PCSK9 (proprotein convertase subtilisin/kexin type 9) promotes degradation of the LDLR [LDL (low-density lipoprotein) receptor] through an as-yet-undefined mechanism, leading to a reduction in cellular LDLc (LDL-cholesterol) and a concomitant increase in serum LDLc. Central to the function of PCSK9 is a direct protein–protein interaction formed with the LDLR. In the present study, we investigated a strategy to modulate LDL uptake by blocking this interaction using specific antibodies directed against PCSK9. Studies using surface plasmon resonance demonstrated that direct binding of PCSK9 to the LDLR could be abolished with three different anti-PCSK9 antibodies. Two of these antibodies were raised against peptide epitopes in a region of the catalytic domain of PCSK9 that is involved in the interaction with the LDLR. Such antibodies restored LDL uptake in HepG2 cells treated with exogenous PCSK9 and in HepG2 cells engineered to overexpress recombinant PCSK9. This latter observation indicates that antibodies blocking the PCSK9–LDLR interaction can inhibit the action of PCSK9 produced endogenously in a cell-based system. These antibodies also disrupted the higher-affinity interaction between the natural gain-of-function mutant of PCSK9, D374Y, and the LDLR in both the cell-free and cell-based assays. These data indicate that antibodies targeting PCSK9 can reverse the PCSK9-mediated modulation of cell-surface LDLRs.

PrP C (cellular prion protein) is located at the surface of neuronal cells in detergent-insoluble lipid rafts, yet is internalized by clathrin-dependent endocytosis. As PrP C is glycosyl-phosphatidylinositol-anchored, it requires a transmembrane adaptor protein to connect it to the clathrin endocytosis machinery. Using receptor-associated protein and small interfering RNA against particular LDL (low-density lipoprotein) family members, in combination with immunofluorescence microscopy and surface biotinylation assays, we show that the transmembrane LRP1 (LDL receptor-related protein 1) is required for the Cu 2+ -mediated endocytosis of PrP C in neuronal cells. We show also that another LRP1 ligand that can cause neurodegenerative disease, the Alzheimer's amyloid precursor protein, does not modulate the endocytosis of PrP C .

In the RAS (renin–angiotensin system), Ang I (angiotensin I) is cleaved by ACE (angiotensin-converting enzyme) to form Ang II (angiotensin II), which has effects on blood pressure, fluid and electrolyte homoeostasis. We have examined the kinetics of angiotensin peptide cleavage by full-length human ACE, the separate N- and C-domains of ACE, the homologue of ACE, ACE2, and NEP (neprilysin). The activity of the enzyme preparations was determined by active-site titrations using competitive tight-binding inhibitors and fluorogenic substrates. Ang I was effectively cleaved by NEP to Ang (1–7) ( k cat / K m of 6.2×10 5 M −1 ·s −1 ), but was a poor substrate for ACE2 ( k cat / K m of 3.3×10 4 M −1 ·s −1 ). Ang (1–9) was a better substrate for NEP than ACE ( k cat / K m of 3.7×10 5 M −1 ·s −1 compared with k cat / K m of 6.8×10 4 M −1 ·s −1 ). Ang II was cleaved efficiently by ACE2 to Ang (1–7) ( k cat / K m of 2.2×10 6 M −1 ·s −1 ) and was cleaved by NEP ( k cat / K m of 2.2×10 5 M −1 ·s −1 ) to several degradation products. In contrast with a previous report, Ang (1–7), like Ang I and Ang (1–9), was cleaved with a similar efficiency by both the N- and C-domains of ACE ( k cat / K m of 3.6×10 5 M −1 ·s −1 compared with k cat / K m of 3.3×10 5 M −1 ·s −1 ). The two active sites of ACE exhibited negative co-operativity when either Ang I or Ang (1–7) was the substrate. In addition, a range of ACE inhibitors failed to inhibit ACE2. These kinetic data highlight that the flux of peptides through the RAS is complex, with the levels of ACE, ACE2 and NEP dictating whether vasoconstriction or vasodilation will predominate.

