Biology of Giardia lamblia

An appropriate classification for Giardia spp. is critical to an understanding of the pathogenesis and epidemiology of infection, as well as the biology of the organism. This process has been difficult for a number of reasons. (i) The (presumed) asexual nature of the organism does not allow mating experiments to allow species designation. For clonal organisms in the same clade, there are no well-defined criteria for species designation; these designations remain controversial. (ii) Many of the earlier descriptions ofGiardia spp. assumed a different species for each host and consequently overestimated the number of species. Subsequent species descriptions based on morphologic differences detected by light microscopy have probably underestimated the differences among isolates, strains, or species. (iii) Cross-transmission experiments ofGiardia from one host to another have yielded inconsistent results. (iv) The available tools for distinguishingGiardia isolates have been inadequate until the recent introduction of molecular and electron micrographic techniques for classifying Giardia spp. In view of these concerns, a review of the history of the description of Giardia and the designation of Giardia species is warranted.

Giardia was initially described by van Leeuwenhoek in 1681 as he was examining his own diarrheal stools under the microscope (66). The organism was described in greater detail by Lambl in 1859, who thought the organism belonged to the genusCercomonas and named it Cercomonas intestinalis(172). Thereafter, some have named the genus after him while others have named the species of the human form after him (i.e.,G. lamblia). In 1879, Grassi named a rodent organism now known to be a Giardia species, Dimorphus muris, apparently unaware of Lambl's earlier description. In 1882 and 1883, Kunstler described an organism in tadpoles (?G. agilis) that he named Giardia, the first timeGiardia was used as a genus name. In 1888, Blanchard suggested the name Lamblia intestinalis (29), which Stiles then changed to G. duodenalis in 1902 (314). Subsequently, Kofoid and Christiansen proposed the names G. lamblia in 1915 (165) and G. enterica in 1920 (166). There continued to be controversy about the number of Giardia species for many years, with some investigators suggesting species names on the basis of host of origin and others focusing on morphology. For example, over 40 species names had been proposed on the basis of host of origin (169). Simon, on the other hand, used morphologic criteria to distinguish between G. lamblia and G. murisand accepted the name G. lamblia for the human form (304). In 1952, Filice published a detailed morphologic description of Giardia and proposed that three species names be used on the basis of the morphology of the median body:G. duodenalis, G. muris, and G. agilis(104). The species name G. lamblia became widely accepted through the 1970s. Since the 1980s, some have encouraged the use of the name G. duodenalis, and in the 1990s, the name G. intestinalis has been encouraged by other investigators (170). At this time there does not appear to be adequate reason to abandon the term G. lamblia, which has been widely accepted in the medical and scientific literature.

G. lamblia is pear shaped and has one or two transverse, claw-shaped median bodies; G. agilis is long and slender (100) and has a teardrop-shaped median body; and theG. muris trophozoite is shorter and rounder and has a small, rounded median body. G. lamblia is found in humans and a variety of other mammals, G. muris is found in rodents, andG. agilis is found in amphibians (Table1).

For the Giardia isolates grouped with G. lamblia on the basis of morphologic criteria discernible by light microscopy, differences that can be detected by electron microscopy have allowed the description of additional species, G. psittaci from parakeets (78) and G. ardeaefrom herons (81). Another species, Giardia microti, has been suggested on the basis of host specificity for voles and muskrats, differences in the cyst as assessed by electron micrography (97), and by differences of the 18S rRNA sequences compared with G. lamblia of human origin (331).

Molecular classification tools have been of great value in understanding the pathogenesis and host range of Giardiaisolates obtained from humans and a variety of other mammals. The first study of the molecular differences of G. lamblia isolates (26) was a zymodeme analysis of five axenized isolates, three from humans, one from a guinea pig, and one from a cat, using six metabolic enzymes. Zymodeme analysis consists of the typing of organisms based on the migration of a set of enzymes on a starch gel in the presence of an electric field. The migration depends on the size, structure, and isoelectric point of these enzymes. Since these properties are a function of the primary amino acid sequence, differences in the zymodemes should reflect differences in the sequences of the genes encoding these enzymes. In 1985, restriction fragment length polymorphism analysis of 15 isolates was performed using random probes (251). These studies resulted in the description of three groups; group 3 was so different from groups 1 and 2 that the suggestion of a separate species designation was made. Subsequently, a number of other molecular classification studies have been performed using zymodeme analysis (2, 16, 47, 212, 214, 234, 277, 316) and restriction fragment length polymorphism analysis (65, 85-87, 90, 137). Pulsed-field gel electrophoresis (PFGE) chromosome patterns have also been studied (7, 42, 143, 167, 291) but are of limited value for classification because of the frequent occurrence of chromosome rearrangements (4, 179). Likewise, classification by surface antigens (250) is limited by antigenic variation of the variant-specific proteins (VSPs) (6, 243). These studies have been very useful, but the conclusions that can be drawn from these types of data are limited by the semiquantitative nature of the data. To allow a more quantitative comparison of Giardiaisolates, sequence comparisons of the small-subunit rRNA, triosephosphate isomerase (tim), and glutamate dehydrogenase (GDH) genes have been utilized in a number of subsequent studies (17, 92, 193, 229-231).

These studies have all confirmed the division of G. lambliahuman isolates into two major genotypes (Table2). The first consists of Nash groups 1 and 2, Mayrhofer assemblage A, groups 1 and 2, and the Polish isolates, while Nash group 3, Mayrhofer assemblage B, groups 3 and 4, and the Belgian isolates form the other major genotype. For example, thetim nucleotide sequences of the group 1 and 2 isolates diverged by 1% in the protein coding region and 2% in the flanking regions, while groups 1 and 3 diverged by 19% and the flanking regions were so dissimilar as to preclude their alignment (193). The small-subunit or 18S rRNA (SS rRNA) sequence shows a 1% divergence between groups 1 and 3 (333), reflecting its more highly conserved sequence. In addition to their marked genetic differences, the two genotypes may have a number of important biologic differences. For example, the GS isolate (group 3) was significantly more pathogenic in infections of human volunteers than was the WB isolate (group 1) (249). Group 3 organisms also appear to grow much more slowly in axenic culture than do genotype 1 organisms (155).

More recently, a number of additional assemblages (genotypes) have been proposed for Giardia isolates from a variety of mammals. These isolates are morphologically identical to human G. lamblia, but sequences of their protein-coding regions differ. These studies have allowed the identification of a dog isolate that is genetically distinct from human G. lamblia. Giardiaisolates from dogs have been notoriously difficult to axenize, in comparison to isolates from humans (213), leading to the proposal that Giardia isolates from dogs were different from cat or human isolates. However, a few dog isolates have been axenized and characterized (17). To further evaluate the zoonotic potential of dog Giardia, suckling-mouse infections were established from 11 consecutive infected dogs. On the basis of sequence analysis, these were assigned to two assemblages (C and D) that were quite distinct from assemblages A and B (230). A PCR-based study of nine fecal isolates from dogs found that one of the nine was similar to human isolates while the other eight were different (138). These results suggest that most dog isolates are genetically distinct from those found in humans and have little or no potential for zoonotic transmission.

Separate assemblages (E through G) have also been proposed for hoofed livestock (92), cat (229), and rat (229) isolates (Table 2). This same sequence-based study demonstrated that G. microti was a member of this compilation of seven assemblages (229). Further studies ofGiardia obtained from cattle have demonstrated that some of the isolates belong to the livestock assemblage (assemblage E) while others belong to assemblage A (genotype 1) and thus may have the potential for human infection (259). Assemblages C through G have not yet been isolated from humans, suggesting the likelihood that some genotypes of G. lamblia have a broad range of host specificity that includes humans while others appear to be more restricted in their host range and may not pose a risk of zoonotic transmission. Whether these seven assemblages should be considered separate species should await further data and consensus. To unify the designation of the G. lambliagenotypes, I propose an approach that takes into account all the currently available data. Priority should go to the first classification suggested by Nash et al. (251) in 1985. However, “group” has been used in different ways by different authors. In addition, it is clear from data published after the initial description that Nash groups 1 and 2 are closely related while group 3 is genetically distinct enough to allow consideration of a separate species or subspecies name as suggested (251). The assignment of species status to clonal organisms is problematic (324) and should be delayed until there is general agreement among investigators, but agreement on genotype designation should facilitate a better understanding and coordination of the literature. Therefore, I propose that genotype A be accepted as the designation for Nash group 1 (A-1), Nash group 2 (A-2), assemblage A, and the Polish isolates. Genotype B would then designate Nash group 3, assemblage B, and the Belgian isolates. It may also be appropriate to use a similar designation for assemblages C through G, since these also reflect significant genetic differences as well as host differences.

