Recent developments in the molecular, biochemical and functional characterization of GPI8 and the GPI-anchoring mechanism [Review]

Abstract

Glycoconjugates are utilized by eukaryotic organisms ranging from yeast to humans for the cell surface expression of a wide variety of proteins and lipids. These glycoconjugates are expressed as enzymes or receptors and serve a diversity of functions, including cell signaling and cell survival. In parasitic protozoans, glycoconjugates play roles in infectivity, survival, virulence and immune evasion. Among the alternate glycoconjugate structures that have been identified, glycosylphosphatidylinositols (GPIs) represent a universal structure for the anchorage of proteins, lipids, and phosphosaccharides to cellular membranes. Biosynthesis of the GPI is a multi-step process that culminates in the attachment of the assembled GPI to a precursor protein. This final step in the transfer of the GPI to a protein is catalyzed by GPI8 of the putative transamidase complex (TAM). GPI8 functions dually to perform the proteolytic cleavage of the C-terminal signal sequence of the precursor protein, followed by the formation of an amide bond between the protein and the ethanolamine phosphate of the GPI. This review summarizes the current aggregate of biochemical, gene-disruption and active site mutagenesis studies, which provide evidence that GPI8 is responsible for the protein-GPI anchoring reaction. We describe recently published studies that have identified other potential components of the TAM complex and that have elucidated their likely role in protein-GPI attachment. Further, we discuss the biochemical, molecular and functional differences between protozoan and mammalian GPI8 and the protein-GPI anchoring machinery. Finally, we will present the implications of these studies for the development of anti-parasite drug therapies.

• Glycosylphosphatidylinositol
• GPI8
• protein-GPI anchoring mechanism
• transamidase complex

Overview of glycoconjugates and their significance

Glycosylphosphatidylinositols (GPIs) attach a diverse group of macromolecules to the plasma membrane of eukaryotes [1]. In higher eukaryotes, GPIs are found on a minor group of proteins. In parasitic protozoans, the majority of the cell surface proteins are anchored by GPIs. A combination of molecular, biochemical, and mutagenesis approaches has demonstrated that the GPI-anchor consists of a basic core glycan structure that is highly conserved among all eukaryotic organisms (reviewed in [1–3]). The structural differences are represented in the extensive variability in the carbohydrate side chain modifications and lipids among the divergent eukaryotic organisms, as well as between species, strains and/or developmental stages of lower eukaryotes [4], [5]. In addition to GPI-anchored proteins, parasitic protozoans also express GPI-related complex glycophospholipids [6]. Thus, GPI-anchored macromolecules are classified into three main groups, i.e., protein anchoring GPIs, glycoinositol-phospholipids (GIPLs) and GPI-phosphosaccharides [3].

Parasitic protozoans of the Kinetoplastida order, i.e., Trypanosoma brucei species, Trypanosoma cruzi, and Leishmania species, are the causative agents of important human and livestock diseases. Numerous studies support the conclusion that glycoconjugates play critical roles in trypanosomes attachment to host cells, as well as in invasion, differentiation and replication. Examples of GPI-anchored proteins in these protozoans include variant surface glycoprotein (VSG) and procyclic acidic repetitive protein (PARP) of Trypanosoma brucei; several proteins expressed in the infective trypomastigote (1G7, SSP-3, mucins, members of trans-sialidase superfamily) and intracellular amastigote (SSP-4, ASP-1, ASP-2) stages of T. cruzi; and the GP63 and PSA-2 of Leishmania. Other unique GPI-macromolecules that are abundantly expressed in trypanosomes include GIPLs, lipophosphoglycan (LPG), and proteophosphoglycan (PPG) of Leishmania, and GIPLs, phosphoglycan (PG), and lipopeptidophosphoglycan (LPPG) of T. cruzi. A detailed discussion of the biological significance of GPI-glycoconjugates in protozoans, as well as in yeast and mammalian cells, can be found in several reviews [3], [4], [6–17].

Several investigators have utilized genetic approaches to evaluate the overall importance of glycoconjugates in parasitic trypanosomes. In early studies, phenotypic GPI mutants of T. cruzi and Leishmania were generated by episomal over-expression of T. brucei GPI-phospholipase C (GPI-PLC) [18–20]. In both T. cruzi and Leishmania, GPI-PLC-mediated cleavage of GPI intermediates resulted in a depletion of GPI-anchored proteins as well as GIPLs. The GPI-PLC-expressing T. cruzi and Leishmania grew well in axenic cultures as epimastigotes and promastigotes, respectively. However, these mutants exhibited an inability to replicate as intracellular amastigotes and a failure to maintain active infection in animal models of Chagas disease and Leishmaniasis, respectively. The loss of virulence in GPI-PLC expressing T. cruzi and Leishmania suggested the importance of GPIs for completion of their life cycle. Yet GPI-PLC also acts on inositolphospholipids (PIs) [21], which are important second messengers in signal transduction-mediated cellular functions, including nuclear regulatory pathways, in eukaryotic organisms [22–24]. In Leishmania expressing GPI-PLC, PIs were not affected while the potential contribution of PI depletion to the observed phenotype was not examined in T. cruzi. Furthermore, given a general depletion of GPI precursors, these studies did not distinguish between the relative significance of GIPLs versus protein-GPIs in parasite replication and survival. Thus, focus has been placed on identification and functional characterization of the genes involved in GPI biosynthesis and/or attachment (see Figure 1).