Cell-specific differences in the utilization of the two N-glycosylation sequons (Asn 180 -Ile-Thr and Asn 196 -Phe-Thr) of the prion protein (PrP) have been proposed to influence the aetiology of the neurodegenerative prion diseases. As the N-glycosylation of PrP is ablated by deletion of the C-terminal glycosyl-phosphatidylinositol (GPI) anchor signal sequence, we have investigated the determinants for PrP sequon utilization in human neuronal cells using the novel approach of restoring N-glycosylation to secreted forms of PrP lacking a GPI anchor. N-glycosylation was restored to an efficiency comparable with that of GPI anchored PrP when the distance of the sequon to the C-terminus was increased so that it was sufficient to reach the active site of oligosaccharyltransferase before chain termination. Our findings indicate that sequon utilization in PrP is a co-translational process that precedes GPI anchor addition and, as such, will be greatly influenced by the dynamics of the translocon—oligosaccharyltransferase complex.

NO is related to the pathological condition acute renal failure, in which we previously observed that the level of soluble dipeptidase in urine was decreased. In this study the role of NO in the shedding of the glycosylphosphatidylinositol (GPI)-anchored form of renal dipeptidase (RDPase) was examined. The NO donors sodium nitroprusside (SNP) and S -nitroso- N -acetylpenicillamine rapidly inhibited the release of RDPase from porcine kidney proximal tubules. The substrate of NO synthase, l -Arg, also inhibited the release of RDPase, and this effect was reversed by the NO synthase inhibitor N ω -nitro- l -arginine methyl ester. Western-blot analyses using antibodies raised against porcine RDPase and the inositol-1,2-cyclic monophosphate moiety formed on phospholipase C cleavage of the GPI anchor demonstrated that SNP mediated its inhibitory effect on the release of RDPase via a GPI-specific phospholipase C (GPI-PLC). Peroxynitrite scavengers (deferoxamine and superoxide dismutase) or reducing agent (dithiothreitol) did not affect SNP's inhibition of the release of RDPase. Exposure to the G-protein activator AlF − 4 mimicked the l -Arg effect in the presence of a low concentration of l -Arg, and the effect was completely reversed by U73122, an intracellular phosphatidylinositol-specific PLC (PI-PLC) inhibitor. These results suggest a signal-transduction pathway involving NO, which is produced by NO synthase(s) following activation of a G-protein-coupled PI-PLC, resulting in inhibition of the GPI-PLC that cleaves and releases RDPase. Therefore, this indicates a role for NO as an inhibitory regulator of the shedding of the GPI-anchored RDPase in acute renal failure.

Mammalian aminopeptidase A (APA; glutamyl aminopeptidase; EC 3.4.11.7) is a type II membrane-spanning protein consisting of a short N-terminal cytosolic domain, a single transmembrane domain and a large extracellular C-terminal domain containing the active site. The extracellular domain consists of a 107kDa domain, containing the zinc-binding motif and all the residues involved in catalysis, separated by a protease-susceptible hinge region from the 45kDa C-terminal domain of unknown function. To investigate the role of the 45kDa domain, a construct of murine APA (G594Δ) lacking this C-terminal domain was expressed in COS-1 cells. This truncated form of APA, although expressed, lacked enzymic activity and failed to reach the cell surface. Confocal immunofluorescence microscopy revealed that G594Δ co-localized with the lectin concanavalin A and had a similar staining pattern as protein disulphide-isomerase, indicating that it was retained in the endoplasmic reticulum. Thus the C-terminal 45kDa domain appears to be acting like a pro-domain and seems to be required for the correct folding and trafficking of APA. In contrast, mutation of cysteine-43 to serine, which is involved in the disulphide-linkage of the APA homodimer, did not affect the enzymic activity, cellular location or rate of trafficking through the secretory pathway of APA.

The insolubility of glycosylphosphatidylinositol (GPI)-anchored proteins in certain detergents appears to be an intrinsic property of their association with sphingolipids and cholesterol in lipid rafts. We show that the GPI-anchored protein membrane dipeptidase is localized in detergent-insoluble lipid rafts isolated from porcine kidney microvillar membranes, and that these rafts, which lack caveolin, are enriched not only in sphingomyelin and cholesterol, but also in the glycosphingolipid lactosylceramide (LacCer). Dipeptidase purified from porcine kidney was reconstituted into artificial liposomes in order to investigate the relationship between glycosphingolipids and GPI-anchored protein detergent-insolubility. Dipeptidase was insoluble in liposomes containing extremely low concentrations of LacCer. In contrast, identical concentrations of glucosylceramide or galactosylceramide failed to promote significant detergent-insolubility. Cholesterol was shown to enhance the detergent-insoluble effect of LacCer. GC–MS analysis revealed dramatic differences between the fatty acyl compositions of LacCer and those of the other glycosphingolipids. However, despite these differences, we show that the unusually marked effect of LacCer to promote the detergent-insolubility of dipeptidase cannot be singularly attributed to the fatty acyl composition of this glycosphingolipid molecule. Instead, we suggest that the ability of LacCer to confer detergent-insolubility on this GPI-anchored protein is dependent on the structure of the lipid molecule in its entirety, and that this glycosphingolipid may have an important role to play in the stabilization of lipid rafts, particularly the caveolin-free glycosphingolipid signalling domains.

