Novel Protein-disulfide Isomerases from the Early-diverging Protist Giardia lamblia *

Protein-disulfide isomerase is essential for formation and reshuffling of disulfide bonds during nascent protein folding in the endoplasmic reticulum. The two thioredoxin-like active sites catalyze a variety of thiol-disulfide exchange reactions. We have characterized three novel protein-disulfide isomerases from the primitive eukaryote Giardia lamblia. Unlike other protein-disulfide isomerases, the giardial enzymes have only one active site. The active-site sequence motif in the giardial proteins (CGHC) is characteristic of eukaryotic protein-disulfide isomerases, and not other members of the thioredoxin superfamily that have one active site, such as thioredoxin and Dsb proteins from Gram-negative bacteria. The three giardial proteins have very different amino acid sequences and molecular masses (26, 50, and 13 kDa). All three enzymes were capable of rearranging disulfide bonds, and giardial protein-disulfide isomerase-2 also displayed oxidant and reductant activities. Surprisingly, the three giardial proteins also had Ca2+-dependent transglutaminase activity. This is the first report of protein-disulfide isomerases with a single active site that have diverse roles in protein cross-linking. This study may provide clues to the evolution of key functions of the endoplasmic reticulum in eukaryotic cells, protein disulfide formation, and isomerization.

Many secreted proteins have disulfide bonds that are crucial for their structure or function. It has long been known that although the necessary information for folding is dictated by the primary structure of a protein, this process can be exceedingly slow (1). In vivo, the actions of specialized enzymes located in specific compartments of both bacterial and eukaryotic cells catalyze the formation and isomerization of disulfide bonds. In Gram-negative bacteria, the crucial function of disulfide bond formation is sequestered within the periplasmic space between the inner and outer cell membranes and is catalyzed by a group of enzymes belonging to the Dsb (disulfide bond) family. These proteins have a single thioredoxin-like active site consisting of a pair of cysteines in a CXXC motif. In eukaryotic cells, the endoplasmic reticulum (ER) 1 is the only cellular compartment that is sufficiently oxidizing for disulfide bond formation. The ER has a high concentration of the enzymes and molecular chaperones involved in the folding and assembly of proteins (2). The ER-resident enzyme, proteindisulfide isomerase (PDI; EC 5.3.4.1), catalyzes thiol-disulfide exchange reactions. Like Dsb proteins, PDI is a member of the thioredoxin superfamily, but has two thioredoxin-like active sites (CGHC) that are involved in disulfide bond formation and rearrangement reactions (3).
In the case of the protozoan parasite Giardia lamblia, it is likely that PDI activity plays a key role in the folding of outer surface proteins that are central to avoidance of the host immune system and survival of this parasite in the external environment (4). Specifically, the plasmalemma of the trophozoite is coated with an extremely cysteine-rich (12-16%) variable surface protein (5,6). Of the variable surface proteins studied to date, all the cysteine residues are in intramolecular disulfide bonds that are not susceptible to reduction in the native protein (7). This presents a formidable challenge for correct disulfide bond formation and variable surface protein folding in the ER. Interestingly, a large proportion of the cysteine residues are in a tetrapeptide CXXC motif that rarely corresponds to the thioredoxin family active-site sequence. 2 In addition, the two major proteins of the extracellular cyst wall are cysteine-rich (8,9), and disulfide bonds are important to cyst wall integrity (10). In contrast to the variable surface proteins, the cyst wall proteins are linked by intermolecular disulfide bonds (9).
Giardia is a biological fossil, with both eukaryotic and prokaryotic characteristics (11,12). This protozoan has an ER structure, but does not contain certain other internal organelles typical of eukaryotes, such as mitochondria and peroxisomes (13). Furthermore, unlike most other eukaryotes, Giardia uses cysteine instead of glutathione as the ER thiol redox buffer (14). To gain insight into the evolution of thiol-disulfide exchange mechanisms in eukaryotes and because of their likely importance in giardial biology, we characterized genes encoding giardial PDIs (gPDIs). The hypothesis underlying this work was that gPDIs might reflect an early-diverging state of these enzymes in eukaryotic cells. We found three very different genes (gPDI-1, gPDI-2, and gPDI-3) that encode proteins of ϳ26, 50, and 13 kDa with only one CGHC active site and so are unlike most other eukaryotic PDIs. However, this active site is identical in sequence to the two active sites found in eukaryotic PDIs and is distinct from the single active site of the Dsb proteins and thioredoxins (15). The gPDIs can catalyze the shuffling of disulfide bonds and, interestingly, also display significant transglutaminase activity. This is the first report of PDIs with only one thioredoxin-like active site that have diverse enzymatic roles in protein cross-linking.