Figure 1.  Biosynthesis of glycosylphosphatidylinositol (GPI)-glycoconjugates. Shown is the GPI biosynthesis pathway in eukaryotes. Steps that are unique to the respective organism are indicated as follows: ^yyeast; ^mmammalian cells (reviewed in [134]). GIPLs and the GPI-phosphosaccharides, e.g., lipophosphoglycan (LPG) or lipophosphopeptidoglycan (LPPG), are formed from distinct intermediates in the GPI biosynthesis pathway, designated, respectively as PI₂ and M1 (Manα1-4GlcN-PI) [135]. Specific enzymatic reactions: (1) GlcNAc transferase complex, addition of GlcNAc to PI; (2) GlcNAc-PI-de-N-acetylase, removal of N-acetyl group from GlcNAc; (2A) inositol-acyltransferase, addition of acyl group to GlcN-PI at position 2^y,m; (3) GPI-α1-4mannosyltransferase (GPI-α1-4MT), transfer of mannose to position 4 of GlcN; (F) Flippase, translocation of GPI intermediate from the cytoplasmic to the luminal face of endoplasmic reticulum; (3A) Ethanolamine phosphotransferase (EtNPT-I), addition of EtNP to the first mannose^y,m; (4) GPI-α1-6MT, transfer of mannose to position 6 of the first mannose; (5) and (5A) GPI-α1-2MT, transfer of mannose to position 2 of the second and third^y mannose, respectively; (6) EtNPT-III, transfer of EtNP to the third mannose at position 6; (6A) EtNPT-II, transfer of EtNP to the second mannose at position 6^y; (7) Transamidase complex, cleavage of C-terminal GPI addition signal sequence (ss) of the precursor protein and attachment of GPI. This Figure is reproduced in color in Molecular Membrane Biology online.

With the elucidation of the GPI biosynthesis pathway(s), the specific importance of GIPLs in comparison to protein-GPIs to growth and virulence of parasitic protozoans could be delineated. Studies of several L. mexicana and L. major mutants with partial or complete defects in expression of protein-GPIs [18], [20], [25–28], LPG [11], [12], [29–36], or the depletion of a combination of protein-GPIs, LPG and/or GIPLs [26], [37–39] present a complex portrait of the requirement for distinct classes of GPI-macromolecules in development and virulence of these parasites. Neither Leishmania species depend upon protein-GPIs for axenic growth as promastigotes [18], [26], [27]. In L. major, mutants with selective loss of protein-GPIs, e.g., reduced expression of GP63, or complete loss of LPG expression were less virulent in mice [11], [12], [25], [34]. However, loss of combined LPG/GIPL expression in L. major did not ultimately affect amastigote survival, as disease development occurred, albeit in a delayed fashion [39]. In L. mexicana, mutants with defects in only one class of GPI-macromolecules were capable of infecting macrophages in vitro remained virulent in mice [27], [32], [33]; and mutants in which all classes of GPI-macromolecules were affected did not exhibit reduced macrophage infectivity or virulence [38]. Thus, LPG appears to be more important to virulence in L. major than in L. mexicana (reviewed in [14]) and the amenability of the protein-GPI pathway as a chemotherapeutic target will likely be dependent upon the Leishmania species. In T. brucei, protein-GPIs were found to be essential to survival of bloodstream forms and were required for procyclics to establish infection in tsetse flies [40], [41]. To date, however, no genetic mutants of the T. cruzi protein-GPI, GPI-phosphosaccharide and/or GIPL biosynthetic pathways have been developed. In comparison to the effects of glycoconjugate deficiency in these trypanosomatids, mammalian cell mutants defective in GPI biosynthesis or protein-GPI anchoring machinery are viable [42–47], although an acquired genetic deficiency in humans [48] or disruption of specific GPI-anchored proteins in mice and/or cell lines results in developmental anomalies and selective defects in hematopoietic cells [49–52]. The results of the genetic studies outlined above have important implications, as enzymes involved in GPI biosynthesis and/or protein-GPI anchoring might serve as potential targets for chemotherapeutic interventions of parasitic diseases.

Characteristics of proteins anchored by GPI

Newly synthesized proteins that are destined to be GPI anchored have two signal peptides (see Figure 2). First, the precursor protein contains an N-terminal signal peptide that directs the translocation of the protein across the endoplasmic reticulum (ER) membrane [53–55]. Second, this protein contains a C-terminal GPI anchor addition signal sequence that is cleaved and replaced by the GPI anchor [56]. The analysis of the primary sequences and the signal sequence mutagenesis of proteins that receive a GPI have been used to determine the characteristic features of the GPI anchor addition signal sequence in different organisms [57–64]. These studies have revealed that a GPI addition signal consists of a stretch of three amino acids (ω, ω + 1, ω + 2), followed by a hydrophilic spacer segment of 10–12 amino acids, and a terminal hydrophobic segment of 12–20 amino acids [65]. The ω residue on the N-terminal side of the cleavage point and GPI attachment is restricted to six amino acids with small side chains, the ω + 2 is restricted to relatively small amino acids, whereas ω + 1 accepts every amino acid except proline and tryptophan [59], [62]. The length of the spacer region and the extent of hydrophobicity in the hydrophobic region govern the efficiency of protein-GPI anchoring. These features have been used to create computer algorithms, e.g., DGPI, to predict the presence of C-terminal GPI addition signal in protein sequences [66].

Figure 2.  Characteristic features of the proteins anchored by GPI. Precursor proteins destined to receive a GPI contain two signal sequences, an N-terminal signal sequence and a C-terminal signal sequence, which respectively direct the translocation of the protein to the endoplasmic reticulum (ER) and the anchoring of the GPI to the protein. The C-terminal signal sequence is composed of cleavage attachment (C/A) site of three amino acids, referred to as ω (omega), ω + 1, ω + 2; a short hydrophilic spacer element (designated ---); and a series of hydrophobic amino acids (hydrophobic domain). Cleavage occurs between the ω and ω + 1 site and is followed by GPI attachment [76].

Recombination and mutagenesis studies have indicated that variability exists in the acceptable GPI addition signal sequence between organisms. The GPI addition signal sequence of T. brucei VSG did not function in mammalian cells. This difference was proposed to be due mainly to the amino acids at the ω and ω + 2 sites, since the amino acids present at these sites in parasite proteins are larger than those found in mammalian proteins. VSG, in which ω and ω + 2 amino acids were replaced by those found in decay accelerating factor (DAF), was appropriately processed, anchored to a GPI and displayed on the mammalian cell surface [67], [68]. These studies have suggested that there are differences in the active site of the enzyme involved in GPI transfer to proteins between higher eukaryotes and protozoans [64], [67]. If such is the case, it should be possible to design specific inhibitors of the protozoan protein-GPI anchoring machinery that do not affect mammalian activity and thereby serve as a selective therapeutic drug to treat parasitic diseases.