Angiotensin-converting enzyme (ACE) is one of a growing number of integral membrane proteins that is shed from the cell surface through proteolytic cleavage by a secretase. To investigate the requirements for ectodomain shedding, we replaced the glycosylphosphatidylinositol addition sequence in membrane dipeptidase (MDP) - a membrane protein that is not shed - with the juxtamembrane stalk, transmembrane (TM) and cytosolic domains of ACE. The resulting construct, MDP–STM ACE , was targeted to the cell surface in a glycosylated and enzymically active form, and was shed into the medium. The site of cleavage in MDP–STM ACE was identified by MS as the Arg 374 -Ser 375 bond, corresponding to the Arg 1203 -Ser 1204 secretase cleavage site in somatic ACE. The release of MDP–STM ACE and ACE from the cells was inhibited in an identical manner by batimastat and two other hydroxamic acid-based zinc metallosecretase inhibitors. In contrast, a construct lacking the juxtamembrane stalk, MDP–TM ACE , although expressed at the cell surface in an enzymically active form, was not shed, implying that the juxtamembrane stalk is the critical determinant of shedding. However, an additional construct, ACEΔC, in which the N-terminal domain of somatic ACE was fused to the stalk, TM and cytosolic domains, was also not shed, despite the presence of a cleavable stalk, implying that in contrast with the C-terminal domain, the N-terminal domain lacks a signal required for shedding. These data are discussed in the context of two classes of secretases that differ in their requirements for recognition of substrate proteins.

Spontaneous enzymic release of renal dipeptidase (RDPase; EC 3.4.13.19), a glycosylphosphatidylinositol (GPI)-linked ectoenzyme, was observed in vitro during incubation of porcine proximal tubules at 37°C. Triton X-114 phase separation of the released RDPase showed that the majority of the enzyme activity partitioned into the aqueous phase, indicating its hydrophilic nature. Immunoblot analyses using an antibody against the cross-reacting determinant (CRD) inositol 1,2-cyclic monophosphate, the epitope formed by phospholipase C (PLC) cleavage of the GPI anchor on a protein, detected the released RDPase. Reprobing the immunoblot with an anti-RDPase serum showed the RDPase band co-migrating with the CRD band. The release of RDPase from the proximal tubules was a Ca 2+ -dependent process and had a pH optimum of 9.0. These results indicate that RDPase is released from the proximal tubules by the action of a distinct endogenous GPI-specific PLC.

The somatic and testis isoforms of angiotensin-converting enzyme (ACE) are both C-terminally anchored ectoproteins that are shed by an unidentified secretase. Although testis and somatic ACE both share the same stalk and membrane domains the latter was reported to be shed inefficiently compared with testis ACE, and this was ascribed to cleavage at an alternative site [Beldent, Michaud, Bonnefoy, Chauvet and Corvol (1995) J. Biol. Chem. 270, 28962-28969]. These differences constitute a useful model system of the regulation and substrate preferences of the ACE secretase, and hence we investigated this further. In transfected Chinese hamster ovary cells, human somatic ACE (hsACE) was indeed shed less efficiently than human testis ACE, and shedding of somatic ACE responded poorly to phorbol ester activation. However, using several analytical techniques, we found no evidence that the somatic ACE cleavage site differed from that characterized in testis ACE. First, anti-peptide antibodies raised to specific sequences on either side of the reported cleavage site (Arg 1137 /Leu 1138 ) clearly recognized soluble porcine somatic ACE, indicating that cleavage was C-terminal to Arg 1137 . Second, a competitive ELISA gave superimposable curves for porcine plasma ACE, secretase-cleaved porcine somatic ACE (eACE), and trypsin-cleaved ACE, suggesting similar C-terminal sequences. Third, mass-spectral analyses of digests of released soluble hsACE or of eACE enabled precise assignments of the C-termini, in each case to Arg 1203 . These data indicated that soluble human and porcine somatic ACE, whether generated in vivo or in vitro , have C-termini consistent with cleavage at a single site, the Arg 1203 /Ser 1204 bond, identical with the Arg 627 /Ser 628 site in testis ACE. In conclusion, the inefficient release of somatic ACE is not due to cleavage at an alternative stalk site, but instead supports the hypothesis that the testis ACE ectodomain contains a motif that activates shedding, which is occluded by the additional domain found in somatic ACE.

