Proliferation of enterotoxigenic Escherichia coli strain TW11681 in stools of experimentally infected human volunteers

qPCR assay validation

In the stool specimens collected on the day of infection, the O19 polymerase gene could either not be detected in the qPCR assay or was present in small amounts. This suggests that the volunteers were not extensively colonized with other O19 E. coli strains prior to ingesting the dose, and that most of the qPCR signals we observe can therefore be attributed to TW11681. The standard curve based on the positive control dilution series had acceptable curve fitting parameters (97% PCR efficiency and a mean squared error of 0.01), the assay has a linear dynamic range for detecting between 13 and 132,000 TW11681 genomes, and none of the no-template controls showed signs of amplification.

We found no indication that any of the qPCR results were affected by PCR inhibitors in the template DNA. Of the sample reactions that were based on the 0.1 ng/µl template DNA, 84, representing 43 different samples, gave results that fell within the dynamic range of the assay. The median within-sample quantitation cycle (Cq) difference between reactions based on 0.1 ng/µl and 1.0 ng/µl template DNA for these 43 samples was 3.10, the differences looked normally distributed (Shapiro–Wilk Normality Test p value: 0.22; skewness coefficient − 0.51), and there were no clear outliers. If the ETEC quantitation estimates presented here were substantially biased by PCR inhibitors from the template DNA preparations, the differences in Cq estimates between the two template concentrations would probably be substantially more skewed towards no Cq difference.

TW11681 proliferation and association to clinical symptoms

For each stool specimen, we isolated DNA and quantified the number of TW11681 genomes by qPCR. The estimated TW11681 concentration in a stool specimen is represented by percent estimated weight of TW11681 genome DNA relative to the weight of all DNA from the specimen. The main weight of other DNA in the stool specimen probably originates from the remaining gut microbiota and from epithelial cells that has been exfoliated from the large intestine [14]. Assuming the production of these other cell populations remains relatively stable during the follow-up period, the changes in TW11681 DNA concentration should reflect actual changes in stool TW11681 proliferation during the infection.

As seen in Fig. 1, there were two main patterns for changes in the stool TW11681 DNA concentration. Six volunteers (EV10, EV11, EV13, EV16, EV17, and EV18) had a rapid and substantial increase and subsequent decline in TW11681 DNA concentration, with mean maximum concentration at 5.3% (range 3.3–8.2%) a median 3.5 (range 2–5) days after dose ingestion. The remaining 3 volunteers (EV12, EV14, and EV15) had low TW11681 DNA concentrations (< 0.5%) throughout the follow up period. The size of the TW11681 proliferation peaks did not seem to increase with inoculum size. Both volunteers who developed diarrhea and all five who experienced abdominal pain or cramps had these peaks, but the peaks did not necessarily coincide with the onset of symptoms. For example, EV11’s diarrheal episode and EV13’s episode with abdominal pain or cramps was over even before the TW11681 DNA concentration peaked. Notably, for EV10, EV16, and EV18, the rapid decline in the TW11681 DNA concentration started before onset of the antibiotic treatment.

Immune responses

All volunteers except EV14 seemed to develop some serum antibody responses to the CfaB subunit of the ETEC Colonization Factor Antigen I (CFA/I) and the E. coli mucinase YghJ (Fig. 2). The three volunteers who did not have the TW11681 proliferation peaks (EV12, EV14, and EV15) had low levels of pre-existing CfaB- and YghJ-specific serum antibodies, similar to most of the other volunteers, but tended to have lower increases in antibody levels to these proteins than the other volunteers. Of the 6 volunteers who had TW11681 proliferation peaks, EV13 had the smallest peak as well as the lowest CfaB- and YghJ-specific serum antibody level increases (Fig. 2).

Strain description

ETEC strain TW11681 (O19:H45; GenBank BioProject: PRJNA59749) was isolated in Guinea-Bissau in 1997 from the stool of a 6 month old girl who had diarrhea [10]. This ciprofloxacin-sensitive strain has genes encoding the STh and the two colonization factors CFA/I and Coli Surface antigen 21 (CS21). TW11681’s genome size is approximately 5.30 Mbp, around 316 kbp of which is plasmid DNA [17]. Phylogenetically, it belongs to an epidemiologically important ETEC lineage that is often found to be associated with childhood diarrhea (ETEC8 family [3] and Lineage L6 [4]).