Biochemical characterization of the protein-GPI anchoring reaction

The early steps in the assembly of GPIs have been demonstrated to occur on the cytoplasmic face of the ER [69–71]. A flippase enzyme is believed to accomplish the change in orientation of GPI intermediates from the cytosolic to the lumenal face of the ER membrane [72–74] followed by assembly of precursor GPI. The subsequent transfer of GPI to protein occurs on the luminal face of the ER where precursor proteins are translocated [55], [70], [75]. The final step of protein-GPI assembly, i.e., the attachment of the preformed GPI to a precursor protein involves: (1) the proteolytic cleavage of the C-terminal GPI addition signal sequence of precursor protein, and (2) the formation of a covalent bond between a carboxyl group of the protein and the amino group of ethanolamine phosphate of the GPI [76]. Based upon biochemical evidence, the two functionalities, endoproteolysis and amidation, are accomplished by a single enzyme, which was later identified to be encoded by GPI8.

Several approaches have been used for the biochemical characterization of the protein-GPI attachment reaction in different organisms. Reporter assays have been utilized to monitor the steps involved in GPI-anchoring using precursor proteins, either endogenous protein substrates [77–80] or reporter proteins such as placental alkaline phosphatase (PLAP), in yeast and mammalian cells [59], [78] or VSG117 in T. brucei [80]. An in vitro translation assay was developed as a convenient system to assess the post- (or co-) translational processing and GPI-anchoring of the precursor protein in rough microsomal membranes (RMs) of yeast [79] and mammalian cells [81]. The kinetics of precursor protein (prepro-α factor in yeast and prepro-mini-PLAP in mammalian cells) processing to a GPI-anchored mature form was observed by pulse-chase/immunoprecipitation and SDS-PAGE. These studies were the first to show that enzymatic cleavage of the N-terminal ER targeting signal sequence and of the C-terminal GPI anchor addition signal sequence, as well as protein-GPI attachment, occur in microsomes. The inability to detect processed PLAP (i.e., PLAP from which both signal sequences have been cleaved but to which GPI is not attached) in these assays led to a suggestion that removal of the C-terminal GPI addition signal sequence and addition of GPI occur either very rapidly or simultaneously, and thus, were most likely catalyzed by a single enzyme [60], [82]. This idea was sustained by the results of assays utilizing RMs from GPI-anchoring defective mutants; only the N-terminal signal sequence was processed, and the pro-PLAP form accumulated, suggesting that the same enzyme performs proteolytic cleavage of the C-terminal GPI addition signal sequence, as well as the linkage of the GPI to protein [44].

In T. brucei, a microsomal assay using procyclic forms was developed to study the processing and GPI-anchoring of a reporter precursor protein, the non-endogenous VSG variant, VSG117. The fate of the VSG reporter protein could be monitored by metabolic labeling, followed by immunoprecipitation with anti-VSG117 antibody. The immunoprecipitated antigen-antibody complex was resolved by SDS-PAGE and visualized by fluorimetry. Along with the observations similar to those made with yeast and mammalian microsomes, the GPI was incorporated into VSG117 in the absence of an energy source [80], as previously described for endogenous VSG [77]. The activation of a protein carboxyl group and its condensation with an appropriate nucleophile in the absence of an energy source is, generally, accomplished by a transamidase [83–85]. It was, therefore, suggested that addition of a GPI to protein most likely involves a nucleophilic attack by the ethanolamine group of the GPI on an activated carbonyl on the ω residue of the pro-protein [86]. Subsequently, hydrazine and hydroxylamine, artificial substrates that can act as nucleophilic acceptors in transpeptidase and transamidase reactions, respectively, were added as co-reactants in the cell-free assay system [80], [86], [87]. Incubation of T. brucei RMs with hydrazine resulted in VSG117 release that was optimal at 37°C, suggestive of enzyme-mediated replacement of GPI by hydrazine. Monitoring of this reaction following a complete inhibition of GPI anchor biosynthesis by mannosamine demonstrated that hydrazine was incorporated into VSG, acting directly as a nucleophilic amine to form a hydrazide with the activated VSG-carbonyl intermediate [80], [88]. Similarly, hydrazine and hydroxylamine competed with GPIs in the prepro-mini-PLAP assay with mammalian cell RMs, leading to the formation of mini-PLAP-hydrazide and -hydroxamate derivatives, respectively [86]. The incorporation of hydrazine and hydroxylamine was enzyme catalyzed, as addition of heated or disaggregated RMs, detergent extraction of RMs, or incubation without RMs did not result in GPI attachment to PLAP and VSG [80], [86]. Collectively, these studies provided evidence that protein-GPI anchoring is an enzyme-catalyzed endoproteolysis/amidation reaction in yeast, mammalian cells and trypanosomes.

Identification of the gene (GPI8) involved in protein-GPI anchoring reaction

Both reverse and forward genetic approaches were utilized in different eukaryotic organisms to identify GPI8, the gene that encodes the enzyme responsible for protein-GPI anchoring. The first indication that GPI8 is essential in a protein-GPI anchoring reaction was provided by studies of the yeast Saccharomyces cerevisiae. Several temperature-sensitive (ts), mutant yeast strains, defective in their ability to express GPI-anchored proteins on their surface, were generated via chemical mutagenesis [89]. By transforming the ts mutants with a yeast chromosomal DNA library, the complementing gene that restored the surface expression of GAS1, was isolated, sequenced and named yGPI8. Deletion of yGPI8 in a parental wild-type strain proved to be lethal [90], as would be hypothesized from studies indicating that GPI-anchored proteins are essential in yeast [91–95]. The human homologue of yGPI8, hGPI8, was able to rescue the growth of these ts yeast mutants [90].

Unlike in yeast, mammalian cells appear to be less dependent in vitro upon protein-GPIs. Thus, viable mutants defective in GPI-anchoring were derived by anti-DAF-based negative fluorescent cell sorting of N-methyl-N'-nitro-N-nitrosoguanidine pretreated human lymphoblast cells [43]. Analysis of the putative GPI precursors in the mutant lines led to the identification of class K mutants that exhibited markedly increased amounts of mature GPIs, but were deficient in a step preliminary to or associated with protein transfer of assembled GPIs [43], [90]. The wild-type hGPI8 and the chimeric yGpi8 in which C-terminal residues were replaced by hGPI8 peptide, but not the wild-type yGPI8, complemented the class K mutants, restoring the surface expression of GPI-anchored DAF [45]. These studies led to a conclusion that: (i) GPI8 plays a role in protein anchoring of GPI, and (ii) hGPI8 is a functional homologue of yGPI8. However, yGPI8 lacks certain features of the human homologue.

