A Saposin-like Domain Influences the Intracellular Localization, Stability, and Catalytic Activity of Human Acyloxyacyl Hydrolase*

Acyloxyacyl hydrolase, a leukocyte enzyme that acts on bacterial lipopolysaccharides (LPSs) and many glyc- erolipids, is synthesized as a precursor polypeptide that undergoes internal disulfide linkage before being pro- teolytically processed into two subunits. The larger subunit contains an amino acid sequence (Gly-X-Ser-X-Gly) that is found at the active sites of many lipases, while the smaller subunit has amino acid sequence similarity to saposins (sphingolipid activator proteins), cofactors for sphingolipid glycohydrolases. We show here that both acyloxyacyl hydrolase subunits are required for catalytic activity toward LPS and glycerophosphatidyl- choline. In addition, mutations that truncate or delete the small subunit have profound effects on the intracel- lular localization, proteolytic processing, and stability of the enzyme in baby hamster kidney cells. Remarkably, proteolytic cleavage of the precursor protein increases the activity of the enzyme toward LPS by 10-20-fold without altering its activity toward glycerophosphati-dylcholine. Proper orientation of the two subunits thus seems very important for the substrate specificity of this unusual enzyme.

subunit could enable the enzyme to recognize LPS as a substrate, while the large subunit would play the major role in catalysis.
In the experiments described here, we studied the properties of wild-type and mutated AOAH using pulse-chase analysis of 35S-labeled AOAH to explore its biosynthesis, immunofluorescence to localize the enzyme intracellularly, and assays for enzymatic activity to measure its function. Because we were unable to establish stable rAOAH-expressing transfections in leukocyte cell lines, we used the BHK570 fibroblast line for these studies. The results indicate that while the large subunit, as expected, plays an important role in catalysis, the small subunit is essential for several properties of the enzyme, including its intracellular localization and its catalytic activity toward LPS and glycerophospholipid. AOAH cDNA-Human AOAH cDNA constructs in vectors pZEM229R (4) and pRS431 were provided by ZymoGenetics, Seattle, WA, as were BHK570 and BHK570 431Acells. Plasmid pZEM229R 4-33 contains the 2162-bp full-length AOAH cDNA cloned into the EcoRI site of pZEM229R (1). AOAH expressed from this plasmid has full enzymatic activity but differs from wild-type AOAH at 3 amino acids (due to 3-base transitions introduced during PCR). Plasmid pRS431 carries a 1744-bp AOAH cDNA fragment cloned into the EcoRI site of pZEM229R.Z This 1744-bp insert comprises the correct coding sequence ofAOAH with very little upstream cDNA sequence and no downstream sequence. BHK570 431Acells are stably transfected with pRS431. Plasmid clone 2.1 is a partial cDNA clone that harbors a 96-bp deletion (from bp 400 to 495 in the wild-type sequence ( Fig. 1) (I)).
AOAH cDNA insert of pZEM229R 4-33 were corrected with AOAH DNA Site-directed Mutagenesis-The 3-base transitions in the 2162-bp from pRS431. This wild-type 2162-bp AOAH cDNA, cloned in pSE-LECT-1 (Promega Biotech Inc.), was used as template for site-directed mutagenesis (Promega Biotech's Altered Sites in vitro Mutagenesis System). The introduced base changes were confirmed by DNAsequencing.
Plasmid Constructs-Plasmid clone 2.1 was digested with BcZI and BgZII, and the 430-bp fragment (which contains the 96-bp deletion) was substituted for the corresponding BcZI-BgZ I1 fragment (526 bp) in pZEM229R 4-33. The new plasmid, pZEM229R A4-33, harbors the 96base pair deletion in the region encoding the small subunit of AOAH.
