Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The organisms known as Pneumocystis carinii are a diverse family of fungi that cause severe pneumonia in the immunocompromised host. P. carinii pneumonia is one of the major opportunistic infections in human immunodeficiency virus-infected individuals, and is also a significant clinical problem in patients undergoing organ transplantation, those receiving chemotherapy for malignant disease, and those with severe combined immunodeficiency disorders. Knowledge of the biology of this family of pathogenic fungi is limited, in part due to the extreme difficulties in the culture of the organism (1, 2). Interest has focused recently on a multigene family that encodes a surface-expressed subtilisin-like serine protease known as PRT1 (3, 4). Characterization of the PRT1 gene family in rat-derived P. carinii (P. carinii f. sp. carinii) has shown that many PRT1 genes reside in the subtelomeric regions of P. carinii f. sp. carinii chromosomes (3, 5). Two other gene families that encode surface proteins have also been identified in the subtelomeric regions, the genes encoding the major surface glycoprotein (MSG) (6), and the MSG-related (MSR) antigen (also known as variant MSG) (18, 19). Regulation of expression of the MSG gene family takes place by the installation of different copies of MSG at a unique subtelomeric expression site, giving rise to variation in the isoform of MSG that is expressed on the surface of the organism (13, 17, 20). The colocalization of copies of MSG and PRT1 genes suggests that PRT1 may also have a role in the generation of antigenic diversity (3, 5, 23). The PRT1 protease belongs to the kexin (KEX) family of subtilisin-like serine proteases and shows homology to other fungal processing proteases, including Saccharomyces cerevisiae KEX2, involved in the processing of the -mating factor and the killer toxin (26). The P. carinii PRT1 protease differs from other kexins in being encoded by multicopy gene family rather than a single-copy gene and, in addition, is expressed on the surface of the organism. The catalytic domain of PRT1 genes is conserved between different gene copies, whereas the proline-rich region, a domain rarely seen in other KEXs, is highly polymorphic between PRT1 family members (3, 4). The PRT1 protease gene family has been found in several formae speciales of P. carinii, including the human pathogen P. carinii f. sp. hominis and also P. carinii f. sp. muris (4, 27).

Surface proteins in eukaryotic microbes can be attached to the cell membrane via a hydrophobic transmembrane polypeptide domain, or alternatively, by a glycosylphosphatidylinositol (GPI) anchor (28). The sequence of the 3'-end of the PRT1 genes suggests that many family members encode a GPI anchor signal sequence at the C-terminus. This type of attachment is typified by the variant surface glycoproteins (VSGs) of Trypanosoma brucei (29), and also the surface protease (gp63) of Leishmania (30). The P. carinii MSG is likely to be linked to the surface of the organism via a GPI anchor. The C-terminal residues of MSG are capable of directing addition of a GPI anchor in a heterologous system (31). Similarly, members of the recently described MSR gene family have C-terminal sequences that appear to be similar to MSG in having a potential GPI attachment site (18, 19).

The overall structure of C-terminal signal sequences responsible for GPI attachment appear to be conserved, although sequences in parasitic protozoa that are known to be successful at directing GPI attachment are not always functional in mammalian systems (32). The conserved structure has allowed the successful use of heterologous systems to study GPI-anchor addition (33). In the present article we report the use of CD14 for the study of GPI anchor signal sequences in a COS-cell expression system. CD14 is a GPI-anchored, glycosylated monocyte antigen (34) that is thought to have a role in the response to lipopolysaccharide (35). Using a truncated CD14 construct, we show that C-terminal sequences from PRT1 can confer the ability to add a GPI anchor, supporting the view that the PRT1 protease is attached to the surface of P. carinii organisms by a GPI anchor.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Generation of Recombinant Plasmids

