Identification of Conserved Candidate Vaccine Antigens in the Surface Proteome of Giardia lamblia

Immunological cross-protection between G. lamblia strains of the two human-pathogenic genetic assemblages.

Successful vaccine antigens should ideally confer protection against different strains of the two genetic assemblages (A and B) of human-pathogenic G. lamblia. Because of marked genetic diversity between the assemblages, particularly among the immunodominant VSPs (32, 42), it is not clear how much immunologically relevant overlap exists in the host response to infection with divergent G. lamblia strains. To address this question, we utilized a giardiasis model in gerbils, which are highly susceptible to a wide range of G. lamblia strains (43), for a controlled cross-protection trial with a primary infection followed by pharmacological clearance and secondary challenge. Gerbils were inoculated with the assemblage A strain, WB, or the assemblage B strain, GS/M, or left uninfected. After 4 weeks, an adequate time for induction of strong adaptive immune responses (44), all animals were treated with metronidazole to eradicate any residual infection and subsequently challenged with either of the two G. lamblia strains. Primary infection with either G. lamblia strain led to complete protection upon secondary challenge with the same strain (Fig. 1). Importantly, secondary challenge with the respective other, assemblage-mismatched strain also conferred marked protection, as indicated by >100-fold reduction in trophozoite load in the small intestine at the peak of infection upon secondary challenge (Fig. 1). Although cross-protection was slightly less complete than protection against assemblage-matched strains, these data strongly suggest that immunologically relevant cross-protection exists between divergent G. lamblia strains and thus provide the rationale for identifying conserved antigens that mediate this protection.

Proteomic analysis of G. lamblia surface proteins.

Our previous studies and those of others had used whole trophozoite extracts of G. lamblia to identify proteins recognized by immune serum as candidate vaccine antigens (20, 39). This strategy could have biased antigen identification toward more abundant proteins while potentially underestimating less abundant surface proteins. Because surface proteins may be important targets of protective IgA responses (19), we pursued an alternative approach focused on identifying conserved cell surface proteins. Trophozoites of the two divergent G. lamblia strains, WB (assemblage A) and GS/M (assemblage B), were labeled with a reactive but non-membrane-permeant biotin (N-hydroxysuccinimide [NHS]-biotin) (Fig. 2). Incubation of biotin-labeled cells with a fluorescently labeled streptavidin revealed strong staining at the trophozoite surface and flagella, which was not observed in controls not labeled with biotin (Fig. 2). Furthermore, immunoblot analysis of total cell lysates with anti-biotin showed a range of biotin-labeled proteins that were not seen in control lysates (Fig. 2), demonstrating the specificity of the biotin labeling.

Next, we immunoprecipitated biotin-labeled proteins from total cell lysates of G. lamblia WB by streptavidin affinity chromatography, digested the precipitates in situ with trypsin, and identified the resulting peptides by sequential liquid chromatography and tandem mass spectrometry (Fig. 2). Comparison of the peptides with simulated digests of all gene products predicted from the G. lamblia WB genome sequence led to the identification of 86 proteins. Of these, 15/86 (17%) were annotated as VSPs, which are known to be abundant at the trophozoite surface (33), thus validating our experimental strategy for identifying surface proteins. Bioinformatics analysis showed that 24 (28%) of the 86 identified proteins, including the VSP proteins, have well-defined transmembrane regions, with (22/24) or without (2/24) signal peptides. An additional 21 (25%) of the proteins have motifs for palmitoylation, which can mediate an alternative mode of membrane binding in G. lamblia (45), whereas glycosylphosphatidylinositol (GPI) anchoring of membrane proteins is rare in the parasite (46). Thus, >50% of the identified proteins have features that are commonly associated with membrane proteins.

The same strategy for surface protein identification was then applied to the GS/M strain of G. lamblia. Mass spectrometry of tryptic digests and comparison of the peptides with simulated digests of the predicted gene products of the G. lamblia GS/M genome led to the identification of 51 proteins. As for the WB strain, a significant number (6/51, 12%) of identified proteins were annotated as VSPs. Comparison of the identified proteins in WB and GS/M revealed that 27 were found in both strains. None of these were annotated as VSPs, which underlines that the VSP repertoires of the two G. lamblia strains are largely nonoverlapping (31, 32, 42). Of the 27 common proteins, 23 showed greater than 90% amino acid sequence identity between WB and GS/M (Table 1), indicating that the vast majority of the identified non-VSP surface proteins are conserved between the two divergent assemblage A and B strains.

