Characterization of the Giardia intestinalis secretome during interaction with human intestinal epithelial cells: The impact on host cells

Background Giardia intestinalis is a non-invasive protozoan parasite that causes giardiasis in humans, the most common form of parasite-induced diarrhea. Disease mechanisms are not completely defined and very few virulence factors are known. Methodology To identify putative virulence factors and elucidate mechanistic pathways leading to disease, we have used proteomics to identify the major excretory-secretory products (ESPs) when Giardia trophozoites of WB and GS isolates (assemblages A and B, respectively) interact with intestinal epithelial cells (IECs) in vitro. Findings The main parts of the IEC and parasite secretomes are constitutively released proteins, the majority of which are associated with metabolism but several proteins are released in response to their interaction (87 and 41 WB and GS proteins, respectively, 76 and 45 human proteins in response to the respective isolates). In parasitized IECs, the secretome profile indicated effects on the cell actin cytoskeleton and the induction of immune responses whereas that of Giardia showed anti-oxidation, proteolysis (protease-associated) and induction of encystation responses. The Giardia secretome also contained immunodominant and glycosylated proteins as well as new candidate virulence factors and assemblage-specific differences were identified. A minor part of Giardia ESPs had signal peptides (29% for both isolates) and extracellular vesicles were detected in the ESPs fractions, suggesting alternative secretory pathways. Microscopic analyses showed ESPs binding to IECs and partial internalization. Parasite ESPs reduced ERK1/2 and P38 phosphorylation and NF-κB nuclear translocation. Giardia ESPs altered gene expression in IECs, with a transcriptional profile indicating recruitment of immune cells via chemokines, disturbances in glucose homeostasis, cholesterol and lipid metabolism, cell cycle and induction of apoptosis. Conclusions This is the first study identifying Giardia ESPs and evaluating their effects on IECs. It highlights the importance of host and parasite ESPs during interactions and reveals the intricate cellular responses that can explain disease mechanisms and attenuated inflammatory responses during giardiasis.


Methodology
To identify putative virulence factors and elucidate mechanistic pathways leading to disease, we have used proteomics to identify the major excretory-secretory products (ESPs) when Giardia trophozoites of WB and GS isolates (assemblages A and B, respectively) interact with intestinal epithelial cells (IECs) in vitro.

Findings
The main parts of the IEC and parasite secretomes are constitutively released proteins, the majority of which are associated with metabolism but several proteins are released in response to their interaction (87 and 41 WB and GS proteins, respectively, 76 and 45 human proteins in response to the respective isolates). In parasitized IECs, the secretome profile indicated effects on the cell actin cytoskeleton and the induction of immune responses whereas that of Giardia showed anti-oxidation, proteolysis (protease-associated) and induction of encystation responses. The Giardia secretome also contained immunodominant and glycosylated proteins as well as new candidate virulence factors and assemblage-specific differences were identified. A minor part of Giardia ESPs had signal peptides (29% for both isolates) and extracellular vesicles were detected in the ESPs fractions, suggesting alternative secretory pathways. Microscopic analyses showed ESPs binding to IECs and partial internalization. Parasite ESPs reduced ERK1/2 and P38 phosphorylation and NF-κB nuclear translocation. Giardia ESPs altered gene expression PLOS

Introduction
Giardia intestinalis is an intestinal protozoan parasite causing 280 million human diarrheal cases (giardiasis) annually [1]. Giardia has two life-cycle stages: infectious cysts and diseasecausing trophozoites [2]. Trophozoites move by flagella and adhere to intestinal epithelial cells (IECs) strongly using a suction cup on the ventral surface [2]. The direct contact of Giardia with IECs induces cellular damage on the apical surface, compromising brush border enzymes function, and affect cellular junctions, leading to an increased intestinal barrier permeability [3]. These changes cause maldigestion, malabsorption and electrolyte imbalance, collectively culminating in causing diarrhea [4]. Symptoms are usually acute but chronicity is not uncommon and some patients develop intestinal and extra-intestinal complications post-infection [1,5]. Many Giardia infections, however, remain asymptomatic and the reason for variable disease outcomes remains obscure. Early studies of Giardia pathogenicity suggested that the parasite releases proteins that can contribute to disease induction. Excretory-secretory products (ESPs) of Giardia affect intestinal absorption and secretion [6], cause histopathological alterations [7] and induce specific antibodies against glycoproteins [8]. A secreted 58 kDa glycoprotein has been shown to react with antibodies from giardiasis patients and induce fluid accumulation in ligated rabbit ileal loop experiments [9,10]. Nevertheless, the identity of this protein remained unknown as tryptic digests did not align with any proteins in the Giardia Database. Cysteine proteases (CPs) are also yet unidentified Giardia ESPs. Their involvement in virulence, however, is well studied, demonstrated by their association with parasite attachment [11], and the degradation of extracellular matrix (ECM) proteins [12], antibodies [13] and proinflammatory cytokines [14,15]. Arginine metabolising enzymes (AMEs) (e.g. arginine deiminase (ADI) and ornithine carbamoyltransferase (OCT)) are the best-studied examples of secreted virulence factors in Giardia. AMEs outcompete host cells for arginine uptake, inhibiting cellular proliferation [16], abrogating nitric oxide production, a compound cytotoxic to Giardia, [17], and inhibiting cytokine production by dendritic cells [18]. Other reports documented the secretion of elongation factor-1 alpha (EF-1α) [19] and the glycolytic enzyme enolase [17] into Giardia growth medium and the medium of interaction with IECs in vitro, respectively. Their functions in parasite virulence, if any, is still unknown. Finally, variable surface proteins (VSPs) are cysteinerich proteins that cover trophozoite surface and they are released in large quantities into the growth medium (%70% within 24h) [20]. VSPs are immunodominant during infection [21] and their constant switching (i.e. antigenic variation) help the parasite evade the host immune system [22].
In view of the simple endomembrane system of Giardia, which lacks a true Golgi apparatus, peroxisomes and endosomal/lysosomal system [2], the secretion of proteins, their sorting and trafficking remain a perplexing aspect of the parasite's biology. Most of our knowledge about secretion and secreted proteins of Giardia derive from experiments targeting the sorting and trafficking of VSPs to the plasma membrane and cyst-wall proteins in encystation-specific vesicles (ESVs) to the cyst wall during encystation [23]. AMEs, EF-1α and enolase all lack signal peptides (SPPs) but they are still secreted by the parasite. This suggests the presence of nonconventional secretory mechanisms in Giardia such as extracellular vesicles (EVs) carrying protein cargo with no SSP. EVs include the exosomes (30-100 nm) and microvesicles (MVs) (100-1000 nm), which have been shown to play roles in parasites communication, host cell colonization and immune modulation [24,25]. A recent study has identified MVs released from Giardia trophozoites and encysting cells [26]. The released MVs affected trophozoite attachment and dendritic cell activation [26].
In this study, we used proteomics to identify the secretome of the Giardia reference isolates WB and GS during growth and interaction with IECs in vitro in serum-free medium. We also show the release of EVs by Giardia as a part of its secretome. We demonstrate that Giardia ESPs alter gene expression in IECs, cell signalling and the production of proinflammatory cytokines. This study highlights the importance of Giardia ESPs in inducing pathological effects in IECs and suggests how this might have immunomodulatory effects on host immune responses.

Ethics statement and chemicals
The cell lines used in this study were all obtained from American Type Culture Collection (ATCC): Giardia intestinalis isolates WB clone C6 (ATCC30957; assemblage A) and GS, clone H7 (ATCC50581; assemblage B) were used in this study in combination with the human colonic adenocarcinoma cell line, Caco-2 clone TC7 (ATCC HTB-37). All the experiments were approved by the Programme Review Board at Department of Cell and Molecular Biology, Uppsala University, Sweden. All chemicals were purchased from Sigma-Aldrich (MO, USA), unless otherwise stated.

Medium for collection of excretory/secretory products (ESPs) and trophozoite viability
To collect Giardia ESPs from axenic cultures we used a modified RPMI-1640 medium containing 55.5 mM glucose, 22.8 mM L-arginine, 11.4 mM L-cysteine hydrochloride monohydrate, 11.4 mM ascorbic acid, 2mM Glutamax (Gibco), 1 mM sodium pyruvate (Gibco) and 1x MEM essential amino acids. The medium was filter-sterilized and placed in a water bath (final pH is 6.85 at 37˚C) until used.
Confluent Giardia culture tubes, prepared as described above, were used for trophozoite viability assessment. Briefly, the TYDK medium was decanted from each tube followed by three washes in Hank's Balanced Salt Solution with glucose (HBSS-G) supplemented with 11.4 mM L-cysteine (pH 6.8, 37˚C). Culture tubes were replenished with the modified medium, closed tightly and incubated for 2h or 6h (37˚C). At the end of each incubation, tubes were placed on ice to detach trophozoites (10 min), followed by counting using the expulsion of Trypan Blue and flagellar motility as viability criteria.