A fragment of encoding the first 335 residues of CD14 complementary DNA (cDNA) (34) was amplified using polymerase chain reaction (PCR) with primers CD14Met/KpnI (5'-CGGGGTACCGCCACCATGGAGCGCGCGTC-3') and CD14G335/ BamHI (5'-CGCGGATCCAGGGACCAGGAAGG-3'). The PCR was performed with 10⁵ U/mL¹ native pfu DNA polymerase (Stratagene, La Jolla, CA) in the buffer supplied, each primer at 0.2 µM and each deoxynucleotide at 30 µM. The amplification products were digested with restriction endonuclease Asp718 and BamHI, and then cloned into the vector pcDNA3 (Invitrogen, Gronengen, Germany), resulting in the construct CD14G335T, which has a synthetic Kozak sequence 5' to the initiation codon and a stop codon 3 base pairs 3' to the BamHI insertion site.

Four fragments from the 3'end of PRT1 were generated by PCR from plasmids containing PRT1-pAGA and PRT1-73j, using the fu polymerase. Primers pcprot48/BamHI (pcprot48/BamHI (5'-CGC GGATCCCCGGAACAAAAACCAACATC-3') and pcprot44/ EcoRI (5'-CCGGAATTCTTAAAAAGAGTAACCCAAAAAT A-3') were used to amplify from both PRT1-pAGA and PRT1-73j. The primer pcprot49/BamHI (5'-CGCGGATCCGGTGCTTCT CATC-3') and pcprot44/EcoRI were used to amplify from PRT1-73j and primers pcrot50/BamHI (5'-CGCGGGATTCAGTGCTTCT CAC-3') and pcprot44/EcoRI from PRT1- pAGA. The four PCR products from PRT1 were cloned into CD14G335T after digesting with BamHI and EcoRI. The 3' ends of the four resulting constructs in pCDNA3 were sequenced using the Big Dye terminator kit, to confirm that all the constructs were in frame. DNA for transfection was prepared using the Nucleobond midiprep kit (Clontech, Basingstoke, U.K.). The cDNA of CD44 (36) was kindly provided by Suneale Banerji Institute of Molecular Medicine, U.K.

Fluorescent Studies of Recombinant Proteins

COS-1 cells were grown in RPMI 1640 medium (Sigma, Poole, U.K.) supplemented with 10% fetal bovine serum (FBS, Paisley, U.K.) (Sigma), penicillin, streptomycin, and glutamine (GIBCO BRL), under standard incubation conditions. Transfections were performed in six-well dishes using Fugene 6 (Roche Molecular Biochemicals, Lewes, U.K.) according to the manufacturer's instructions, using 0.8 µg supercoiled plasmid DNA and 2 µL Fugene per well. After 24 h, the cells were stained with monoclonal antibody (mAb). Nonpermeabilized cells were washed with phosphate-buffered saline (PBS) containing 1% FBS and 0.1% sodium azide. They were then stained with a 10-fold dilution of tissue culture supernatant containing the mAbs 60bca (ATCC) or UCHM-1 (Sigma) for 1 h in PBS, with 5% FBS and 0.1% sodium azide. After further washing, the cells were incubated with the fluorescein isothiocyanate (FITC)-conjugated goat F(ab')₂ antimouse immunoglobulin (Ig) secondary antibody (Dako, Ely, U.K.) for 1 h. Cells were fixed with 2% paraformaldehyde before scanning using a Millipore Fluoscan 2350 fluorimeter with an excitement wavelength of 485 nm and a detection wavelength of 530 nm. A total of 20 data points from the center of each well were averaged to assess the level of fluorescence, and each well was examined by fluorescent microscopy to confirm the morphology of the cells. When required, the cells were permeabilized using PBS/ 0.1% Triton X-100 before staining (37). Phosphatidylinositol-specific phospholipase C (PIPLC) (Glyko, Bicester, Oxfordshire, U.K.) was used to cleave GPI-anchored proteins. Whole cells were incubated with 1 U/mL¹ PIPLC in a low-salt buffer (50 mM triethanolamine-HCl, pH 7.5, and 300 mM sucrose) for 1 h at 37°C. Microscopy was also performed on cells grown on coverslips and prepared as described earlier. Photographs were taken using a Zeiss Photomicroscope III.