In L. mexicana, T. brucei, and T. cruzi, the GPI8 genes were identified by a forward genetics approach. A L. mexicana GPI8 null mutant (LmΔGPI8), generated by targeted disruption of GPI8 alleles, was shown to be deficient in surface expression of the major GPI-anchored protein, GP63, that was secreted in unanchored form [27], [96]. Expression of the non-protein-linked glycoconjugates (LPG and GIPLs) was unaffected. This mutant was viable in the promastigote stage and was infective to animals, which showed that GPI-anchored proteins are not essential for L. mexicana development. Disruption of GPI8 in T. brucei (TbΔGPI8) resulted in a loss of GPI-anchored proteins (e.g., procyclin) from the cell surface of procyclic forms and lead to the accumulation of precursor GPIs. The ΔGPI8 procyclics, while viable in culture, were unable to establish infections in the tsetse mid-gut, confirming that GPI-anchored proteins are essential for insect-parasite interactions [41]. In T. brucei bloodstream forms, the double-stranded RNA interference (dsRNAi) approach was used for functional studies. In this system, TbGPI8 mRNA depletion was achieved by tetracycline induction of dual T7 promoters driving the production of dsRNA. GPI8-depleted bloodstream forms of T. brucei showed an accumulation of precursor GPIs, unanchored VSG and severe growth deficiency associated with multiple nuclei and kinetoplasts, indicative of cytokinesis arrest. Similar growth arrest phenotype was observed in bloodstream form of T. brucei mutants defective in other GPI-biosynthesis genes (e.g., GPI10 [40]). These results indicate that GPIs are essential for the viability of bloodstream-form trypanosomes and imply that the GPI pathway is a valid drug target in T. brucei [27], [41].

A search of public genomic databases revealed that GPI8 homologues exist in apicomplexan parasites (Plasmodium falciparum, Toxoplasma gondii), plants (Arapidopsis thalania), and other eukaryotes (Caenorhabditis elegans, Drosophila melanogaster) [97]. These genes have not, as yet, been functionally characterized. Further investigations of their activity will likely increase our understanding of the protein-GPI anchoring mechanism, substrate specificity as well as the overall function and essentiality of GPI-anchored proteins in other organisms.

GPI8 protein: its role in protein-GPI anchoring reaction

The yeast and human GPI8s contain a hydrophobic N-terminal signal sequence, a luminally-oriented middle segment, and a C-terminal transmembrane sequence [45], [90]. The L. mexicana, T. brucei and T. cruzi GPI8s exhibit ∼31% sequence homology to yeast and human GPI8s. An N-terminal ER targeting signal sequence is present; however, protozoan GPI8s lack the predicted C-terminal transmembrane helical domain found in yeast and human GPI8s. The comparative features of GPI8s in different organisms are summarized in Table I. These protein predictions suggest that GPI8s may comprise two general groups: (1) type I transmembrane proteins, e.g., yeast and human GPI8s, and (2) soluble proteins, e.g., L. mexicana, T. brucei and T. cruzi GPI8s.

Table I.  Characteristics of GPI8s.

Organism	Accession no.	ORF (amino acids)	Predicted (detected) protein size in kDa	Predicted (detected) N-glycosylation sites	Transmembrane domain	Reference
S. cerevisiae	P49018	395	47 (50)	3 (3)	Yes	[90]
H. sapiens	Q92643	411	47 (45)	0	Yes	[45]
L. mexicana	AJ242865	349	38 (42)	3 (nd)	No	[96]
T. brucei	AJ308106	319	37 (nd)	1 (nd)	No	[115]
T. cruzi^1	na	325	37 (nd)	1 (nd)	No	[133]

¹TcGPI8 sequence is identified as ‘GPI-anchor transamidase subunit 8, putative’, location in genome at 8643:107468-108445. Abbreviations: ORF, open reading frame; nd, not demonstrated; na, not available.

On the basis of amino acid homology and motifs, GPI8s are classified among the CD clan of cysteine proteases, and are further subdivided into the C13 family, based upon their mechanism of peptide bond hydrolysis [98–100]. This family includes peptidases found in legumes [101–104], mammals [105], [106] and nematodes, e.g., Schistosoma mansoni [107]. The asparaginyl endopeptidases [108], originally identified in leguminous plants, and so named legumains, have been shown to participate in the proteolytic processing of concanavalin A and of other pro-proteins present in their seeds [109–111]. More recently, mammalian homologues of legumain have been identified in lysosomes [105], [106] and may function in the proteolytic processing of bacterial antigens, and possibly, of pro-enzymes [112]. A group of homologues in Schistosoma species, cathepsin B, D, and L and hemoglobinase, cooperate in the degradation of hemoglobin [107]. The presence of conserved residues in the predicted active site of these proteases dictates their catalytic activity (Figure 3). Endoproteolysis, the cleavage of a protein within the peptide chain, by cysteine proteases is believed to occur via general acid: base catalysis. In this mechanism, the conserved histidine (His) and cysteine (Cys) residues defining the catalytic diad of the active site of C13 peptidases function in concert to cleave the peptide bond. First, His acts as a general base (proton acceptor), deprotonating the sulfhydryl group of cysteine protease. Secondly, the sulfur of the Cys's thiol residue mounts a nucleophilic attack on the carbonyl carbon of the peptide substrate [98]. The transfer of acyl from the cysteine protease to a water molecule, which serves as a nucleophile, generates the free acid on the cleaved peptide. In the mechanism proposed for GPI-protein anchoring, the transfer of an acyl group from the thio-carbonyl to the nucleophilic amine of the GPI's ethanolamine results in the formation of an amide bond between the protein and GPI and the release of the free C-terminal signal sequence [87].