EXPERIMENTAL PROCEDURES
Identification and Sequencing of PDI Genes-A fragment of the gPDI-1 gene (clone cLM-12a-12t; GenBank TM /EBI accession number AQ049081) was identified from a random genome sequence survey (16). A segment was amplified from genomic DNA with primers PDI-5 (5Ј-GCC AGA GTA CGC TAA GGC C-3Ј) and PDI-3 (5Ј-TTG CGC CGT-TAT AGT CGA G-3Ј), purified, and labeled with 32 P by random priming (Prime It II kit, Stratagene, La Jolla, CA). A full-length clone was identified from a genomic DNA library in Zap II (17). A clone expressing full-length gPDI-2 was isolated by screening a genomic DNA expression library (1-3-kilobase pair inserts) in Zap II with polyclonal antibodies prepared against a protein fraction of ϳ49 kDa from a polyacrylamide gel segment (18). The gPDI-3 gene was identified by 3Ј-rapid amplification of cDNA ends (RACE) with a degenerate oligonucleotide designed against the PDI active site. The complete gene was isolated by screening a Zap II genomic library (3-5-kilobase pair inserts) with a 183-base pair probe generated from genomic DNA with primers PDI-S:1 (5Ј-GAT CTG TCT GAT GAC GCT CCC GAG-3Ј) and PDI3-3Ј (5Ј-CAC GTG GAA TCA CAG CCT CTG-3Ј). The double strand sequences of gPDI-1 and gPDI-2 were obtained by primer walking using the T7 Sequenase Version 2.0 sequencing kit (Amersham Pharmacia Biotech) and that of gPDI-3 by automated sequencing.
Analyses of gPDI Expression-G. lamblia trophozoites (strain WB, ATCC 30957, clone C6) were cultivated and encysted as described (19). Total RNA was isolated from G. lamblia at the indicated stages of differentiation by extraction with RNAzol B (Tel-Test Inc., Friendswood, TX). Samples of total RNA (15 g/lane) were fractionated on 1.5% formaldehyde-agarose gels, downward capillary-blotted in 20ϫ SSC, and immobilized onto nylon membranes (Zeta-Probe, Bio-Rad) by baking in a vacuum oven for 1 h at 80°C. For Northern hybridization, probes corresponding to the open reading frames of gPDI-1, gPDI-2, and gPDI-3 were polymerase chain reaction-amplified, gel-purified, and radiolabeled by random priming (Prime It II kit). Blots were prehybridized in 6ϫ SSC, 5ϫDenhardt's solution, 0.5% (w/v) SDS, and 20 g/ml salmon sperm DNA for 1 h at 65°C. Hybridization at 65°C was con-tinued overnight in the presence of a mixture of the three gPDI probes. The membrane was washed twice in 2ϫ SSC and 0.1% (w/v) SDS at room temperature for 15 min and then once at 60°C for 15 min in 0.2ϫ SSC and 0.1% (w/v) SDS. The washed membrane was exposed to film overnight.
Heterologous Expression of gPDI Gene Products for Enzymatic Studies and Antibody Production-gPDIs were overexpressed as C-terminal glutathione S-transferase (GST) fusion proteins lacking the predicted N-terminal signal peptides. Each gene was amplified from Giardia genomic DNA (QIAGEN Blood and Cell Culture kit) with Pfu polymerase (Stratagene) and subcloned into pGEX-KG (21) using XbaI and HindIII linkers. gPDI-1 was amplified with PDI1Q5Ј (5Ј-CAT CTA-GAG GTT GTC GAG TTA GGC-3Ј) and PDI1Q3Ј (5Ј-TGA AGC TTA T-GG AGC CAC TTC TC-3Ј). For gPDI-2, oligonucleotides PDI2-GEX-5Ј (5Ј-TAT TCT AGA GGT CTT GGT TCT CA-3Ј) and PDI2-GEX-3Ј (5Ј-ATA AAG CTT AGA AGT TCT CAT TGA GCA T-3Ј) were used. For gPDI-3, oligonucleotides PDI3MAL5Ј and PDI3MAL3Ј (see below) were used. The GST-PDI fusion proteins were overexpressed and purified on glutathione-agarose (Amersham Pharmacia Biotech). The purified preparations contained some free GST that did not interfere with the enzymatic activities. Antibodies against each recombinant GST-gPDI fusion protein were raised in New Zealand White rabbits as described (22). Purification of anti-gPDI-1 and anti-gPDI-3 antibodies over a protein A-agarose column did not affect their reactivity. Giardia has neither glutathione nor GST (14), and anti-GST antibodies did not interfere with detection of gPDIs in immunoblots or immunocytochemistry.