Analysis of Recombinant Proteins by Western Blotting

Detergent-soluble proteins were extracted by lifting the cells in PBS/5mM ethylenediamenetetraacetic acid, followed by lysis in 0.1% Triton X-100 buffer containing sodium azide and protease inhibitors (Roche). Cellular debris and detergent-insoluble material were removed by centifugation, and the proteins separated using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), with prestained molecular weight markers (New England Biolabs) as standards. Proteins were transferred onto Hybond C-extra nitrocellulose membrane (Amersham Pharmacia Biotech). Recombinant CD14 proteins were detected by reacting with mAb 60bca, a peroxidase-conjugated goat-antimouse Ig secondary antibody (Dako), and were visualized using enhanced chemiluminescence (Amersham Pharmacia Biotech).

Labeling of Recombinant Proteins with ³⁵S-Methionine

Transfected cells were washed in PBS before being placed in methionine-deficient Minimal Essential Medium (MEM) (Sigma) for 1 h. Methionine-deficient MEM was supplemented with 2% FBS and 25 pmol (0.925 MBq) of ³⁵S L-methionine (Amersham Pharmacia Biotech), and the cells were incubated for 16 h, washed, and treated with PIPLC as described earlier. The PIPLC buffer was removed and a total cell lysate prepared as described earlier. Samples were precleared by incubation with protein A-sepharose (Sigma) for 1 h, and the sepharose beads were removed. Approximately 1 µg of mAb 60bca was added and incubated for 2 h, protein A-sepharose was added, and the mix was incubated for a further 2 h. The sepharose beads were washed four times in PBS/0.2% Triton X-100 and a further time in PBS alone. The proteins were eluted by boiling in sample buffer and separated by SDS-PAGE. Radiolabeled proteins were visualized using Kodak Biomax MR film.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Predicted GPI Signal Sequence in PRT1 Genes

The composition of the C-terminal domain of the deduced amino acid sequence of PRT1 genes suggested that they had the potential to encode GPI-anchor attachment. Although PRT1 family members were highly polymorphic in some domains, particularly the proline-rich region, the sequence of the C-terminus of the protein was more conserved (3). The PRT1 sequences conformed to the characteristics of a protein which is processed to a GPI form, having a hydrophobic signal sequence at both the N-terminus and the C-terminus. The GPI anchor signal is characterized by the GPI attachment site, known as the  site, followed by a short hydrophilic stretch, followed by a strongly hydrophobic region, usually of 15 to 30 residues, with similarities to signal peptides (38). Using the values generated by Kodukula and colleagues, we were able to make predictions about the potential  sites found within copies of PRT1 (39). The most likely  site was one of three adjacent residues, between 31 and 33 residues from the C-terminus of the nascent protein (Figure 1). A number of PRT1 sequences were examined from a genomic library and a cDNA library (4) and from the expressed sequence tags from the Pneumocystis genome project (http:// http://biology.uky.edu/Pc/). More than 85% of the 72 sequences examined had predicted amino acid sequences identical to either PRT1-pAGA or PRT1-73j in the region of the  site.

View larger version (11K): [in this window] [in a new window]   Figure 1.   Alignment of the 3' end of PRT1-73j and PRT1-pAGA, showing the predicted  sites (in bold). The values show the relative probability that the residue indicated is the  site, and the underlined residues are predicted to form the hydrophobic tail essential for GPI-anchor addition.