Whenever members of a grouping of parasitic organisms parasitize different hosts, it is appropriate to address the question of coevolution of the host and parasite. For Giardia spp., the ability to do so has been limited because of the uncertainty about the host specificity of a specific Giardia species or genotype as well as the problem of appropriate classification of isolates. The development of good molecular classification tools and a better understanding of host specificity makes it reasonable to address the possibility of coevolution of different species or genotypes ofGiardia with their hosts. The greater difference betweenG. lamblia and G. ardeae than among the G. lamblia genotypes supports the idea that the divergence betweenG. lamblia and G. ardeae accompanied the divergence between birds and mammals. However, G. lambliaand G. ardeae are closer to each other than to G. muris, the opposite of the expected finding, since mice diverged from other mammals more recently than from birds. Sequence information is not yet available from G. agilis to allow a similar comparison of the amphibian and mammalian Giardiaspecies. The closer-than-expected relationship between G. ardeae and G. lamblia could perhaps be explained if transmission of Giardia between birds and mammals was followed by divergence of the two lineages with their hosts.

Traditionally, all living organisms have been classified as prokaryotes or eukaryotes, and some still argue for retaining the two major divisions (208). However, the most widely accepted classification now utilizes three major divisions, Archaea (archaebacteria), Bacteria (eubacteria), and Eukarya (eukaryotes) (353), which can then be divided into kingdoms. With either classification system, G. lamblia is clearly a eukaryotic organism and has been considered a member of the protozoa, the more “animal-like” of the unicellular eukaryotes. These protozoan organisms have traditionally been classified by their morphology into flagellates, ciliates, amebae (rhizopods), and sporozoa. Thus, G. lamblia was classified with the flagellated protozoans, including the kinetoplastids (e.g.,Leishmania spp. and Trypanosoma spp.), parabasalids (e.g., Trichomonas vaginalis), andDientamoeba (e.g., Dientamoeba fragilis) (185). Giardia has been placed in the order Diplomonadida (two karyomastigonts, each with four flagella, two nuclei, no mitochondria, and no Golgi complex; cysts are present, and it can be free-living or parasitic) and the family Hexamitidae (six or eight flagella, two nuclei, bilaterally symmetrical, and sometimes axostyles and median or parabasal bodies), along with the mole parasiteSppironucleus muris (145) and the free-living organism Hexamita inflata. Some of the higher orders of classification do not appear to be phylogenetically valid, such as the placement of all flagellated protozoans together. However, the family Hexamitidae does appear to be a monophyletic group (see below).

One recent classification system with one prokaryotic and five eukaryotic kingdoms has retained Protozoa as one of the six kingdoms (45), but most recent proposals have suggested abandoning the term “protozoa” in favor of the more general but at the same time more precise term “protista” (56, 57, 350). This review will use the term “protista,” but it should be noted that in the clinical context, G. lamblia is usually referred to as a protozoan.

Recent classifications of the eukaryotic microbial organisms have depended primarily on molecular comparisons. An ideal molecular classification system would be based on a gene(s) that is required for all life and that is sufficiently highly conserved across all forms of life that accurate alignments allow comparison and classification of all organisms. In many ways, rRNA has been the most useful gene for molecular comparisons, because rRNA sequences are highly conserved across life and because the function of the rRNA is so central to the biology of the organism. Therefore, the most widely accepted classification scheme has been based on SS (18S) rRNA sequences. Based on comparisons of SS rRNA sequences, G. lamblia was proposed as one of the most primitive eukaryotic organisms (308), along with T. vaginalis and the microsporidia. The use of SS rRNA sequences to place Giardia as an early-branching eukaryote has been criticized because of the high G+C content of the SS rRNA of Giardia (75%) and becauseGiardia spp. are parasitic organisms; artifacts may be introduced by high rates of mutation accompanying host adaptation. An analysis of the early-branching eukaryotes suggested that the basal position of Giardia was an artifactual result of long-branch attraction due to a greater evolutionary rate ofGiardia (315). Analysis of the large subunit of RNA polymerase II and reanalysis of the eF1 and eF2 sequences also supported the idea of the long-branch artifact (133). However, no such effect was shown for the eRF3 tree (142). In addition, the absence in G. lambliaof the highly conserved N-terminal domain of eRF3 found in other eukaryotes including T. vaginalis suggested the divergence of G. lamblia before the acquisition of the N-terminal domain (142). It should also be noted that the phylogenetic placement of Hexamita, a free-living organism in the same family (as determined by morphologic criteria), avoids artifacts due to high G+C content or parasitism. The G+C content of theHexamita inflata SS rRNA is 51%, and H. inflatawas found to be monophyletic with Giardia (183, 332). The classification of H. inflata with G. lamblia is also supported by a comparison of the glyceraldehyde-3-phosphate dehydrogenase sequences of G. lamblia, Trepomonas agilis, H. inflata, and Spironucleus sp. (287). A comparison of the SS rRNA sequences ofGiardia, Hexamita, Trepomonas, andSpironucleus, all diplomonads, showed that all were phylogenetically related and that the last three comprised one clade while Giardia occupied another clade (46). These sequence-based comparisons have led to the following proposed classification system for Giardia: kingdom Protozoa, subkingdom Archezoa (includes the phyla Metamonada and Microsporidia), subphylum Eopharyngia, class Trepomonadea, subclass Diplozoa, and order Giardiida (includes the families Octomitidae and Giardiidae). In this classification system, the diplomonads are considered to comprise the subclass, Diplozoa, and Hexamita, Trepomonas, andSpironucleus are all members of the other diplozoan order, Distomatida (46).

An examination of the different genes used for the phylogenic classification of Giardia shows that the genes used in transcription and translation generally yield a basal position forGiardia, although artifact due to long-branch attraction has not been ruled out (Table 3). Interestingly, a phylogenetic tree of one of the genes thought to be of mitochondrial origin (cpn60) also suggests thatGiardia is an early-branching eukaryote (107, 286). Of the other groups of genes, the metabolic genes do not yield a clear position, but some suggest a eubacterial origin forGiardia. Cytoskeletal genes give mixed results, with actin giving an early divergence but tubulin yielding a divergence that is later than that of Entamoeba histolytica. Among other classes of genes, HSP70 and cathepsin B phylogenies suggest an early divergence for Giardia. Thus, although the answers are not conclusive, most of the data suggest that G. lamblia and the other diplomonads are among the most basal of the extant eukaryotes.

G. lamblia is a typical eukaryotic organism in that it has a distinct nucleus and nuclear membrane, cytoskeleton, and endomembrane system, but it lacks other organelles that are nearly universal in eukaryotes, such as nucleoli and peroxisomes. In addition,G. lamblia is anaerobic, lacking mitochondria or any of the components of oxidative phosphorylation.

The hypothesis that eukaryotes arose through endosymbiotic events, resulting in the origin of plastids and mitochondria and the use of oxidative phosphorylation as the major source of energy production, has become widely accepted (120). One version of this hypothesis is that a common ancestor of the archaebacteria and eukaryotes evolved a nucleus and cytoskeleton. A descendant then endocytosed a eubacterium, resulting in the development of a mitochondrion. The observation that certain amitochondriate eukaryotes (e.g., G. lamblia and other diplomonads as well asTrichomonas vaginalis) were basal on an 18S rRNA tree led to the suggestion that those organisms descended from Archezoa before the plastid symbiotic event (44, 46). However, the subsequent recognition of mitochondrial genes in the genomes of T. vaginalis (39) and G. lamblia(286) has led to the suggestion that the amitochondrial protists have secondarily lost their mitochondria.

An alternative version of the symbiotic hypothesis is that eukaryotes developed through a symbiotic association of an archaebacterium with a eubacterium (207). In this proposal, the eubacterial endosymbiont provided the pathways for organisms depending on the metabolism of pyruvate by mitochondrial respiration as well as those depending on pyruvate metabolism by pyruvate:ferredoxin oxidoreductase (PFOR), either in the cytosol (G. lamblia) or in hydrogenosomes (T. vaginalis). According to this proposal, the organism leading to G. lamblia subsequently lost the enzymatic components of mitochondrial respiration.

G. lamblia trophozoites obtained from a rabbit, chinchilla, and cat were first grown axenically (in the absence of exogenous cells) in 1970 (222). HSP-1 medium, a subsequent modification reported in 1976, contained phytone peptone, glucose,l-cysteine HCl, Hanks solution, and human serum (223). This study reported the first human G. lamblia isolate, Portland-1 (P-1). P-1 and WB, an isolate obtained from a symptomatic human who probably acquired his infection in Afghanistan (307), belong to the same genotype and have been used for many of the studies of G. lamblia. The growth medium has subsequently been modified, and currently the most commonly used medium is modified TYI-S-33 (159) (Table4). Among the notable requirements for axenic growth are the absolute requirements for a low O₂concentration, a high cysteine concentration, and the requirement for exogenous lipids, which are obtained from the serum component. When kept at 37°C, the trophozoites adhere to the glass wall of the tube in which they are grown. This adherence is dependent on glycolysis and on contraction of the proteins of the ventral disk (99). To date, the other species of Giardia (e.g., G. muris and G. agilis) have not been grown axenically.