PDI Enzymatic Assays-Purified recombinant GST-gPDI-1, GST-gPDI-2, and GST-gPDI-3 fusion proteins were used in each assay. The ability of the recombinant gPDIs to catalyze the refolding of "scrambled" bovine pancreatic ribonuclease type III-A (Sigma; a measure of disulfide isomerization) was determined as described by Hawkins et al. (23). The PDI-catalyzed folding of reduced bovine pancreatic trypsin inhibitor (BPTI; type I-P, Sigma; oxidative formation of native disulfide bonds) was measured as described (24) in 0.2 M Tris-Cl containing 0.1 mM GSSG and 0.2 mM GSH. PDI purified from chick liver (25) was used as a positive control in these assays.
The ability of PDIs to reduce the disulfide bond between insulin chains with 0.33 mM dithiothreitol (DTT) or 1 mM L-cysteine was measured as ⌬A 650 (26). Bovine PDI (PanVera, Madison, WI) and purified Escherichia coli DsbA protein (provided by Dr. J. Bardwell) (27) were used as positive controls.
Peptide Affinity Binding-Photoaffinity labeling of gPDI was performed using an affinity probe specific for the proposed peptide-binding site of rat PDI (residues 451-476): 125 I-labeled monoiodo-N ␣ -3-(4-hydroxyphenylpropionyl)-Asn-Lys-(N ⑀ -p-azidobenzoyl)-Ala-NH 2 (29). Each gPDI was preincubated with the probe in 100 mM Tris-HCl buffer (pH 7.5) for 10 min at room temperature. The cross-linking moiety was activated by exposure to a short-wave source (254 nm, 8 watts) at 1 cm for 2 min. Affinity labeling was assessed by autoradiography of the samples after separation on 10% SDS-polyacrylamide gel. Immunoreactivity to anti-rat PDI antisera was characterized after transfer to polyvinylidene difluoride membrane. Binding was identified using alkaline phosphatase-conjugated goat anti-rabbit IgG, followed by colorimetric detection with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium.
Transglutaminase Assays-Transglutaminase (TGase) activity of recombinant gPDIs was measured in a microtiter plate assay containing 200 l of 100 mM Tris-HCl (pH 8.5), 10 mM DTT, and 10 mM CaCl 2 or EDTA (30). The reaction was incubated at 55°C for 1 h, and TGase conjugation of the amino group donor 5-(biotinamido)pentylamine to dimethylcasein was measured using alkaline phosphatase-labeled streptavidin and p-nitrophenyl phosphate as a reporter.

RESULTS
Identification and Sequencing of PDI Genes-Our initial identification of putative gPDI gene fragments was based on the presence of a classical PDI active site: F(Y/F)APWCGHCK (15,31). Three PDI genes were isolated from Giardia (Fig. 1A). Interestingly, the three gPDIs have only one active site, each of which corresponds exactly to that of "classical" PDIs rather than to the active sites of thioredoxin or bacterial Dsb proteins (Fig. 1B) (15). Other short regions (VDCT and G(Y/F)PT) were also common to many other PDIs (Fig. 1A). 3 Little similarity between gPDI-1 and gPDI-2 was evident after the terminus of gPDI-3 (Fig. 1A). Additionally, the three gPDI sequences have rather similar predicted N-terminal signal peptides, but only gPDI-3 has a potential KDEL-type signal for retention in the ER (KQRL). However, the C-terminal KRKK motif of gPDI-2 conforms to KKXX ER retention signal criteria (Fig. 1A) (32). The gPDIs differ greatly in the sizes of their predicted open reading frames and encoded polypeptides: gPDI-2 (50,431 Da) is similar in size to PDIs with two active sites from higher organisms (15,25,31). With a predicted molecular mass of 25,846 Da, gPDI-1 is slightly smaller than the yeast Mpd1 and Mpd2 proteins (36.4 and 32.4 kDa) (33,34) and similar to the bacterial Dsb proteins (DsbA is 21.5 kDa), which also have one active site (15,27). gPDI-3, with a predicted molecular mass of 12,648 Da, is the smallest eukaryotic protein containing the PDI active site reported to date.