Generation of CD14-PRT1 Fusion Protein

We used a heterologous expression system to test whether the predicted GPI signal sequence of PRT1 directed surface expression of the protein. This approach was chosen in preference to biochemical studies because of the problems of obtaining a sufficient quantity and quality of this organism, for which in vitro culture still remains highly problematic. The transient transfection of COS-1 cells has been used in studies of GPI anchors, and the monocyte antigen CD14, a well characterized GPI-anchored glycoprotein (34), was used to construct the recombinant fusion protein. The functional  site of CD14 has not been determined, and we predicted three potential  sites: at residues G₃₃₅, N₃₄₅, and S₃₅₈. A truncated CD14 molecule with no  sites was constructed to eliminate the potential to direct GPI-anchor addition. The primer pair CD14met/KpnI and CD14G335/BamHI were used to amplify the sequence from the CD14 cDNA clone using pfu polymerase. The amplification product was cloned into the vector pCDNA3, which resulted in a truncated CD14 construct, CD14G335T. This truncated CD14 construct, lacking the GPI signal sequence, was used as the backbone for the PRT1 constructs.

Two different C-terminal PRT1 sequences were chosen, to test whether the C-terminus of PRT1 could direct the addition of a GPI anchor. Expression constructs were made using the truncated CD14 and the C-terminal sequence of the PRT1-73j (a cDNA clone) and PRT1-pAGA (a genomic clone) (4). Two fragments were generated from PRT1-73j; the longer fragment was amplified using primers pcprot48/BamHI and pcprot44/EcoRI and encoded 89 residues, including the serine/threonine-rich region and the potential residues for the GPI-anchor addition. The shorter fragment was amplified using primers pcprot49/ BamHI and pcprot44/EcoRI and encoded 32 residues containing the potential residues for the GPI-anchor addition (Figure 2a). Two similar fragments were generated from PRT1-PRT1-pAGA, one amplified using pcprot48/BamHI and pcprot44/EcoRI (80 residues), and the other using pcprot50/BamHI and pcprot44/EcoRI (32 residues). These four fragments were cloned into the vector pCDNA3-CD14G335T to create the four constructs CD14G335-73j48-44, CD14G335-73j49-44, CD14G335-pAGA48-44, and CD14G335-pAGA50-44 (Figure 2a). All four of these constructs had the potential to direct addition of a GPI anchor (Figure 2b), although the two shorter constructs contained only the minimal number of residues. DNA sequencing was used to verify that the cloning procedure had occurred successfully.

View larger version (18K): [in this window] [in a new window]   View larger version (11K): [in this window] [in a new window]   Figure 2.   (a) Design of recombinant proteins based on CD14. The top line shows wild-type CD14 with the potential  sites marked above the sequence, with the relative probability that they are the native  site shown in brackets. Truncated CD14 (CD14G335T) has all the potential  sites removed. C-terminal sequences from PRT1 were cloned in frame and downstream of the truncated CD14 to give the other constructs. The blocks with vertical stripes represent serine-threonine-rich regions, and the stippled regions represent the C-terminal potential GPI attachment signal. The arrows show the position of the predicted  site in the constructs. (b) C-terminal sequences from the CD14 constructs. The residues in CD14 that have potential to be the  site of CD14 are shown above the sequence, and the residues in the recombinant proteins that have the potential to be  sites are shown in bold. Residues predicted to form the hydrophobic tail essential for GPI-anchor addition are underlined.

Expression of Recombinant CD14-PRT1 Fusion Proteins

A total of seven plasmidsthe four CD14-PRT1 fusion constructs, the full-length CD14, CD14G335T, and a vector control (pCDNA3)were transiently transfected into COS-1 cells. The detergent-soluble proteins from transfected cells were extracted with Triton X-100. These proteins were separated using SDS-PAGE and Western-blotted, and the recombinant protein was detected using the anti-CD14 antibody 60bca (Figure 3). The antibody reacted with the CD14 protein, migrating as multiple bands of relative molecular mass (M_r) 40 to 55 kD (measured M_r of fully glycosylated wild-type CD14 is 55 kD). Multiple protein bands were observed with each of the constructs, where the lowest M_r seen corresponded approximately with the predicted M_r of the nascent protein. The bands of higher M_r indicated the presence of several glycosylated forms, or a mixture of proteins at various stages of post-translational modification. These results indicate that the transfection of cells resulted in the expression of recombinant CD14 glycoproteins which were recognized by the 60bca antibody. Transfected cells were permeabilized using Triton X-100 and stained to visualize both intracellular and cell-surface CD14 recombinant protein. Cells transfected with CD14G335T showed little surface labeling but intense staining around the nucleus (data not shown), which was likely to be the endoplasmic reticulum and golgi apparatus. This indicates that although the cells transfected with CD14G335T were expressing the protein in amounts comparable with the other constructs, little of it was reaching the surface.