Confirmation of surface protein localization.

To confirm the surface localization of the identified proteins by immunofluorescence staining, G. lamblia WB trophozoites were allowed to attach to glass slides and were fixed with paraformaldehyde but left nonpermeabilized for selective staining of surface targets accessible to antibodies from the outside. As a control, fixed cells were permeabilized with Triton X-100 to permit antibody access to all proteins regardless of location. For staining, we used previously validated antibodies (47, 48) or generated new mouse polyclonal antibodies against the recombinant proteins. Immunoblotting of total G. lamblia lysates and staining with the new antibodies revealed the expected single bands, confirming their specificity (Fig. 3A). Staining and immunofluorescent imaging of the fixed trophozoites revealed that all five putative surface proteins for which we had antibodies, i.e., β-giardin, α1-giardin, α11-giardin, uridine phosphorylase-like protein 1 (UPL-1), and protein 21.1, were detectable without permeabilization (Fig. 3B). As a control, a known intracellular protein, calmodulin kinase (CAMK) (47), could not be stained under the same conditions, demonstrating the selectivity of our approach (Fig. 3B). Staining patterns varied from what appeared to be specific or preferential to the adhesion disc (β-giardin, protein 21.1) to localized (α1- and α11-giardin) or generalized (UPL-1) surface staining. By comparison, in permeabilized cells, staining for all five proteins was generally strong and widespread (Fig. 3B), suggesting that the proteins were not exclusively present at the surface.

To localize proteins for which we did not have specific antibodies, we opted for expressing them in G. lamblia WB as fusion proteins with a hemagglutinin (HA) epitope tag and staining with an anti-HA antibody. All five targeted proteins (pyruvate-flavodoxin oxidoreductase, pyruvate-phosphate dikinase, dynamin, peroxiredoxin-1, and ribosomal protein S8) could be stained without permeabilization, while a known intracellular control protein, NEK kinase, showed no staining (Fig. 3C). Staining patterns and intensity, again, ranged from fairly selective, such as the strong flagellar and head staining of pyruvate-phosphate dikinase, to more generalized and weaker (e.g., dynamin) effects. After permeabilization, all five proteins and the NEK kinase control showed strong staining across the cells (Fig. 3C), suggesting that these proteins were not limited to the surface. Together, the staining data confirm that all 10 of the tested surface proteins identified by selective biotin labeling and mass spectrometry were expressed at the trophozoite surface and accessible to external antibodies.

Immunogenicity and protective capacity of selected conserved surface antigens.

Candidate antigens that have limited similarity to human proteins are generally preferable for vaccines since the risk of adverse immunological cross-reactions upon immunization should be reduced. In evaluating the similarity of the identified G. lamblia surface proteins to human proteins, we selected a subset of 12 (of 23) that had <20% similarity (and E values of >10⁻³⁰) relative to any predicted human protein (Table 1). Out of these, we produced four representative examples, α1-giardin, α11-giardin, UPL-1, and protein 21.1, as recombinant His-tagged proteins in Escherichia coli and purified them by nickel column chromatography to >80% purity in soluble form (Fig. 4A). A fifth protein, β-giardin, was also expressed in E. coli but could not be readily purified in soluble form.

To test the immunogenicity of the candidate antigens, we administered individual purified antigens together with the prototype mucosal adjuvant cholera toxin to mice by intranasal administration. Immunization induced high titers of serum IgG for all four antigens (Fig. 4B), whereas significant mucosal secretory IgA was elicited for only two of them, α11-giardin and UPL-1 (Fig. 4C). As a further evaluation of the immunogenicity of the selected antigens, we tested antibody titers in patients with laboratory-confirmed giardiasis (49). Healthy and presumed-unexposed individuals were used as controls. Infection induced significant increases in serum IgG titers against three of the four antigens (α1-giardin, α11-giardin, and protein 21.1) (Fig. 4D) and in serum IgA titers against all four antigens (Fig. 4E). Together, these data show that the four selected antigens are immunogenic after mucosal immunization or natural infection in mice and humans.