Figure 3.  The postulated catalytic mechanism for GPI-anchor attachment by GPI8. Cleavage of the C-terminal signal sequence of the precursor protein is believed to occur in an endoproteolytic reaction in which the peptide bond is cleaved between the ω and ω + 1 site within the protein chain, followed by GPI attachment to the protein in an amidation reaction. Cysteine proteases (enzyme, Enz) perform peptide bond hydrolysis via a general acid: base catalysis, as follows: (1) The sulfhydryl (SH) group of the cysteine (Cys) residue is deprotonated by the histidine (His) residue, which acts as a ‘general base’ or proton acceptor; (2) Nucleophilic attack of the Cys's sulfur on the carbonyl carbon results in the cleavage of the peptide and the formation of an acyl-enzyme intermediate. The transfer of acyl to water (H₂O) that serves as a nucleophile releases the Enz and formation of carboxylic acid on the cleaved peptide. When GPI serves as the nucleophile, an amide bond (*) is formed between the nitrogen in the ethanolamine phosphate (EtNP) of the GPI and the carbonyl group of the protein, completing the attachment of a GPI to the protein. Alkyl groups of the protein are represented as R; R₁ corresponds to the portion of the precursor protein that is GPI-anchored; R₂ corresponds to the peptide or, in the case of protein-GPI anchoring, the C-terminal signal sequence that is released. (Diagram created in ChemSketch 5.0, Advanced Chemistry Development Inc.).

The enzymatic properties of GPI8 were confirmed in the reporter protein-GPI anchoring assay systems. GPI anchoring was lost in T. brucei membranes depleted of soluble ER contents by incubation under high pH conditions [88]. Reconstitution of T. brucei RMs with recombinant LmGPI8 restored the production of VSG-GPI. Upon incubation with hydrazide, reconstituted RMs produced VSG-hydrazide. These results suggested that TbGPI8 was a soluble protein and LmGPI8 functions as a homologue of TbGPI8 to perform protein-GPI anchoring. Further, this activity was sensitive to treatment of either recombinant LmGPI8 or of T. brucei RMs with iodoacetamide, indicative of the importance of disulfide bond formation or of catalysis by sulfhydryl residue(s) in the anchoring reaction [88]. Others showed the physical association between mammalian GPI8 and precursor protein to be anchored by GPI. In vitro translation of a prepro-mini-PLAP variant in mammalian RMs in the presence of N ^ε-(5-azido-2-nitrobenzoyl)-Lys-tRNA resulted in the incorporation of photo-reactive lysine residues. Upon photolysis, the pro-mini-PLAP cross-linked to ER proteins in its immediate vicinity. The demonstration of a photo-adduct between recombinant epitope-tagged (FLAG)-GPI8 and pro-PLAP provided the first direct evidence of an interaction between a precursor protein substrate and the GPI8 [113]. A similar conclusion that GPI8 associated with mini-PLAP was obtained using semi-permeabilized cells treated with the cross-linking agent bismaleimidohexane. While this assay resulted in the cross-linking of PLAP to GPI8 in the parental K562 line, no association was detected in GPI8-deficient class K mutants, providing further strength for the evidence that GPI8 is a necessary component of the protein-GPI anchoring reaction [114]. Finally, enzymatic cleavage of a fluorogenic peptide, acetyl-S-V-L-N-aminomethyl-coumarine, that mimics GPI anchor addition signal sequence by T. brucei lysates and recombinant TbGPI8 provided evidence that GPI8 is indeed directly involved in the proteolytic processing of the GPI addition signal sequence [115].

Taken together, the biochemical studies performed to-date have established that GPI8 catalyzes the endoproteolysis and amidation steps involved in transfer of GPI to proteins. The two steps most likely occur in concert as a transamidation reaction, leading to the suggestion that GPI8-encoded protein is a transamidase [76]. However, in vivo evidence of GPI8 transamidase activity awaits future studies.

Active-site studies of GPI8

At present, no 3D structure has been constructed for any member of the C13 family to facilitate a molecular model for understanding the catalytic and other domains of GPI8. Alternatively, functional studies of purified enzymes as well as site-directed mutagenesis of conserved residues in the putative active site have been used to characterize this group of cysteine proteases. Mutation of the conserved Cys and His residues in legumain resulted in a loss of protease activity, establishing them as critical to the active site of this protein [101]. The role of GPI8 in protein-GPI anchoring based upon mutation of putative active-site residues has been examined for yGPI8 [116], hGPI8 [117] and LmGPI8 [96]. Among potential active-site residues present in yGPI8, two histidine and two cysteine residues were conserved in a majority of C13 family members, including the GPI8s. The experimental studies demonstrating the functional importance of the residues, highlighted in Figure 4, in GPI anchoring are detailed below.

Figure 4.  CLUSTAL W (1.82) multiple sequence alignment of GPI8 in T. cruzi, T. brucei, L. mexicana, S. cerevisiae, H. sapiens, T. gondii and P. falciparum and of legumain in Canavalia ensiformis. Accession numbers: TbGPI8, AJ308106; LmGPI8, AJ242865; yGPI8, P49018; hGPI8, Q92643; TgGPI8, AJ507036; PfGPI8, AJ401201; Legumain, JX0344. Arrows indicate the conserved cysteine (C) and histidine (H) residues. The postulated active site residues for GPI8, based on mutagenesis studies of legumain (plant endopeptidase), are indicated by •. Specific residues that have been examined by mutagenesis and overexpression in the respective organisms are underlined.

Two approaches were applied to examine the potential active-site histidine (H) and cysteine (C) residues in yGPI8 [116]. The conserved residues in yGPI8 were individually mutated to alanine (A), and these alleles were investigated for their ability to: (i) restore viability and/or protein-GPI attachment in conditional, ts ygpi8 mutants or yGPI8 deletion mutants (yΔGPI8), or (ii) produce dominant-negative phenotypic effects on wild-type (parental) yeast strains, e.g., growth impairment, lethality, and/or GPI-anchoring defects [90]. Using the first approach, expression of the yGpi8^C199A and yGpi8^H57A did not re-establish the viability of yΔGPI8 mutants or growth of the ts ygpi8 mutants at 37°C. Using the second approach, expression of these alleles resulted in an accumulation of the precursor GPI lipid CP2, delayed maturation of GAS1 protein, and arrested cell growth. yGpi8^C85A and yGpi8^H54A complemented the growth defect of yΔGPI8 [116].