View larger version (32K): [in this window] [in a new window]   Figure 3.   Expression of CD14 recombinant proteins in COS cells. The anti-CD14 mAb 60bca reacted with all forms of the recombinant proteins. Transfected cells were lysed with Triton X-100 and analyzed by Western blotting. Lane 1: CD14; lane 2: CD14G335T; lane 3: vector control; lane 4: CD14G335-73j48-44; lane 5: CD14G335-73j49-44; lane 6: CD14G335-pAGA48-44; lane 7: CD14G335-pAGA50-44. The molecular weight standards (left) are shown in kD.

Addition of C-Terminal Sequences from P. carinii PRT1 Direct Surface Expression of Trunctated CD14

Transfected, nonpermeabilized cells were stained with the anti-CD14 antibody 60bca, followed by a FITC-labeled secondary antibody. The amount of antibody binding was examined by scanning the plates using a fluorimeter (Figure 4), and the pattern of staining examined using fluorescent microscopy (Figures 4b-4g). Cells expressing the full-length CD14 showed a significant amount of surface labeling compared (Figure 4b) with cells transfected with vector, which showed no surface labeling by microscopy (not shown). The pattern of expression in the CD14 cells was indicative of staining due to a protein on the surface, with a halo of dense staining surrounding a paler interior. The staining was largely even, although the cell surface appeared rough. Cells expressing CD14G335T showed a small amount of surface labeling, and under the microscope the pattern of staining was more punctate than with the full-length CD14 (Figure 4c). All four constructs showed amounts of staining significantly higher than that for CD14G335T, indicating that the addition of the C-terminus from PRT1-PRT1-73j or PRT1-pAGA resulted in more of the CD14-derived protein reaching the surface (Figures 4a, d-g). Similar results were obtained when the cells were stained with a different anti-CD14 mAb, UCHM-1. The shorter constructs (CD14G335-73j49-44 and CD14G335-pAGA50-44) were able to direct more protein to the cell surface than were the longer constructs (CD14G335-73j48-44 and CD14G335-pAGA48-44). The pattern of the surface of the cells resembled that of CD14 with an even and intense surface labeling. Cells expressing the CD14-PRT1 constructs were then permeabilized, stained, and examined by fluorescent microscopy as before. These cells showed both surface labeling and intracellular staining, whereas cells transfected with the vector control did not show any reactivity (data not shown).

View larger version (14K): [in this window] [in a new window]   Figure 4.   Surface expression of truncated CD14 constructs using C-terminal sequences from PRT1, as measured by fluorimetry. COS cells were transfected with CD14 constructs or a vector control, and stained using anti-CD14 mAb 60bca, 24 h after transfection, and without permibilizing the cells. Three wells were used for each construct, two transfected with 0.5 µg plasmid DNA per well, and one transfected with 0.8 µg plasmid DNA. The filled bars show the average of the two wells transfected with 0.5 µg DNA/well, and the open bars show the average of all three wells. The average fluorescence of 20 data points was taken for each well; the error bars show standard deviation between the values for each well.