Lipid rafts are regions of the plasma membrane that are enriched in cholesterol, glycosphingolipids and acylated proteins, and which have been proposed as sites for the proteolytic processing of the Alzheimer's amyloid precursor protein (APP). Lipid rafts can be isolated on the basis of their insolubility in Triton X-100 at 4 °C, with the resulting low-density, detergent-insoluble glycolipid-enriched fraction (DIG) being isolated by flotation through a sucrose density gradient. The detergent-insolubility of APP in mouse cerebral cortex relative to a variety of DIG marker proteins (alkaline phosphatase, flotillin, F3 protein and prion protein) and non-DIG proteins (alkaline phosphodiesterase I, aminopeptidase A and clathrin) has been examined. Alkaline phosphatase, flotillin, F3 protein and the prion protein were present exclusively in the DIG region of the sucrose gradient over a range of protein/detergent ratios used to solubilize the membranes and displayed a characteristic enrichment in the low-density fraction as the protein/detergent ratio was decreased. In contrast, most of the APP, alkaline phosphodiesterase I, aminopeptidase A and clathrin was effectively solubilized at all of the protein/detergent ratios examined. However, a minor proportion of these latter proteins was detected in DIGs at levels which remained constant irrespective of the protein/detergent ratio. When DIGs were isolated from the sucrose gradients and treated with excess Triton X-100, both the DIG marker proteins and APP, alkaline phosphodiesterase I and clathrin were predominantly resistant to detergent extraction at 37 °C. These results show that, although a minor proportion of APP is present in DIGs, where it is detergent-insoluble even at 37 °C, it behaves as an atypical lipid raft protein and raises questions as to whether lipid rafts are a site for its proteolytic processing.

Mammalian angiotensin-converting enzyme (ACE; EC 3.4.15.1) is one of several proteins that exist in both membrane-bound and soluble forms as a result of a post-translational proteolytic processing event. For ACE we have previously identified a metalloprotease (secretase) responsible for this proteolytic cleavage. The effect of a range of structurally related zinc metalloprotease inhibitors on the activity of the secretase has been examined. Batimastat (BB94) was the most potent inhibitor of the secretase in pig kidney microvillar membranes, displaying an IC 50 of 0.47 μ M, whereas TAPI-2 was slightly less potent (IC 50 18 μ M). Removal of the thienothiomethyl substituent adjacent to the hydroxamic acid moiety or the substitution of the P2′ substituent decreased the inhibitory potency of batimastat towards the secretase. Several other non-hydroxamate-based collagenase inhibitors were without inhibitory effect on the secretase, indicating that ACE secretase is a novel zinc metalloprotease that is related to, but distinct from, the matrix metalloproteases. The full-length amphipathic form of ACE was labelled selectively with 3-trifluoromethyl-3-( m -[ 125 I]iodophenyl)diazirine in the membrane-spanning hydrophobic region. Although trypsin was able to cleave the hydrophobic anchoring domain from the bulk of the protein, there was no cleavage of full-length ACE by a Triton X-100-solubilized pig kidney secretase preparation when the substrate was in detergent solution. In contrast, the Triton X-100-solubilized secretase preparation released ACE from pig intestinal microvillar membranes, which lack endogenous secretase activity, and cleaved the purified amphipathic form of ACE when it was incorporated into artificial lipid vesicles. Thus the secretase has an absolute requirement for its substrate to be inserted in a lipid bilayer, a factor that might have implications for the development of cell-free assays for other membrane protein secretases. ACE secretase could be solubilized from the membrane with Triton X-100 and CHAPS, but not with n-octylβ- D -glucopyranoside. Furthermore trypsin could release the secretase from the membrane, implying that like its substrate, ACE, it too is a stalked integral membrane protein.