To assess protective capacity, antigen-immunized mice were challenged 2 weeks after the last of three immunizations by oral gavage with G. lamblia GS/M trophozoites, and parasite load in the small intestine was determined after 5 days, the time of peak parasite burden in control mice (19). Immunization with all four antigens conferred significant protection, as evidenced by a 50- to 500-fold reduction in trophozoite numbers compared to those in mock-immunized mice (Fig. 5A). Furthermore, immunization with one of the antigens, α1-giardin, shortened the duration of infection since most immunized mice had eradicated infection by day 20, while none of the nonimmunized controls were cleared at that time (Fig. 5B). Despite strong induction of antigen-specific IgG after immunization (Fig. 4B), protection was not significantly correlated with IgG titers for any of the four antigens (Fig. 5C), suggesting that IgG is not likely to act as an antigiardial effector in this system. Similarly, mucosal IgA, which had been shown to play a role in clearance of G. lamblia infection (19, 20), was not important for protection since immunization of IgA-deficient mice with two of the antigens, α1-giardin or UPL-1, conferred significant protection that was similar to that in wild-type mice (Fig. 5D).

Finally, immunization with α1-giardin also induced significant protection in gerbils against subsequent G. lamblia infection, indicating that protection was not host specific (Fig. 5E). Together, these data show that the newly identified surface antigens are promising candidates for a protective vaccine against giardiasis.

G. lamblia isolates and cultivation.

We used G. lamblia isolates WB clone C6 (ATCC 50803; human, assemblage A) (71) and GS/M clone H7 (GS/M; ATCC 50581; human, assemblage B) (9). Parasites were grown in modified TYI-S33 medium to no more than 80% confluence (72).

Biotin labeling of surface proteins.

Unattached parasites were decanted and attached, intact cells were resuspended in ice-cold phosphate-buffered saline containing 0.2 M sucrose, pH 7.4 (PBS-S). Parasites were placed on ice for 5 min, concentrated by centrifugation for 5 min at 600 × g and 4°C, and cells were combined into new tubes. Cells were washed two times in PBS-S (50 ml/tube) and then a third time with PBS-S containing a protease inhibitor cocktail (Complete Mini-EDTA; Roche) plus 1 μM E-64 (PBS-S+; Sigma). Cells were counted and resuspended in cold PBS-S+ at a density of 1 × 10⁸ cells/ml, placed into 1.5-ml microcentrifuge tubes, and labeled with 2 mM EZ-Link NHS-PEG4-biotin (Thermo Scientific) for 2 h with continuous rotation under refrigeration. Unlabeled controls were processed under identical conditions but without added NHS-PEG4-biotin. Following the labeling reaction, viability of trophozoites was assessed with fluorescein diacetate (25 μg/ml) and propidium iodide (6 μg/ml), and excess biotin was quenched with 0.5 M Tris-HCl (pH 8.0) to a final concentration of 50 mM. Parasites were washed three times with ice-cold PBS-S in new 50-ml tubes.

Following the biotinylation and washing steps, biotin-labeled and unlabeled control cells were assessed by immunofluorescence assay for localization of the biotin label. Ten microliters of cells was allowed to attach to prewarmed eight-well slides at 37°C for 10 min, which were then fixed for 10 min at 37°C with 2% paraformaldehyde in PBS. Cells were rinsed in PBS, permeabilized in 0.5% Triton X-100 for 10 min, blocked (10% goat serum, 1% glycerol, 0.1% bovine serum albumin [BSA], 0.1% fish gelatin, and 0.04% sodium azide in PBS) for at least 30 min at room temperature, and probed with streptavidin-Alexa 488 (1:800 diluted in blocking buffer; Molecular Probes) for 30 min. Cells were washed four time in PBS, postfixed in 4% paraformaldehyde (PFA) for 5 min, rinsed with PBS, and mounted in Prolong Gold with 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes). Biotin conjugates associated with whole cells were imaged on an E800 Nikon (New York City, NY) microscope with an EXFO (Vanier, Canada) X-cite fluorescent 120-W metal halide illuminator and imaged with a DMX 1200F Nikon fluorescence sensitive digital camera.

