Diarrheal patients hospitalized at the International Centre for Diarrheal Disease Research, Bangladesh (icddr,b) in Dhaka, Bangladesh were screened for ETEC, Vibrio cholerae O1, Shigella, Salmonella and enteric parasites on the first day of hospitalization. Adult male and female patients (n=30) with stool specimens positive for ETEC, as determined by the presence of LT and/or ST enterotoxin genes using polymerase chain reaction (PCR) (32), but negative for other enteric pathogens, were enrolled ( Table 1 ). Patients had experienced diarrhea for 6-96 hours (median 24 hours) prior to hospitalization and most had severe dehydration (83%). Expression of ETEC-specific CFs by the isolated ETEC strains was determined by a dot blot assay (33). Among the enrolled patients, 77% were infected with ETEC strains expressing both LT and ST, 10% were LT only and 13% were ST only ( Table 1 ). A majority of isolated ETEC strains expressed known CFs; CS5+CS6+ and CFA/I+ strains were most common (37% and 20%, respectively), whereas strains expressing other CFs were less frequent (20%). In 20% of the strains, no CF was identified. In addition, healthy adult participants (n=21) who had no history of gastrointestinal disorders, diarrhea, febrile illness or antibiotic use during two weeks before enrollment, based on the clinical assessment by study physicians, were recruited from a similar urban area in Dhaka ( Table 1 ). Informed written consent was obtained from each participant. The study was approved by the research review and the ethical review committees of icddr,b.

Heparinized venous blood was collected from ETEC infected patients at the acute stage of illness (the second day after hospitalization, day 2), at early convalescence (day 7, range day 6-10), and late convalescence (30 ± 7 days and 90 ± 7 days after hospitalization). From healthy individuals, blood specimens were collected once after enrollment. Plasma and peripheral blood mononuclear cells (PBMCs) were separated by density-gradient centrifugation using Ficoll-Isopaque (Pharmacia, Sweden). Plasma specimens were stored at -20°C. From subsets of patients, CD4+ cells were depleted from PBMCs using magnetic beads (Dynabeads, Invitrogen, USA), according to the manufacturer´s instructions. The frequency of CD4+ T cells among lymphocytes after depletion was <3% (mean 1%), as determined by flow cytometric analysis.

For analysis of T cell responses, PBMCs were cultured in DMEM F12 medium (Thermo Fisher Scientific, USA) supplemented with 50 mg/ml gentamicin (Sigma, USA) and 5% human AB+ serum (Sigma-Aldrich) at 37°C in a 5% CO₂ incubator. Whole PBMCs and PBMCs depleted of CD4+ cells (10⁵ cells/well) were cultured in duplicate wells in U-bottomed 96-well plates. Cells were left untreated or stimulated with phytohaemagglutinin (PHA,1 µg/ml, Remel, USA), dmLT (10 µg/ml), LTB (10 µg/ml), LTE112K (10 µg/ml), LT (10 µg/ml), CFA/I (1 µg/ml), CS5 (5 µg/ml) or CS6 (1 µg/ml). After 5 days culture, supernatants were collected and stored at -80°C for cytokine analysis. The indicated stimuli concentrations were found to induce maximal cytokine responses in initial set-up experiments testing a range of different concentrations. Details regarding production of different CF antigens, LTB and dmLT have been previously described (27). LTE112K, LT and LTB used in control experiments were kindly provided by J. Clements, Tulane University, USA.

For analysis of ETEC-specific ASCs, the antibodies in lymphocytes secretions (ALS) assay was used. Several studies have shown that results from the ALS assay, in which the levels of antigen-specific antibodies spontaneously secreted by peripheral blood ASCs in vitro are measured by ELISA, closely correlate with results from the more traditional ELISPOT assay where the numbers of antigen-specific ASCs are determined by counting spots developed on a membrane (13, 34–36). In the ALS assay, PBMCs were cultured at 10⁷ cells/ml in RPMI complete medium (Invitrogen, USA) supplemented with 10% fetal bovine serum (FBS, Invitrogen), Na-pyruvate (1mM, Invitrogen), L-Glutamine (2mM, Invitrogen), 100 IU/ml Penicillin (Invitrogen) and 0.1mg/ml Streptomycin (Invitrogen). Cells were incubated in 96-well flat-bottomed tissue culture plates (2 x 10⁶ cells/well) for 48 hours at 37°C with 5% CO₂. Supernatants were collected and stored at -80°C.