The cDNA encoding the AOAH signal sequence and propeptide was cloned upstream of the large subunit as follows. A 116-bp region from R. Seale, ZymoGenetics, Seattle, WA, personal communication.  the 5' EcoRI site to the end of the propeptide coding sequence was amplified by PCR. The 5' primer (5' TGGGAATTCGTCGACCAC) included the EcoRI restriction site, and the 3' primer (5' ACITCTA-GAGAGGCTGGGCCTGGA) was constructed to introduce an XbaI site at the end of the amplified product. The 122-bp product was digested with XbaI and ligated to a 1272-bp XbaI-EcoRI fragment encoding the entire large subunit. The 1394-bp linear product was then digested with EcoRI and ligated to the EcoRI site of pBluescript I1 KS (Stratagene Cloning Systems, La Jolla, CA). The DNA sequence of the amplified 122-bp PCR product was found to be in frame with the beginning of the large subunit. The 1394-bp fragment was then cloned into plasmid pZEM229R (making pZEMl38) at the EcoRI sites and transfected into BHK570 cells.
Dansfections-Cells were transfected using calcium phosphate precipitation (6). Methotrexate-resistant cells were cultured in increasingly higher concentrations of methotrexate to amplify the transfected DNA (7). Transfected cells were cloned by limiting dilution to obtain subclones with high AOAH activity.
Antisera-Polyclonal antibodies to rAOAH or to synthetic peptides (synthesized by Lynn Deogny and Clive Slaughter, Howard Hughes Medical Institute, UT-Southwestern Medical Center) were raised in New Zealand White rabbits (8). IgG fractions were prepared using standard methods (8).
Immunofluorescence-The location of intracellular rAOAH in transfected cells was studied by indirect immunofluorescence (9). Incubations were done on ice and the block solution contained 1% cold-water fish gelatin (Sigma). F(ab),' goat anti-rabbit IgG conjugated to fluorescein isothiocyanate (TAG0 Immunologicals, Camarillo, CA) was the secondary antibody.
BHK570 cells producing rAOAH were also pulsed with T~-an~~S-label for 5 h followed by a chase of 1,3,5, and 7 h with complete medium. The media and cells were collected at each time point and processed as above. Some experiments included 10 m M NH,C1 in the labeling and chase media (11).
Purification of rAOAH-Stably transfected BHK570 cells were plated into 150 cm2 flasks and grown in complete growth medium for 3 days. The medium was removed, the cells were rinsed with phosphatebuffered saline, and the medium was replaced with serum-free medium (DMEM with 5 mgfliter insulin, 5 mgAiter transferrin, 5 pg/liter selenium, and 10 mg/liter fetuin). After another 4-5 days, the medium was passed over a 2.5 x 6-cm column of hydroxylapatite (Bio-Gel HTP; Bio-Rad) at 4 "C. Proteins were then eluted at room temperature with a linear sodium phosphate gradient (5-400 mM), and the AOAH in the eluted fractions was quantitated using enzymatic and immunoblot (8) assays. Fractions containing the highest amounts of enzymatic activity and/or rAOAH protein were pooled and concentrated (Centricon-30; Amicon, Beverly, MA). Partially purified AOAH was stored at -70 "C in 15% glycerol.
Quantitation ofrAOAH-Samples were subjected to SDS-PAGE and Western transfer onto Immobilon-P membrane (Millipore Corp., Bedford, MA). rAOAH was detected by using affinity-purified rabbit IgG raised to the COOH-terminal 19 amino acids of AOAH, or polyclonal rabbit IgG raised to intact rAOAH, followed by affinity-purified labeled goat anti-rabbit IgG (gift of Ellen Vitetta, this institution). Sections of the membrane containing 1261-immunolabeled rAOAH polypeptides were removed and their radioactivity was counted. The amount of mutant rAOAH present was determined relative to known amounts of pure wild-type rAOAH, processed in the same manner.
The 35S content of immunoprecipitated proteins from pulse-chase experiments was quantitated using a PhosphorImager SF (Molecular Dynamics, Inc., Sunnyvale, CA).