Collection and processing of ESPs from axenic cultures and host parasite interactions
Tubes with Giardia trophozoites in modified RPMI-1640 medium (50 ml, two tubes of each isolate) were used to collect Giardia ESPs in axenic cultures. The tubes and media, following counting and viability assessment, were centrifuged (930 xg, 10 min, 4˚C) to pellet trophozoites. Collected media were filter-sterilized (0.22 μM) and treated with protease inhibitor cocktail tablets (cOmplete, Mini EDTA-free, Roche, Sigma-Aldrich). Media were concentrated down to 200-300 μl using the Amicon Ultra 15 mL centrifugal filters with 3kDa cutoff (Merck-Millipore, Darmstadt, Germany). To further eliminate any traces of residual serum, concentrates were processed through the AlbuminOUT kit (G-Biosciences, MO, USA) according to the manufacturer's instructions. Three biological replicates (2h and 6h) were prepared and stored at -80˚C until used for MS analysis. The whole experimental setup is described in Fig 1A. Differentiated Caco-2 cells (75 cm 2 flasks) were washed with warm PBS (3x) and incubated with 6 x 10 7 trophozoites resuspended in 25 ml of DMEM (no FBS) for 2h or 6h. Trophozoites were washed (3x, HBSS-G plus 11.4 mM L-cysteine) prior addition to cells. Caco-2 cells treated as above and incubated alone in DMEM (2h and 6h) served as controls ( Fig 1A). All incubations were performed in 10% CO 2 humidified tissue culture incubator at 37˚C. At the end of each interaction, flasks were placed on ice (10 min) to detach trophozoites for counting and viability assessment. Media were then collected, processed as described earlier and stored at -80˚C until used for MS analysis. Three biological replicates were prepared (2h or 6h including controls).
USA) according to the manufacturer's instructions. Giardia EVs were pelleted by centrifugation (1500 xg, 30 min) and resuspended in 100 μl PBS buffer and processed immediately for electron microscopy (EM). Samples (5μl) were adsorbed onto a Formvar-coated 200 mesh grid (15 seconds), followed by treatment with 2% molybdenum (10 seconds) and contrasting using 2% uranyl acetate (10 seconds). Grids were dried and visualized using a transmission electron The collection of Giardia intestinalis, isolates WB and GS, excretory-secretory products (ESPs) in a serum-free medium from axenic culture and coculture with differentiated intestinal epithelial cells (IECs), Caco-2, for 2h and 6h. Samples were processed as described in Materials and Methods followed by liquid chromatography/mass spectrometry to identify secreted proteins. (B) Studies of IECs exposed to G. intestinalis ESPs using differential gene expression analyses by RNA sequencing, quantitative real time PCR, Western blots, ELISA and confocal microscopy. The panel illustrates the experimental setup (left to right) showing IECs exposed to ESPs of either Giardia isolates through a Transwell insert (direct exposure) or different amounts of ESPs collected from axenic culture (i.e. the absence of IECs, 1 μg, 5 μg or 10 μg per ml of medium microscope at the BioVis Core Facility, Uppsala University, Uppsala, Sweden. A negative control (media alone) was included and processed similarly.

Liquid chromatography and mass spectrometry (LC/MS) analysis
Before digestion, the buffer in the collected samples was exchanged to a urea-containing lysis buffer (20 mM HEPES, 9 M urea, Complete Mini EDTA-free protease inhibitor cocktail (Roche) using 3 kDa molecular weight cut off filters (Thermo Fisher Scientific, MA, USA). Proteins were then subjected to dithiothreitol (DTT, 12.5 mM) reduction and iodoacetamide (IAA, 6.9 mM) alkylation before digested in-solution with trypsin (1 μg added per sample). The samples were desalted using Pierce C 18 Spin Columns (Thermo Fisher Scientific). Eluates were dried and resolved in 15 μL 0.1% formic acid of which 5 μL were injected. The nanoLC-MS/MS experiments were performed using a Q Exactive Orbitrap mass spectrometer (ThermoFisher Scientific, Bremen, Germany) equipped with a nano-electrospray ion source. The peptides were separated by C 18 reversed phase liquid chromatography using an EASY-nLC 1000 system (Thermo Fisher Scientific). A set-up of pre-column and analytical column was used. The pre-column was a 2 cm EASYcolumn (ID 100 μm, 5 μm particles) (Thermo Fisher Scientific) while the analytical column was a 10 cm EASY-column (ID 75 μm, 3 μm particles, Thermo Fisher Scientific). Peptides were eluted with a 150-min linear gradient from 4% to 100% acetonitrile at 250 nL min-1. The mass spectrometer was operated in positive ion mode acquiring a survey mass spectrum with resolving power 70,000 (full width half maximum), m/z 400-1750 using an automatic gain control (AGC) target of 3×10 6 . The 10 most intense ions were selected for higher-energy collisional dissociation (HCD) fragmentation (25% normalized collision energy) and MS/MS spectra were generated with an AGC target of 5×10 5 at a resolution of 17,500. Mass spectrometer worked in datadependent mode.
Database searches were performed using the Sequest algorithm, embedded in Proteome Discoverer 1.4 (Thermo Fisher Scientific). Proteins in axenic cultures were identified by performing separate searches against the fasta files of the Giardia proteome for each isolate (GS clone containing 4470 entries and WB clone 9667 entries) (available at www.giardiadb.org; giardiadb.org/common/downloads/release5.0/). The fasta file for human proteins was downloaded from Uniprot 2015-05 and contained 20192 entries. Proteins in co-cultures were searched against a combined database of Giardia and human proteins in a MudPIT search combining all three replicates. For both searches, the following parameters were used for data processing: Maximum 10 ppm and 0.02 Da error tolerances for the survey scan and MS/MS analysis, respectively, trypsin as digesting enzyme, carbamidomethylation of cysteins as fixed modifications, oxidation of methionine and deamidation of asparagine and glutamine as variable modifications, maximum of two miss cleavages sites. The target decoy PSM validator was used to calculate false discovery rate. The search criteria for protein identification were set to at least two matching peptides of 95% confidence level per protein.
Identified proteins were subject to gene ontology (GO) enrichment analysis to identify their biological as well as proteins class. The analysis was performed using the PANTHER algorithms provided within the Gene Ontology Consortium website (http://geneontology.org/ page/go-enrichment-analysis). The results were processed through the GO slim option for biological GO terms, with Bonferroni correction applied for multiple testing (P < 0.05). In addition, the STRING Database (http://string-db.org/) was also used to search for InterPro domains, applying Bonferroni correction at P < 0.05 as above. Identified proteins were also analyzed for the presence of secretion signal peptides (SSPs) using the SignalP program embedded within the GiardiaDB (http://giardiadb.org/giardiadb/queries_tools.jsp). The search criteria were set to include any of the following; a SignalP-NN conclusion score of 3, SignalP-NN D-score of 0.5 and SignalP-HMM signal probability of 0.5.