The concentrations of IL-17A and IFN-γ were determined using sandwich ELISA (eBioscience, USA) following the manufacturer’s instructions. IgA antibodies specific for LTB, CFA/I, CS5 and CS6 in ALS and plasma specimens were analyzed by ELISA. Briefly, as described previously (26), to measure LTB-specific IgA antibodies in ALS/plasma specimens, 96-well ELISA plates (Nunc, USA) were coated with GM1 ganglioside (0.3 nmol/mL) overnight at room temperature followed by recombinant LTB subunit (1 μg/mL). To determine CF specific IgA antibody responses, high binding 96-well ELISA plates (Greiner, Austria) were coated with CFA/I (0.5 µg/mL), CS3 (0.3 µg/mL), CS5 (0.5 µg/mL) or CS6 (0.5 µg/mL). Plasma/ALS specimen was added to each well; initially diluted 1:10 (plasma LTB and CFs), 1:4 (ALS LTB) or 1:2 (ALS CFs) in PBS with 0.1% bovide serum albumin (BSA) and 0.05% Tween and a serially 3-fold dilution (plasma LTB and CFs and ALS LTB) or 2-fold (ALS CFs) was performed. The presence of antigen-specific antibodies was detected using horseradish peroxidase (HRP)-conjugated anti-human IgA (Jackson ImmunoResearch; 1:1500 dilution in 0.1% BSA-PBS-Tween) and ortho phenylenediamine (Sigma-Aldrich) in 0.1 M sodium citrate buffer (pH 4.5) and 30% hydrogen peroxide (Merck, USA). For plasma, the reactions were stopped after 20 minutes and endpoint titers were determined as the reciprocal interpolated dilutions of the samples at 450 nm that were 0.4 above background. For ALS, the reactions were stopped after 20 minutes by adding 1 M H₂SO₄ and endpoint titers were determined as the reciprocal interpolated dilutions of the samples at 492 nm that were 0.2 above background. All plasma, ALS and T cell supernatant samples from one individual collected at different time points were analyzed on the same plate in each type of assay and reference samples were included in all analyses to verify consistency between different runs.

Fresh PBMCs (10⁶) were stained with Aqua fluorescent live/dead (Fixable Dead Cells Stain Kit, Invitrogen) to exclude dead cells. To analyze the frequency of different Th cell subsets, PBMCs were stained with anti-CD3-PETR (clone S4.1, Invitrogen), anti-CD4-PerCP (SK3, Becton Dickinson, USA), anti-CD45RO-PECy7 (UCHL1, BD), anti-CXCR3-PE (1C6/CXCR3, BD) and anti-CCR6-BV421 (11A9, BD). To analyze the purity of CD4-depleted PBMCs, cells were stained with anti-CD3-PETR (clone S4.1, Invitrogen) and anti-CD4-PerCP (SK3, BD). After staining, cells were acquired (100,000 total events) using FACSAria™ Fusion (BD, USA). Results were analyzed using FlowJo software (version 10, FlowJo LLC, BD, USA). Among the lymphocytes, 99% cells were found to be alive as determined by live/dead staining. Different subsets of live T cells were analyzed following the gating strategy shown in Supplementary Figure 1 .

Statistical analyses were performed using GraphPad Prism 6 (GraphPad Software, Inc., USA). The Kruskal Wallis test with Dunn’s multiple comparison post-test was used to compare immunological responses in healthy controls and individual sampling time points of ETEC patients. The Mann-Whitney test was used to compare maximum responses at any time point between ETEC patients and healthy individuals. The increases (fold-rises) in magnitudes of immune responses in patients were calculated either as the antibody or cytokine levels found in patient samples on day 2, 7, 30 or 90 after hospitalization divided by the corresponding mean+2SD values in samples from healthy controls or as the antibody or cytokine levels found in patients on day 7, 30 or 90 after hospitalization divided by the corresponding levels on day 2, as indicated for each analysis. A responder was defined as having ≥2-fold increase in the magnitude of response. The Fisher’s Exact test was used to compare the responder frequencies at different time points. Correlation analyses were performed using the Spearman test. P values <0.05 were considered significant.