An AOAH Precursor Is Secreted by BHK570
Cells-Wild-type rAOAH biosynthesis in transfected BHK570 431A cells was studied using SDS-PAGE and autoradiography to characterize biosynthetically labeled 35S-labeled AOAH. As shown in Fig. 2, anti-AOAH antibodies immunoprecipitated two labeled proteins, of apparent M , = 70,000 and M , = 60,000, from cell lysates. In contrast, the cell media contained only one immunoprecipitable protein, ofM, = 70,000. When immune precipitates were treated with 2-ME prior to SDS-PAGE, a protein of M, = -14,000 was detected in the cell lysates, and a protein of M , = 52,000 was found instead of the 60,000-Da peptide seen in the unreduced lysates. When reduced, the 70,000-Da peptides in the cell lysates and media migrated as 65,000-Da proteins. Two-dimensional gel analysis (first dimension, unreduced; second dimension, reduced) of the cell lysate showed that the small polypeptide found following reduction derived from the M, = 60,000 peptide in the unreduced sample (data not shown): the increase in mobility of the 70,000-Da polypeptides in the presence of 2-ME was possibly due to the loss of intrachain disulfide bonds. These data suggest that the larger protein (70 kDa) is the precursor form of AOAH; as shown schematically in Fig. 1, the two subunits of the protein are not separable by disulfide bond reduction unless the precursor has been proteolytically cleaved between the subunits.
Some of the rAOAH produced in BHK570 cells is appropriately processed (mature rAOAH) and can be separated into its two subunits by disulfide bond reduction (Fig. 2, lane 3), and a large fraction of the unprocessed AOAH is secreted into the media.
Pulse-chase analysis (Fig. 3) confirmed that the precursor polypeptide is partly processed to mature intracellular AOAH and partly secreted into the media. Little mature AOAH was found in the media. When ammonium chloride (10 mM) was present during the pulse and chase, very little of the 60 kDa peptide was found (Fig. 41, in keeping with the conclusions that 1) this protein is derived from the 70-kDa protein and 2) proteolytic cleavage of the enzyme into its subunits may occur in an acidic intracellular compartment.
The fate of the 35S-labeled 70-kDa precursor was quantitated after a 5-h chase in nonradioactive medium (Table I). A large fraction (44.8%) of the precursor was secreted during the 5-h period, whereas approximately one-fourth was processed to the 60-kDa peptide. The total recovery of labeled rAOAH was lower following ammonium chloride treatment and much less of the precursor was found in the media at 5 h (Table I).
Proteolytic Processing in Vitro Using Tkypsin or Chymotrypsin-The rAOAH precursor was partially purified from serum-free medium. Chymotrypsin or trypsin could cleave it between the two subunits; when reduced with 2-ME prior to SDS-PAGE analysis, the trypsinor chymotrypsin-generated large subunit had the size of the native large subunit (Fig. 5A). The NH,-terminal amino acid sequence of the trypsin-generated large subunit was Ser-Gly-Ser-Asp-Ile. . . , indicating that trypsin treatment resulted in a large subunit that was two amino acids longer than the native large subunit (N-Ser-Asp-Ile. . .I. The AOAH propeptide sequence (Ser-Pro-Ala-Asn. . .) was found at the amino terminus of the small subunit after trypsin cleavage, indicating that the propeptide was retained in the secreted protein.
Proteolytic Processing of the Precursor Protein Increases Its Enzymatic Activity toward LPS without Altering Its Activity toward GPC"Trypsin or chymotrypsin treatment of the precursor greatly increased its ability to attack LPS (Fig. 6). Activation was associated with a large increase in V, , , with no change in K, (Fig. 6, inset). In contrast, the activity of the enzyme toward GPC was not increased (Fig. 6). The preference of the enzyme for removing palmitate, rather than oleate, from the sn-2 position of GPC (16) was also not changed by protease treatment (data not shown). Neutrophil-derived (native) AOAH also has greater in vitro activity toward LPS than GPC (16).
The fatty acids released from LPS after incubation with activated and non-activated rAOAH were examined by TLC. Before and after activation, the enzyme removed the two secondary fatty acids (myristate, laurate) at similar rates (data not shown); there was no preference for attacking a particular acyloxyacyl bond on the lipid A backbone.