Controlled human infection model study

The volunteer study is based on experimental infection of nine healthy 23–28 years old students (1 man and 8 women) living in Norway who had not traveled to LMICs within 12 months prior to study start. In brief, the volunteers were included in the study in successive groups of three during the fall of 2016. The volunteers had fasted overnight before orally ingesting the dose at 11:00 a.m. at the Infectious Diseases Ward at Haukeland University Hospital, Bergen, Norway. They drank 120 ml 1.33% bicarbonate buffer, followed after 1 min by 30 ml of the bicarbonate buffer containing 1 × 10⁶, 1 × 10⁷, or 1 × 10⁸ CFUs of TW11681. The volunteers were allowed to drink and eat normally 1 h after that. The TW11681 doses were prepared by culturing the strain on animal product-free LB agar plates before harvesting, and the cells were re-suspended and subsequently washed three times in phosphate-buffered saline before being administered to the volunteers. Clinical signs and symptoms were self-reported by the volunteers during the daily clinical evaluations, stool samples were obtained daily, and blood samples were collected on the day of the infection and 7, 10, and 28 days after. Stool consistency was graded on a scale from 1 to 5, where the grades seen in this study were 1 and 2 (firm and soft formed stools, respectively) and 3 (viscous opaque liquid or semiliquid which assumes the shape of the container). The volunteers were considered to have diarrhea if they pass grade ≥ 3 stools with a combined weight of ≥ 200 g over a 48-h period or a single grade ≥ 3 stool of ≥ 300 g. To clear the infection, ciprofloxacin treatment (500 mg orally two times daily for 3 days) was started in the morning 5 days after they ingested the dose. None of the volunteers became ill enough to warrant earlier treatment.

Microbiological detection of ETEC

Stool specimens were acquired daily starting on the first day after the dose had been ingested. The specimens were streaked onto Enterobacteriaceae-selective Lactose Agar (Haukeland University Hospital, Bergen, Norway), followed by over-night incubation at 35 °C. A sample was considered to be negative for ETEC if no E. coli-like colonies were seen on the plate. If colonies were present, a representative selection of colonies from the most confluent part of the plate was collected with a 1 µl inoculation loop, DNA was then extracted by boiling and centrifugation followed by testing the supernatant for the presence of the ETEC toxin genes by PCR as described earlier [13].