We first investigated the mucosal and plasma IgA responses to ETEC infection in adult hospitalized ETEC infected diarrheal patients. Mucosal responses were analyzed by measuring LTB- and CF-specific IgA antibodies secreted from intestine-derived ASCs migrating in peripheral circulation using the ALS method. Analysis of ASCs using ALS or ELISPOT has been shown to be a good correlate of mucosal B-cell responses in many studies of enteric infections and vaccines (11–13, 37). IgA antibodies specific for LTB and CFs most commonly expressed by the infecting strains (ETEC positive for CS6, CS5 and CFA/I) were analyzed in samples from both diarrheal patients infected with ETEC strains expressing the corresponding antigens and healthy controls. Two days after hospitalization, ALS IgA responses were low and infrequent in most patients, but 8-33% of the patients responded to LTB, CS6 and CFA/I with antibody levels at least 2-fold higher than those observed in healthy controls already at this early time point ( Figure 1 and Table 2 ). On day 7 after hospitalization, significant increases in levels of ALS IgA responses were observed against both LTB and all CFs compared to healthy individuals ( Figure 1 and Table 2 ). CF-specific ALS responses were more frequent (82-100%) than LTB responses (56%) at this time point ( Table 2 ). Responses were low in most participants on day 30 except for IgA against CFA/I, which remained elevated in about half of the analyzed participants (P>0.05). The ALS antibody levels on day 7 and day 30 were also compared to levels measured in samples collected two days after hospitalization (corresponding to 6 hours to 4 days after onset of symptoms) ( Supplementary Table 1 ). This control analysis gave similar responder frequencies as comparisons to the healthy controls for most antigens, but slightly lower responder frequencies to CFA/I.

ETEC patients also mounted significant IgA responses in plasma. Magnitudes of IgA responses to LTB were increased already on study day 2 in 15% of the participants compared to healthy controls, whereas few patients responded to CFs in plasma at this early time point ( Figure 2 and Table 2 ). On day 7, magnitudes of responses to both LTB and all CFs were significantly higher in patients compared to healthy controls and reached the highest levels detected during the study period ( Figure 2 ). Among the patients, 68% had at least 2-fold increased levels of LTB-specific IgA and 67-83% CF-specific IgA in plasma on day 7 compared to healthy controls ( Table 2 ). The responder frequencies were slightly higher for most antigens when comparisons were made to levels measured on day 2 instead of to healthy controls ( Supplementary Table 1 ). Plasma responses against all antigens declined until day 30, but LTB- (both magnitudes and responder frequencies) and CFA/I-specific (only magnitudes) responses still remained significantly elevated compared to levels in healthy controls at this time point ( Figure 2 and Table 2 ). All plasma responses declined to levels comparable to those observed in controls by day 90.

Collectively, these results show that a majority of the ETEC infected patients enrolled in the study mounted strong mucosal and plasma IgA responses to the LT and CFs expressed by the infecting ETEC strains, with maximal antibody levels detected on day 7 after hospitalization.

To investigate the Th response to ETEC, we first analyzed the total frequencies of memory Th (CD4+CD45RO+) cells in fresh PBMCs collected from ETEC patients infected with ETEC strains expressing LT, CFA/I, CS5 or CS6 and healthy controls. Frequencies of memory Th cells among lymphocytes were comparable in ETEC patients and controls both in the acute and convalescence phase of infection ( Figure 3A ). However, when maximum frequencies of memory Th cells at any time point were compared with proportions in the healthy controls, a small but significant increase was observed in ETEC patients ( Figure 3B ), which was not seen when analyzing levels at each time point separately ( Figure 3A ).