Differential gene transcription of differentiated Caco-2 cells in response to Giardia ESPs
We used RNA sequencing (RNA Seq) or quantitative PCR (qPCR) to identify differential gene expression on the RNA level in IECs (i.e. differentiated Caco-2 cells) exposed to Giardia ESPs in two different experimental setups ( Fig 1B). First, IECs (6-well plate) were exposed to parasite ESPs via a Transwell insert (0.4 μm pore polyester membrane, Corning, Sigma Aldrich). Here, IECs were physically separated from parasite but directly exposed to their ESPs. For this experiment, we used RNA Seq to identify global changes in gene expression. Second, IECs were exposed to 1, 5 or 10 μg/ml of ESPs from axenic cultures. This aimed at assessing changes in pro-inflammatory gene expression in IECs exposed to ESPs released in axenic culture (i.e. no IECs) and assess how IECs respond to different amounts of Giardia ESPs. Changes of gene expression were studied using qPCR. An illustration of the experiments above is depicted in Fig 1B. The experiments below were performed using complete DMEM with 10% HI-FBS and all incubations were performed in a 10% CO 2 humidified incubator at 37˚C. Before the beginning of all experiments, tissue culture flasks or well-plates were washed twice with warm PBS and incubated with fresh media for 2h. Trophozoite cultures (50 ml) were washed once with warm PBS then cold PBS, incubated on ice (10 min), counted and pelleted by centrifugation (930 xg, 10 min, 4˚C). Media collected in the insert experiment were centrifuged (930 xg, 10 min, 4˚C), filter-sterilized (0.45 μM) and stored at -20˚C to be used later for cytokines measurement. Three biological replicates were analyzed with RNA Seq for the insert experiment, whereas four biological replicates were used for qPCR in the other experiment. Each biological replicate had a control (IECs alone in media) and IECs exposed to ESPs of either WB or GS isolate. For RNA collection, IECs were lyzed directly in the tissue culture flasks or well-plates using Trizol reagent (Ambion, Thermo Fisher Scientific) and stored at -80˚C until processed later.
In the insert experiment, IECs were washed and replenished with 4 ml of fresh DMEM added directly into the well. Trophozoites were processed as stated above and resuspended in DMEM, from which 2 ml (1 x 10 7 trophozoites) were added into the Transwell inserts and incubated for 2h or 6h. Media were collected at the end of each time point. In the second experiment, IECs (25 cm 2 flasks) were washed and replenished with 10 ml of fresh DMEM. The amounts of protein in ESPs from either the WB and GS isolates were quantified using the Qubit protein kit (Thermo Fisher Scientific) and added to flasks at a concentration of 1, 5 or 10 μg/ml followed by incubation for 2 h or 6h.

Enzyme linked immunosorbent assay (ELISA) for cytokines in collected media
Media collected from Caco-2-trophozoite interactions were analyzed by ELISA to measure the concentration of cytokines released by IECs in response to parasite ESPs. The Human CXCL8/ IL-8 Quantikine ELISA Kit, Human CCL20/MIP-3 alpha Quantikine ELISA Kit and Human CXCL1/GRO alpha Quantikine ELISA Kit were purchased from R&D Signalling and the samples were processed for cytokine measurements following the manufacturer's recommendations. The results were plotted and analyzed using a four Parameter Logistic ELISA curve fitting provided at http://www.elisaanalysis.com.

RNA extraction, library preparation and RNA sequencing
Samples collected as described earlier were processed using the PureLink RNA Mini Kit (Ambion, Thermo Fisher scientific) according to the manufacturer's instruction. During RNA extraction, a DNaseI treatment step (PureLink DNase Set, Ambion, Thermo Fisher Scientific) was performed on the spin columns to remove genomic DNA. The quality of extracted RNA was assessed by checking the 260/280 and 260/230 ratios using a NanoDrop 1000 Spectrometer (Thermo Fisher Scientific) and by running samples (500 ng) on a 1.5% Tris-Borate-EDTA (TBE) agarose gel prepared with 20 mM of guanidium isothiocyanate (GITC). For cDNA library preparation and RNA Sequencing, RNA (50 ng each) from each sample was reverse transcribed according to the instructions provided in the Ion AmpliSeq Transcriptome Human Gene Expression kit (Revision A.0, Thermo Fisher Scientific) and the generated cDNA was amplified using the Ion AmpliSeq Transcriptome Human Gene Expression core panel (Thermo Fisher Scientific). Primer sequences were partially digested followed by adaptors ligation (Ion P1 Adapter and Ion Xpress Barcode Adapter, Thermo Fisher Scientific) and purification using the Agencourt AMPure XP reagent (Beckman Coulter Inc., CA, USA). Adaptor-ligated amplicons were eluted into the amplification mix (Platinum PCR SuperMix High Fidelity and Library Amplification Primer Mix, Thermo Fisher Scientific) and amplified. Amplicons were subject to size-selection and purification using Agencourt AMPure XP reagent (Beckman Coulter) and were quantified using the Fragment Analyzer instrument (Advanced Analytical Technologies Inc., Ankeny, IA, USA) with DNF-474 High Sensitivity NGS Fragment Analysis Kit (Advanced Analytical Technologies Inc.). Samples were then pooled, followed by emulsion PCR on the Ion OneTouch 2 system with Ion PI Template OT2 200 Kit v3 (Thermo Fisher Scientific) and enrichment using the Ion OneTouch ES (Thermo Fisher Scientific). Samples were loaded on an Ion PI chip Kit v3 and sequenced in the Ion Proton System using the Ion PI™ Sequencing 200 Kit v3 (Thermo Fisher Scientific).

Bioinformatic analysis of RNA sequencing data
Acquired reads from RNA Seq were analyzed using the hg19 AmpliSeq RNA plugin in the Torrent Suite Server with default settings. This analysis tool produces tables with raw read counts as well as normalized expression values for all genes in the Ion AmpliSeq Human Transcriptome panel. For each experiment, the RPKM values were normalized across samples using the moose2 software (https://github.com/grabherr/moose2). Differential expression was estimated based on Laplace/Normal distributions calculated from replicate samples, computing individual significance values as Genes with significant change in the level of transcription were also subject to GO Terms analysis (http://geneontology.org/page/go-enrichment-analysis) to identify biological or molecular functions as well as protein class. We used the GO slim option with Bonferroni correction applied for multiple testing (P < 0.05).

Quantitative polymerase chain reaction (qPCR)
For qPCR experiments, RNA was extracted as above and cDNA was synthesized from 1μg RNA using the RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) following the manufacturer's instructions. qPCR was performed on the genes, il-8, cxcl1-3, ccl2 and ccl20 with gapdh used as an endogenous control. The Maxima SYBR Green/ROX qPCR Master Mix (Thermo Fisher Scientific) was used in the qPCR reactions with 250 nM of each primer and 10 ng template. Cycling conditions included an initial polymerase heat activation step (10 min, 95˚C), and 40 cycles of 15 sec denaturation (95˚C), 30 sec annealing (60˚C) and 30 sec extension (72˚C). The change in gene transcription was calculated using the ΔΔCt method.
Activator protein 1 (AP-1) reporter plasmid luciferase assay AP-1 is a transcription factor involved in inducing the transcription of inflammatory genes [28,29]. It is activated by MAPKs and could work in conjunction with NFκB to activate gene transcription [28,29]. To detect AP-1 activation in response to Giardia ESPs, Caco-2 cells were transfected with a plasmid pGL4.44[luc2P/AP1 RE/Hygro] expressing the luciferase reporter gene luc2P (Photinus pyralis) under the control of a promoter that has the transcription response element of AP-1. The pGL4.75 Vector (encoding Renilla luciferase) was used as a normalisation control and was co-transfected into Caco-2 cells with the pGL4.44 plasmid. Both plasmids were purchased from Promega (WI, USA). Transfections were performed in a 24-wells plate format according to the supplier's protocol, using 1.1μg of plasmid DNA/10 5 cells and the FuGENE HD reagent (Promega). In the beginning of experiment, both parasite and Caco-2 cells were washed and processed as described earlier and the wells were replenished with 500 μl DMEM (no phenol red as it interferes with luminescence reading). Trophozoites were re-suspended in DMEM (no phenol red) and 10 6 trophozoites were added into the Transwell insert in a total volume of 100 μl and incubated for 2h or 6h (37˚C, 10% CO 2 ). Three wells were used for each time point and the controls. At the end of each incubation, the inserts were removed and the cells were washed (3x) with 200 μl PBS at RT, and lyzed with an equal amount of the Dual-Glo Luciferase Assay reagent. A 100 μl was dispensed into the well of 96-well plate in triplicates and luminescence was read from both plasmids following the instructions of Dual-Glo Luciferase Assay reagent and using an Infinite M200 Pro Tecan machine (Tecan Group Ltd., Männedorf, Switzerland). Cells with no plasmid (background fluorescence), cells with plasmid alone (negative control), and cells incubated with Phorbol 12-myristate 13-acetate (PMA) (positive control) served as controls. Three biological replicates were performed in this experiment.