Structure-Function Analysis: Enzymatic Activity and Intracellular Localization of Wild-type and Mutant Forms of rA0A.H-Intracellular AOAH was localized in BHK570 cells by indirect immunofluorescence. Whereas no background fluorescence was seen in cells transfected with vector DNA (pZEM229R), cells transfected with wild-type AOAH-expressing plasmids showed bright fluorescence in large vesicular structures (Fig. 7, Panel B ) . When added to the culture medium, fluorescein isothiocyanate-dextran accumulated in the same structures; anti-AOAH antibodies were found in large vacuoles when studied by immunogold electron microscopy (data not shown). We consider these structures lysosomes. Fluorescence was also noted in a more diffuse, punctate pattern that involved most of the cytoplasm. Ammonium chloride treatment of the cells did not change this distribution (Fig. 7,  Panel C).
Analysis of mutated rAOAH derivatives suggested the following conclusions: 1) N-linked glycosylation of the small subunit is not essential for enzymatic activity or intracellular localization. The small subunit contains a single site for N-linked glycosylation. This site is glycosylated in wild-type rAOAH, as was shown by a decrease in the apparent size of the small subunit following treatment with N-glycanase (Fig. 5, Panel B ) . (An essentially identical result was obtained previously using native neutrophil AOAH (1)). Elimination of this site by replacing ThrG1 with Ala did not alter the catalytic activity of the enzyme toward LPS (Table II), nor did it change the intracellular distribution of AOAH (Fig. 7, Panel D ) . The small subunit of the mutated enzyme had the same apparent size as N-glycanase-treated wild-type rAOAH small subunit, indicating that the mutagenesis had the desired effect (Fig. 7, Panel B). .~. . . Fig. 3. Molecular size markers are shown at the left. The processed AOAH seen (as a 60,000-Da band) in the reduced samples was less than 10% of the total immunoprecipitated AOAH, and this percentage did not change with increasing chase time (compare to untreated cells (Fig. 3)).

FIG. 4. Effect of ammonium chloride on the processing of wildtype AOAH in BHK570 cells. Wild-type AOAH in transfected BHK570 cells was radiolabeled with ["Slmethionine and [35Slcysteine in the presence of 10 mnl ammonium chloride. AOAH in cell lysates was immunoprecipitated at different chase times and analyzed by SDS-PAGE and autoradiography. The lane designations are as in
Glycosylation seems to account for much of the apparent size of the small subunit when it is analyzed using an alternate system of SDS-PAGE that resolves proteins in the 5-20 kDa range. The deglycosylated small subunit had an apparent M, of 8,100, compared with 14,000 for the glycosylated peptide. Since the predicted mass of the small subunit, deduced from the amino acid sequence, is 13,702 Da (l), this finding suggests that intracellular proteolytic processing may remove approximately 5,600 Da (48 amino acids) from the carboxyl terminus of the small subunit. When deglycosylated, the small subunit of human neutrophil AOAH had apparent sizes of 7,900 and 8,100 Da (data not shown), suggesting that in vivo processing in neutrophils may remove variable amounts of the region that links the small and large subunits in the precursor.
2) The complete small subunit is not required for enzymatic activity toward LPS, but may be necessary for proteolytic processing, secretion, and intracellular targeting. rAOAH that lacked an internal 33-amino acid region of the small subunit had approximately 42% of the wild-type enzymatic activity (Table 11); its activity toward LPS could be enhanced by treatment with chymotrypsin. This variant form of AOAH did not localize to the large vesicles (Fig. 7, Panel E ) , was poorly secreted, and was less stable than wild-type rAOAH (Table I). Immune precipitates of cell extracts and cell media contained one peptide of M , = 60,000 that could not be reduced into two subunits by 2-ME (data not shown). The absence of this 33 amino acid component of the small subunit apparently prevents normal intracellular targeting, processing, and secretion of rAOAH by BHK570 cells.
3) The large subunit contains a component of the catalytic site. AOAH has the sequence Gly-X-Ser-X-Gly; the serine is known to participate with His and Asp in a catalytic triad in many lipases (17). When AOAH Se?'j3 was replaced by Leu, the activity of the enzyme toward LPS decreased by 90-fold (Table  II), in keeping with a key role for this serine in the catalytic activity of AOAH.