Next, we analyzed the proportions of Th1, Th2 and Th17 subsets within memory Th cells based on the expression of the chemokine receptors CXCR3 and CCR6 in both patients and controls. The frequencies of Th17 (CCR6+CXCR3-) cells were significantly increased in patients already on day 2 compared to the healthy controls and Th17 responses remained significantly elevated until day 7 of infection ( Figure 3C ). The proportions of Th17 cells decreased on day 30-90. In contrast, ETEC diarrhea did not induce any increases in proportions of Th1 (CCR6-CXCR3+) or Th2 (CCR6-CXCR3-) cells at any time point ( Figures 3D, E ). Analysis of maximum frequencies of cells at any time point during infection gave the same result; strongly increased proportions of Th17 cells, whereas frequencies of the Th1 and Th2 subsets were comparable to the controls ( Figure 3F ). These flow cytometric results suggest that ETEC infection induces a Th response that is predominantly of the Th17 type, but did not reveal the antigen specificity of the responding cells.

To determine if the increased proportions of Th17 cells in the circulation of ETEC infected patients were due to LT- and/or CF-specific T cell responses, PBMCs from both patients and controls were stimulated with LTB, dmLT or CFs matching the infecting strain and the secretion of IL-17A and IFN-γ was analyzed in the culture medium.

T cell responses to LT were seen more clearly after stimulation with dmLT than LTB. A few patients with LT+ ETEC infection had IL-17A responses against dmLT or LTB on day 2 (9-13%), but significant responses to dmLT were observed first on day 7 after hospitalization, when 41% of the participants showed elevated IL-17A levels compared to controls ( Figure 4A and Table 3 ). Responses peaked on day 7, but remained elevated until day 30 in 22% of patients (P>0.05 compared to day 2). In most patients, IL-17A responses returned to the levels observed in healthy controls on day 90. A trend for increased IL-17A responses after stimulation with LTB was also observed on day 7 ( Figure 4C ) when 33% patients responded in comparison to healthy controls ( Table 3 ). When the maximum levels of IL-17A responses at any time point of infection were analyzed, increased IL-17A responses were found against both dmLT (56%) and LTB (45%) ( Figure 4E and Table 3 ). Depletion of CD4+ Th cells from PBMCs isolated on day 7 after infection resulted in almost complete loss of IL-17A production in response to stimulation with dmLT and LTB ( Supplementary Figures 2A, B ), supporting that the IL-17A responses to LT were derived from CD4+ Th cells.

We also analyzed IFN-γ responses to dmLT and LTB in patients and controls. Neither dmLT- nor LTB-stimulation induced significant magnitudes of IFN-γ responses at any individual time point after infection ( Figures 4B, D ), but the responder frequencies to dmLT were significantly higher on day 7 compared to day 2 (36 vs 4%, Table 3 ). When maximum magnitudes of responses were evaluated at any time point, patients had significantly higher IFN-γ responses against both dmLT and LTB compared to healthy individuals ( Figure 4F ), although the median magnitudes of responses were lower than for IL-17A. Maximum increases in magnitudes (fold-rises) of IL-17A responses to dmLT or LTB did not correlate significantly with increases in magnitudes of IFN-γ responses to the same antigens ( Supplementary Figure 3 ).