Western blot analyses
To correlate RNA Seq results with those at protein level, we studied the effects of Giardia ESPs on differentiated Caco-2 cells using Western blotting. We tested for the activation/ inhibition of mitogen activated protein kinases (MAPKs) and the nuclear translocation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), both of which are involved in inflammatory gene transcription [30,31]. Briefly, the insert experiment was repeated as described earlier with the inclusion of three washes with warm PBS (37˚C) at the end of each incubation. In the last wash, cells were scraped off and pelleted by centrifugation (930 xg, 10 min, 4˚C) and the pellet was lyzed using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) according to supplier's instructions. Both Halt Protease and Phosphatase Inhibitor Cocktails (Thermo Fisher Scientific) were added to lysis buffers during extractions. Cytoplasmic and nuclear fractions were aliquoted and stored at -80˚C until later use. To confirm that Giardia ESPs can modulate/inhibit immune signaling in IECs, we repeated the experiment above but IECs were previously treated with lipopolysaccharide (LPS, Sigma-Aldrich) (10 μg/ml) or tumour necrosis factor-alpha (TNF-α, Sigma-Aldrich) (100 ng/ml) for 2h prior adding trophozoites into the inserts. Trophozoites in inserts and IECs, in the presence of inflammatory signal, were left to interact for 2h and 6h, upon which IECs were processed as above for protein collection. IECs treated with LPS and TNF-α were tested for MAPKs activation/inhibition and the nuclear translocation of NF-κB. Protein samples collected as described above were used for SDS-PAGE and Western blots. Briefly, 30 μg of total proteins were electrophoresed on AnykD-stain-free gels (Bio-Rad) (100 v, 4˚C) and transferred onto a PVDF membrane (100 v, 90 min, 4˚C). Membranes were washed with Tris-Buffered Saline with 0.1% Tween-20 (TBST) and blocked with 5% non-fat dry milk in TBST for 1h at RT, followed by incubation with primary antibodies overnight (4˚C). Blots were then washed with TBST (3x) and incubated for 1h at RT with a secondary anti-mouse HRP conjugated or anti-rabbit HRP conjugated antibodies (1:5000 dilution, Cell Signaling Technology, MA, USA). Blots were developed using the Clarity ECL Western Blotting Substrate (Bio-Rad). For re-probing, blots were stripped using a mild (200 mM glycine, 0.1% w/v SDS, 1% Tween 20, pH 2.2) or harsh stripping buffer (2% SDS, 62.5 mM Tris, pH 6.8, and 114.4 mM beta-mercaptoethanol) and probed for the loading control, which is either an anti-human GAPDH (cat no. MAB5718, R&D Systems, Inc., MN, USA), anti α-tubulin antibody (Cell Signaling Technology) or TATA box binding protein (TBP) for nuclear samples (MA5-14739, Thermo Fisher Scientific). To detect the activation of MAPKs (ERK1/2, P38 and JNK), the following antibodies were used against the phosphorylated and non-phosphorylated forms of kinases: phospho-p44/42 MAPK (Erk1/2) (cat no. 8544) and p44/42 MAPK (Erk1/2) (cat no. 4348), Phospho-p38 MAPK (cat no. 4511) and p38 MAPK (cat no. 8690), Phospho-SAPK/JNK (cat no. 4671) and SAPK/JNK (cat no. 9258), and NF-κB1 p105/p50 (cat no. 13586). Antibodies were purchased from Cell Signaling Technology and used at the dilution recommended by the supplier.

Immunofluorescence of labelled ESPs and confocal microscopy
For ESPs labelling with Alexa Fluor 488, proteins collected from axenic culture were washed (3x) with cold PBS during sample concentration to remove free amino acids and antioxidants that might interfere with protein labelling (recommended in the kit). ESPs (45μg) were incubated with 11.3 nmol/μl of Alexa Fluor-488 and excess dye was removed by gel filtration following the instructions in the Alexa Fluor 488 Microscale Protein Labelling Kit (Thermo Fisher Scientific).
To visualise ESPs interaction with host cells, chamber slides were washed and replenished with fresh DMEM as above, followed by adding labelled ESPs to the wells at 1 or 5 μg/ml and incubation for 2 h or 6 h (37˚C, 10% CO 2 ). The amounts of labelled proteins were quantified using a NanoDrop 1000 Spectrometer (Thermo Fisher Scientific). Caco-2 cells incubated in DMEM only (no Labelled ESPs added) served as a control. Slides were washed, fixed (4% paraformaldehyde in PBS) and mounted (Vectashield anti-fade reagent with DAPI, Vector Laboratories, CA, USA), excluding the blocking step. The fluorescent label on IECs was visualised using a Zeiss LSM700 confocal microscope (Zeiss, Oberkochen, Germany) at the BioVis Core Facility, SciLifeLab, Uppsala, Sweden.

Statistical analyses
The results from the cytokine measurements and differential gene expression for qPCR results were analysed using a one-way analysis of variance (ANOVA) to assess overall significance among the tested groups at α < 0.05. When significant, the values from the replicates were processed through the Bonferroni-Holm's and significance was marked at P < 0.05.

Results
The secretome of Giardia trophozoites in the absence and presence of host cells In order to identify proteins in the Giardia secretome, we collected ESPs from Giardia trophozoites grown axenically or in the presence of IECs ( Fig 1A). The secretomes of axenic WB (assemblage A) and GS (assemblage B) Giardia trophozoites were analyzed in a modified serum-free medium (see Materials and methods) that could support the full viability of trophozoites for 2h and estimated viabilities of 98.74% ± 1.03 (WB, n = 5) and 98.32% ± 0.81 (GS, n = 5) under 6h of incubation. In total, 196 proteins were identified in the secretome of the WB isolate with 149 proteins detected in 2h culture and 180 in 6h culture; 133 proteins overlapped in both time points (Fig 2A). For the GS isolate, 155 proteins were identified including 141 proteins from the 2h culture and 83 proteins from 6h culture; 69 proteins overlapped in both time points (Fig 2A). The reproducibility of the secretome profiles was conserved in the three biological replicates. Full lists of identified proteins, including tryptic peptides, score, coverage and number of peptide spectrum matches (PSM) are presented in S1 Table (WB) and  S2 Table (GS). The secretomes of both isolates were also compared leading to the identification of 122 common secreted proteins (S3 Table).
We also collected and analyzed the medium from Giardia trophozoites incubated with differentiated Caco-2 cells (Fig 1A). In total, 248 WB proteins were identified in the co-culture medium, 156 at 2h, 240 at 6h and 148 in both time points (Fig 2B, S4 Table). For the GS isolate, 152 proteins in total were identified, 122 at 2h, 128 at 6h and 98 at both time points (Fig 2B, S5  Table). The top 50 proteins, based on their peptide score, are listed in Table 1. These proteins overlap with those in axenic culture (S1 and S2 Tables). A comparison was performed between the secretomes of WB trophozoites incubated with (S4 Table) and without IECs (S1 Table) through which 161 common proteins and 87 interaction-specific proteins were identified (Table 2). Similarly, an analysis of the secreted proteins from GS isolate with (S5 Table) and without IECs (S2 Table) identified 111 common proteins and 41 proteins specific to the interaction ( Table 2). Amongst the proteins specific to interactions, 13 proteins overlap between the two isolates (denoted in Table 2) whereas 74 WB trophozoite proteins and 28 GS trophozoite proteins are exclusive to the interaction process and represent an isolate-specific response to IECs. This overall shows that most of the secreted Giardia proteins during interactions with IECs are conserved between the WB and GS isolates but there are assemblage-specific differences.
Secreted Giardia proteins in co-culture were analyzed for the enrichment of certain biological functions and for protein class. For the GS isolate, since the current online algorithms do not support GO analysis, orthologous genes in WB were used to perform the quest. The results are presented in Fig 2D and 2E. The largest number of detected proteins have functions associated with metabolism (85 proteins in WB and 64 in GS) with different subgroups in the child  linages associated with protein metabolism, amino acid catabolic and metabolic processes, glycolysis, monosaccharide metabolic process and the generation of precursor metabolites and energy ( Fig 2D). These metabolic groups were also seen in the secretome of axenic cultures (S1 Fig). Other metabolic functions such as the hexose monophosphate shunt and arginine metabolism could also be identified when secreted proteins were manually delineated into different metabolic pathways (S1 Fig). Furthermore, a nuclease with endonuclease I activity (Table 1), possibly involved in nucleic acids salvaging, and proteins involved in pyrimidine salvage (cytidine deiminase and uridine phosphorylase (UPL-1)) as well as in lipid metabolism were identified in secretome of co-culture (S1 Fig). The lipid metabolic proteins included enzymes involved in phospholipids (PLs) remodelling and salvaging fatty acids (phospholipase B, PLB), producing monoacylglycerol (lysophospholipid phosphatase) and generating inositol-3 phosphate (I3P, I3P synthase), the precursor for all inositol containing compounds including PLs [32]. These result show that the major part of Giardia secretome is involved in different metabolic functions. The majority of proteins in the two isolates metabolo-secretome fell under four major categories: the hydrolase, dehydrogenase, oxidoreductase and protease groups. The oxidoreductase group contained 25 proteins in WB and 14 proteins in GS (Fig 2E). Compared with axenic culture (14 proteins) (S2A Fig), the WB isolate released 11 additional proteins with oxidoreductase activity including peroxidases when incubated with IECs (e.g. PFOR, NADH oxidase lateral transfer candidate, peroxiredoxin 1, FixW protein alcohol dehydrogenases (ALDs), glutamate dehydrogenase and glutamate synthase and a HP GL50803_9355, Table 2). This suggests a response to counteract IECs oxidative defences such as reactive oxygen species. It has been suggested that oxidative stress provides the trigger to initiate encystation [33] and proteins involved in cyst wall synthesis (e.g. glucosamine 6-phosphate deaminase and Nacetylglucosamine 6-phosphate mutase) were released by WB isolate interacting with IECs, corroborating this notion. For GS, on the other hand, peroxidases and antioxidant proteins were already secreted by trophozoites without exposure to IECs (S2 Table) and thus the results above show that the two isolates exhibit different anti-oxidative responses during interaction with IECs.
The protease group contained 13 secreted proteins in both isolates ( Fig 2E) and harboured the cysteine-(cathepsins B and L), serine-(Alanyl dipeptidyl peptidases) and metallo-type (Aminoacyl-histidine dipeptidases and Xaa-Pro dipeptidases) of proteases. For WB isolate, two secreted metalloproteases (GL50803_8407 and GL50803_3822) were released in response to IECs whereas the rest of proteases were the same as those in axenic culture. For GS isolate, the same proteases were detected in both co-culture and axenic culture with two serine proteases (GL50581_3181 and GL50581_2704) being non-orthologous to any WB isolate proteins.
Many of the proteins detected in our analysis have been previously reported as immunoreactive proteins during human giardiasis (S6 Table) [21]. Amongst these proteins, the variable  Characterization of the Giardia secretome surface proteins (VSPs) are immunodominant during infection [2] and together with the high cysteine membrane proteins (HCMPs) they occupied a large portion of the WB isolate secretome (23 and 12, respectively, S1 Table). In the GS isolate secretome, lower numbers of VSPs and HCMPs were detected (4 and 5, respectively, S2 Table), more likely due to the fragmented genome used in protein identification. VSPs and HCMPs are cysteine-rich proteins and those detected in our analysis have an epidermal growth factor (EGF)-like domain in common (InterPro:IPR000742, 39 in WB and 10 in GS) which is also found in other secreted cysteinerich proteins (e.g. tenascins, high cysteine proteins (HCPs) and neurogenic locus Notch protein precursor, S1 and S2 Tables). EGF-like domains are often present in secreted protein [34] and thus, this promoted us to look for proteins with secretion signal peptides (SSPs). A genome-wide screen in the Giardia database showed that 1361 proteins of the WB isolate have SSPs, amongst which 82 proteins were detected in our co-culture secretome analysis (~29%, S7 Table) with 41 of those being N-glycosylated [35] (S7 Table). For the GS isolate, 422 proteins in Giardia DB contain SSPs, 44 were detected in our analysis (~29%, S7 Table). The majority of proteins with SSPs (S7 Table)