Expression of gPDIs during the Life Cycle of Giardia-Although vegetative trophozoites and encysting cells secrete different cysteine-rich outer surface proteins that must be folded in the ER (15), the levels of gPDI-1, gPDI-2, and gPDI-3 proteins did not change during encystation ( Fig. 2A). Northern analyses showed that early in encystation (5 h), the mRNA levels of each gPDI increased slightly, and they decreased late in encystation (48 h) (Fig. 2B). The sizes of the gPDI-1, gPDI-2, and gPDI-3 transcripts (ϳ0.7, 1.3, and 0.35 kilobase pairs, respectively) were similar to those predicted by the open reading frames, suggesting that the untranslated regions were short, as is typical of other giardial transcripts (35). Indeed, the 5Ј-untranslated regions for gPDI-1, gPDI-2, and gPDI-3 were 3, 4, and 14 nucleotides, respectively, and each transcript begins in an AT-rich region, as do many other giardial genes (36). For gPDI-2, it overlapped the TAA stop codon, which is not unusual for Giardia. gPDI-3 had a slightly divergent polyadenylation signal, and 3Ј-RACE analysis showed that the transcript was polyadenylated 20 nucleotides past the TGA stop codon (TGAttccacgtgtaaactatgctAAAAA . . . ). This supports the idea that the first position is the least stringent nucleotide in the polyadenylation signal (37).
gPDI Enzymatic Activities-The particular thiol-disulfide oxidoreductase activity displayed by PDI is dependent on both the redox status of the substrate and the cellular milieu. We investigated the oxidant, reductant, and isomerization activities of the gPDIs. The ability to refold RNase that has been scrambled (reduced, denatured, and randomly refolded by reoxidation in air) is a common measure of protein-disulfide isomerization (23,28,38). Recombinant gPDI-1, gPDI-2, and gPDI-3 each had significant activity in catalyzing the restoration of RNase activity (Fig. 3A).
The catalysis of disulfide bond formation or oxidation by PDI can be quantitated by measuring the folding of reduced BPTI as the inhibition of trypsin hydrolysis. The rate-limiting step in the folding of BPTI in vitro is the isomerization of two-disulfide intermediates (24). Recombinant gPDI-2 showed a time-dependent increase in BPTI folding, whereas gPDI-1 and gPDI-3 had low to negligible activity (Fig. 3B).
In addition, microsomal PDI and cell-surface PDI can reduce disulfide bonds in the presence of a reducing agent as measured by the precipitation of reduced insulin chains (26). gPDI-2 and bovine PDI catalyzed the reduction of insulin with DTT more rapidly than DsbA (Fig. 3C). Since Giardia lacks glutathione and uses cysteine as its major low molecular mass thiol (14), we asked if cysteine could also support insulin reduction. gPDI-2 catalyzed the reduction of insulin with either DTT or cysteine. 4 gPDI-3 had a low level of activity only with DTT (Fig. 3C), and gPDI-1 was inactive (Fig. 3C). 4 DsbA is an extremely efficient catalyst of disulfide bond formation in the bacterial periplasm, but has poor isomerase activity. E. coli dsbA Ϫ mutants have defects in the functioning of many proteins (including alkaline phosphatase, ␤-lactamase, and OmpA) that require disulfide bond formation in the periplasm for activity (15,27). Alkaline phosphatase activity in the dsbA Ϫ mutant JCB571 is diminished between 5-and 58fold (33). The gPDIs were cloned into the pMAL-P2 vector for targeting to the periplasmic space. gPDI-2 was able to partially complement the dsbA Ϫ mutant JCB571 (Fig. 4). In addition, the assembly of bacterial flagella is dependent on disulfidebonded proteins; hence, dsbA mutants are also defective in motility (27). gPDI-2 was also able to restore motility to a dsbA Ϫ mutant. 4 However, neither gPDI-1 nor gPDI-3 complemented dsbA Ϫ mutants (Fig. 4). Since a eukaryotic PDI must FIG. 3. Recombinant gPDIs have PDI enzymatic activities. Purified recombinant GST-gPDI-1, GST-gPDI-2, and GST-gPDI-3 fusion proteins were used in each assay. f, PDI controls (chick PDI in A and B; bovine PDI in C); q, gPDI-1; E, gPDI-2; ϫ, gPDI-3; Ⅺ (in C only), DsbA; OE, buffer control minus enzyme. A, the ability of the recombinant gPDIs to catalyze the refolding of scrambled bovine pancreatic ribonuclease type III-A (a measure of disulfide isomerization) was determined as described (23). The data shown are from a single experiment that is representative of at least three separate repeats. Incubation times were not the same for all three experiments. B, the oxidative formation of native disulfide bonds was measured as the folding of reduced and denatured BPTI (type I-P) (24). Purified chick liver PDI (25) was used as a positive control in these assays. Each time point was the average of four separate samples for each gPDI or control, and the experiments were conducted twice. C, the ability of PDIs to reduce the disulfide bond between insulin chains was measured as described (26). The data shown are from one experiment that is representative of at least three separate experiments. be capable of reoxidation by DsbB to observe complementation (39), these results for gPDI-1 and gPDI-3 are consistent with the enzymatic data in Fig. 3B.