Most eukaryotic organisms depend primarily on aerobic metabolism for their energy production. However, certain eukaryotes, includingTrichomonas spp., Entamoeba spp., andGiardia spp., are characterized by their lack of mitochondria and cytochrome-mediated oxidative phosphorylation. They rely on fermentative metabolism (even when oxygen is present) for energy conservation. Glycolysis and its brief extensions generate ATP, with generation dependent only on substrate level phosphorylation. Glucose is not completely oxidized to CO₂ and H₂O as in aerobic metabolism but is incompletely catabolized to acetate, ethanol, alanine, and CO₂. The balance of end product formation is sensitive to the O₂tension and glucose concentration in the medium. The proposed pathways of energy production from glucose and aspartate are shown in Fig. 1 to3.

The metabolism of trophozoites is markedly affected by small changes in oxygen concentration. Under strictly anaerobic conditions, alanine is the major product of carbohydrate metabolism (72, 263, 265) (Fig. 2). Even with the addition of minimal amounts of O₂ (i.e., concentrations of <0.25 μM), ethanol production is stimulated and alanine production is inhibited (263). With further increases in O₂concentration, ethanol and alanine production are inhibited. At O₂ concentrations of >46 μM, alanine production is completely inhibited and acetate and CO₂ are the predominant products of energy metabolism. These oxygen concentrations are likely to be relevant to the intestinal milieu in which the trophozoites replicate since the oxygen concentration in this environment is estimated to vary between 0 and 60 μM. Thus, the pathway of metabolism of pyruvate appears to be altered for differing anaerobic or microaerophilic environments.

Despite the anaerobic metabolism, trophozoites do produce some oxygen free radicals, and a mechanism for detoxification of oxygen is necessary. Detoxification of oxygen free radicals in aerobic organisms is typically accomplished by an enzymatic pathway beginning with superoxide dismutase. Superoxide dismutase activity has been detected by some (188), but other investigators did not detect superoxide dismutase, catalase, or peroxidase activities in G. lamblia (35) and proposed an H₂O-producing NADH oxidase as the major route of oxygen detoxification in G. lamblia (36).

Routine Giardia medium contains 50 mM glucose, and many of the metabolic studies of trophozoites have been performed with that medium. When the glucose concentration is reduced to 10 mM, there is little effect on trophozoite growth (299). At glucose concentrations below 10 mM, replication rates are reduced by nearly 50% and ethanol production is markedly reduced, alanine production is moderately reduced, and acetate production is unaffected. Thus, glucose promotes trophozoite growth but is not absolutely essential. The growth enhancement by glucose was not provided by other sugars.

In T. vaginalis, glycolytic enzymes (glucose to pyruvate) are found in the cytosol but those involved in pyruvate oxidation are localized in a separate organelle, the hydrogenosome (152). In Giardia, there is no metabolic compartmentalization; rather, all the reactions occur in the cytosol or on the cytosolic surfaces of membranes for PFOR-mediated reactions (188).

Glucose supplies the major source of energy derived from carbohydrates (147). Glucose is converted to pyruvate by the Embden-Meyerhof-Parnas and hexose monophosphate shunt pathways (Fig.1). For most eukaryotic and prokaryotic organisms, the conversion of fructose-6-phosphate to fructose-1,6-bisphosphate is an irreversible and regulated step that is catalyzed by an ATP-dependent phosphofructokinase. However, for Giardia(219), as well as for T. vaginalis(221) and E. histolytica (279), this reaction is catalyzed by a pyrophosphate-dependent phosphofructokinase. In contrast to the ATP-dependent enzyme, this enzyme catalyzes a reversible reaction and is not a regulated enzyme (220). The pyrophosphate-dependent phosphofructokinase from Giardia has been cloned and characterized (274, 289). Triosephosphate isomerase, an enzyme that catalyzes a reversible conversion between dihydroxyacetone phosphate and d-glyceraldehyde-3-phosphate, was cloned by complementation in Escherichia coli and characterized (238).

Two enzymes that convert phosphoenolpyruvate into pyruvate, ATP-dependent pyruvate kinase (271), and pyrophosphate-dependent pyruvate phosphate dikinase (PPDK) have been identified (37, 132, 141). There is a potential energy advantage in the reaction mediated by PPDK, since two molecules of ATP can be generated by a coordinated reaction involving PPDK and adenylate kinase. Adenylate kinase (288) converts two ADP molecules into ATP + ADP by a reaction that is essentially energy neutral, and PPDK converts phosphoenolpyruvate plus AMP into pyruvate + ATP, resulting in the net generation of two ATP molecules during the conversion of phosphoenolpyruvate to pyruvate, rather than the one ATP produced by the pyruvate kinase reaction. Their relative roles in glycolysis have not yet been determined, but the higher specific activity of pyruvate kinase suggests that it may play a major role in glycolysis (271).

The conversion of pyruvate to acetyl coenzyme A (acetyl-CoA) (Fig. 3) is catalyzed by PFOR, which utilizes ferredoxin rather than NAD as the electron acceptor (188, 326, 327), in place of the pyruvate dehydrogenase complex found in aerobic eubacteria and eukaryotes. PFOR is also found in E. histolytica(280) and T. vaginalis (352).

Metronidazole is an antimicrobial agent with a broad spectrum of activity against anaerobic bacteria and protozoa. It is activated when its 5-nitro group is reduced by ferredoxin that has in turn been reduced by PFOR, generating a toxic nitro radical. Both metronidazole and sodium nitrite, another respiratory inhibitor thought to act by destroying the iron-sulfur center of PFOR, were toxic to G. muris trophozoites, but only sodium nitrite was toxic to cysts. It was proposed that they had different effects because of the inability of metronidazole to enter the cyst, but this proposal was not directly tested (262). In a study of different G. lamblia isolates, some of which were fully susceptible to metronidazole and some of which were relatively resistant, decreased ferredoxin levels were associated with drug resistance (192).

Acetyl-CoA can be converted directly to acetate by ADP-forming acetyl-CoA synthetase (293), resulting in the production of ATP from ADP as acetyl-CoA is converted to acetate. Alternatively, acetyl-CoA is converted to ethanol, using acetaldehyde as an intermediate, by the bifunctional enzyme alcohol dehydrogenase E (59, 292, 295). Alcohol dehydrogenase E has an acetaldehyde dehydrogenase activity in the amino terminus that catalyzes the conversion of acetyl-CoA to acetaldehyde and an alcohol dehydrogenase activity in the carboxy terminus that converts the acetaldehyde to ethanol.

Amino acids are becoming increasingly recognized as important components of the energy metabolism of G. lamblia. The uptake of aspartate, alanine, and arginine from the extracellular medium, as well as the documentation of glucose-independent metabolism, suggests the potential importance of amino acid metabolism for energy production in Giardia (215, 297, 299).

The arginine dihydrolase pathway is one potential source of energy (74, 297, 300) (Fig 4). This pathway is present in a number of prokaryotic organisms, but among eukaryotes it has been documented only in T. vaginalis(191) and G. lamblia. In the arginine dihydrolase pathway, arginine is converted to ornithine and ammonia with the generation of ATP from ADP by substrate-level phosphorylation. Ornithine is subsequently exported in exchange for extracellular arginine by a transporter mechanism (298).

Aspartate is another potential source of energy. It is converted to oxaloacetate by aspartate transaminase, entering the intermediary pathway, where it is converted to pyruvate via a malate intermediate (215) (Fig. 2).

Alanine also appears to play an important role in allowing trophozoites to adapt to hypoosmotic challenge. With an isoosmolar extracellular environment, the intracellular alanine concentration is 50 mM (161, 270), and with a hypoosmolar challenge, the concentration of alanine rapidly decreases by an active transport mechanism. (In addition to alanine, potassium appears to play a role in osmoregulation of trophozoites [204].) The secretion of alanine occurs via an alanine transporter that also transportsl-serine, glycine and l-threonine,l-glutamine, and l-asparagine (73, 258). This transporter acts as an antiport, exchanging intracellular alanine for these other amino acids from the extracellular environment (298).

Aside from the synthesis of alanine as a by-product of energy metabolism, the only other amino acid for which de novo synthesis has been documented is valine. Thus, Giardia lacks synthesis of most amino acids and depends on scavenging them from the intestinal milieu in which the trophozoite replicates.

One of the notable requirements for axenic growth of G. lamblia trophozoites is the absolute requirement for a relatively high concentration of cysteine (16 mM). Cysteine also provides a partial protection from the toxicity of oxygen that is not seen with other reducing agents, including cystine, and therefore appears to be a specific effect of cysteine (109, 110, 113). Cysteine is not synthesized de novo and is not synthesized from cystine (202). It appears to be imported into the cell by passive diffusion, although active transport may account for some of the acquisition of cysteine (202). The importance of free thiol groups on the surfaces of trophozoites was demonstrated by the toxicity of thiol-blocking agents that are unable to penetrate intact cells (116). This toxicity suggests that these agents are reacting with thiol groups on the trophozoite surfaces, killing the trophozoites. Cysteine appears to be the major thiol group present (34). When trophozoites are metabolically labeled with radiolabeled cysteine, most of the label is incorporated into the VSPs (6, 11), suggesting that these surface proteins may play a role in protection of the trophozoite from oxygen toxicity (see “Structure and biochemistry of the VSPs” below).