4) The small subunit also contributes to catalysis. Expression of the large subunit by itself produced a polypeptide that had 51% of the wild-type activity toward LPS (Table 11) and GPC (data not shown), suggesting that the small subunit in some way contributes to the ability of the enzyme to attack both substrates. Immune precipitates of cell extracts of BHK570 cells harboring pZEM138 contained one polypeptide of 57,000 Da that was poorly secreted and could not be reduced into smaller subunits (data not shown). Although this construct included the AOAH signal and propeptide sequences, the large subunit had the same localization pattern as the rAOAH that lacked 33 amino acids of the small subunit (Fig. 7, Panel F), again consistent with a role for the small subunit in the intracellular targeting of AOAH by these cells.

DISCUSSION
A lipase with a preference for removing saturated (or short) acyl chains from glycerol-based lipids (16) and LPS (15), AOAH Partially purified secreted AOAH was incubated with chymotrypsin for 0, 1, and 10 min a t 37 "C. The reaction was stopped by adding an equal volume of 2 x sample buffer containing 5% 2-ME. After heating a t 100 "C for 5 min, the samples were subjected to SDS-PAGE followed by Western transfer onto Immobilon P membrane. The AOAH protein bands were detected by incubating the membrane with rabbit anti-rAOAH IgG followed by goat anti-rabbit horseradish peroxidase-conjugated IgG (TAG0 Immunologicals). The membrane was subsequently developed with 4-chloro-1-naphthol. The arrows point to the large and small subunits. B, N-glycanase treatment. Radiolabeled wild-type and glycosylation mutant rAOAHs were immunoprecipitated from cell lysates as described under "Experimental Procedures." The washed pellets were suspended in denaturing buffer (50 mM Tris-C1, pH 7.6,0.5% SDS, 0.35% %ME), boiled, and centrifuged. Half of each supernatant was treated with 0.01 unit ofN-glycanase (Genzyme Corporation, Cambridge, M A ) in the presence of 0.75% Nonidet-40 a t 37 "C for 18 h. The reaction products were analyzed by SDS-PAGE and autoradiography. The region of the gel containing the small subunits is shown. I , wildtype AOAH, untreated; 2, wild-type AOAH, treated with N-glycanase; 3, mutant AOAH with nonglycosylated small subunit, treated; 4, mutant AOAH, untreated.
has been found only in cells that are thought to interact with LPS or Gram-negative bacteria in vivo. These studies were carried out to explore the contributions of the protein's two subunit domains to its enzymatic activities (acyloxyacyl hydrolase and phospholipase), stability, and intracellular localization.
AOAH is synthesized in BHK570 cells as a 70-kDa precursor. This precursor has two fates: secretion, and proteolytic processing to yield mature AOAH, a protein that has two disulfidelinked subunits. Very little mature AOAH is found in the cell media, and only in cells that produced mature rAOAH was it possible to find intense immunofluorescent staining in lysosomes. Processing ofAOAH to the 60-kDa form could be blocked almost completely by ammonium chloride. Taken together, these results suggest that proteolytic processing occurs in lysosomes; AOAH that bypasses these structures remains unprocessed and may be secreted.
The AOAH small subunit bears a striking sequence similarity to the proteins known as saposins (3). Saposins (sphingolipid activator proteins) are small, heat-stable glycoproteins that greatly enhance the hydrolysis of glycosphingolipids by specific lysosomal hydrolases. Given the structural resemblance of LPS to glycosphingolipids, we hypothesized that the small subunit of AOAH facilitates the recognition of its unique target, LPS. Several of our observations indicate that, in fact, the small subunit contributes not only to LPS recognition but also to the intracellular targeting and catalytic function of the enzyme.