PDI can have chaperone activity (3,30,40), which seems to facilitate productive folding, or anti-chaperone activity, which diverts substrate proteins into inactive cross-linked aggregates. A peptide-binding site that binds peptides and unfolded PDI substrates with little regard for sequence has been identified by photoaffinity labeling of mammalian PDI (29). This peptide-binding site appears to participate in substrate recognition for both the oxidoreductase and chaperone activities of PDI (40). The affinity probe labeled gPDI-2, but did not label gPDI-1, gPDI-3, or free GST. The labeling was faint compared with that of chick liver PDI, and antibodies against the latter did not react with any of the gPDIs (data not shown). However, the affinity labeling of gPDI-2 suggests that it is similar to the mammalian PDIs in its ability to bind denatured proteins or peptides.
Transglutaminase Activity of gPDIs-Recently, TGase activity of filarial worms was purified, and its encoding gene was analyzed (41). Surprisingly, the gene did not resemble other TGases, but had two PDI active sites. The recombinant filarial protein had both TGase and PDI activities. Mammalian PDI also has TGase activity. Recombinant gPDI-1 and gPDI-3 had high TGase activity, whereas gPDI-2 was less active. In all cases, the TGase activity was calcium-dependent (Fig. 5).
Localization of PDIs in Giardia-Since the three gPDIs all have a putative N-terminal signal peptide, they can be predicted to enter the secretory pathway. gPDI-3 has a potential ER retention/retrieval signal (KQRL), and gPDI-2 has a KRKK motif following a predicted membrane-spanning region (32). Frozen section immunoelectron microscopy with specific antibodies showed that each gPDI localized to the ER (Fig. 6). The protein chaperone BiP uses a KDEL motif to localize to the ER in Giardia (13). These studies suggest that Giardia may have additional mechanisms for protein retention in the ER, as is also true of higher organisms (42). DISCUSSION The evolutionary appearance of the complex endomembranemediated system of protein folding, modification, sorting, and transport in eukaryotes remains a biological mystery in part because of a paucity of useful models. G. lamblia is a valuable model for gaining basic biological insights because it belongs to one of the earliest known lineages to diverge from the eukaryotic line of descent (11). Most protein trafficking pathways are conserved between yeast and man (43,44). However, in evolutionary terms, the divergence time between Giardia and yeast is at least as great as that between yeast and man (11).
Extensive biochemical evidence showed that Giardia lacks mitochondria and peroxisomes (45, 46). However, the statement that this protist "lacks conventional eukaryotic ER" is premature (42). Giardia does have an ER structure that functions in protein transport and secretion (10,13,47). A giardial signal recognition particle receptor subunit (47) and a BiP ER FIG. 4. gPDI-2 can complement a dsbA ؊ mutant. gPDI-1, gPDI-2, and gPDI-3 were expressed as maltose-binding protein fusion proteins in the periplasm of the E. coli dsbA Ϫ mutant strain JCB571. Alkaline phosphatase activity for dsbA ϩ JCB570, dsbA Ϫ JCB571, and JCB571 transformed with the maltose-binding protein-gPDI fusions was measured as described (27). Results are means Ϯ S.D. from four to six separate experiments.  (30). The reaction was incubated at 55°C for 1 h, and TGase conjugation of the amino group donor 5-(biotinamido)pentylamine to dimethylcasein was measured using alkaline phosphatase-labeled streptavidin and p-nitrophenyl phosphate as a reporter. The data shown are means Ϯ S.D. from four separate experiments. HPDI, human PDI.