CD14-PRT1 Fusion Proteins Are Attached to the Surface via a PIPLC-Labile Linkage

GPI-specific phospholipase C (PIPLC) cleaves GPI anchors specifically, leaving the lipid moiety in the membrane, and releasing the protein with a terminal cyclic phosphoinositol. Cells transfected with the different constructs were incubated with PIPLC to release any GPI-linked proteins. Cells treated with PIPLC were compared with untreated cells by staining nonpermeablized cells with the anti-CD14 mAb 60bca followed by a FITC secondary antibody. The cells were scanned, and also examined under the microscope as before, to check that they had not lifted during their incubation in PIPLC buffer. Full-length CD14 showed a significant reduction in the fluorescence intensity (Figure 5), confirming that in this system CD14 is linked to the surface via a PIPLC-labile moiety. The vector control again showed no cell staining, but CD14G335T did show a significant reduction in cell staining. The four constructs CD14G335-73j48-44, CD14G335-73j49-44, CD14G335-pAGA48-44, and CD14G335-pAGA50-44 all stained significantly more than did CD14G335T, and the majority of the fluorescence was PIPLC-labile (Table 1). As an additional control, a construct containing CD44, a leucocyte antigen which has a transmembrane spanning domain, was transfected into COS-1 cells in an identical experiment. To demonstrate that PIPLC was not causing a wholesale release of membrane-associated proteins, cells expressing CD44 were treated with PIPLC; there was no significant reduction in the amount of surface labeling (Table 1), as assessed by scanning or microscopy.

View larger version (35K): [in this window] [in a new window]   Figure 5.   Surface expression of truncated CD14 constructs using C-terminal sequences from PRT1, as measured by fluorescent microscopy. Nonpermeabilized, transfected COS-1 cells were stained 24 h after transfection using anti-CD14 mAb 60bca and examined using fluorescent microscopy. (a) CD14; (b) CD14G335T; (c) CD14G335-73j48-44; (d) CD14G335-73j49-44; (e) CD14G335-pAGA48-44; ( f ) CD14G335-pAGA50-44.

To test whether the fusion proteins that were released by cleavage of the GPI anchor by PIPLC were intact and that the release was not due to protein degradation, cells transfected with CD14, vector control, CD14G335T, and CD14G335-pAGA50-44 were metabolically labeled with S³⁵-methionine. The tissue-culture supernatants were collected and the cells treated with PIPLC as before. The buffer containing proteins cleaved by the action of PIPLC was collected, and a cell lysate of detergent-soluble proteins was prepared from the remaining cells. All samples were immunoprecipitated with anti-CD14 antibody 60bca and protein A-sepharose beads, and the proteins separated with SDS-PAGE. The cell supernatant proteins secreted by cells transfected with CD14, CD14G335T, and CD14G335-pAGA50-44 (Figure 6a) were compared with the proteins from the total cell lysate (Figure 6b). CD14G335T secreted a very high level of a protein (~ 44 kD) seen in the total cell lysate. A protein corresponding to a larger form of CD14 (~ 51 kD) and CD14G335-pAGA50-44 (~ 48 kD) was found to be removed by PIPLC, showing that the PIPLC-labile protein was present on the cell surface in a glycosylated form.

View larger version (18K): [in this window] [in a new window]   Figure 6.   Release of CD14 recombinant proteins as measured by fluorimetry. Cells were transfected with the CD14 constructs or a vector control, and 24 h after transfection stained using mAb 60bca while nonpermeabilized. Two wells were used for each measurement, both transfected with the 0.8 µg plasmid DNA per well. The open bars show the mean fluorescence with no PIPLC treatment, the filled bars show mean fluorescence with PIPLC treatment. The error bars show standard deviation between the values for each well.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Many proteins of the fungal surface are linked via a GPI anchor either to the membrane or to -glucan moieties of the cell wall (40). This is also the case in a number of protozoa, where constituents of the outer envelope are also GPI-anchored. Leismania spp. have a family of GPI- anchored surface proteases (leishmanolysin or gp63) which make up a large proportion of the outer cell envelope (30). Trypansoma spp. VSG is also GPI-anchored and is involved in antigenic variation (41).