The corresponding conserved amino acids, Cys92, His164, and Cys206, were investigated as potential active-site residues in hGPI8. The expression of hGpi8^H164A and hGpi8^C206A did not functionally complement the class K mutants to restore the protein transfer of GPIs and cell surface expression of CD59, unlike hGpi8^C92A. The in vitro PLAP assay demonstrated that hGpi8^H164A and hGpi8^C206A alleles were unable to process pro-mini-PLAP to GPI- or hydrazide-linked forms, unlike in class K mutant cells complemented with hGPI8 [117].

Similar to yeast and mammalian cells, putative active-site residues are conserved in Leishmania, T. brucei, and T. cruzi. L. mexicana null mutants (LmΔGPI8) transfected with LmGPI8 or LmGpi8^H63A exhibited decreased secretion and increased cell surface expression of GP63, whereas cells complemented with LmGpi8^H174A or LmGpi8^C216G secreted high levels of GP63. Similar observations were made when LmGpi8^C216G allele was expressed in wild-type L. mexicana; with increased concentrations of selective drug, surface expression of GP63 was diminished in a dose-dependent fashion. The LmΔGPI8/LmGpi8^C94G parasites exhibited an intermediate phenotype, with GP63 being processed and transported to the cell surface as well as secreted into the medium, suggesting that Cys94 residue may be necessary for optimal folding of the GPI8 and/or is directly involved in its interaction with other proteins of the complex [27], [96]. Mutation of His54 in yGPI8 [116] and Cys92 in hGPI8 [117] also led to a partial loss of protein-GPI anchoring activity, suggesting that these residues directly or indirectly affect protein-protein interactions.

Other components of protein-GPI anchoring machinery

While GPI8 is the catalytic component involved in protein-GPI anchoring reaction, it functions in cooperation with other essential proteins as a multi-subunit complex, named the transamidase (TAM) complex. Genetic and biochemical approaches have identified, in addition to GPI8, two protein subunits of this complex, namely GAA1 and GPI16/PIG-T, which are common to all eukaryotic organisms that have been studied to date. GPI17/PIG-S and CDC91/PIG-U are detected in yeast and mammalian TAM, but are absent from trypanosomatids, which, alternatively, possess TTA1 and TTA2. PIG-T is a type I transmembrane protein, and GAA1, GPI17/PIG-S and CDC91/PIG-U have multiple transmembrane domains (Figure 5). We discuss the characteristics of the TAM components in yeast, mammalian cells and trypanosomes.

Figure 5.  The transamidase (TAM) complex. A model of GPI8 interaction with other proteins in the TAM complex is presented for (A) yeast and human cells, and (B) T. brucei. Structures are not drawn to scale. Identification and functional analysis of the indicated proteins is described in the text [124]. Adapted by permission from Macmillan Publishers Ltd: EMBO Journal 20(15):4088–4098, copyright 2001.

Table II.  Other components of GPI-anchoring machinery.

Gene	Organism	Accession no.	Reference
GPI17	S. cerevisiae^1	NP_010722^a	[124]
GPI16	S. cerevisiae	NP_012058^a	[123]
GAA1	S. cerevisiae	X79409^b	[93]
CDC91	S. cerevisiae	L31649^b	[128]
PIG-S	H. sapiens^2	AB057723^b	[124]
PIG-T	H. sapiens	NM_015937^b	[124]
GAA1	H. sapiens	NP_003792^b	[118]
PIG-U	H. sapiens	NM_080476^b	[128]
PIG-U	R. norvegicus^3	AB086841^b	[128]
GAA1	T. brucei	AB111857^b	[120]
TTA1	T. brucei	AB111858^b	[120]
TTA2	T. brucei	AB111859^b	[120]

The National Center for Biotechnology (NCBI) protein^a or nucleotide^b database accession numbers are indicated for ¹yeast, ² human, ³rat or trypanosome genes.

GAA1

In yeast, functional complementation of the GPI-anchoring deficient end2 mutant with a cDNA expression library which resulted in restoration of GAS1 maturation and anchoring to GPI, led to the identification of yGAA1, named for its role in GPI-anchor attachment [93]. GAA1s in other organisms were identified based upon homology to yGAA1. Human GAA1 is 25% identical and 57% homologous to yeast protein and 21% identical and 38% similar to T. brucei GAA1. In yeast and human cells, GAA1 is predicted to encode a ∼68-kDa protein, with an uncleaved, cytoplasmically oriented N-terminal signal sequence, and a large, hydrophilic, lumenal domain, followed by several transmembrane domains (six in yeast and seven in human cells). A C-terminal KEKQS ER retention sequence and two putative, N-linked glycosylation sequences (one confirmed by site-directed mutagenesis/endoglycosidase treatment) are identified in yGAA1. Immunofluorescence staining showed co-localization of GAA1 with an authentic ER resident protein WBP1. Targeted disruption of yGAA1 resulted in a lethal phenotype associated with accumulation of GPI precursor CP2 and GAS1 protein; this phenotype was rescued by episomal expression of GAA1 [93].

In mammalian cells, over-expression of anti-sense GAA1 or disruption of GAA1 resulted in decreased expression of the GPI-anchored reporter proteins (CD8-DAF and Thy-1) and an accumulation of mature GPI anchors [117], leading to the conclusion that protein and GPI biosynthesis are not altered by GAA1 deficiency. Microsomal membranes of GAA1-defective cells were capable of cleaving the N-terminal signal peptide to form pro-mini-PLAP. However, the mini-PLAP-GPI or the mini-PLAP-hydrazide forms were not detected [118], [119], a phenomenon characteristic of defects in protein-GPI anchoring activity.

The T. brucei GAA1 is noted to encode a smaller (461 amino acids) protein, as compared to human and yeast GAA1 (621 amino acid) proteins. However, like human and yeast GAA1s, TbGAA1 contains several transmembrane domains. Disruption of TbGAA1 resulted in a loss of cell surface expression of the major GPI-anchored protein, procyclin, which was restored upon the episomal expression of TbGAA1 [120]. Together, these results provide evidence that GAA1 plays an essential role in the anchoring of proteins to GPI in all organisms studied to date.