Like the four saposins, the small subunit ofAOAH has a site for N-linked glycosylation. Although a mutation that destroyed this site in saposin B was associated with a variant form of metachromatic leukodystrophy (18), the function of the saposin carbohydrate chain is poorly understood. It does not seem to be necessary for saposin activity or for resistance to proteolytic attack, and a role in protein folding has been suggested (19). Removing this glycosylation site from AOAH did not alter either the enzymatic activity of the protein (Table 11) or its intracellular location (Fig. 71, nor did the intracellular or extracellular protein become more unstable (data not shown). The significance of glycosylation at this site thus is uncertain. On the other hand, both the rAOAH large subunit and the AOAH variant that lacked a 33-amino acid region within the small subunit (including 2 of the 7 cysteines and the glycosylation site) could not be found in the lysosomes that contained wildtype rAOAH and they were poorly processed, suggesting that (this region of) the AOAH small subunit is critical for targeting the intracellular enzyme. The small subunit deletion mutant was also considerably less stable intracellularly than wild-type rAOAH ( Table I). The extent t o which the instability of the protein leads to artifactual estimates of its other properties is unknown; its secretion rate would be underestimated, for example, if the protein were more susceptible to proteolytic degradation in the culture medium.
Elimination of the presumed active site serine (Ser263) in the large subunit reduced enzymatic activity toward LPS by greater than 99%, in keeping with the conclusion that this serine plays a key role in catalysis. The large subunit itself had greatly reduced enzymatic activity when expressed alone, however, again suggesting that the small subunit is required for enzymatic activity. Although the possibility that the large subunit was synthesized in an inactive form cannot be excluded, its DNA sequence, size on SDS-PAGE gels, and presence in immune precipitates with specific anti-AOAH antibodies all suggest that the desired polypeptide was produced.
The amino acid sequence similarity between saposins, surfactant protein B, and the small subunit ofAOAH was noted by Hagen et al. (1). These molecules may constitute a family of small proteins that function at lipid-water or lipid-air interfaces. There are some important differences, however. First, although the AOAH small subunit amino acid sequence, deduced from the cDNA, includes the 6 Cys residues that are found in the other proteins, if the migration of the mature AOAH small subunit on SDS-PAGE gives a true estimate of molecular size (8,100 Da), this subunit has only 4 or 5 cysteines. At least one of these residues must be involved in disulfide linkage to the large subunit.
Second, unlike the saposins, which are independently synthesized cofactors for the glycohydrolases they serve and which must be present in excess for optimal enzymatic activity (20), the AOAH small subunit is covalently linked to the large subunit and both peptides must be present for catalysis to occur.
Proper orientation of the two subunits, as presumably occurs following proteolytic cleavage of the rAOAH precursor, greatly enhanced the activity of the enzyme toward LPS but not toward GPC. This selective activation for enhanced activity toward a particular substrate evidently has little precedent; it is possible that this mechanism is used in vivo to direct the activity of AOAH toward LPS when, for example, phagocytic cells are activated by bacterial stimuli. Against this notion is the observation that substantial amounts of the precursor form were not isolated from human leukocytes using a monoclonal antibody to the large subunit (121, although the use of relatively weak protease inhibitors may have allowed cleavage of the precursor during the purification procedure. We have also found that LPS and various cytokines elicit relatively little augmentation of AOAH activity toward LPS in human monocytes and neutrophil^.^ The extent to which AOAH biosynthesis in phagocytic cells matches that observed here in BHK570 cells is uncertain. In cells that naturally produce it (monocytes, macrophages, neutrophils, and endothelial cells), AOAH is present in such low abundance that immunolocalization has not been possible. In human neutrophils, the enzyme is not in the (lysosome-like?) azurophilic granules, although the ability of ammonium chloride to inhibit LPS deacylation by these cells suggests that AOAH is in an acidic compartment (21). Only mature AOAH has been purified from HL-60 cells and human neutrophils (i.e. little precursor form has been found) and these cells secrete relatively little of the AOAH that they produce. On the other hand, proteolytic processing of the precursor produces small subunits of similar sizes in both BHK cells and human neutrophils. It seems possible that, when rAOAH is produced in large amounts, as in transfected BHK cells, the mechanism for targeting the precursor to the site of proteolytic processing is overwhelmed, so that much of it is secreted. Whether or not the biosynthesis ofAOAH in all of these cell types is identical, these studies in fibroblasts have revealed valuable information about the structure-function relationships of this unusual enzyme.