In this report, we demonstrate that C-terminal sequences from PRT1 were sufficient to direct linkage to the outer cell membrane via a GPI anchor in a heterologous system. This approach, the expression of CD14/PRT1 fusion proteins in COS-1 cells, was taken because of the difficulty in performing the experiments directly on P. carinii organisms, and was possible because the GPI attachment signal is an independent functional domain and is well conserved among eukaryotes. Some protozoal GPI anchor signals do not operate well in the COS cell system (32), but it is not thought that GPI anchors can be added erroneously. We demonstrated that the CD14/PRT1 fusion proteins reside on the surface of COS-1 cells by showing that anti-CD14 antibodies could access the proteins on nonpermeabilized cells and that the proteins were also accessible to PIPLC. The recombinant proteins, including the truncated CD14, were also shed continually into the cell culture medium. Two different antibodies were used during these experiments, both producing the same results, and mAb 60bca was shown to react with proteins in the correct size range. The specificity of PIPLC in cleaving the proteins provided additional support to the linkage of the recombinant proteins to the surface of COS cells via a GPI anchor. A small amount of residual surface expression was seen with the truncated CD14 construct, but it had a distinctly different pattern of expression when visualized by microscopy. This construct resulted in a high-level expression of the truncated protein, which was largely secreted. This truncated molecule is known to interact with other cell-surface proteins, and the reason for the small amount of PIPLC-labile surface labeling was likely to be due to nonspecific associations with other GPI-anchored proteins.

The GPI attachment signal in PRT1-pAGA and PRT1-73j differed slightly, but both functioned in directing surface expression. Many genomic and cDNA PRT1 sequences were determined, and the GPI signal sequence appeared highly conserved. The GPI attachment signal was predicted to be localized to one of three residues, which may be functionally redundant in P. carinii. Russian and associates (5) reported a cDNA that has homology to PRT1, but its 3' end may not form a GPI anchor. It is not possible to say whether this cDNA shares some of the other features of PRT1, such as a subtelomeric location in the genome and surface localization of the protein. Fungi such as P. carinii have a complex cell wall, and it is thought that GPI-anchored proteins can either reside attached to the cell membrane via a fatty acid, or be transferred and linked to the -glucan of the cell wall (42). Signals which control the targeting of GPI-anchored proteins to either the plasma membrane or the cell wall have been identified in S. cerevisiae (43). Residues immediately N-terminal of the  site appear to be responsible, but it is not known whether such a system operates in P. carinii. This system cannot predict whether PRT1 is membrane-associated or cell wall-associated, although Russian and coworkers have localized PRT1 to the electron-lucent layer between the cell membrane and cell wall in cysts (5). Both mechanisms may exist in P. carinii, with membrane association occurring in the trophic form and linkage to -glucan in the cystic form, inasmuch as glucan is a major component of this life-cycle stage (46).

Members of the family of PRT1 proteases may have several functions in P. carinii. Those family members that are located on the surface of the organism may be involved in host-parasite interaction. PRT1 may act on host proteins to hinder the immune response, but it may also form a role in maintaining the integrity of the cell wall. The function of the GPI-linked protease of Leishmania, GP63, is thought to be protection against complement- mediated lysis, but deletion of six of the seven gp63 genes does not seem to affect survival significantly in the mammalian host (47). Other PRT1 family members may be involved in the processing of the nascent P. carinii MSG to its mature form (48). The genomic organization of PRT1 genes in the subtelomeric regions, contiguous with gene families encoding surface antigens which undergo switching suggests that PRT1 may also be involved in antigenic variation.

The PRT1 family is unique amongst the KEX-like proteases in having both a high proline content and being GPI-anchored. The treatment of P. carinii organisms maintained in short-term culture with protease inhibitors appears effective (49), though it is not known how this effect is mediated. The biologic role of this diverse and intriguing family of proteases is likely to play a pivotal role in the host-parasite interaction, and thus provides fruitful avenues of inquiry into anti-P. carinii therapies.