GAA1 is not homologous to any known proteins in the global databases to suggest its specific function. In effort to characterize the functional role of GAA1 in protein-GPI anchoring machinery, recombinant FLAG-GAA1 was co-expressed with GST-GPI8 in CHO cells. Co-precipitation of the FLAG-GAA1 and GST-GPI8 by either of the epitope-specific antibodies suggested the interaction of these proteins. A similar observation was made in TbΔGAA1 mutants expressing FLAG-GAA1; GPI8, TTA1, and GAA1 were co-precipitated by anti-FLAG antibody, thus providing evidence for the association of GAA1 with other components of the TAM complex in trypanosomes [120]. To determine which segment of GAA1 is essential for this interaction, a truncated Gaa1 lacking the C-terminal transmembrane segment was expressed in ΔGAA1 CHO cells. Truncated Gaa1 interacted with other subunits (GPI8, PIG-S and PIG-T) and a pro-protein substrate, but failed to co-precipitate a GPI precursor and formed a non-functional TAM complex. In subsequent experiments, a large lumenal domain between 1st and 2nd transmembrane domains was shown to be sufficient for GAA1 interaction with TAM subunits [121]. A proline residue in the C-terminal transmembrane domain of GAA1 was identified to serve as a dynamic hinge, providing a structural basis for GPI interaction with TAM [122], and leading to the suggestion that GAA1 may play a role in recruiting GPI to the TAM complex. In vivo demonstration of this GAA1 interaction has not as yet been published.

GPI16/PIG-T and GPI17/PIG-S

To identify additional components of the yeast TAM complex, Fraering et al. [123] expressed GST-tagged yGPI8 in ΔGPI8 yeast mutants. The multi-protein complex containing yGPI8-GST was extracted from microsomal membranes in digitonin, purified via GST-sepharose affinity chromatography and resolved by two-dimensional, blue-native gel electrophoresis. Mass spectrophotometric analysis of tryptic digests of the proteins co-precipitated with yGPI8 identified GAA1, GPI16 (earlier named YHR188c) and GPI17 [123]. A similar approach led to the identification of PIG-S and PIG-T in class K mutant cells expressing FLAG-GST-hGPI8. The two proteins were co-precipitated with the TAM complex, and later identified to be homologues of GPI17 and GPI16, respectively [124].

GPI16 is predicted to encode a type I membrane protein with two N-glycosylation sites, an N-terminal ER targeting signal sequence, large lumenal sequence, and a single C-terminal transmembrane domain. GPI16 homologues identified in several organisms predict a similar size (531–639 amino acids) and a similar hydrophobicity profile. Only the yGPI16s have a small C-terminal cytosolic extension with an ER retrieval motif, but the proteins of other species may be retained in the ER by other signals or interaction with GAA1 or GPI8. GPI17 and its mammalian homologue PIG-S encode N-glycosylated protein (534–555 amino acids) with a cytoplasmically oriented N-terminal hydrophobic signal sequence, and a large luminal domain. A putative transmembrane domain is predicted at the C-terminus of PIG-S and PIG-T proteins.

Traditional biochemical and molecular approaches were employed for functional characterization of GPI16 and GPI17. Triton X-100 treatment of yeast microsomal membranes released GPI16 in soluble fractions. The sub-cellular fractionation of GPI16 with ER proteins (GPI8 and WBP1), sensitivity of GPI16 to protease-mediated digestion in detergent-treated yeast microsomes, and a decrease in molecular mass by endoglycosidase H treatment provide evidence for GPI16 being an integral ER membrane glycoprotein, with the bulk of the protein oriented toward the lumen of the ER. Similar biochemical approaches concluded GPI17 (∼63-kDa) is an N-glycosylated integral membrane glycoprotein with its N-terminal hydrophobic signal sequence not cleaved. Depletion of either GPI16 or GPI17 resulted in an accumulation of precursor GPI lipid CP2 and unanchored pro-proteins (e.g., GAS1 and CWP1) [123], [125], a phenomenon characteristic of GPI-anchoring-deficient yeast mutants. Deletion of GPI16 or GPI17 was lethal [126]. Both GPI16 and GPI17 were essential for GPI8 stability. While GPI16 was unstable in absence of GPI8, GPI17 was destabilized upon GAA1 depletion [123].

Like yeast, deletion of PIG-S or PIG-T in F9 embryonal carcinoma cells resulted in a loss of cell surface expression of the GPI-anchored protein Thy-1, which was restored by the episomal expression of the respective gene. ΔPIG-S and ΔPIG-T microsomes were unable to process pro-mini-PLAP, i.e., they failed to develop the GPI-anchored or hydrazide-protein forms, and accumulated mature GPIs. Co-immunoprecipitation experiments suggested that PIG-S is a stable subunit of the TAM complex, independent of the expression levels of the other components, while PIG-T was critical for stable expression and association with GAA1 and hGPI8 [124]. Others provided direct evidence of a disulfide linkage between PIG-T and GPI8; GST-tagged cysteine-to-serine (S) mutants of GPI8, mutated in conserved Cys92 and Cys206 (active-site residues) were expressed in GPI8-deficient class K mutant cells. Immunoprecipitation under native conditions, followed by western blotting, identified that GPI8^C92S mutant was incapable of interacting with PIG-T. Similarly, PIG-T^C182S, but not PIG-T^C139S, mutant failed to interact with GPI8, leading to a conclusion that Cys92 of GPI8 and Cys182 of PIG-T are involved in the formation of the disulfide bond. Notably disulfide linkage was not essential for the generation of the tetrameric complex, since the PIG-T^C182S or GPI8^C92S were co-precipitated with other components of the TAM complex. However, PIG-T^C182S and GPI8^C92S alleles only partially restored the surface expression of GPI-anchored proteins in the respective deletion mutant (ΔPIG-T and ΔGPI8) cells. In microsomal membranes of the ΔPIG-T/ PIG-T^C182S and ΔGPI8/GPI8^C92S cells, only trace amounts of GPI-anchoring activity assessed by a mini-PLAP assay was restored. These studies concluded that the disulfide linkage between GPI8 and PIG-T is important, but not essential, for normal GPI transfer activity [127].

Despite identification of GPI17 in affinity-purified TAM complexes [123], others have shown that a majority of GPI17 is not associated with a yeast TAM complex [125]. No stable stochiometric interaction of GPI17 with the GPI8-GAA1-GPI16 complex was noted in yeast detergent extracts, leading to the suggestion that GPI17 may exert its GPI anchoring role without interacting in a consistent manner with other TAM components [125]. A similar conclusion was reached in F9 cells, as in the absence of PIG-S, a stable GPI8-GAA1-PIG-T complex is formed [124].