Stool DNA purification

For this study, we designed a DNA purification method that enables extracting genomic DNA equally well from all stool microbes and minimizing the amount of co-purified PCR inhibitors while allowing for purification in a microplate format. Up to one stool specimen for each day from each volunteer was included in the study, including the day the volunteers ingested the dose. Stools were stored for up to 20 h at 4 °C before being mixed by stirring and a representative selection was stored aliquoted at − 70 °C. After thawing, around 50 mg of the specimen was suspended in 200 µl Lysis Buffer (100 mM Tris–Cl, pH 9.0, 40 mM EDTA, 1% SDS, 1 µg/µl Proteinase K) followed by incubation at 60 °C for 60 min with ~ 160 mg 1:1 mix of acid-washed 0.1 mm and 0.5 mm Ø glass beads in a 2 ml deep-well microplate covered with an aluminum foil seal for cold storage (Thermo Fisher Scientific, Waltham, MA). After a 2 × 1 min vigorous vortex of the sealed plate, 70 µl 6 M ammonium acetate was added to each well, followed by re-sealing and mixing with a gentle vortex, 20 min incubation at − 20 °C to precipitate SDS, a 60 s vigorous vortex, and a 6000×g centrifugation for 20 min at 4 °C in a Heraeus Multifuge X3 centrifuge fitted with a HIGHPlate 6000 Microplate Rotor (Thermo Fisher Scientific, Waltham, MA). Note that the integrity of the aluminum foil seal should be checked before vortexing, and that a fresh foil should be added after adding the ammonium acetate. Up to 200 µl of the resulting supernatant was incubated agitated in room temperature for 30 min in a Nunc 96-well Fritted Deep Well Plate (Thermo Fisher Scientific) that contained approximately 90 µl powdered polyvinylpolypyrrolidone (110 µm particle size; Merck, Darmstadt, Germany), 50 mg Amberlite-XAD4 (Merck), and 50 mg Dowex-1X8 (Merck) resins. Prior to adding the supernatants, the resin, which reduces the amounts of potential PCR inhibitors, had been individually distributed to wells where they subsequently soaked for 2–3 h in 200 µl Equilibration Buffer (75 mM Tris–Cl, pH 9.0, 30 mM EDTA, and 1.5 M ammonium acetate) before buffer removal by centrifugation at 800×g for 1 min. After incubation, the filtrates were collected by centrifugation into a 1.2 ml low-profile microplate, and nucleic acids were precipitated by adding 200 µl isopropanol followed by 20 min agitation at room temperature and centrifugation at 6000×g for 20 min at 4 °C. After pipette aspiration, 50 µl Digest Buffer (20 mM Tris–Cl, pH 8.0, 4 mM EDTA, 0.8 µg/µl RNAse A, 1.6 µg/µl Proteinase K, 20 mM TCEP, and 0.05% Tween-20) that had been prepared immediately before use was added while the DNA pellets were still wet, and the plate was incubated agitated for 30 min at 37 °C. A mix of 20 µl AMPure XP (Beckman Coulter Inc., Brea, CA) and 70 µl Binding Buffer (10 mM Tris–Cl, pH 8.0, 1 mM EDTA, 20% PEG 8000, 2.5 M NaCl, and 0.05% Tween-20) was added to each sample and mixed by pipetting, followed by transfer to a 350 µl round-bottom polypropylene microplate, incubation for ≥ 10 min at room temperature, and incubation on a microplate magnet for ≥ 5 min. After discarding the supernatant, the magnetic beads were washed by re-suspending them in 90 µl Binding Buffer before washing twice with 70% ethanol on the magnet. After 2 min air drying, 50 µl Dilution Buffer (10 mM Tris–Cl, pH 8.0, 0.05% Tween-20) was added to the beads, and, after 30 min incubation with agitation at 37 °C, the supernatant containing the DNA was stored at − 20 °C.

qPCR oligonucleotides

A probe-based qPCR assay was developed for quantitating TW11681 bacteria. The qPCR primers were designed to target the E. coli O19-specific O-antigen polymerase gene (wzy, 1,179 bp; GenBank accession no.: LC223608.1), which is present as a single copy on the TW11681 chromosome. Gene sequences from 37 E. coli strains having > 90% sequence identity with wzy in TW11681, 35 of which had 100% identity, were downloaded from the whole-genome shotgun contigs database in GenBank and included in the alignment. The gene appears to have little sequence similarities with other known prokaryote and eukaryote sequences. The forward (GATGGTTAGTTTTATGACTGG; O19-wzy-TF) and reverse (GAAGAGACTAAGAACTTAGTTG; O19-wzy-TR) primers bind at gene nucleotide positions 941 and 1022, respectively, producing an 81 bp PCR fragment, and the probe (AGCACTCTTCTCGATTCCGACA; O19-wzy-TP), which binds at nucleotide 993, was labelled with 6-FAM (5′-end) and BHQ1 (3′-end). Primer3 [18] was used to test the suitability of the selected primer and probe sequences. The primers have 100% identity with the genes used in the above-mentioned alignment, except the forward primer has one mismatch in position 5 against wzy of O19 E. coli strain FCP1.

qPCR assays

Immediately before running the qPCR assays, the purified stool DNA was quantified by using the QuantiFluor dsDNA System (Promega Corporation, Madison, WI) and diluted in Dilution Buffer to 1.0 and 0.1 ng/µl for use as template DNA in the qPCR assay. The assay was performed on four replicates, two for each template dilution, in white 384-well PCR plates on a LightCycler 480 machine (Roche Life Science, Penzberg, Germany). Each 9 µl reaction contained 1X ABsolute qPCR mix (Thermo Fisher Scientific), 0.4 µg/µl each of O19-wzy-TF and O19-wzy-TR primers, 0.2 µg/µl O19-wzy-TP probe, and 1.5 µl template DNA. The plates were incubated for 15 min at 95 °C, followed by 45 cycles of 20 s at 95 °C and 90 s at 60 °C.

