIs for luciferase reporter expression and for TNF production (Fig. 4). UV-inactivated Gram-positive bacteria (S. aureus, group B streptococci, and L. monocytogenes) potently stimulated RAW-κB cells, although higher MOIs were required in comparison to Gram-negative bacteria (Fig. 5). Our previous data showed that rmBPI could strongly antagonize LPS-mediated activation and maturation of macrophages and DCs. To determine the effect of rmBPI on the stimulation of macrophages with complex Gram-negative bacteria, we cocultured Gram-negative and Gram-positive bacteria with macrophages in the presence of graded amounts of rmBPI. Whereas exogenous rmBPI potently inhibited stimulatory activity of Gram-negative bacteria on RAW-κB macrophages, rmBPI did not antagonize the activity of Gram-positive bacteria to stimulate RAW-κB macrophages (Figs. 4 and 5). Notably, rmBPI potently neutralized the immunostimulation by P. aeruginosa, demonstrating that rmBPI-mediated inhibition of macrophage function extends even to structurally diverse LPS isoforms as Pseudomonas LPS. Collectively, these data demonstrate that rmBPI inhibits endotoxic activity of all Gram-negative bacteria tested, likely reflecting neutralization of LPS.
Exogenous rmBPI inhibits endotoxic activity, but not growth of E. coli
Based on our data obtained with UV-treated Gram-negative bacteria, we sought to determine whether rmBPI has also the capability to inhibit the activation of cells by proliferating bacteria. Because E. coli is a strong activator of RAW macrophages, we chose these bacteria as our model system. A 3- to 3.5-h infection with live E. coli at a MOI of 0.1 was sufficient to induce significant TNF release from RAW macrophages and led to the activation of the MAPK p38 as well as ERK1/2 (Fig. 6, B and C). Unlike human BPI (25), exogenous rmBPI had no significant effect on the proliferation of the bacteria (Fig. 6,A), but potently inhibited TNF production by RAW macrophages (Fig. 6,B). Furthermore, TLR4-mediated signal transduction after LPS stimulation was also inhibited in the presence of rmBPI as shown by reduced phosphorylation of p38 and ERK1/2 (Fig. 6 C). These data demonstrate that rmBPI is able to neutralize the immunostimulatory capacity of proliferating E. coli yet it does not prevent bacterial growth.
Human, but not mouse, BPI, inhibits the growth of P. aeruginosa
Since we observed no antibacterial activity of rmBPI against E. coli, we tested its activity against another Gram-negative bacterium, P. aeruginosa. For human BPI, we recently demonstrated a strong and uniform antibacterial activity against various strains of P. aeruginosa that differed in their susceptibility to antibiotics (26). As expected, at concentrations of 10 μg/ml human BPI, we observed a strong antibacterial activity. In contrast, rmBPI tested under the same experimental conditions did not interfere with the bacterial growth at concentrations as high as 100 μg/ml (Fig. 7). From these data, we conclude that rmBPI lacks antimicrobial activity against P. aeruginosa.
Using an epitope tag and affinity chromatography, we have generated highly purified rmBPI protein. Mouse BPI is biologically active and is able to neutralize LPS and prevent the activation of APCs. In addition, rmBPI inhibits the immunostimulatory effects of Gram-negative bacteria, likely reflecting neutralization of LPS of Gram-negative bacteria without interfering with their replication, but does not prevent the stimulation of immune cells by Gram-positive bacteria.
To prevent the overwhelming toxic effects of LPS, the body is equipped with a large family of LPS-neutralizing proteins. For example, cathelicidin, secretory leukocyte protease inhibitor, and acyloxyacyl hydrolase (AOAH) have all been shown to inhibit LPS and block the induction of septic shock (27, 28, 29). These proteins use distinct molecular mechanisms to exert their functions. For example, AOAH cleaves the acyl chain from the lipid A part of LPS, which in turn inactivates the biological activity of LPS (30). In this article, we report that mouse BPI is able to neutralize the LPS stimulatory activity. The precise mechanism for this action is presently unknown. However, from the crystal structure of human BPI, it was predicted that hydrophobic pockets might be able to accommodate LPS (31). Binding of LPS to these hydrophobic pockets leads to an enhanced elimination of LPS from the system presumably by a macrophage-dependent phagocytic process (11). On the molecular level, it was demonstrated that human BPI binds to the lipid A portion of LPS with nanomolar affinity and thereby outcompeting other LPS-interactive proteins (e.g., LBP) and preventing the engagement of MD-2/TLR4 (32).