The secretome of differentiated Caco-2 cells during parasite interactions
The secretome of differentiated Caco-2 cells incubated with WB trophozoites was analyzed for proteins released in co-culture. Overall, 384 Caco-2 cell proteins were identified in the medium of interaction with 21 proteins secreted at 2h, 136 at 6h and 227 at both time points (Fig 2C). The identified proteins, together with the number of peptides, score, coverage and PSMs are presented in S4 Table. A comparison between the secretomes of Caco-2 cells incubated alone (i.e. control, S8 Table) and with WB trophozoites (S4 Table) identified 308 overlapping proteins (S9 Table) and 76 proteins released in response to WB trophozoites ( Table 3). The secretome of Caco-2 cells incubated with GS trophozoites was also analyzed for the proteins released by host cells in response to interaction. In total, 355 proteins were secreted in co-culture, 60 proteins were secreted at 2h, 78 at 6h and 217 at both time points (Fig 2C). The proteins identified, together with the number of peptides, score, coverage and PSMs are presented in S5 Table. The top 50 proteins, based on their peptide scores, are presented in Table 4. A comparison between the secretomes of co-culture (S5 Table) and control (S8 Table) identified 310 common proteins (S10 Table) and 45 proteins that were specifically released in response to GS trophozoites (Table 3). Amongst the proteins specific to interaction (45 proteins) 31 proteins overlapped in the response to both isolates. GO enrichment analyses of parasitized Caco-2 cell secretomes (both isolates, Fig 2F and  2G) revealed that the majority of proteins have metabolic functions (152 protein in response to WB and 135 in response to GS) such as protein metabolism (including proteolysis), carbohydrate metabolism (including monosaccharides and polysaccharide metabolism as well as glycolysis), nucleobase-containing compounds metabolic process and lipid metabolism (only Table 3 WB). The same metabolic groups were also seen in the control Caco-2 cell secretome (S2C Fig). One of the big groups that emerged in our analysis was the cytoskeletal proteins (41 proteins in response to WB isolate and 32 proteins in response to GS). This group included the intermediate filament and the actin family cytoskeletal proteins (including non-motor actin binding protein in the child lineage). Except for the actin family cytoskeletal protein group which was not enriched in control Caco-2 cell secretome, the rest of cytoskeletal protein

Table 4. Top 50 identified proteins of secreted by differentiated Caco-2 cells co-incubated with Giardia intestinalis isolates, WB and GS in vitro.
The proteins are ranked based on peptide score. Note that all listed proteins were also detected in Caco-2 cells culture without Giardia. groups contained more proteins compared to the control secretome. This clearly indicates the effect of Giardia colonization on IECs actin cytoskeleton. An interesting finding is that seven microvilli proteins were detected in the secretome of Caco-2 cells incubated with WB trophozoites (GO:0005903, intestinal epithelial brush border) (e.g. villin-1, intelectin-1, myosin heavy chain-9 and -14, beta 1,4-galactosyltransferase polypeptide 1, plectin and LIM domain and actin binding 1). These proteins might have been cleaved off/dissociated from damaged microvilli as a result of trophozoite attachment. Two of the proteins that we identified in the parasitized Caco-2 cell secretome are associated with immunological functions. The first protein is a cytokine called macrophage migration inhibitory factor (MIF) ( Table 3). MIF is a pleiotropic cytokine that regulates the expression of inflammatory genes (e.g. TNF-α, cyclooxygenase 2 and inducible NO synthase (iNOS)) [39], indicating its involvement in innate immune responses. The second protein is complement c3 identified in the secretome of Caco-2 cell incubated with GS only ( Table 3), suggesting differences in Giardia isolates abilities to activate the complement system.
In conclusion, we have identified the major secreted proteins of host and parasite during interaction. Our analysis showed that both, the host and parasite, release proteins with similar functions (sugar, protein, lipid and nucleic acid metabolism). Both assemblages/isolates (A and B) induced similar responses but there are assemblage-specific differences that might explain observed differences in infectivity and symptoms.