Purification of biotinylated proteins.

Washed biotinylated and control cells were concentrated by centrifugation and transferred to new 1.5-ml microcentrifuge tubes. Biotinylated membrane proteins were processes and purified essentially as described previously (73) with some modifications. Briefly, cell pellets were directly lysed, and membrane proteins were solubilized in 1 ml of lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM EDTA, 2% SDS, 0.1% Igepal CA-630, 0.5% sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride [PMSF], 1 μM E-64, Complete Mini protease inhibitor cocktail), and insoluble material was removed by centrifugation at 13,000 × g for 5 min at 4°C. Lysis buffer was used to preequilibrate streptavidin-agarose by batch method under refrigeration and continuous rotation for 10 min. Approximately 1 mg of cellular protein (estimated at ∼5 μg of membrane protein per 5 × 10⁷ cells) was added per 150 μl of equilibrated streptavidin-agarose slurry. Biotinylated proteins were incubated with streptavidin-agarose for 2.5 h under refrigeration and with continuous rotation. Subsequent washing to remove unbound proteins was done as described previously (73).

Mass spectrometry.

Trypsin-digested peptides were analyzed by high-pressure liquid chromatography coupled with tandem mass spectroscopy (HPLC-MS/MS) using nano-spray ionization as described previously (74) with the following modifications. The nanospray ionization experiments were performed using a QSTAR-Elite hybrid mass spectrometer (AB Sciex) interfaced with nano-scale reversed-phase HPLC (Tempo) using a 10-cm glass capillary (100-μm inside diameter), packed with 5-μm C₁₈ Zorbax beads (Agilent Technologies, Santa Clara, CA). Peptides were eluted from the C₁₈ column into the mass spectrometer using a linear gradient (5 to 60%) of acetonitrile (ACN) at a flow rate of 400 μl/min for 1 h. The buffers used to create the ACN gradient were buffer A (98% H₂O, 2% ACN, 0.2% formic acid, and 0.005% trifluoroacetic acid [TFA]) and buffer B (100% ACN, 0.2% formic acid, and 0.005% TFA).

MS/MS data were acquired in a data-dependent manner in which the MS1 data were acquired at m/z of 400 to 1,800 and the MS2 data were acquired from m/z of 50 to 2,000. Peptide identifications were made using the paragon algorithm executed in Protein Pilot, version 2.0 (AB Sciex). A comprehensive library of G. lamblia protein sequences was obtained from GiardiaDB (https://giardiadb.org/giardiadb/). Proteins were identified by at least two peptides with a confidence of >95% for the peptide identification. The MS was set to disregard ions from certain high-abundance nongiardial proteins. The following programs were used to evaluate putative membrane association: Giardia DB, release 36 (www.giardiadb.org/giardiadb/); SignalP Server, version 3.0 (http://www.cbs.dtu.dk/services/SignalP/); TMHMM Server, version 2.0 (http://www.cbs.dtu.dk/services/TMHMM-2.0/), and CSS-Palm for palmitoylation site prediction (http://csspalm.biocuckoo.org/online.php).

Cloning of epitope-tagged proteins and immunofluorescence assays.

Using genomic DNA from G. lamblia WB/C6, full-length cDNAs and the corresponding promoter of the respective target genes were amplified by PCR and cloned into a vector that introduces a C-terminal triple hemagglutinin (HA) tag and allows for stable expression of the target gene under its own promoter (75). Trophozoites were transfected with each vector and selected for stable expression by puromycin selection as described previously (76, 77).