FIG. 6. Ultrastructural localization of gPDIs to the giardial ER. Trophozoites were processed for cryosection immunoelectron microscopy as described previously (10). Polyclonal antibodies against gPDI-1, gPDI-2, or gPDI-3 were detected with goat anti-rabbit antibodies labeled with 5-nm gold (arrowheads) (10). A and D, gPDI-3; B, gPDI-1; C, gPDI-2. er, endoplasmic reticulum; ad, adhesive ventral disc; pm, plasmalemma; pv, lysosome-like peripheral vacuoles. Bar ϭ 0.1 m. chaperone gene (13) localize to the ER and clearly resemble their counterparts in higher eukaryotic cells. In contrast, the gPDIs, which are also involved in secretory protein maturation, are unlike their counterparts in higher eukaryotic cells. They are novel in being very small and having only one canonical active site. Although sequence similarity is largely restricted to the active site region, all three gPDIs have disulfide isomerase activities characteristic of PDIs, and gPDI-2 can also oxidize and reduce disulfide bonds.
Although the existence of PDIs has been documented for many years, their size, duplicated active sites, and multiple domains have delayed the in depth understanding of structurefunction relationships. A central question is, why do PDIs need two active sites? It has been reported that the two active sites have different catalytic properties (49) and that both sites are not required for oxidative activity (50). However, isomerase activity is greatest in the presence of both active sites, although a 21-kDa C-terminal PDI fragment can unscramble RNase (51). This is interesting because the three gPDIs, each with only one active site, all showed isomerase activity. From nonreducing Western analysis, the gPDIs do not appear to be disulfide-bonded homodimers, 5 unlike other PDIs (1,52).
In addition, proteins with three CGHC active sites are known (31); and recently, two yeast proteins with a single CGHC (yeast Mpd1) and CQHC (yeast Mpd2) active site have been shown to partially rescue lethal PDI deletions (33,34). All of the gPDIs have only one thioredoxin-like active site. gPDI-2 appears to be a "true" PDI in the sense that it can make, break, and shuffle disulfide bonds. gPDI-3 is especially intriguing because, although it is only 13 kDa, it could both renature scrambled ribonuclease (isomerization activity) and slowly reduce disulfide bonds. Perhaps gPDI-3 is equivalent to bacterial DsbC, which is primarily involved in disulfide bond isomerization (15). Alternatively, in Giardia, the transglutaminase activities of gPDI-1 and gPDI-3 may be more important than their disulfide exchange properties because PDIs can serve multifunctional roles in the cell (e.g. Refs. 2, 3, and 53). Expressing the gPDIs in a bacterial system may be responsible for the low thiol-disulfide activities of gPDI-1 and gPDI-3. It is not known why multiple PDI-like proteins are present in Giardia, although in yeast, one PDI and three PDI-related proteins exist (Eug1, Mpd1, and Mpd2), and Gram-negative bacteria have multiple thiol-disulfide oxidoreductases (15,53).
PDI biological activities depend on redox cycling of the vicinal active-site cysteines between dithiol and disulfide states (44). In vitro, the balance is maintained by the redox buffer, usually GSH/GSSG. In vivo, the GSH/GSSG balance of the ER in mammalian cells, which promotes disulfide bond formation, is more oxidizing than that of the cytosol (54). However, yeast strains unable to synthesize GSH can form disulfide bonds in the presence of DTT. A protein called Ero1p (55, 56) may play a role similar to that of DsbB from E. coli. It will be of interest to determine if a protein resembling Ero1p is also present in Giardia. Many bacteria and virtually all eukaryotic cells use GSH (or related cysteine conjugates) as their major low molecular mass thiol redox buffer. In contrast, only certain bacteria use cysteine (57). However, it is striking that among eukaryotic cells, the absence of GSH and the use of free cysteine are known only for the amitochondriate protozoa, Entamoeba, Trichomonas, and Giardia (14,58), which are not closely related. This may be part of a larger pattern of convergent evolution since these parasites colonize human mucosa and share other unusual metabolic pathways. Our studies may help reveal whether these are early-diverging pathways that have been selectively retained or later adaptations to a similar parasitic niche. The use of cysteine, rather than glutathione, by gPDIs may also have functional implications because cysteine is much less stable than GSH.