Transcriptional response of IECs to Giardia ESPs
Interaction between Giardia trophozoites and IECs in vitro induces the transcription of a large number of genes including the chemokines ccl2, ccl20, and cxcl1-3 that are up-regulated up to 100-fold at the RNA level [40]. To test whether Giardia ESPs are capable of inducing similar effects, differentiated Caco-2 cells were exposed to Giardia ESPs through Transwell inserts (Fig 1B) for 2h and 6h. Differentially expressed genes (DEGs) were identified by RNA sequencing (RNA Seq) (S11 Table). Overall, 120 and 87 genes were differentially expressed (DE) in IECs exposed to ESPs of WB and GS trophozoites, respectively, with a global gene expression profile showing differences between the two time points and between the two isolates (S11 Table, Fig 3A).
At 2h, seven genes were DE in IECs exposed to ESPs of the WB isolate, six of which were up-regulated and one gene (il8) was down-regulated ( Table 5). The up-regulated genes encode immediate early genes [41], including the protein constituents of AP-1 transcription factor (FBJ murine osteosarcoma viral oncogene homolog, fos and jun B proto-oncogene, junb). To verify AP-1 activation, Caco-2 cells we transfected with a plasmid expressing a luciferase gene under the control of AP-1 transcription response element and luminescence was measured upon cell exposure to Giardia ESPs. At 2h, no significant change was observed in the AP-1 promoter activity with ESPs of either isolates (1.07 for ESPs of WB isolate and 1.12 for ESPs of GS isolate, P > 0.05) (Fig 3B). At 6h, on the other hand, despite the slight increase in AP-1 promoter activity (1.17 for ESPs of WB isolate and 1.15 for ESPs of GS isolate), the average fold change was significant (n = 4, P < 0.05) (Fig 3B). These findings are in line with the RNA Seq results and indicate that AP-1 is activated in response to Giardia ESPs. The gene encoding DUSP1 was the highest up-regulated gene at 2h and the only up-regulated gene in IECs exposed to ESPs of GS trophozoites (Table 5). Its transcriptional level increased at 6h together with dual specificity phosphatases 4 and 5 (dusp4 and dusp5), and tribbles pseudokinase 1 (trib1) ( Table 5). The products of these genes play important roles in the regulation of MAPKs' activity [42][43][44]. The majority of DEGs (120 in WB and 87 in GS) were up-regulated at 6h (S11 Table) with a higher fold change for DEGs compared to 2h (Table 5, S11 Table). One exception was that the increase in RNA levels il8 in of IECs exposed to GS trophozoite (1.95 fold) was not significant. Insulin induced gene 1 (insig1) was the highest up-regulated gene in IECs in response to ESPs of both isolates at 6h (Table 5) and this gene encodes a protein that regulates cholesterol metabolism, lipogenesis, and glucose homeostasis [45][46][47]. More genes associated with glucose uptake and response to glucose starvation (Table 5) were DE at 6h, indicating that Giardia releases ESPs that interfere with glucose homeostasis. Nuclear receptor subfamily 4 group A member 1 (nr4a1) was the second top up-regulated gene (ESPs of both isolates, Table 5) and this gene has functions associated with cell cycle, inflammation, and induction of apoptosis [48][49][50]. Another two cell cycle genes associated with response to DNA damage, were amongst the top ten up-regulated genes and these include BTG family member 2 (btg2) [51] and growth arrest and DNA damage inducible alpha (gadd45a) [52] (ESPs of both isolates, Table 5). This suggests that Giardia ESPs induce damaging effects on host DNA and arrests IECs in the cell cycle. The tight junction proteins, claudin 3 and 4 (cldn 3 and cldn 4), were also up-regulated (Table 5) and these proteins play essential roles in the maintenance of the intestinal epithelial barrier (IEB) [53].
To obtain a general overview on the function of DEGs, they were processed through the GO Database for biological and molecular functions. For IECs exposed to ESPs of the WB isolate, cell proliferation (GO:0008283) and MAPK cascade (GO:0000165) constituted the biggest biological groups encompassing 11 genes each. Both chemokines (e.g. CXCL1-3 and IL-8) and growth factors (e.g. amphiregulin, proepiregulin, Proheparin-binding EGF-like growth factor) were contained within the cell proliferation group. All chemokine genes, including ccl2 and ccl20, were listed under chemokine activity (GO:0008009) and cytokine receptor binding (GO:0005126) by molecular function. In addition to these cytokines, the gene clcf1 encoding cardiotrophin-like_cytokine_factor_1 was DE in IECs exposed to ESPs of WB trophozoites. CLCF1 is an IL-6 family cytokine capable of activating NFκB and exerts stimulatory functions on B cells [54]. For MAPK cascade, it was interesting to see genes encoding RNA binding proteins (zfp36 and zfp36l2) within this group, both of which are associated with anti-inflammatory functions [55]. Another DE gene is Regnase 1 (ZC3H12A), an RNase that controls inflammatory responses by inducing mRNA decay of specific cytokine transcripts [56]. The rest of functional groups in this analysis included response to external stimulus (GO:0009605), locomotion (GO:0040011), localization (GO:0051179), behaviour (GO:0007610) and cell death (GO:0008219). For IECs exposed to ESPs of the GS isolate, the groups cell proliferation, MAPK cascade, cell death and endoderm development (GO:0007492) emerged by biological functions and no specific groups emerged by molecular function analysis. Only cxcl2, cxcl3 and clcf1 were significantly induced in IECs exposed to ESPs of the GS trophozoites. Overall, based on the transcriptomic profile we could identify a general response in IECs exposed to Giardia ESPs, which involves inducing and regulating inflammatory responses, glucose homeostasis, apoptosis, cell cycle, and maintenance of IEB.
Next, we examined which amount of ESPs could induce the transcription of pro-inflammatory genes in IECs. ESPs collected from axenic culture (1, 5 and 10 μg/ml) were incubated with IECs for 2h or 6h (Fig 3C) and RNA levels were assessed with qPCR. At 2h, 5 μg/ml of WB isolate ESPs induced a 2 to 2.9-fold increase in RNA levels of il8, ccl20, and cxcl1-3 but not ccl2, interleukin-8 (IL-8), chemokine (C-X-C motif) ligand 1 (CXCL1) and chemokine (C-C motif) ligand 20 (CCL20) in the medium of differentiated human colon carcinoma cell line, Caco-2, exposed to ESPs of G. intestinalis WB or GS isolate through a Transwell insert. No statistical significance was found between controls and treatments (P > 0.05). https://doi.org/10.1371/journal.pntd.0006120.g003 Characterization of the Giardia secretome Table 5. Selected groups of differentially expressed genes in differentiated human colon carcinoma Caco-2 cells exposed to excretory-secretory products of Giardia intestinalis isolates WB and GS via Transwell insert. Bold fonts indicate a significantly up-regulated gene whereas italic fonts indicate a significantly down-regulated gene.

Symbol
Gene scoring the highest for cxcl3 (Fig 3C). The fold change, however, was statistically insignificant for all the genes tested (P > 0.05). The fold change in RNA levels remained below 2 for 1 μg/ml (not shown) and 10 μg/ml of ESPs (Fig 3C). At 6h, all genes were significantly up-regulated (2.84-8.46 fold) in IECs incubated with 5 μg/ml WB isolate ESPs, scoring the highest for ccl20 (P < 0.05) and the genes ccl2, ccl20, cxcl1 and cxcl3 remained significantly up-regulated (2.84-4.76 fold) when IECs were incubated with 10 μg/ml of ESPs (Fig 3C). For ESPs of the GS isolate, the fold change at RNA level for all the genes tested at the three concentrations of ESPs was below 2 (P > 0.05) (Fig 3D). Nevertheless, at 6h, il8, ccl20, cxcl1 and cxcl3 were significantly up-regulated (4.23-4.71 fold) with il8 being significantly up-regulated at all ESPs concentrations tested and ccl20 and cxcl1 at 10 μg/ml (Fig 3C). The results show that ESPs of both isolates induce variable transcriptional responses of inflammatory genes in IECs and these responses differ with the amounts of ESPs IECs are exposed to. However, direct host-cell interactions have a much stronger effect on the induction of chemokine gene transcription [40].
To assess whether the up-regulated chemokine genes were translated into their protein products, we selected three chemokines (IL-8, CXCL1 and CCL20) and measured their levels in the collected media from the insert experiment (Fig 3D). The amounts of IL8, CXCL1 and CCL20 were very low and their concentration was similar or lower than their controls at 6h (50.2, 42.1 and 57.9 pg/ml, respectively). A slight increase in the amounts of CXCL1 and CCL20 amounts could be seen at 2h upon IECs exposure to ESPs of the WB trophozoites (12.9 pg/ml in control versus 18.2 pg/ml) and GS trophozoites (24.7 pg/ml in control versus 27.1 pg/ ml), respectively (P > 0.05) (Fig 3D). Overall, the levels of measured chemokines were either close to control or slightly lower and no significant differences could be seen between IECs exposed or unexposed to ESPs (P > 0.05) (Fig 3D). The low levels of measured chemokines indicate that their mRNA levels are post-transcriptionally regulated or the chemokines being degraded upon release from IECs.