The growth of trophozoites predominantly in the duodenum and jejunum initially suggested the possible importance of bile in the growth of Giardia trophozoites. Short-term axenic growth in the absence of serum can be supported by bile (111). The biliary lipids cholesterol and phosphatidylcholine and the bile salts glycocholate and glycodeoxycholate will also support this growth (111). Serum is required for longer-term axenic growth, but it has been shown that the Cohn IV-1 fraction of bovine serum (enriched in alpha globulins, lipoproteins, and growth factors) can substitute for whole serum (194). In fact, insulin-like growth factor II, which is present in fraction IV-1, stimulates trophozoite growth and cysteine uptake. The same fraction from a number of other mammals was also effective in supporting growth, although in some cases antibody depletion was required.

G. lamblia trophozoites do not have the capacity of de novo synthesis of fatty acids (149), with the possible exception of certain minor fatty acids (75). However, free fatty acids are toxic to trophozoites, demonstrating 50% lethal doses of 2 to 12 μM (283). In fact, the toxicity of human milk for G. lamblia trophozoites appears to be mediated through products of milk lipolysis (129, 283). The trophozoites appear to satisfy their lipid requirements by obtaining cholesterol and phosphatidylcholine from the external environment (94, 111, 200). The cholesterol and phospholipids are supplied by lipoproteins, β-cyclodextrins, and bile salts, with transfer of lipids to the parasite surface being facilitated by bile salts (200). It has also been suggested that a low level of endocytosis of lipids occurs (200). Conjugated bile acids appear to be taken up by a carrier-mediated mechanism that includes different carriers for cholyltaurine and cholylglycine (60).

The major fatty acids found in axenically grown trophozoites are palmitic acid, stearic acid, and oleic acid (75). Fatty acid desaturase activity, including desaturation of oleate to linoleate and linolenate, has been documented (75). Arachidonic acid is incorporated into neutral lipids, phospholipids, and a wide variety of cellular lipids (108), while palmitic acid, myristic acid, and oleic acid are transesterified primarily into phospholipids (28), including cellular phospholipids, (e.g., phosphatidylglycerol, phosphatidylcholine, phosphatidylethanolamine, and phosphatidylinositol) (149, 154, 228, 313). Interesterification also occurs with incorporation of conjugated fatty acids into phosphatidylglycerol (108). The toxicity of certain analogs of phosphatidylglycerol for trophozoites has been documented, although the mechanism of this toxicity has not been determined (108).

Isoprenoids are lipids derived from mevalonate that are commonly found in eukaryotic cells. The most notable end product is cholesterol, but isoprenoids are also incorporated into proteins such as the GTP-binding proteins by posttranslational modification. Isoprenylation of proteins has been demonstrated by the incorporation of radiolabeled mevalonate into trophozoite proteins (197). Incorporation of mevalonate and cell growth were inhibited in a reversible manner by competitive inhibitors of 3-hydroxy-3-methylglutoryl (HMG)-CoA reductase. Inhibitors of later steps of isoprenylation permanently inhibited cell growth.

Many of the pathogenic protozoa, including G. lamblia (339), depend on salvage pathways for obtaining purine nucleosides. In addition, G. lamblia(13), as well as T. vaginalis(340) and Tritrichomonas foetus(341), lack pyrimidine synthetic pathways and depends on salvage pathways for obtaining both purine and pyrimidine nucleosides.

Studies of purine metabolism using radiolabeled precursors have demonstrated the incorporation of adenine, adenosine, guanine, and guanosine into nucleotides but no incorporation of components of the de novo synthetic pathway, such as formate, glycine, hypoxanthine, inosine, or xanthine (23, 224, 339). The likely scenario is that the purine nucleosides (adenosine and guanosine) are imported by a transporter with broad specificity for nucleosides as well as deoxyribonucleosides (19, 63). The purine nucleosides are then broken down to the bases by their respective hydrolases (Fig.5). Phosphoribosyl 1-pyrophosphate is synthesized by phosphoribosyl 1-pyrophosphate synthetase (171) and reacts with the salvaged purine bases to produce the nucleoside-5′-monophosphate in a reaction catalyzed by the respective monophosphate phosphoribosyltransferase (PRTase). The GPRTases from most eukaryotes utilize hypoxanthine or xanthine as substrate, but the G. lamblia GPRTase is highly specific for guanine (261, 311), indicating that guanine is the only source for guanine nucleotides. The amino acid sequence of the GPRTase enzyme shows less than 20% identity to the human enzyme (311). The crystallographic structure of the G. lamblia GPRTase has suggested possible reasons for the unique substrate of the G. lamblia enzyme, such as an aspartate substitution for leucine at position 181 (303).

G. lamblia trophozoites also depend on the salvage of exogenous thymidine, cytidine, and uridine for the synthesis of the pyrimidine nucleotides (13, 190) (Fig.6). Uridine is probably imported by the thymidine transporter (62, 89) and broken down into uracil by a hydrolase (13) or phosphorylase (181). Cytidine and uridine can also be imported by a broad-specificity transporter (63), but most cytidine probably enters the synthetic pathway for UTP synthesis rather than being converted more directly to CTP (Fig. 6). The majority of CTP is probably generated from UTP by CTP synthetase (151). A low-affinity nucleobase transporter has also been documented (89). Although the transporter's function has not been determined, it may be involved in export of unusable (e.g., thymine) or excessive quantities of nucleobases.

DNA synthesis also depends on the salvage of exogenous deoxynucleotides. G. lamblia trophozoites lack ribonucleotide reductase and do not incorporate purine bases or nucleosides into DNA. Rather, they depend on the salvage of exogenous purine deoxynucleosides (20) (Fig.7). A single purine deoxynucleoside kinase for the synthesis of deoxynucleotides from their respective deoxynucleosides (deoxyadenosine and deoxyguanosine) and a pyrimidine deoxynucleoside kinase catalyze the production of the pyrimidine deoxynucleotides from the deoxynucleosides thymidine and cytosine (176).

The G. lamblia trophozoites are pear-shaped and are approximately 12 to 15 μm long and 5 to 9 μm wide. The cytoskeleton includes a median body, four pairs of flagella (anterior, posterior, caudal, and ventral), and a ventral disk (Fig.8 and 9). Trophozoites have two nuclei without nucleoli that are located anteriorly and are symmetric with respect to the long axis (Fig. 9). Lysosomal vacuoles, as well as ribosomal and glycogen granules, are found in the cytoplasm. Golgi complexes become visible in encysting trophozoites but have not been confirmed to be present in vegetative trophozoites (117). However, stacked membranes suggestive of Golgi complexes have been demonstrated (174, 309).

G. lamblia trophozoites have two nuclei that are nearly identical in appearance (Fig. 8 and 9). They replicate at approximately the same time (351) and are both transcriptionally active (153) as determined by uridine incorporation into nuclear RNA. Both nuclei have approximately equal numbers of rDNA genes as determined by in situ hybridization using the rDNA probe (153). Both have approximately equal amounts of DNA as determined by the intensity of nuclear staining with 4′,6-diamidino-2-phenylindole (DAPI) (153) or propidium iodide (25), although the possession of equal amounts of DNA by the two nuclei has been questioned (M. Frisardi and J. Samuelson, Woods Hole Molecular Parasitology Meeting, abstr. 263A, 1999, and abstr. 202C, 2000). It is generally assumed that the two nuclei have the same complement of genes and chromosomes, and results from our laboratory using fluorescence in situ hybridization with single-copy genes support this assumption (L. Yu, C. W. Birky, and R. D. Adam, unpublished observations). These features all differ from those of the ciliated protista that are binucleate, such as Tetrahymena and Paramecium(276). These organisms have a smaller micronucleus that contains the genomic DNA but is not transcriptionally active. DNA from the micronucleus is amplified into multiple copies, forming the macronucleus, from which transcription occurs. Giardiaspp. and the other diplomonads appear to be unique in their possession of two nuclei that are identical by the above parameters.

The nuclei of higher eukaryotes have readily visible nucleoli, which are the sites of rRNA transcription. Fibrillarin is a major component of nucleoli and is required for pre-rRNA processing as well as for viability in Saccharomyces cerevisiae (144, 296, 325) Nucleoli have not been identified in G. lamblianuclei, and antibody to G. lamblia fibrillarin diffusely stains the nuclei of G. lamblia, suggesting that rRNA transcription and processing are not localized to certain regions of the nuclei (240).

Axenic cultures have not been synchronized, and the division time is very brief, so studies of cellular and nuclear division have been limited to observations of trophozoite replication based on static microscopy. On the basis of these observations, it has been proposed that during cellular division, the nuclei move laterally followed by nuclear replication, resulting in trophozoites with four nuclei. The trophozoites then divide in the longitudinal plane in such a way as to maintain the left-right asymmetry (48, 104, 153). Electron micrographs of excysting organisms undergoing cytokinesis have shown cells that appear to be joined laterally or with the ventral disks opposing each other (130). Laterally joined cells would be consistent with the above mechanism of cytokinesis, but replication that results in opposing ventral disks would invert the left-right asymmetry with each division. Thus, it is not yet possible to be certain of the morphology of cytokinesis.