We have previously shown that dmLT enhances IL-17A responses to both antigens and mitogens/superantigens in vitro (29–31). To determine if the strong IL-17A response observed after stimulation with dmLT compared to LTB was related to the adjuvant effect of dmLT, cells from patients infected by LT+ ETEC were also stimulated with native LT and the LT-derivative LTE112K, which has a single mutation in the A-subunit that inhibits the ability of the molecule to induce cAMP and act as an oral adjuvant (38, 39). First, the adjuvant activity of dmLT and LTE112K was evaluated by stimulating PBMCs with the mitogen PHA with or without addition of dmLT or LTE112K. Only dmLT enhanced IL-17A responses to PHA ( Supplementary Figure 4A ), confirming that dmLT but not LTE112E has adjuvant function. We did not measure IFN-γ responses, since our previous studies have shown that dmLT cannot enhance IFN-γ responses in vitro (29–31). Next, the responses to dmLT on day 2 and day 7 were compared to responses induced by native LT or the other homologues LTE112K and LTB. dmLT, LT and LTE112K all induced significantly increased levels of IL-17A on day 7 compared to day 2 ( Supplementary Figure 4B , P=0.02, P=0.01, P=0.02, respectively). IFN-γ responses were weaker and more variable between patients and only significant for dmLT and LT (P=0.01, P=0.03, respectively; Supplementary Figure 4C ). Magnitudes of IL-17A and IFN-γ responses to LT and LTE112K were comparable to the responses induced by dmLT, whereas responses to LTB were significantly lower ( Supplementary Figures 4B, C ). Responses to LTB produced in the same laboratory (denoted by LTB2) as LT and LTE112K were also comparable to responses induced by stimulation with the LTB used in all other T cell and antibody analyses in the study (denoted by LTB1). The similar cytokine responses to dmLT and LTE112K, but lower responses to LTB, indicate that the strong response to dmLT is not directly related to the enzymatic effect of the A subunit included to the dmLT molecule.

Taken together, these results suggest that ETEC infected patients mount Th responses to LT that are predominantly of the Th17 type and can be detected at maximal levels 7 days after hospitalization.

We also analyzed Th responses to CFs expressed by the infecting strains on day 7, 30 and 90 after hospitalization. We could not measure CF-specific T cell responses on day 2, since the CF expression of the infecting ETEC strain was not known at this time point and the cell numbers did not allow stimulation with several different CFs. Patients infected with CS6+ ETEC had significantly increased magnitudes of IL-17A responses against CS6 on day 7 compared to healthy controls (30% responder frequency) but responses remained elevated in only a few patients on day 30-90 ( Figure 5A and Table 3 ). Some patients responded with increased levels IL-17A production to stimulation with CS5 and CFA/I on day 7, however, the number of patients tested for these antigens was small (n=3-7) and responses were not significant ( Figures 5C, E ). Considering the maximum IL-17A responses at any time point, increased responses were again only observed against CS6 ( Figure 5G ). Depletion of CD4+ Th cells from PBMCs isolated on day 7 after infection substantially reduced IL-17A production in response to stimulation with CS6 and CS5 ( Supplementary Figure 2C ), supporting that the observed IL-17A responses to CFs were derived from CD4+ Th cells. Patients had almost no detectable CF-specific IFN-γ responses at any time point after infection ( Figures 5B, D, F ). However, a very small but significant increase in IFN-γ was found for CS5 when maximum responses at any time point were considered ( Figure 5H ).

We next evaluated the relationship between the IgA and T cell cytokine responses induced by ETEC infection. We first investigated the correlation between the maximum increases in magnitudes (fold-rises between maximum responses at any time point of infection versus day 2) of LTB-specific ALS and plasma IgA responses versus the maximum increases in magnitudes of IL-17A and IFN-γ responses induced by stimulation with dmLT, resulting in several significant correlations, as described below. Correlation analyses were also performed using the maximal response level minus the level measured on day 2, but these analyses did not reveal any significant correlations (data not shown). The magnitudes (fold-rises) of ALS IgA responses to LTB correlated significantly with the IL-17A responses (r=0.51, P=0.02) ( Figure 6A ) as well as IFN-γ responses (r=0.56, p=0.01) ( Figure 6C ) to dmLT ( Figure 6C ). In contrast, plasma IgA LTB antibody responses did not correlate with either the IL-17A or IFN-γ responses to dmLT ( Figures 6B, D ). We also analyzed the correlation between the weaker IL-17A and IFN-γ responses to LTB versus ALS and plasma IgA antibody responses to the same antigen. Significant correlations were observed between ALS IgA and IFN-γ responses to LTB (r=0.50, p=0.02) ( Supplementary Figure 5 ). These results support a relation between both Th17 and Th1 responses and ASC-IgA responses, most likely reflecting mucosal IgA responses to ETEC.