The effect of ESPs on cell signaling
Giardia ESPs from axenic cultures were labelled with Alexa Fluor488 and added to the medium to visualize their interaction with IECs. Confocal images taken upon 2h and 6h of incubation with IECs showed that the labelled ESPs of WB and GS trophozoites bind to the IECs surface as well as the intercellular space (i.e. intercellular junctions) and some are internalized into differentiated Caco-2 cells (Fig 4). This indicates the presence of ligands or receptors on IECs surface and a potential of the internalized ESPs to modulate cellular processes and signaling.
We consequently examined MAPK signaling in IECs exposed to ESPs via inserts because this pathway emerged in the functional analysis of RNA Seq data. Western blot analyses showed a slight decrease in the phosphorylation levels of ERK1/2 and P38 at 2h but an increased level of nuclear translocation of NFκB compared to control (Fig 5A). Despite the Characterization of the Giardia secretome   Characterization of the Giardia secretome slight decrease in ERK1/2 and P38 phosphorylation, their roles in inducing the nuclear recruitment of NFκB cannot be excluded and it indicates a possibility of other factors contributing to NFκB activation (e.g. growth factors, stress, hypoxia or nutrient depletion). At 6h, we could see a marked reduction in the phosphorylation levels of ERK1/2 and P38 compared to control and this coincided with a decrease in the levels of NFκB nuclear translocation ( Fig 5A). The nuclear translocation of NFκB, however, was not completely abolished. The phosphorylated form of JNK could not be detected (S3A Fig). This shows that Giardia ESPs actively modulate MAPK signaling to avert the induction of strong inflammatory responses and this effect is built up upon prolonged exposure of IECs to ESPs. To test whether Giardia ESPs can also inhibit inflammatory signaling, we incubated IECs with LPS or TNF-α for 2h, then added Giardia trophozoites into Transwell inserts in the persistence of inflammatory stimuli for 2h or 6h. While the controls exhibited increased phosphorylation of ERK1/2 and P38 as well as nuclear recruitment of NFκB, these levels were reduced in inflamed tissue upon exposure to Giardia ESPs at both time points (Fig 5B). These results show that Giardia ESPs can attenuate inflammatory signaling mediated by MAPK phosphorylation and NFκB recruitment. It is possible that the parasite senses IECs inflammatory signals and release specific protein factors to subvert such effects.