Membrane dipeptidase (MDP; EC 3.4.13.19) enzymic activity that was inhibited by cilastatin has been detected on the surface of 3T3-L1 cells. On differentiation of the cells from fibroblasts to adipocytes the activity of MDP increased 12-fold. Immunoelectrophoretic blot analysis indicated that on adipogenesis the increase in the amount of MDP preceded the appearance of GLUT-4. MDP on 3T3-L1 adipocytes was anchored in the bilayer by a glycosyl phosphatidylinositol (GPI) moiety as evidenced by its release into the medium in a hydrophilic form on treatment of the cells with bacterial phosphatidylinositol-specific phospholipase C and the appearance of the inositol 1,2-cyclic monophosphate cross-reacting determinant. Incubation of 3T3-L1 adipocytes with either insulin or the sulphonylurea glimepiride led to a rapid concentration- and time-dependent release of MDP from the cell surface. The hydrophilic form of MDP released from the cells on stimulation with insulin was recognized by antibodies against the inositol 1,2-cyclic monophosphate cross-reacting determinant, indicating that it had been generated by cleavage of its GPI anchor through the action of a phospholipase C.

Membrane dipeptidase (EC 3.4.13.19) is a plasma membrane zinc peptidase that is involved in the renal metabolism of glutathione and its conjugates, such as leukotriene D 4 . The enzyme lacks the classical signatures of other zinc-dependent hydrolases and shows no homology with any other mammalian protein. We have used site-directed mutagenesis to explore the roles of five histidine residues in pig membrane dipeptidase that are conserved among mammalian species. When expressed in COS-1 cells, the mutants H49K and H128L exhibited a specific activity and K m for the substrate Gly- D -Phe comparable with those of the wild-type enzyme. However, the mutants H20L, H152L and H198K were inactive, but were expressed at the cell surface at equivalent levels to the wild-type, as assessed by immunoblotting and immunofluorescence. These three mutants were compared with regard to their ability to bind to the competitive inhibitor cilastatin, which binds with equal efficacy to native and EDTA-treated pig kidney membrane dipeptidase. Expressed wild-type enzyme and mutants H20L and H198K were efficiently bound by cilastatin–Sepharose, but H152L failed to bind. Thus His-152 appears to be involved in the binding of substrate or inhibitor, whereas His-20 and His-198 appear to be involved in catalysis. Membrane dipeptidase shares some similarity with a dipeptidase recently cloned from Acinetobacter calcoaceticus . In particular, His-20 and His-198 of membrane dipeptidase are conserved in the bacterial enzyme, as are Glu-125 and His-219, previously shown to be required for catalytic activity.

Membrane dipeptidase (EC 3.4.13.19) enzyme activity that is inhibited by cilastatin has been detected in pancreatic zymogen granule membranes of human, porcine and rat origin. Immunoelectrophoretic blot analysis of human and porcine pancreatic zymogen granule membranes with polyclonal antisera raised against the corresponding kidney membrane dipeptidase revealed that the enzyme is a disulphide-linked homodimer of subunit mass 61 kDa in the human and 45 kDa in the pig. Although membrane dipeptidase was, along with glycoprotein-2, one of the only two major components of carbonate high pH-washed membranes, no enzyme activity or immunoreactivity was detected in the zymogen granule contents. Digestion with bacterial phosphatidylinositol-specific phospholipase C (PI-PLC), and subsequent recognition by antibodies specific for the cross-reacting determinant, revealed that membrane dipeptidase in human and porcine pancreatic zymogen granule membranes is glycosyl-phosphatidylinositol-anchored. Membrane dipeptidase was released from the pancreatic zymogen granule membranes by an endogenous hydrolase, and the released form migrated as a disulphide-linked dimer on SDS/PAGE under non-reducing conditions. Under reducing conditions it migrated with the same apparent molecular mass as the membrane-bound form, and was still a substrate for bacterial PI-PLC. Treatment of kidney microvillar membranes with phospholipase A 2 resulted in the release of membrane dipeptidase in a form that demonstrated electrophoretic and cilastatin–Sepharose binding properties identical to those of the endogenously released form of the enzyme from zymogen granule membranes. These results indicate that the glycosyl-phosphatidylinositol anchor on the pancreatic membrane dipeptidase is cleaved by an endogenous hydrolase, probably a phospholipase A, and that this cleavage may promote the release of the protein from the membrane.