The positive control DNA template was prepared from frozen aliquots of purified genomic DNA from three strains (TW11681, ETEC strain TW10722 [O115:H5], and E. coli laboratory strain BL21(DE3) [O7:H-]) that had been diluted in Dilution Buffer and pooled to a final concentration of 0.5 ng/µl of each strain. Tenfold dilutions ranging from 0.5 ng/µl to 50 fg/µl, as well as 25 and 12.5 fg/µl were used in triplicates as template DNA for positive controls in the assay, while Dilution Buffer was used for the corresponding no template controls. Since each TW11681 genome is 5.30 Mbp and, therefore, weights 5.7 fg, 1.5 µl of the positive control template dilutions added to the qPCR ranges from 132,000 to 3 TW11681 genome copies.

qPCR curves were analyzed by using the LightCycler 480 Software, version 1.5.1.62 (Roche Life Science), where the quantitation cycle (Cq) for each curve was determined by using the Second Derivative Maximum Method. The regression curve of the results from the positive control dilution series was generated by using the same software, and the linear dynamic range of the assay was subsequently set to span the dilutions that gave Cq results that closely conformed to this regression curve.

We tested the suitability of the qPCR annealing and elongation temperature by running the assay on a temperature gradient spanning 50–63 °C. When separated on a 4% agarose gel, a single distinct band corresponding to the expected 81 bp PCR product was seen for all reactions (data not shown), suggesting that 60 °C is suitable for the assay. We chose to use a long (90 s) annealing and elongation time to allow for more flexibility if multiplexing the assay with primers for additional qPCR targets.

To check for any negative effects of PCR inhibitors in the sample template DNA, we assessed whether the reactions containing 0.1 ng/µl template DNA tended to have higher PCR efficiencies than those containing 1.0 ng/µl template DNA. We calculated the mean Cq difference between the two replicates that were based on the 1.0 ng/µl template DNA and those that were based on the 0.1 ng/µl template DNA, excluding results that fell outside the estimated linear dynamic range of the assay. Histograms and Normal Q–Q Plots were then drawn to visually assess whether the distribution of these differences was normally distributed and symmetric, and Shapiro–Wilk Normality Test and Sample Skewness based on the traditional Fisher-Pearson Coefficient of Skewness test were performed to test these assumptions. We used the moments package [19] in R, version 3.4.2 [20] for these analyses.

Immunological assays

Serum was prepared from blood collected 0, 10, and 28 days after the TW11681 cells had been ingested. We used multiplex flow-cytometry bead assays to measure IgA and IgG/IgM antibody responses to the two proteins CfaB (UniProtKB ID: P0CK93) and YghJ (UniProtKB ID: A0A080EY22). CfaB is the major structural protein of the plasmid-encoded fimbrial ETEC colonization factor CFA/I [5], while YghJ is a highly conserved enzyme involved in degrading the major mucins in the small intestine during ETEC colonization [21]. The genes encoding CfaB and YghJ, excluding the sequence encoding the signal peptides, were amplified by PCR and cloned into the pET-30 (Merck Millipore, Burlington, MA) expression vector. The 16 and 167 kDa, respectively, proteins were expressed in ClearColi BL21(DE3) cells (Lucigen, Middleton, WI) and purified by His-tag capture and dialysis against 20 mM phosphate buffer, pH 7.4 containing 0.3 M NaCl. The proteins were cross-linked to Cyto-Plex carboxylated polystyrene beads (Thermo Fisher Scientific) by using carboxyl-reactive chemistry, incubated with human volunteer serum and, subsequently, with fluorescently labeled anti-IgA or anti-IgG/IgM secondary antibodies (Jackson ImmunoResearch Inc, West Grove, PA). Each bead’s secondary antibody fluorescence intensity was measured on a flow cytometer, and the median intensity minus the median intensity of the negative control, which was beads labeled with His-tag labeled glutatione s-transferase, was used in the analyses.