In trypanosomes, immunoprecipitation with anti-FLAG antibody from TbGPI8-FLAG-expressing parasites enabled identification of a partial amino-acid sequence of a ∼70 kDa protein that was 13% identical and 25% similar to GPI16 [120]. No functional analysis of TbGP16 has been reported to date.

CDC91/PIG-U

Yeast CDC91 (394 amino acids) exhibits 28% identity to PIG-U of mammalian cells, but has not yet been functionally characterized. PIG-U was identified by a novel selection approach. Chemically mutagenized CHO cells were isolated by their resistance to cytolysis by aerolysin, a GPI-anchored protein-binding toxin. One aerolysin-resistant clonal line exhibited reduced surface staining with fluorescent aerolysin and with antibodies to DAF and CD59 proteins. Microsomes isolated from this CHO line, though capable of synthesizing mature GPIs, were unable to form GPI-anchored mini-PLAP. Transfection of hGPI8, hGAA1, PIG-S and PIG-T did not restore GPI anchoring in this CHO mutant line. These results indicated that this mutant was defective in a new TAM complex gene, which was designated class U. Transfection of the class U mutant with a rat cDNA expression library, followed by restoration of aerolysin sensitivity and DAF surface expression, led to the identification of the complementing gene, thereby named PIG-U [128].

PIG-U encodes a highly hydrophobic protein (435 amino acids) with nine potential transmembrane domains. A sequence of 21 amino acids spanning residues 269–289 in PIG-U is highly similar to that in a corresponding region in CDC91 and consists of a 17-residue motif present in fatty acid elongases. Transfection of PIG-U in class U mutants restored the cell surface expression of the GPI-anchored proteins CD59 and DAF. Subsequently, co-immunoprecipitation experiments showed that FLAG-tagged PIG-U associates with other TAM subunits. PIG-U is not essential for the formation of the TAM complex, as the high-molecular weight complexes of GPI8, GAA1, PIG-T and PIG-S were purified from class U mutants. However, the PIG-U-deficient TAM complex is non-functional and lacks protein-GPI anchoring activity [128].

Yeast CDC91, like PIG-U, retains the predicted characteristics of a highly hydrophobic protein with multiple transmembrane domains. Transfection of CDC91 into class U mutants partially restored protein-GPI anchoring, suggesting that it is a functional homologue [128]. When the conserved aromatic amino acids in the elongase motif were mutated to leucine, the mutant recombinant protein failed to restore protein-GPI in PIG-U mutants. It was postulated, but not demonstrated experimentally, that PIG-U is involved in the recognition of long-chain fatty acids in GPI [128].

TTA1 and TTA2

T. brucei procyclics expressing GPI8-FLAG-GST were subjected to Nonidet P-40 treatment, and affinity-purified complexes containing GPI8 were resolved by SDS-PAGE. N-terminal sequencing of the co-purified 40- and 35-kDa proteins led to the identification of two new subunits of the TAM complex, named trypanosomatid transamidase 1 (TTA1) and 2 (TTA2), respectively. TTA1 and TTA2 sequences were identified in the T. brucei genome, and the predicted proteins contained 2 and 6 transmembrane domains, respectively. Targeted disruption of TTA1 or TTA2 in T. brucei procyclic forms resulted in a loss of the surface expression of the GPI-anchored protein, procyclin, and slower growth, which was normalized by transfection of corresponding cDNA [120]. The biosynthesis of GPI precursors was not affected in ΔTTA1 and ΔTTA2 mutants. These studies provide evidence that TTA1 and TTA2 are essential components of the T. brucei TAM complex [120].

TTA1 and TTA2 have no significant sequence homology to PIG-S and PIG-U or any other proteins of known function. Nevertheless, hydrophobicity profiles and membrane orientations of TTA1 and TTA2 are similar to those of PIG-S and PIG-U, respectively [120], [124], [128]. Given that between mammalian and T. brucei cells, the GPI-addition signal sequence of the precursor proteins are not interchangeable and the structures of the precursor GPIs are different in the glycan and inositolphospholipid parts, it is likely that the TAM components that recognize the GPI-addition signal sequence, C/A site, and the precursor GPI in T. brucei are different from those in mammalian cells. Whether TTA1 and TTA2, the two different components of T. brucei TAM complex, are involved in such functions would hopefully be addressed in future studies.

With the publication of the complete genome sequences of L. major, T. brucei and T. cruzi on 15 July 2005 [129–132], other gene homologues of the described TAM complex, e.g., GAA1 in T. cruzi, can now be identified by searching the public databases. We anticipate that this information will facilitate additional genetic or functional studies to evaluate the importance of these published proteins in the TAM mechanism of the respective organisms.

Conclusion

The collection of genetic and biochemical studies presented in this review suggest that GPI8 is responsible for anchoring GPIs to proteins and that GPI8 functions in association with other proteins in the TAM complex. Much is to be learned about the biochemical and physical interaction among the multiple components of the TAM complex, their exact function, and the requirement for such a large complex for GPI-protein anchoring. Nevertheless, substantial differences are noted in substrate specificity and function of the subunits of the TAM complexes of mammalian cells and parasitic trypanosomes. Further, trypanosomes are highly dependent upon protein-GPI anchoring machinery for their survival and virulence in mammalian host. Future studies are, thus, likely to exploit these differences for the development of novel therapeutics against parasitic diseases.

The work in NG's laboratory has been supported, in part, by grants from the American Heart Association, John Sealy Memorial Endowment Fund for Biomedical Research, American Health Assistance Foundation, and National Institutes of Health (AI053098-01, AI054578-01). MZ was awarded the James W. McLaughlin Predoctoral Fellowship at the University of Texas Medical Branch and has been awarded an appointment as a Postdoctoral Research Fellow in the Emerging Infectious Diseases (EID) Fellowship Program administered by the Association of Public Health Laboratories (APHL) and funded by the Centers for Disease Control and Prevention (CDC). MZ thanks Drs Anant K. Menon and Paul J. Brindley for valuable critical input and discussions. Our thanks are due to Mardelle Susman for proofreading and editing of the manuscript.