A number of drugs that arrest the cell cycle of mammalian cells have been tested for their effect on trophozoites. Colchicine (mammalian G₂+M) and gamma irradiation (mammalian G₂+M) had no effect, while hydroxyurea (mammalian G₂+M) and razoxane (mammalian G₁/S, sometimes G₂+M), arrested the cell cycle in the G₂+M phase (140). To date, axenic cultures of G. lambliahave not been synchronized.

Trophozoites colonize the small intestine of their host, predominating in the mid-jejunum. They attach by their concave ventral surfaces (ventral adhesive disk) to the intestinal wall, where they obtain the necessary nutrients and avoid transport beyond the jejunum. The ventral disk mediates a mechanical attachment not only to the intestinal wall but also to the surface of the container used for axenic growth. Neither cellular invasion nor receptor-mediated attachment has been documented for Giardia spp. Therefore, the cytoskeleton and especially the ventral disk play a key role in the survival of the organism in the intestine of the host.

The ventral disk is a unique and important component of the G. lamblia cytoskeleton. Grossly, it appears as a concave structure with a maximum depth of 0.4 μm covering the entire ventral surface. The edge narrows into a lateral crest, and a more flexible ventrolateral flange (Fig. 8) surrounds the disk (82). The disk contains the contractile proteins actinin, α-actinin, myosin, and tropomyosin (103) as the biochemical basis for the contraction of the disk involved in adherence. Attachment depends on active metabolism and is inhibited by temperatures below 37°C, increased oxygen levels, or reduced cysteine concentrations (109, 113). Ultrastructurally, the ventral disk includes a set of microtubules containing 13 protofilaments and linked to the ventral membrane. These microtubules form the base of the microribbons (dorsal ribbons) that extend nearly perpendicular from the membrane (Fig. 8 and10). The protein constituents of the dorsal ribbons (microribbons) include a set of giardins, which are alpha-coiled-helix proteins approximately 29 to 38 kDa in size. The giardins line the edges of the microribbons but are not found in the microtubules (273). Although 23 forms of these proteins have been separated by two-dimensional electrophoresis, the N-terminal sequences of some variants are identical, suggesting that post-translational modification accounts for some of the forms (273). Several giardins, α₁-giardin (273), α₂-giardin (15), β-giardin (10, 134), and γ-giardin (257), have been cloned, and their sequences confirm the alpha-helical structure of these proteins. The α₁-giardin and α₂-giardin sequences demonstrate approximately 80% identity at the nucleotide and amino acid levels (15), while the other giardins have no significant sequence similarity. Of the two sequences published for the β-giardin gene, the one found in reference (134) is the correct one except that the start codon is 13 amino acids upstream from the reported cDNA sequence (2 amino acids upstream from the reported genomic sequence) (10), yielding a slightly higher predicted molecular mass (R. Adam, unpublished results). A protein with immunologic cross-reactivity with β-giardin, but significantly larger, has also been identified as a cytoskeletal protein (head stalk protein; HPSR2) (206). This 183-kDa protein has a long coiled-coil stalk and an N-terminal hydrolytic domain. Whether it has the same cellular distribution as the giardins has not been reported.

Tubulins are not found in the microribbons but are found in the microtubules (310). The microtubules of the ventral disk, as well as those of the flagella, are presumably composed of αβ-tubulin. Posttranslational modifications of the G. lamblia tubulins include acetylation and polyglycylation (310, 345, 346). Benzimidazoles are compounds used primarily as anti-helminthic agents, but they also have significant in vitro activity against G. lamblia as well as in vivo efficacy for giardiasis (49, 68, 70, 122, 233). Their effect is thought to be mediated by their interaction with β-tubulin.G. lamblia trophozoites exposed to albendazole (or other benzimidazoles) lose their ability to adhere despite normal flagellar movement (70). This suggests a difference between the tubulin found in the ventral disk and that found in the flagella. This, as well as the observation that adherence can occur in the absence of flagellar function (103), suggest that the ventral disk is important for adherence but the flagella are not. With more prolonged exposure (e.g. 24 h) to albendazole, the ventral disk becomes fragmented, with dislocation of the microribbons and microtubules of the ventral disk (49). Substantial electron-dense precipitates can be found in the microtubules and microribbons and to a lesser extent in the other tubulin-containing structures, such as the median body and flagella (49). The pronounced effect of albendazole on the microribbons, despite the apparent lack of tubulin in the microribbons, raises the possibility that the benzimidazoles are reacting with proteins other than tubulins, such as the giardins. Vinculin is a protein that binds α-actinin and mediates the attachment of actin filaments to membrane sites. Its location in attachment regions of the ventral disk has led to the suggestion that it may be involved in the attachment process (241). Further functional studies are required to confirm or refute this proposal.

The trophozoite has four pairs of flagella that begin at two sets of basal bodies that are near the midline and anteroventral to the nucleus. They emerge from the anterior, posterior, caudal, and ventral regions of the trophozoite. Paraflagellar rods extend along one side of the two ventral flagella (102, 135). Nine pairs of microtubules encircle two microtubules to form the flagella (Fig. 8). The flagella appear to be important for motility but not for attachment. In addition, their early emergence through the cyst wall during the process of excystation suggests their importance in excystation (38) (see “Excystation” below).

The median body is a component of the cytoskeleton that is located in the midline and dorsal to the caudal flagella and consists of a group of microtubules in a tight bundle. It is unique toGiardia spp., and its morphology helps define the morphologic characteristics of the different Giardiaspecies (Table 1). G. lamblia trophozoites typically have two median bodies that are shaped like claw hammers. The median body has been proposed as the assembly site for microtubules to be incorporated into the ventral disk (216). Caltractin/centrin, a calcium binding protein that is responsible for basal-body orientation, has been identified in the basal bodies as well as the paraflagellar rods and median bodies (21, 216). A 101-kDa coiled-coil protein has also been localized to the median body by immunofluorescence microscopy (205).

Encystation occurs after organisms have undergone nuclear replication but before cytokinesis; therefore, cysts contain four nuclei. They are approximately 5 by 7 to 10 μm in diameter and are covered by a wall that is 0.3 to 0.5 μm thick and composed of an outer filamentous layer and an inner membranous layer with two membranes. The outer portion of the cyst wall is covered by a web of 7- to 20-nm filaments (79, 80). Four major proteins have been identified in the outer cyst wall, 29, 75, 88, and 102 kDa in size (80). The sugar component of the outer portion is predominantly galactosamine in the form ofN-acetylgalactosamine (GalNAc) (148). Earlier claims that the cyst wall is composed of chitin (N-acetylglucosamine) have been refuted (3).

As the environmentally stable form of the life cycle, cysts have a metabolic rate only 10 to 20% of that found in trophozoites (262). The respiration of both cysts and trophozoites is stimulated by ethanol, while trophozoites are stimulated only by glucose.

The genome of G. lamblia has the features expected of eukaryotic cells, including linear chromosomes flanked by telomeres that are similar in sequence to those of other eukaryotes (TAGGG) (8, 178). Chromosomal DNA of eukaryotes forms chromatin by associating with four core histones (H2a, H2b, H3, and H4) and a linker histone (H1) to form nucleosomes. G. lambliahas all four of the core histones, and these histones are very similar to those of other eukaryotes and have no particular similarity to the histone-like proteins of the Archaea (354). The GC-rich rDNA genes (31, 69), as well as certain protein-coding genes that were sequenced in some of the initial molecular work onG. lamblia, suggested a GC-rich genome (3). However, the G+C content of the contigs that have been assembled as part of the genome project have yielded a GC content of 49%, excluding the rDNA genes (210).

PFGE separations of chromosomal DNA have been used to demonstrate thatG. lamblia trophozoites have five chromosome linkage groups ranging in size from approximately 1.6 to 3.8 Mb (7, 177) (Fig 12). The initial report of PFGE separations of G. lamblia chromosomes identified four intensely staining ethidium bromide-stained chromosomal bands for most isolates (including ISR and WB) and five bands for JH (7). Probes specific for each of the five chromosomal bands of the JH isolate were used to demonstrate that the smallest of the four bands of the ISR and WB isolates represented two comigrating chromosomes.

Bands that stained more faintly with ethidium bromide were shown to be size variants of chromosome 1, first by using several chromosome-specific probes (7) and then by more detailed mapping of the various size variants (4, 139). These studies showed that chromosome 1 varied in size from 1.1 to 1.9 Mb and that the size variation occurred in both subtelomeric regions (4, 139). Previous studies had shown that tandem repeats of the rDNA (ribosomal DNA) unit were adjacent to the telomeric repeat region (TAGGG) (8, 179, 180) and were frequently involved in subtelomeric rearrangements. The rDNA repeat is found on different chromosomes among isolates of the same genotype (8). The rDNA repeats are found on one end of chromosome 1 of the ISR isolate, and about 30% of the size difference is accounted for by the difference in rDNA copy number (4). The remainder of the size difference appears to be due to differences in repetitive DNA near the other telomere. These studies of chromosome 1 have identified at least three (4) or four (139) size variants of chromosome 1 within cloned isolates, leading to the suggestion that trophozoites are tetraploid at least for chromosome 1. Even though the central regions of theGiardia chromosomes appear to be fairly constant (Adam, unpublished), there is substantial variability at the telomeres.