Discussion
Several recent studies have shed the light on the secretomes of different protozoan parasites and their role in parasite physiology, adaptation to host environment, pathogenicity and immune modulation [36,37,[57][58][59]. To date, very few Giardia ESPs have been identified, and thus we devised experiments to characterise Giardia secretome, identify secreted proteins that may be involved in virulence and better understand disease mechanism (Fig 1). We also assessed the effect of Giardia colonization on host cells and identified IECs secretory responses.
Our analysis highlighted metabolic functions as a dominant secretory response when Giardia trophozoites interact with IECs (Fig 6). These functions involve carbohydrate (including glycolysis), protein (including arginine metabolism), lipid (phospholipid remodelling) and nucleic acid (purines and pyrimidines salvage) metabolism. One striking observation is that the same types of metabolic proteins are released by Giardia and host cells during interactions, indicating competition for acquiring nutrients (Fig 6).
The release of glycolytic enzymes has been previously reported in Giardia [17] and in many parasites [60] including Trichomonas vaginalis [61], T. brucei [36] and L. donovani [37]. Secreted glycolytic enzymes of parasites have functions associated with virulence, but those of Giardia require further investigation.
The protein metabolic group contained proteins similar to those secreted by L. donovani and T. brucei [36,37]. In this group, a large number of CPs were identified whose activity is associated with disease induction and immune modulation [12,15,62]. Herein, we report the identity of secreted CPs (Fig 6) as well as other proteases (e.g. serine and metalloproteases). We also showed a variation in protease numbers between the isolates; an aspect underlying virulence [63]. Little is known about the serine and metalloproteases of Giardia but studies in Entamoeba invadens [64], Acanthamoeba castellini [65], Acanthamoeba lugdunensis [66] and T. brucei [67] have identified some roles in differentiation [64,68], pathogenesis (e.g. degradation of ECM and immunoglobulins) [64][65][66][67] and invasion [69]. It will be interesting to characterize the function of these secreted proteases during Giardia infections.
Giardia trophozoites cannot synthesize purines and pyrimidines de novo and rely on exogenous sources for their acquisition [70]. UPL-1 is secreted by Giardia and its gene is up-regulated during interaction with IECs [33,71]. This enzyme is involved in pyrimidine salvage and it is highly active in the small intestines [72] corroborating the notion of host-parasite competition for nutrients (Fig 6). UPL-1 is also an immunogenic protein recognized by sera from giardiasis patients, indicating its potential as a diagnostic and vaccine target [21]. Interacting Giardia and IECs also secrete nucleases but which nucleases' activity can surpass the other, it is unknown. Nucleases harness nucleic acids for salvaging purines and pyrimidines as reported for E. histolytica [73], L. donovani [37] and T. brucie [36]. Nucleases of T. brucei, have also been proposed to attenuate nucleic acids-induced inflammatory responses [36].
Giardia trophozoites cannot synthesize lipids de novo and acquire those from the host [32]. PLB is an enzyme secreted by the parasite and IECs into the interaction medium (Fig 6). PLB is involved in PLs remodeling in Giardia and it localizes to brush border membranes of mature enterocytes in humans [74]. In virulent Candida albicans strains, PLB is suggested to promote host cell damage and lysis, and the deletion of plb attenuated Candida virulence in murine models [75]. I3PS is another ESP secreted by Giardia and Caco-2 cells. I3PS synthesises I3P, the precursor for all inositol containing compounds including PLs [76]. The deletion of i3ps in Leishmania renders the parasite avirulent in vivo [77] but whether it contributes to Giardia's virulence this requires investigation. Since the above proteins are required for parasite nutrition and thus survival, it will be interesting to test their potential targets for drugs and/or vaccines.
Giardia releases proteins with antioxidant functions during growth and interaction with IECs (Fig 6). This highlights the microaerophilic nature of Giardia [78] and the mechanism by which the parasite maintains a low redox potential extracellularly leading to enhanced intracellular homeostasis. When readily released by the parasite or in response to interaction with IECs, antioxidant proteins can promptly attenuate host ROS during infection as suggested for Characterization of the Giardia secretome E. histolytica [79][80][81]. We previously showed that the transcriptional levels of antioxidant genes vary between the WB and GS isolates during oxidative stress [82] and herein we show differences in the array of secreted antioxidant proteins, further pointing out the differences in oxidative stress responses between assemblage A and B Giardia parasites.
Giardia escapes host and environmental stresses by cysts formation and thus encystation is regarded as a mean of survival for the parasite. In response to IECs, both Giardia isolates released proteins involved in cyst wall formation (Fig 6) and these proteins were not detected in axenic culture indicating that host cells provide the signal to initiate encystation. Interestingly, in a recent study that investigated transcriptional changes in bioluminescent WB isolate trophozoites from high density foci in the mice intestines, the authors reported the up-regulation of encystation, oxidative stress and metabolic genes (carbohydrate, protein, lipid and nucleoside base metabolism) [83]. Herein, we show that the protein products for many of upregulated genes in that study are being released into the interaction medium. Nevertheless, we did not see any cyst formation in our in vitro model, indicating that stress alone might not be a sufficient stimuli to complete encystation successfully and possibly requires other factors like bile [84].
Giardia VSPs, HCMPs and tenascin were abundant proteins detected in the secretome of both isolates (Fig 6). VSPs are constantly released from trophozoite surface and they are involved in immune evasion during infection [2,20]. HCMPs are similar to VSPs but lack the c-terminal tail. So far, there is no reports on HCMPs function in Giardia except for being transcriptionally up-regulated during interaction with IECs [33,40,71]. Tenascins are also upregulated during interaction with IECs [85]. Despite the lack of true similarity with human tenascins, the presence of an EGF-like domains in these glycoproteins suggests their involvement in protein-protein interaction, cell signaling or adhesion [86,87]. Tenascins are also able to bind lectin [88,89] and thus a role for Giardia tenascins in host innate immunity have been suggested previously [85]. The fact that 8 out 10 WB isolate "tenascins" are secreted, glycosylated and up-regulated during interaction with IECs [85] might suggest interesting roles during Giardia infection.
Only a small percentage of proteins in the parasite's secretome were identified with SPP. Recently, several arguments have supported the hypothesis that many parasitic proteins are released in EVs, containing cytosolic and plasma membrane proteins but not proteins from intracellular organelles [60]. In this study, we provided EM evidence on the presence of EVs from Giardia trophozoites (S2 Fig). The sorting of EVs cargo, however, is not well understood but the suggestion is that posttranslational modifications (e.g. glycosylation, ubiquitination, phosphorylation or acylation) could target proteins for vesicular secretion [36,60,90]. Whether this is the case in Giardia, this requires further investigation.
Functional analysis of parasitized IECs secretome indicated that Giardia trophozoites induce cytoskeletal changes in IECs. It is known that Giardia adheres to IECs very strongly, leaving marks on cell surface and affecting the organisation of the actin cytoskeleton [91][92][93]. One interesting finding is that Giardia effect on microvilli structure was pronounced by the detection of seven microvilli proteins in the co-culture medium, one of which is villin. Villin plays an essential role in reorganizing microvilli actin filaments as well as actin bundles assembly, capping and severing [94,95]. Villin binds phosphatidylinositol 4,5-bisphosphate (PIP2), enhancing actin bundling and lysophosphatidic acid preventing all actin modifying activities [96,97]. This protein, however, is cleaved during Giardia infection, uncoupling its protective role from the actin cytoskeleton [62].
The secretome of parasitized IECs contained host proteins with immunological function. Examples are the MIF and complement factor c3. MIF is a pleiotropic cytokine that induces inflammatory responses mediated by TNF-α, interferon-γ (IFNγ), IL-1β, IL-12, IL-6 and, IL-8, among others [39]. MIF plays a role in resistance to infection with Trypanosoma cruzi [98], T. gondii [99] and Leishmania major [100] but many studies have also reported its contribution in disease pathology [101][102][103]. It will be interesting to identify its role during Giardia infections. The complement factor c3 was specifically released by IECs in response to GS isolate, which was previously shown to activate the complement system [104]. In fact, the complement system contribute to trophozoite killing both in vitro and in vivo [105,106] and mice with nonfunctional component system (deficient in mannose-binding lectin 2 or complement factor 3a receptor [107]) exhibit delayed parasite clearance [104]. Therefore, while these results show the induction of innate immune responses by parasitized IECs, it also shows that these responses might vary in infections with different isolates/assemblages.
To our knowledge, this is the first study that has investigated transcriptional changes in IECs exposed to Giardia ESPs. Interestingly, transcriptome data brought about and highlighted important pathological processes previously shown to occur during Giardia infection. Specifically, the disturbance in glucose homeostasis [108,109] and IEB integrity [91,92], cell cycle arrest [27] and apoptosis [93,110,111]. It could also indicate another pathomechanism that is Giardia ESPs might interfere with cholesterol and lipid metabolism based on the differential expression of insig1 gene. Of note is that transcriptional changes in IECs were not immediate and were only pronounced at 6h, possibly until enough ESPs had accumulated in the media to induce an effect. By comparison, a similar transcriptional profile was seen in parasitized IECs, but transcriptional changes occurred as early as 1.5h upon trophozoite attachment [40]. Therefore, while Giardia ESPs could also play a role in disease induction, these effects might be slower and less apparent compared to those induced upon trophozoite attachment to IECs.
Giardia ESPs induced the transcription of chemokines genes previously shown to be induced upon trophozoite attachment (e.g. cxcl1-3, ccl2 and ccl20) [40]. When produced locally, chemokines attract immune cells (e.g. B cells, T cells, dendritic cells and macrophages) to the infection site [112]. Giardia ESPs also induced the transcription of il8, an important neutrophils attractant produced during the early phase of infection. NFκB and AP-1 are TFs involved in the induction of inflammatory genes transcription [28,29] and their activity was evident in our results and previously reported by others in response to Giardia ESPs [113]. Despite the activation of both TFs, the upstream signaling pathway leading to their activation (i.e. MAPK phosphorylation) appeared to be attenuated but not abolished, even in the presence of inflammatory stimuli. Our transcriptomic data suggests that this attenuation could be mediated by DEGs encoding, dusp1, dusp4, dusp5 and trib1, all of which are known to regulate/inhibit MAPKs' activity [42][43][44][114][115][116]. Although the results of ERK1/2 and P38 phosphorylation were inconsistent with a previous report [113], these discrepancies might be due to the different experimental setup used to study ESPs effects on MAPK signaling. In our model, IECs were exposed to parasite ESPs released directly into the medium whereas in the other study ESPs were added directly to IECs (HT29). In fact, we got the same result when ESPs collected in the absence of IECs were added directly to Caco-2 cells (S3B Fig). This could suggest that when Giardia senses inflammatory signals it actively releases ESPs to attenuate the pathways leading to cytokines production whereas ESPs produced in the absence of IECs exert an opposite effect.
The differential up-regulation of inflammatory cytokines/chemokines genes did not correlate with the amounts of cytokines measured in medium of insert experiment. Indeed, in some instances the levels of IL8, CXCL1 and CCL20 were lower than their controls, indicating degradation of their mRNAs post-transcriptionally or upon synthesis and release into the medium. Tristetraprolin (i.e. zfp36 or TTP) among DEGs, encodes a protein that binds the AU rich element (ARE) in the 3'UTR of cytokine transcripts inducing their decay and its role in controlling inflammation is well established in the literature [117,118]. Which parasite factors induce TTP transcription are to be identified. Second, in a previous report, Giardia ESPs, specifically CPs, were able to degrade IL8 and this has been shown to attenuate neutrophils infiltration into mice intestines even in the presence of inflammatory stimuli [14,15]. Based on our findings it is plausible to propose that ESPs mediate the attenuation of inflammatory responses by promoting the decay of cytokine transcripts and/or cytokines degradation.
To conclude, an interesting picture of host-parasite interactions during giardiasis emerge from this and other studies. Parasites produce ESPs that consist of secreted proteins [11,17,19,85], released surface proteins [20] and EVs [26]. These ESPs affect gene expression, secretion, signaling, metabolism and immune responses in IECs [11,16,40,85,113]. Nevertheless, parasite attachment to IECs induces stronger and more complex responses compared to when IECs are exposed to ESPs only [40]. Concurrently, there is a secretory response by parasitized IECs where similar factors are released affecting parasite attachment, metabolism and gene expression [16,85]. So far, most studies of Giardia-host cell interactions have been performed with simplified models using a small selection of human cell lines but it will be important to follow up these results using more complex in vitro host systems like enteroids [119]. This can be complemented by in vivo experiments of secreted proteins in mice where trophozoites are mixed with early encysting cells [83]. Thus, Giardia interaction with IECs is an intricate process and further understanding of this cross-talk could be the key for understanding giardiasis. Western blot analyses of differentiated Caco-2 cells exposed to G. intestinalis ESPs through a Transwell insert for 2h or 6h in the absence of inflammatory stimuli. Blots were probed for the phosphorylated/non-phosphorylated forms of JNK. (B) Western blot analyses of differentiated Caco-2 cells exposed to G. intestinalis ESPs collected in axenic culture (i.e. in the absence of intestinal epithelial cells). Blots were probed for total proteins (phosphorylated/ non-phosphorylated forms) of mitogen activated proteins kinases (MAPKs, ERK1/2 and P38) and nuclear factor kappa beta (NFκB). (TIF) S1 Table. All proteins identified in the secretome of Giardia intestinalis WB isolate in the serum-free medium, RPMI-1640. Proteins are compared between the two time points, 2h and 6h, which are indicated by colours (orange, 2h) and (blue, 6h). Identification criterion for proteins is based on two peptides per protein. Click on (+) sign in each row to see the tryptic peptides identified. (XLSX) S2 Table. All proteins identified in the secretome of Giardia intestinalis GS isolate in the serum-free medium, 1640. Proteins are compared between the two time points, 2h and 6h, which are indicated by colours (orange, 2h) and (blue, 6h). Identification criterion for proteins is based on two matching peptides per protein. Click on (+) sign in each row to see the tryptic peptides identified. (XLSX) S3 Table. Common and isolate-specific proteins identified in the secretome of Giardia intestinalis WB and GS isolates. Open reading frames together with the protein names are presented in columns based on their annotation in the Giardia Database. (XLSX) S4 Table. All secreted proteins identified in in the serum-free medium of the differentiated human colon carcinoma cell line, Caco-2, incubated with Giardia intestinalis WB isolate. Proteins are compared between the two time points, 2h and 6h, which are indicated by colours (orange, 2h) and (blue, 6h). Identification criterion is two peptides per protein. Click on the (+) sign in each row to see the tryptic peptides identified.  Table. Transcriptional profile of differentiated human colon carcinoma cell line Caco-2 exposed to excretory-secretory products (ESPs) of Giardia intestinalis isolates WB and GS through a Transwell insert placed into tissue culture plates. The table shows transcriptional changes including fold change of genes transcriptions in Caco-2 cells exposed to ESPs of WB or GS isolate for 2h or 6h. The table also contains differentially expressed genes, which are in blue fonts (up-regulated)