We have recently shown that mouse immature BMDCs express mRNA for BPI at low levels, which is strongly induced upon stimulation with LPS (15). In this study, we demonstrate that the addition of exogenous recombinant BPI inhibits the induction of DC maturation driven by LPS. Following activation of the maturation process, transcriptional activation of BPI may, in part, explain the reduced responsiveness of mature DCs to LPS (33). The differential expression of BPI by immature and mature DCs would ensure that immature DCs expressing low-level BPI first become activated before they start to neutralize the activation stimulus (LPS). Another LPS-neutralizing protein, AOAH, was also shown to be present in DCs and its transcription is inducible by LPS stimulation (30). Together, mouse BPI and AOAH may also contribute to the phenomenon of LPS tolerance, both in DCs and possibly in vivo.
rmBPI is able to neutralize LPS from different Gram-negative species. The LPS structure of P. aeruginosa is highly acetylated and exhibits poor immunostimulatory activity through TLR4 in comparison to the classical LPS structure of E. coli, S. typhimurium, and Neisseria (34). Notably, BPI does not appear to discriminate among these structural variations in LPS, since neutralization was observed for P. aeruginosa as well as other Gram-negative organisms, although P. aeruginosa stimulation is not completely inhibited by rmBPI. The residual activation of the cells by P. aeruginosa might be due to the engagement of multiple TLR ligands by the bacteria. Therefore, the mechanism of LPS recognition by BPI appears to be distinct from the requirements for TLR4 activation by LPS subtypes. Most likely this is due to the interaction of BPI with the lipid A region which is common in the LPS of all Gram-negative bacteria as it was demonstrated for human BPI (32, 35). Classical LPS is known as the strongest TLR- stimulating component derived from Gram-negative bacteria, which explains the pronounced inhibition of cell activation by these microorganisms in the presence of rmBPI (36). The residual NF-κB activity observed after the stimulation with Gram-negative bacteria in the presence of rmBPI (Fig. 4) in our study is most likely due to other TLR ligands such as flagellin and bacterial DNA, which cannot be neutralized by BPI.
The recombinant N-terminal fragment of human BPI was shown to neutralize LPS and LPS-mediated shock in animal models and exhibit bactericidal activity in infection models with Gram-negative bacteria (37, 38). Full-length mouse recombinant BPI neutralizes LPS both as an isolated stimulus as well as LPS on the surface of bacteria but does not interfere with the bacterial growth. Our results do not exclude the possibility that BPI has bactericidal properties; however, we were not able to demonstrate this in our experimental conditions. Because BPI is expressed and secreted in concert with other antimicrobial proteins, the combinatorial action of these molecules in vivo may lead to an overall enhancement of their microbicidal spectrum or potency (39). This was demonstrated for the enhanced antimicrobial activity of BPI with other defensins in vitro and the influence of other extracellular components (40).
Recently, it was shown that a human N-terminal fragment of BPI was able to bind to the Gram-positive bacterium Streptococcus pneumoniae as well as to pneumolysin, a pore-forming protein derived from these bacteria. This binding led to an enhanced immune response and thereby modestly enhanced the survival of mice lethally challenged with S. pneumoniae (41). In our study, we did not detect a major influence of recombinant full-length mouse BPI on group B Streptococci, a Gram-positive organism that does not secrete pneumolysin. We also observed that the immune stimulatory capacity of the Gram-positive organism analyzed in this study was generally very low.
In summary, we report for the first time recombinant expression and affinity purification of mouse BPI using an eukaryotic expression system. This protein exerts strong LPS- neutralizing activity both on the isolated LPS level as well as on Gram-negative bacteria. Further characterization of the activity and mechanism of action of BPI and other LPS-neutralizing proteins will advance our understanding and ability to control the powerful immunostimulatory activity of bacterial LPS.
We thank Christian Bogdan, Salman Qureshi, Andreas Diefenbach, and Martin Röllinghoff for critical reading of this manuscript.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported in parts by grants from the German Research Foundation (to M.S.) and the Interdisziplinäres Zentrum für Klinische Forschung Erlangen (to A.G.).
Abbreviations used in this paper: BPI, bactericidal/permeability-increasing protein; LBP, LPS-binding protein; rmBPI, recombinant mouse BPI; DC, dendritic cell; MOI, multiplicity of infection; BMDC, bone marrow-derived DC; AOAH, acyloxyacyl hydrolase.