Other studies have shown greater numbers of ethidium bromide-stained bands on PFGE separations (167, 328), but these probably represent size variations of the five chromosomes. One report suggesting that Giardia contained eight chromosomes and that a chromosome of approximately 650 to 800 kb (called chromosome 3) was duplicated (329) most probably represents a description of the size variants of chromosome 1.

The sum of the sizes of the chromosomes (3.8, 3.0, 2.3, 1.6, and 1.6 Mb) (7) yields an estimated haploid genome size of 12.3 Mb, similar to the size estimate of 10.6 to 11.9 Mb obtained by densitometric analysis of the NotI fragments of G. lamblia DNA (93). These values of approximately 12 Mb are compatible with the estimates obtained by analysis of the G. lamblia genome data (www.mbl.edu/Giardia ). As of December 2000, the sum of the lengths of the contigs was approximately 11 Mb, not including the rDNA repeats or the vsp genes, which have not yet been assembled into the contigs. The reason for the discrepant values of 3.0 × 10⁷ (251) and 8.0 × 10⁷ (31) obtained by C₀T analysis is not clear.

In addition to the identification of size variants of chromosome 1, other evidence for polyploidy includes the identification of multiple alleles of repeat-containing vsp genes. The vspA6gene is located on chromosome 4, and at least three different alleles have been identified (356); these three alleles vary in the copy number of a 195-bp repeat but have identical flanking regions and map to a single chromosomal location, indicating the presence of at least three copies of chromosome 4. Similarly, three and four alleles of the vspC5 gene have been identified in different lines of the WB isolate (Fig. 13) (355), indicating a ploidy of at least 3 or 4 for chromosome 5. Thus, the qualitative data suggest a ploidy of at least 4 (or aneuploidy). In contrast, some of the quantitative data have suggested a higher ploidy. Semiquantitative data based on densitometric scanning of PFGE separations suggested a total of 30 to 50 chromosomal DNA molecules (6 to 10 copies of each chromosome) (7). Similarly, the total DNA content of 1.34 × 10⁸ bp (31, 84) divided by the genome complexity of 1.2 × 10⁷ bp yields an estimated ploidy of 10 to 12. A possible clarification of the ploidy of G. lamblia has been provided by fluorescence-activated cell sorter analysis of vegetatively growing and stationary-phase trophozoites as well as cysts, using E. coli chromosomal DNA as a control (25). The results of the fluorescence-activated cell sorter analysis indicate that trophozoites alternate between 4N and 8N (N is the haploid genome), with most of the stationary trophozoites in the 4N state. These results suggest the possibility that trophozoites each contain two diploid nuclei and that DNA replication occurs shortly after nuclear division so that most of the trophozoites are in the G₂ phase of DNA replication. If this interpretation is correct, trophozoites have an effective ploidy of 4, but quantitative measurements of stationary trophozoites would yield a ploidy of 8.

For any organism with a ploidy greater than 1, a certain degree of allelic heterozygosity is expected. The degree of heterozygosity is generally low in sexual organisms because of meiotic recombination but can become very high in asexual organisms, as has been demonstrated for the rotifers (347). Since Giardia spp. are assumed to be ancient asexual organisms, one might expect a very high degree of allelic heterozygosity. However, the degree of allelic heterozygosity in G. lamblia is actually quite low. Although heterozygosity of repeat copy number for the vsp genes has been shown (356, 358), as well as eight nucleotide differences between two different alleles of the vspA6 gene, the degree of heterozygosity identified in 12 kb of sequence of the triosephosphate isomerase (tim) gene from multiple isolates was less than 0.02% (17). The degree of allelic heterozygosity identified in the genome project is also very low (A. McArthur and G. Olson, personal communication). The reason for this low degree of heterozygosity has not been determined. Potential reasons could be unrecognized sex in Giardia or intermittent loss of a nucleus.

The development of transient- and stable-transfection systems has contributed substantially to our understanding of the genetics ofG. lamblia. The initial description of a transfection system consisted of transient expression of luciferase flanked at the 5′ end by a short region of the GDH gene and at the 3′ end by the putative polyadenylation signal (361). In this and in subsequent transfection systems, DNA was introduced by electroporation. Subsequently, stable episomal transfection was performed using the puromycin-resistance pac gene (305) or the neomycin resistance neo gene (318) as selectable markers. Apparently, the bacterial plasmids used as the framework for the transfection vectors contained sequences that functioned as origins of replication in G. lamblia.

An interesting difference between the WB isolate (genotype A) and the GS isolate (genotype B) in the disposition of transfected DNA was identified (305). DNA introduced as a supercoiled plasmid into WB replicated as an episomal vector, while linearized vector was integrated into the WB genome by homologous recombination (305). In contrast, transfection of GS with linearized or circular vector DNA resulted in homologous integration, and stable episomal replication of the vector did not occur. Homologous introduction of DNA into the genome did not result in knockouts with total elimination of the wild-type alleles, presumably because of the polyploid nature of the genome. In addition to the expression of drug resistance markers, modified GFP was expressed (305) and has subsequently been used to study protein transport in trophozoites (76, 127). High levels of luciferase expression have been obtained by using the G. lamblia virus as a vector for introduction of foreign DNA (364, 365). Use of the virus vector resulted in very high levels of luciferase expression for at least 30 days in the absence of drug selection.

Transcription in G. lamblia is distinctly eukaryotic in nature; nonetheless, it has a number of features that are more characteristic of prokaryotes. As in all eukaryotes, the transcript is produced in the nucleus and transported to the cytoplasm for translation. The polyadenylation of transcripts is typical for eukaryotes, but the short 5′ UTRs and general lack of introns are more characteristic of prokaryotes (although introns may be relatively infrequent in the unicellular eukaryotes). Also, in contrast to other eukaryotes, most G. lamblia transcripts do not appear to be capped at their 5′ ends.

Analysis of the transcript of the α- and β-tubulin genes revealed that the 5′ UTR was only 6 nucleotides in length for these genes (160). Subsequent analyses of other G. lambliatranscripts performed by primer extension analysis, S1 nuclease protection, and 5′ rapid amplification of cDNA ends have revealed that other transcripts also have short 5′ UTRs, ranging in size from 0 to 14 nucleotides (5). The one exception to date has been the glucosamine-6-phosphate isomerase B gene, which has two transcripts, a constitutive transcript with a 5′ UTR of 3 to 4 nucleotides and a transcript with a 146-nucleotide 5′ UTR that is induced during encystation (164).

The short 5′ UTRs suggested the possibility that G. lambliapromoters might be located near the beginning of the open reading frames. This possibility was supported by the observation of a set of highly conserved motifs in the 5′-flanking regions of a number of cytoskeletal genes (136). The consensus sequence AATTAAAAA was identified at the site of transcription initiation with the actual initiation site located at the TA (or CA). A second region 20 to 35 nucleotides upstream from the transcription initiation site, CAAAAA(A/T)(T/C)AGA(G/T)TC(C/T)GAA was proposed as a promoter region, and a third consensus sequence, CAATTT, was found at −40 to −70. When the upstream region was tested in a transfection assay, a 44-bp sequence of the GDH gene provided maximal transcriptional activity (305), confirming that the GDH promoter was short and located near the coding region. A deletion and mutational analysis of the GDH promoter region confirmed the importance of the AT-rich region at the initiation site as well as a CAAAT region 34 bp upstream (360). Specific binding of the immediate 51-bp upstream region to a 68 kDa nuclear extract protein was demonstrated by band shift analysis and by UV cross-linking (360). A deletion and mutational analysis of the ran promoter demonstrated that maximal promoter activity was present in the region from -51 to -2 and that regions further upstream did not contribute to promoter function (319). The most important component for promoter activity was the -51 to -20 region; smaller portions of that region gave reduced promoter activity.

In addition to the promoter immediately upstream of the initiation codon, a downstream region of 8 to 13 nucleotides has been suggested as a potential E. coli-like Shine-Dalgarno sequence. The identification of a 15-base sequence in the G. lamblia SS rRNA (GUCCGCGCCCCCGAG) led to a search of the database for complementary sequences in the 5′ regions of protein-coding sequences. Most of the 59 sequences evaluated had a complementary sequence within 300 bp of the initiation codon. The consensus (e.g., CCGGGGGGGGCUU) markedly the increased translation efficiency (up to 5,000-fold) without significantly affecting the rate of transcription when included in a transfection assay (363).