In contrast to disulfide bonds, certain proteins are crosslinked by isopeptide bonds that cannot be reduced and are extremely resistant to proteolysis. Isopeptide bonds are formed by TGases that catalyze a Ca 2ϩ -dependent acyl transfer reaction in which an amide bond is formed between the ␥-carboxyamide group of a glutamine residue in a specific target protein and certain primary amines. If the physiological amine substrate is a lysine of a second protein, the reaction results in formation of ␥-glutamyl-⑀-aminolysine protein cross-links. In metazoan organisms, such isopeptide bonds are crucial to structures as diverse as thrombin clots, cataracts, and skin (59). TGase activity of filarial worms is required for both egg formation and molting and may lead to the cross-linking of worm proteins to host proteins. Surprisingly, purified filarial worm TGase is encoded by a gene with homology to PDIs, and mammalian PDI also exhibit TGase activity (41). Since Giardia is such an early-diverging eukaryote, our finding that three PDIs with very different sequences also have substantial TGase activity suggests that most, if not all, PDIs may be able to catalyze the formation of two very distinct types of protein cross-links. The gPDIs do not have the consensus TGase active site: YGQCWVF (59). Because of its small size, gPDI-3 may be a good candidate for investigating the novel TGase activity of PDIs. We speculate that isopeptide bonds may play a crucial role in the resistance of the giardial cyst wall to degradation.
The existence of PDIs with a single active site influences overall phylogenetic analyses of the thioredoxin-like domains of eukaryotic PDIs. Without yet resolving the phylogenetic root, it is simplest to hypothesize that two-and three-domain PDIs arose from a single domain ancestor (53), of which gPDI-2 and gPDI-3 appear to be direct descendants. In contrast, our preliminary phylogenetic analyses 6 suggest that the yeast single-domain PDIs and gPDI-1 appear to be secondarily derived from ancestors with two domains, i.e. these genes branch within a clade of two-domain PDIs. However, because gPDI-1 is so divergent from the other gPDIs, its placement is not entirely clear at present. We, like Kanai et al. (53), find loss and duplication of thioredoxin-like domains common during evolution of the various eukaryotic PDIs. The single-domain PDIs of G. lamblia appear key to understanding the origin of two domain PDIs within the Eukaryotae.
One of the major functions of the ER is the folding and formation of disulfide bonds of membrane and secreted proteins. If correct disulfide bond formation is prevented by treating living cells with the membrane-permeable reductant DTT, thereby interfering with the ER redox environment (60), further protein modifications and transport are blocked or delayed. In yeast and higher eukaryotic cells, protein glycosylation in the ER is dependent upon correct S-S bond formation (60). Giardia may be a simpler model for the study of endomembrane systems, as it appears to have relatively little glycosylation of secreted proteins (35). Nonetheless, DTT prevents disulfide bond formation in cyst wall proteins and blocks differentiation into the cyst form in a reversible manner. 7 This supports a critical role for PDI activity in the ER during giardial differentiation.
Our analyses of a novel group of ER protein-folding enzymes has revealed significantly less complexity than the corresponding enzymes from later-diverging eukaryotes. Studies of the gPDIs may give valuable insights into the divergence of the Dsb and PDI members of the thioredoxin superfamily and into the appearance of the eukaryotic endomembrane system. It will also be of great interest to examine the functions of gPDI sequences outside the active-site regions because there is great diversity between the three protein sequences. The crystal structures of both DsbA (61) and thioredoxin (62) have been resolved. Despite little homology between the primary amino acid sequences, the two proteins have a strikingly similar "thioredoxin fold" formed by discontinuous parts of the primary amino acid sequence. The CXXC active site lies at the base of this fold, with the N-terminal cysteine thiol group able to form a covalent intermediate with the substrate. Apart from the thioredoxin-like active site, PDIs share little homology with thioredoxins or Dsb proteins. Since the gPDIs are smaller than other PDIs, especially gPDI-3, which is only ϳ13 kDa, they may be good candidates for crystallization and detailing structure-function relationships of eukaryotic disulfide exchange proteins. gPDIs may be "minimal" enzymes and possibly represent an ancestral PDI. Therefore, our studies could provide clues to the evolution of the ER and its specific functions.