For staining, live trophozoites were allowed to attach to poly-l-lysine-coated coverslips for 10 min at 37°C, washed in warm PBS, fixed with 4% paraformaldehyde for 30 min, and left untreated or permeabilized with 0.5% Triton X-100 for 10 min. Fixed cells were blocked for 30 min in 5% goat serum, 1% glycerol, 0.1% bovine serum albumin, 0.1% fish gelatin, and 0.04% sodium azide and stained for 1 h at room temperature with primary antibodies against the target proteins. Antibodies against calmodulin (47) and β-giardin (48) were validated previously, while those against α1-giardin, α11-giardin, protein 21.1, and UPL-1 were generated by immunizing mice with recombinant forms of these proteins (see below) and validating them by immunoblotting with whole-cell extracts of G. lamblia WB. Antibodies against the HA tag were obtained commercially (Sigma). After the primary staining step, slides were washed and further incubated for 1 h with Alexa 488-labeled secondary antibody against the Ig isotype of the primary antibody. After a rinse with water, coverslips were mounted with Prolong Gold with DAPI (Molecular Probes). Staining was examined using a Nikon Eclipse E800 microscope equipped with an X-Cite 120 fluorescence lamp.

Production of recombinant proteins.

Sequences of candidate antigens were obtained from the published genome sequence of G. lamblia GS/M (ATCC 5058), and cDNAs were generated by total gene synthesis with codon optimization for E. coli (GenScript Biotech). Genes were inserted into NdeI-XhoI cloning sites of the bacterial expression vector pET-15b (Millipore Sigma), and vectors were transformed into T7 Express lysY/Iq Competent E. coli (NEB) per the manufacturer’s protocol. Cells were grown at 37°C in LB medium with carbenicillin until they reached an optical density at 600 nm (OD₆₀₀) of 0.4 to 0.6, and protein expression was induced by addition of 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 2 to 3 h. Recombinant proteins were purified from bacterial lysates using HisPur Ni-nitrilotriacetic acid (NTA) resin (Thermo Scientific), concentrated using Amicon Ultra centrifugal filters (EMD Millipore), and stored at a concentration of >1 mg/ml at 4°C and in a buffer of 300 mM NaCl, 25 mM HEPES, 100 mM arginine, 20 mM imidazole, 10% glycerol, 0.1% Tween 20, and protease inhibitors. Purity of the soluble recombinant proteins was analyzed by SDS-PAGE. Endotoxin contamination in the antigen preparations was determined with a ToxinSensor Chromogenic LAL Endotoxin assay kit (GenScript) and found to range between 0.79 and 0.99 endotoxin units/ml (equivalent to 79 to 99 pg/ml, or approximately 0.4 to 0.5 pg per dose upon intranasal administration in mice). Preliminary tests showed that endotoxin had no impact on G. lamblia infectivity in mice since three intranasal administrations of a 5,000-fold-higher dose (2.5 ng of lipopolysaccharide [LPS] from E. coli O127:B8) than the amounts found in the antigen preparations over 4 weeks yielded similar numbers of trophozoites in the small intestine compared to numbers in mice not given LPS (7.35 ± 0.10 versus 7.63 ± 0.18, n = 5 and 6, respectively; P, not significant by Mann-Whitney U test). Furthermore, PCR analysis with universal 16 rRNA primers did not detect any contaminating bacterial DNA in the antigen preparations.

Human serum samples and multiplex beads immunoassays.

Serum samples from 31 Norwegian adult patients with laboratory-confirmed giardiasis were collected (49). Most participants acquired infection by traveling in areas where G. lamblia is endemic, including Africa (n = 11), Europe (n = 7), Asia (n = 6), and the Americas (n = 4), and three were of uncertain origin. Control sera were obtained from 19 presumed-unexposed individuals. Patients and controls matched for age (means and standard deviations [SD] of 42.3 ± 20.4 and 39.6 ± 16.8 years for infected patients and controls, respectively) and sex (45% and 37% female, respectively). The study was approved by the Regional Committee for Medical Research Ethics of Western Norway.