Transcripts have been analyzed to determine whether they have 7-methylguanosine or other caps at the 5′ ends. Total polyadenylated RNA was analyzed to determine if the 5′ ends were susceptible to T4 RNA ligase (363). They were resistant to 5′ phosphorylation unless pretreated with calf intestinal alkaline phosphatase to remove the 5′ phosphates. These results suggested that the RNA was phosphorylated but not capped. Treatment with the decapping enzyme tobacco acid pyrophosphatase did not increase phosphorylation, also indicating a lack of 5′ capping. These results suggest that most polyadenylated RNA does not have a 7′ methylguanosine cap. However, since the studies were done using total rather than transcript-specific mRNA, it is still possible that individual transcripts could be capped. In fact, for the differentially processed transcript of the glucosamine-6-phosphate isomerase B gene, the constitutive transcript with a short 5′ UTR is not capped while the transcript with the longer (146-nucleotide) 5′ UTR expressed during encystation does have a cap, as demonstrated by RNA ligation after treatment with tobacco acid pyrophosphatase (164). Whether this was a 7-methylguanosine or another type of cap was not determined.

Transcripts have relatively short 3′ UTRs with short poly(A) tails typically beginning approximately 10 to 30 nucleotides beyond the stop codon. The sequence, AGTPuAAPy, precedes the poly(A) tail by approximately 10 nucleotides and has been proposed as a polyadenylation signal (3, 273).

Small nuclear RNA molecules are involved in the splicing of nuclear pre-mRNA (snRNA U1, U2, U4, U5, and U6) and pre-rRNA (sno-RNA U3, U8, U14, snR10, and snR30) to produce the mature RNA (256). Many of these molecules from other eukaryotes have a trimethylguanosine cap at the 5′ end. A number of candidate snRNAs from G. lamblia were immunoprecipated by anti-trimethylguanosine antiserum; caps were confirmed by the susceptibility of the antibody reactivity to the decapping effect of tobacco acid pyrophosphatase. The exact roles and identities of the candidate snRNAs have not yet been determined. It will be of especial interest to determine whether splicing of mRNA occurs since introns have not been identified in theG. lamblia genes characterized to date.

The initial characterization of the rRNA and rDNA genes revealed that the rRNA molecules were much smaller than the typical eukaryotic 18S (SS rRNA) or 28S or large-subunit rRNA (LS rRNA) molecules and in fact were even slightly smaller than the 16S and 23SE. coli rRNA molecules (31, 69). The rDNA repeats were correspondingly small, but had an organization and sequence consistent with that of eukaryotes (see “Giardia and other diplomonads as early-branching eukaryotes” above). In addition, other components of the translation apparatus, including elongation factor 1α, elongation factor 2α, eRF1, eRF3, RPB1, and RNA polymerase III (Table 3), are distinctly eukaryotic.

A double-stranded RNA virus was identified in G. lamblia trophozoites following the observation of a 6.2-kb double-stranded RNA contaminating nucleic acid preparations (64, 337). The virus is nonenveloped and icosahedral, with a diameter of 33 nm, and it infects the nuclei with about 200 copies per nucleus (337). G. lamblia virus (GLV) has a 100-kDa major capsid protein which depends on a cysteine protease for cleavage into the mature protein (362). A second open reading frame encodes a 190-kDa RNA-dependent RNA polymerase (338, 349). The majority of axenic isolates either harbor or are susceptible to GLV infection (including isolates from genotypes A and B), while a minority of isolates are highly resistant to infection (225). GLV infection occurs by endocytosis (323), and susceptibility to infection depends on a specific receptor found of the cell membrane surface (302). A second virus has also been identified; it also has a 6.2-kb double-stranded RNA genome but encodes a slightly smaller capsid protein (95 kDa), which differs significantly from the 100-kDa capsid protein of GLV (322).

G. lamblia trophozoites undergo antigenic variation of a family of immunodominant cysteine-rich surface antigens in vitro and in vivo (6, 12, 243). The initial studies of the surface antigens of G. lamblia showed differences among strains by crossed immunoelectrophoresis and enzyme-linked immunosorbent assay (306) and marked differences in the molecular masses of “excretory-secretory products” from different surface-iodinatedG. lamblia isolates (247, 250). These surface antigens varied in number and in size from approximately 50 to 200 kDa in a study of 19 isolates (250). A monoclonal antibody (MAb 6E7) for a 170-kDa surface antigen (initially called CRP170 but now called VSPA6) from the WB isolate was cytotoxic for WB trophozoites but not for isolates expressing other surface antigens (242). WB trophozoites were doubly cloned by limiting dilution and incubated with MAb 6E7; some organisms survived and were totally resistant to the cytotoxic effect of MAb 6E7 (243). One of these cloned lines of organisms (WB1269) had a 68-kDa surface-labeled antigen (initially called CRP68 but now called VSP1269), while another (WB1267) had a 64-kDa surface antigen (VSP1267). A monoclonal antibody (MAb 5C1) reactive with the surface antigen of WB1267 was cytotoxic for WB1267 trophozoites. These trophozoites were incubated with MAb 5C1, and organisms surviving the cytotoxic antibody reacted with neither of the initial MAbs and expressed surface antigens in the 60- to 100-kDa size range. Subsequent data have confirmed that individual organisms express only one VSP at a time (245); the detection of multiple surface-labeled bands in some of the studies resulted from subpopulations of trophozoites expressing several different VSP types. In axenic culture, variation occurs approximately once every 6 to 12 generations for a frequency of 10⁻³ to 10⁻⁴(244).

Antigenic variation has subsequently been confirmed in animal models (12, 118, 119) and in infected human volunteers (248). Cloned WB trophozoites expressing the VSPA6 were inoculated into gerbils, and trophozoites collected from their intestines 7 days later demonstrated populations of trophozoites expressing multiple VSPs ranging in size from 50 to 170 kDa (12). Trophozoites collected 28 days after infection were similar to those collected at 7 days. The lack of apparent change from 7 to 28 days argued against a role for acquired immunity in selecting antigenic variants. However, the results were somewhat different in a mouse model of G. lamblia infection. G. lambliatrophozoites in athymic nude mice and in heterozygousnu/⁺ mice changed VSP type coincident with the development of a humoral anti-vsp antibody response (119). In contrast, scid mice did not develop an antibody response and trophozoites did not undergo antigenic variation. Thus, in the mouse model, the antibody response to the VSP may have been the selective force for antigenic variants.

When human volunteers were inoculated with cloned GS trophozoites by duodenal intubation, four of four became infected with trophozoites expressing VSPH7 (72 kDa). Only 1 of 13 became infected with organisms expressing VSPB6 (200 kDa), suggesting that the ability to survive in the human intestine is greater with one VSP than with another. Whether this is due to selection against one VSP or positive selection for the other has not been determined. The subsequent pattern of change during 22 days after infection showed gradual disappearance of VSPH7 followed by the appearance of other VSPs with different immunoreactivity. The loss of VSPH7 was accompanied by the development of serum antibodies to VSPH7, suggesting the possibility that the immune response led to the antigenic variation.

An alternative or additional possible biological reason for antigenic variation might be adaptation to different intestinal environments. Some evidence for this possibility is provided by the difference in protease susceptibility for different VSP antigen types (252). Trypsin and chymotrypsin were toxic to WB trophozoites expressing VSPA6 (reactive with MAb 6E7). Some organisms survived after exposure to these proteases, and when they were subsequently grown in the presence of trypsin and chymotrypsin, they were no longer susceptible to the cytotoxic effects and expressed alternative VSPs that were not reactive with MAb 6E7.

Antigenic variation has also been documented during the cycle of encystation and excystation (217, 320). In vitro encystation followed by excystation of trophozoites expressing the VSP TSA417 resulted in the loss of TSA417 from the plasma membrane followed by its appearance in the lysosome-like peripheral vacuoles (211, 320). During excystation, the TSA417 transcript was replaced by a variety of other vsp transcripts. This occurred too rapidly to be the result of selection of antigenic variants surviving excystation and may have represented an induced antigenic shift. This antigenic variation during excystation could promote survival by evasion of an intestinal immune response or by adaptation to other important intestinal factors.

Thus, the two major hypotheses regarding the purpose of antigenic variation are (i) evasion of the host immune defense and (ii) enabling of the organisms to survive in different intestinal environments. It should be emphasized that these hypotheses are not mutually exclusive. A better understanding of the biological reasons for antigenic variation would be facilitated if the role of the VSPs was known. It would also be informative to know if the free-living diplomonads such as Hexamita have proteins homologous to the VSPs and if they undergo antigenic variation.

The initial biochemical studies of the VSPs (then called ES products) indicated that they were protease susceptible. They did not bind to a series of lectins, suggesting that they were not glycosylated (247). A portion of the gene for one of these surface antigens was subsequently cloned from an expression library using MAb 6E7. Sequence analysis revealed a cysteine-rich (12%) protein with frequent CXXC motifs (6). Subsequent sequences of a number of vsp genes have shown a number of characteristics common to all vsp genes. They are all cysteine rich (12%), and most cysteines are present in CXXC motifs. When trophozoites are metabolically labeled with radiolabeled cysteine, most of the label is incorporated into the VSPs (6, 11). It has been suggested that these cysteines are present in the form of disulfide bonds, since free thiol groups have not been detected in purified VSPs (14, 266). However, the presence of thiol groups on the trophozoite surfaces suggests that the VSP cysteines may not all be disulfide bonded (116). Whether the level of disulfide bonding changes in vivo with different oxygenation levels of the surrounding environment has not been determined.