A diverse range of membrane proteins of Type I or Type II topology also occur as a circulating, soluble form. These soluble forms are often derived from the membrane form by proteolysis by a group of enzymes referred to collectively as ‘secretases‘ or ‘sheddases’. The cleavage generally occurs close to the extracellular face of the membrane, releasing physiologically active protein. This secretion process also provides a mechanism for down-regulating the protein at the cell surface. Examples of such post-translational proteolysis are seen in the Alzheimer's amyloid precursor protein, the vasoregulatory enzyme angiotensin converting enzyme, transforming growth factor-α, the tumour necrosis factor ligand and receptor superfamilies, certain cytokine receptors, and others. Since the proteins concerned are involved in pathophysiological processes such as neurodegeneration, apoptosis, oncogenesis and inflammation, the secretases could provide novel therapeutic targets. Recent characterization of these individual secretases has revealed common features, particularly sensitivity to certain metalloprotease inhibitors and up-regulation of activity by phorbol esters. It is therefore likely that a closely related family of metallosecretases controls the surface expression of multiple integral membrane proteins. Current knowledge of the various secretases are compared in this Review, and strategies for cell-free assays of such proteases are outlined as a prelude to their ultimate purification and cloning.

The Triton-insoluble complex from porcine lung membranes has been separated into two distinct subfractions visible as discrete light-scattering bands following buoyant density-gradient centrifugation in sucrose. Both of these detergent-insoluble complexes were enriched in the glycosyl-phosphatidylinositol (GPI)-anchored ectoenzymes alkaline phosphatase, aminopeptidase P and 5´-nucleotidase, and both complexes excluded the polypeptide-anchored ectoenzymes angiotensin-converting enzyme, dipeptidyl peptidase IV and aminopeptidases A and N. The GPI-anchored proteins in both complexes were susceptible to release by phosphatidylinositol-specific phospholipase C. Both complexes were also enriched in cholesterol and glycosphingolipids, and in caveolin/VIP21, although only the higher-density fraction was enriched in the plasmalemmal caveolar marker proteins Ca 2+ -ATPase and the inositol 1,4,5-trisphosphate receptor. Among the annexin family of proteins, annexins I and IV were absent from the two detergent-insoluble complexes, annexin V was present in both, and annexins II and VI were only enriched in the higher-density fraction. When the metal chelator EGTA was present in the isolation buffers, annexins II and VI dissociated from the higher-density detergent-insoluble complex and only a single light-scattering band was observed on the sucrose gradient, at the same position as for the lower-density complex. In contrast, in the presence of excess calcium only a single detergent-insoluble complex was isolated from the sucrose gradients, at an intermediate density. Thus the detergent-insoluble membrane complex can be subfractionated on the basis of what appears to be calcium-dependent, annexin-mediated, vesicle aggregation into two distinct populations, only one of which is enriched in plasmalemmal caveolar marker proteins.

Aminopeptidase P (AP-P; X-Pro aminopeptidase; EC 3.4.11.9), a key enzyme in the metabolism of the vasodilator bradykinin, has been cloned from a pig kidney cortex cDNA library following the use of the PCR to identify sub-libraries enriched in AP-P clones. The complete primary sequence of the enzyme has been deduced from a full-length cDNA clone. This predicts a protein of 673 amino acids with a cleavable N-terminal signal sequence and six potential N-linked glycosylation sites. A stretch of mainly hydrophobic amino acids at the C-terminus is predicted to co-ordinate the attachment of a glycosyl-phosphatidylinositol (GPI) membrane anchor. Although AP-P is a zinc metallopeptidase, the predicted primary sequence does not contain any recognizable zinc-binding motif. Transient expression of AP-P cDNA in COS-1 cells resulted in enzymic activity characteristic of AP-P, namely apstatin- and EDTA-sensitive hydrolysis of bradykinin and Gly-Pro-Hyp. The expressed protein was recognized as a polypeptide of M r 91000 under reducing conditions, following immunoblotting of COS-1 membranes with a polyclonal antibody raised against purified pig kidney AP-P. The presence of a GPI anchor on expressed AP-P was established by demonstrating release of the enzyme from a membrane fraction following treatment with bacterial phosphatidylinositol-specific phospholipase C and its corresponding conversion from an amphipathic to a hydrophilic form, as assessed by phase separation in Triton X-114. Sequence comparisons confirm that AP-P is a member of the proline peptidase family of hydrolytic enzymes and is unrelated in sequence to other brush-border membrane peptidases.