For titer determinations, recombinant antigens or glutathione S-transferase (GST) as a control protein was separately coupled by sodium N-hydroxysulfosuccinimide and N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide HCl cross-linking chemistry to 4-μm Cyto-Plex polystyrene beads with different levels of internal fluorescence (Thermo Fisher Scientific, Waltham, MA). Bead density was determined by microscopic counting. Each well of a MultiScreen HTS HV filter plate was prewetted with 200 μl of assay buffer (PBS containing 1% BSA and 0.05% Tween 20), and 10⁴ beads of each protein-labeled type were added. Diluted human serum (1:50 in assay buffer) was added (100 μl/well), and plates were incubated for 1 h at room temperature and washed with assay buffer; Alexa Fluor 488-labeled AffiniPure goat anti-human IgA or Alexa Fluor 555-labeled goat anti-human IgG(H+L) antibody (Jackson Immunoresearch, Ely, UK) was added to the wells (1:400 or 1:500 dilution, respectively, in 200 μl assay buffer), and samples were incubated for an additional hour at room temperature. Wells incubated with Alexa Fluor 488-labeled mouse anti-His antibody (1:100 in 200 μl; Novus Biologicals, Abingdon, UK) were used as positive controls since all recombinant antigens carried the His epitope tag. GST-tagged beads were used as negative controls. After incubation, beads were washed twice in fluorescence-activated cell sorting (FACS) buffer (PBS containing 1% BSA), resuspended in 160 μl/well of FACS buffer, transferred into a round-bottom microplate, and analyzed in an LSRFortessa flow cytometer (BD Biosciences, San Jose, CA). IgA and IgG responses were quantified by determining the median fluorescence intensity (MFI) for each protein-coupled bead population using FlowJo software, version 10.4.2 (FlowJo LLC, Ashland, OR).

Antigen immunizations.

Adult (6- to 8-week-old) female C57BL/6 mice (The Jackson Laboratory), IgA-deficient mice on a C57BL/6 background (19), and adult (10-week-old) female Mongolian gerbils (Meriones unguiculatus; Charles River Laboratories) were used for these studies. For intranasal immunizations, animals were placed under anesthesia with ketamine and xylazine, and a small volume (5 μl in mice; 10 μl in gerbils) of a mixture of antigen (50 to 100 μg) and cholera toxin (10 μg) was instilled into one of the nares. Cholera toxin alone was used as a control. Animals were left recumbent for at least 1 min and then allowed to recover in their cages. Immunizations were done three times at least 2 weeks apart. Mice were bled via the tail vein to obtain plasma, or the small intestine was removed and rinsed with PBS and protease inhibitors to obtain mucosal washes. Antibody levels against the immunizing antigens were determined by enzyme-linked immunosorbent assay (ELISA) in plasma or mucosal washes (20).

G. lamblia infections in rodents.

For infection challenges we used wild-type (C57BL/6) mice, IgA-deficient mice on a C57BL/6 background (19), and Mongolian gerbils. Infection challenges were done at least 2 weeks after the last immunization. Mice were treated for 2 days before and throughout the infection with an antibiotic cocktail (1.4 mg/ml neomycin, 1 mg/ml ampicillin, and 1 mg/ml vancomycin in the drinking water), while gerbils were not pretreated. G. lamblia trophozoites (WB/C6 or GS/M) were grown to mid-logarithmic phase and administered in growth medium by oral gavage to gerbils (10⁷ trophozoites in a 400-μl volume) or mice (10⁶ trophozoites in a 200-μl volume). At different times after infection, the small intestine was removed, opened longitudinally, placed into 2 to 5 ml of PBS, and cooled on ice for 10 min. After vigorous shaking of samples, live trophozoites were counted in a hemocytometer. Euthanasia of mice and gerbils was performed with CO₂ inhalation followed by removal of a major organ. All animal studies were reviewed and approved by the University of California San Diego Institutional Animal Care and Use Committee.

Statistical analysis.

An unpaired t test, Mann-Whitney U test, or Kruskal-Wallis test with Bonferroni correction and the appropriate post hoc tests were used to compare results between animal groups as appropriate. Data are presented as means ± standard errors (SE) or as individual data points and geometric means. For the human samples, differences in specific antibody responses between individuals in the exposed and presumed-unexposed group were analyzed by a Mann-Whitney U test using SPSS Statistics, version 24. Differences with P values of <0.05 were considered statistically significant for all experiments.