After synthesis, the VSPs are transported through the ER (211) to the membrane, where they diffusely coat the membrane (112, 275). Smaller amounts of VSP can also be detected in the lysosome-like peripheral vacuoles, suggesting the possibility that these vacuoles are involved in recycling of the VSPs (211). The VSPs have an approximately 14- to 17-amino-acid signal peptide that presumably represents a signal peptide for its transport through the ER. The signal peptide is cleaved, 14 amino acids from VSPH7 of the GS isolate (201) and 17 amino acids from TSA417 of the WB isolate (14).

The 38 C-terminal amino acids of the VSPs are >90% conserved at the amino acid and nucleotide levels. Each VSP reported to date ends with the amino acid motif CRGKA in both genotype A and B organisms. The conserved C terminus may represent a membrane-anchoring domain. Evidence for this proposal has been reported in work showing a difference in the C termini of membrane-bound and secreted VSP (267). A 3,500-kDa fragment is cleaved from the C terminus of the membrane-associated VSP prior to secretion. The suggested cleavage site is between the K and S of the highly conserved NKSGLS motif (267), which comprises the amino acids 33 to 38 from the C terminus (235). This NKSGLS motif is generally conserved in the genotype A organisms (88) but is not found in the vspH7 gene from GS, a genotype B organism (253).

The vsp 3′ UTRs have the putative polyadenylation signal common to all G. lamblia strains, AGTPuAAPy (3, 273), immediately preceded by the sequence ACTPyAGPuT, which sometimes begins within the termination codon (88, 320) and (Y. M. Yang and R. Adam, unpublished observations).

On the basis of sequences in vsp genes similar to those shown in zinc binding by nuclear proteins, it has been proposed that the VSP genes may encode zinc binding proteins. The ability to bind zinc was shown for VSPH7 and other VSPs but was also found in regions of the VSP that did not contain the “zinc binding” motifs (254). In addition to zinc, VSPH7 was able to bind other metals including iron. Another laboratory, using VSP4A1 (CRISP90) from a genotype A isolate, demonstrated the ability of reduced but not oxidized VSP to bind zinc. However, the degree of binding was substantially lower than an equimolar zinc:protein amount and was thought to be biologically insignificant (266).

All of the VPSs reported have potential N-linked glycosylation sites (201); biochemical analysis of VSP4A1 demonstrated O-linked GlcNAc rather than the more commonly found N-linked glycosylation (268). VSP4A1 was also palmitoylated in the membrane-anchored portion of the C-terminal region (267, 268). Subsequent work has extended the finding of glycosylation and palmitoylation to other VSPs, including VSPA6 and VSPH7 (131).

In addition to their marked 3′ similarity, thevsp genes demonstrate different degrees of identity in other regions. In some cases, there is a high degree of similarity throughout the entire gene. For example, there are two identical convergent copies of the vsp1267 gene approximately 3 kb apart (235) and there is greater than 90% identity in the nonrepeat regions of vspA6 and vspA6-S1(357) as well as CRP136 and CRP65 (52). These examples suggest that the vsp gene repertoire has been expanded by gene duplication followed by divergence. The sequences for one of these vsp genes (vspA6-S1 orvspG3M) from two geographically separated isolates (WB from Afghanistan and G3M from Peru) were identical over the regions of comparison (237, 357), suggesting that this divergence is not an extremely rapid process. A number of vsp genes related to TSA417 (vsp417) have been identified from genotype A-1 and A-2 isolates. Comparisons have shown that the differences between genotypes are smaller than the differences between members of this vsp gene family, suggesting that the divergence among the various members of the vsp417 gene family occurred before the divergence between genotypes A-1 and A-2. The vsp genes from the genotype B isolate GS also appear to exist in families (246, 253) but demonstrate similarity to the vsp genes of genotype A only in the conserved 38-amino-acid C terminus and in the CXXC motif.

There are also examples of near identity followed by abrupt divergence, suggesting that in some cases, the vsp repertoire has been expanded by recombination among the vsp genes. For example,vspC5 demonstrates near identity to vspC5-S2 from approximately −150 to +94 and to vspC5-S1 from −150 to +130 (355). A more complete understanding of thevsp repertoire and gene organization will be available when the sequences obtained as part of the Giardia genome project have been assembled.

The vsp genes are present on most if not all the chromosomes and are reasonably dispersed throughout the genome. There does appear to be some clustering of genes as demonstrated by the linkage of somevsp genes (e.g., vsp1267) (Table5) and by the proximity of vspgenes on some cosmid clones (Yang and Adam, unpublished).

Antigenic variation in African trypanosomes is frequently associated with genome rearrangements, such as duplicative transposition to a telomeric location (32, 290). Therefore, it is of interest to know whether the vsp genes are telomeric and whether there are genomic rearrangements associated with antigenic variation. Current data indicate that most of the vsp genes are not telomere associated (355-357). The vspA6 gene is not found in the 150- and 240-kb NotI telomeric fragments of chromosome 4 (356), and the vsp conserved region does not hybridize to these regions (Adam, unpublished). Genome rearrangements are not directly associated with antigenic variation ofvspA6 (356) or vspC5(358).

Giardia trophozoites have a ploidy of at least 4, so there are at least four alleles of each gene. Since the usual definitions of allele cannot be used in a presumably asexual organism, genes have been considered to be allelic when they map to the same genomic location. For most genes, the alleles are identical or nearly identical to each other, sometimes containing heterozygosities at one or two sites (17). However, in the case of the repeat-containing vsp genes, such as vspA6 andvspC5, the different alleles can be distinguished by different repeat copy numbers (356, 358) (Fig. 13). Three distinct alleles of the vspA6 gene have been documented which differ in the copy number (i.e., 8, 9, or 23) of the 195-bp repeat (356). In addition to the differences in repeat copy number, the allele with the greatest repeat copy number contains 8 nucleotide substitutions within the coding region. Some of these substitutions change the amino acid composition, but none change the reading frame. The flanking regions of the three alleles are identical.

Despite the open reading frames and identical flanking regions, a steady-state transcript is expressed only from the allele with the largest repeat copy number as documented by the correct size of band on a Northern blot in addition to direct RNA sequencing, which identified the nucleotide substitutions specific to the allele with the largest repeat copy number. Thus, a transcript is detectable from only one of the alleles despite the absence of apparent sequence alterations that might explain the difference in expression.

The vspC5 gene is another repeat-containing gene, containing 17 to 26 copies of a 105-bp repeat. In this case, the WBA6 genome contains four distinct alleles of the vspC5 gene (17, 20, 21, or 26 repeats) while the WBC5 genome (derived from WBA6) contains only three distinct alleles (26, 21, or 20 repeats). There are no sequence differences among the different alleles, either in the coding or in the flanking regions. However, Northern analysis has identified only a single band that is most consistent with a transcript from the allele containing 26 repeats. (Because of the short 5′ and 3′ UTRs, the transcript size closely matches that of the coding region.)

Initial studies suggested significant recombination involving thevspA6 (CRP170) gene (6) and deletion of an expression-linked copy of vspA6 in an isolate no longer expressing vspA6 (9). However, subsequent data have shown that these recombination events can be explained by changes in repeat copy number for the repeat-containing vsp genes (355-358) Other than changes in repeat copy number, chromosome-mapping studies of the vspA6 and vspC5genes have revealed no changes in genomic location and organization in isolates where a particular vsp gene is expressed or not expressed (356, 358). This indicates that antigenic variation does not occur by moving the vsp genes into expression sites.

The allele-specific expression of the vsp genes and the absence of sequence alteration or DNA rearrangements associated with antigenic variation suggest the possibility of an epigenetic form of control for vsp genes expression. There is no methylated or “J” DNA in Giardia (335), arguing against altered DNA as a means of controlling vsp gene expression. Whether changes in chromosome structure or histone acetylation status are involved in the control of vsp gene expression has not been studied.

Taglin is a surface antigen that migrates as a 28- and 30-kDa doublet on Western blots (343) and has protease-induced lectin activity (184). After trypsin treatment it binds to mannose-6-phosphate. This antigen is nonvariable, and expression is constant throughout encystation and excystation (320).

G. lamblia trophozoites also have a 49-kDa antigen (GP49) that is attached to the membrane surface by a glycosylphosphatidylinositol (GPI) anchor (61). During synthesis of the GPI anchor, exogenous phosphatidylinositol is incorporated, but myo-inositol is converted to phosphatidylinositol prior to incorporation (317). The GPI anchor is not subject to cleavage by phospholipase C, in contrast to a variety of other GPI-anchored surface molecules, including the trypanosome variant surface glycoprotein (43). GP49 is constant among different isolates and does not demonstrate variation within single isolates (61).