Anti-Vi specific total IgG were measured in serum samples using Vi from Citrobacter freundii (22) (equivalent to the Vi produced by S. Typhi (23)) as coating antigen (52-53% acetylation (24)), adapting a method previously described (22, 25, 26). Plates were coated with Vi in PBS at a final concentration of 1 µg/mL overnight (16 h) at 4°C of Maxisorp 96-well round bottom plates (Nunc Maxisorp, 449824), followed by aspiration (without wash) and blocking with 5% PBS milk (Millipore, 70166) for 1 h at 25°C. Plates were then washed 3 times with PBS-Tween 0.05% (Sigma-Aldrich, P1379) (PBS-T), before addition of primary antibodies (sera samples) diluted in 5% PBS milk, and were incubated for 2 h at 25°C. Each human serum sample was run in triplicate at different dilutions (1:100, 1:4000, and 1:160,000, respectively) in PBS milk 5% in different plates. Standard sera made of 10 calibrators two-fold apart diluted in PBS milk 5% plus two blank wells with no sera were run in duplicate on each plate. High control and low control (HC and LC, respectively) were run on each plate at dilution able to give as result a range equivalent to 0.21-0.89 and 0.39-1.73 EU/mL, respectively, for LC and HC. Plates are then washed 3 times with PBS-T and incubated for 1 h at 25°C with alkaline phosphatase goat anti-human IgG (Sigma-Aldrich, A3187) diluted 1:5000 in PBS-T plus 0.1% BSA (Sigma-Aldrich, A9418), before the final 3 washes with PBS-T and addition of p-Nitrophenyl phosphate substrate (Sigma-fast N2770, Sigma-Aldrich), for 1 h at 25°C. Absorbances were read with a spectrophotometer at OD 405 and 490 nm (Biotek automatic plate reader), maintaining strict timing between plates.

ELISA units are expressed in relation to a five-parameter human antigen-specific antibody standard serum curve composed of 10 standard points and 2 blank wells (run in duplicate on each plate). One ELISA unit is defined as the reciprocal of the dilution of the standard serum that gives an absorbance value (optical density measured at 405 nm subtracted to optical density measured at 490 nm—the latter being the background wavelength of the plastic of the plate) equal to 1.

The primary anti-Vi standard serum was calibrated and antigen concentration was set at saturation of the signal. Several QC criteria were applied on each run, in particular, the R-square value for the 5 PL curve fit of the standard dilution series maximum background OD, the minimum value of OD maximum, range of acceptance in terms of OD for 1 EU/mL, deviation to the expected EU/mL both for high and low controls. If at least one of the above-mentioned criteria was not met, the entire layout was repeated under the same experimental conditions. A sample is instead considered valid if the EU/mL determined as average from the values obtained in the three independent plates have a CV% < 30% at the dilution selected for obtaining results (the ones in which OD values obtained fall within the linear part of the standard curve); if not met, the sample was re-run under the same assay conditions (frequency of re-run due to failure of this parameter is <2%).

The VaccZyme ELISA was carried out according to the manufacturer’s instructions (Binding Site). Vi of S. Typhi was used as coating antigen. Five calibrators, covering a range from 7.4 to 600 U/mL, with a high and a low positive control included in each run. The assay comes with a sensitivity of 7.4 U/mL. Four independent replicates of serum samples were run at 1:800, 1:1600 or 1:3200 dilutions each tested in triplicates and plates were read at OD_450nm. Results were obtained by interpolating the OD_450nm values against the standard curve (as per manufacturer instructions) and were represented by the average U/mL of the triplicate dilution falling within the quantifiable range of the curve, taking into account the dilution factor.

Vi Human Standard Serum and five samples with different IgG concentrations have been generated by pooling sera from high responder subjects originally enrolled in Vi-CRM₁₉₇ (TCV) Phase 1 trial in Belgium (25). Working aliquots of the standard sera were stored at −80°C until use. Human IgG-depleted serum (Molecular Innovations cod. HPLA-SER-GF) was used as a negative matrix.

Samples used to assess standard curve accuracy: Vi Human Standard Serum has been used to prepare 24 standard curves composed of 10 two-fold dilutions starting at 10 EU/mL and including 2 blanks.

Samples used to assess precision, and dilution linearity: Vi Human Standard Serum neat or diluted ½, ¼, 1/8, 1/16, 1/32, 1/64 in negative matrix (IgG depleted sera from molecular diagnostic) were assayed independently in a standard assay by two operators working on the same days, in four independent replicates on each plate, on three different days (24 measurements in total for each individual serum).

Samples to assess specificity: human standard sera were preincubated overnight at 4°C with an equal volume of homologous competitor (Vi) at the final concentrations of 50, 20, 5, 1, 0.1 and 0.01 µg/mL in PBS + 5% milk, in comparison to sera diluted overnight 1:100 in PBS + 5% milk (negative control). The lowest concentration of Vi able to cause a reduction of the ELISA Units ≥ 80% was then used to determine the homologous and heterologous specificity, assessed with samples incubated with OAg from a different Salmonella species (S. Typhimurium) and a different pathogen (Shigella sonnei) in comparison to undepleted control. All samples were incubated overnight (16–18 h) at 4°C prior to being tested.

Test results have been analyzed using Excel, GraphPad PRISM software version 7 and Minitab 18 (Minitab Inc., Chicago, IL, USA). Geometric and arithmetic mean, standard deviation, and coefficient of variation were the major statistics.

To calculate the standard curve accuracy, percentage Residual Error RE% ((recalculated value-nominal value)/nominal value*100 (back calculated error)) was plotted in function of the nominal concentration. Lower and upper limit of standard curve accuracy (LLSCA and ULSCA, respectively) were set at the nominal value of the standard curve at the last and the first dilutions, respectively, with a 90% prediction interval of RE% within the acceptance range of [-25%; 25%].

Linearity was assessed by plotting resulting titers (EU/mL after multiplication per dilution factor) divided by the median of all results vs each dilution with fitted model and 95% CI with acceptance criteria on each datapoint to fit on the 0.7-1.3 range. In the dilution linearity and precision analysis eight outliers were excluded using the Hampel filter methods (all found in the highest dilution of tested samples). The Hampel filter method consists of considering outliers the values outside the interval I = M − k ∗ b ∗ MAD; M+k ∗ b ∗ MAD equivalent to [19.84, 2019.53] where MAD (Median Absolute Deviation) is defined as the median of the absolute deviations from the median M, b is a constant linked to the distribution of the data and the factor k was set to 2. Linearity of the assay has been determined, in addition, by calculating log2-transformed GM (Geometric Mean) plotted against log2-transformed samples dilution and the coefficient of determination (R2) of the regression and 95% confidence interval of the slope.

Precision was assessed by calculating inter and intraassay precision using Minitab; ANOVA with variance component analysis (mixed effect model with random factors) was used to estimate the intermediate precision, and to evaluate the contributions of the operator and day of analysis to the variability. Acceptance criteria were set at CV% resulted ≤ 20% for intra-assay precision and ≤ 25% for inter-assay precision.

The specificity was expressed as % of inhibition, defined as = (EU/mL Control – EU/mL Inhibited sample)/EU/mL control x 100.

We developed and qualified a new ELISA method to quantify anti-Vi IgG in human samples, based on a method we previously used for preclinical and clinical testing (22, 26–30). In this assay, one ELISA unit (EU) is defined as the reciprocal of the dilution of the standard serum that gives an absorbance value equal to 1. Each plate contains high and low control sera with assigned EU/mL, a calibrated standard curve composed of 10 serial dilution points (in duplicate), and 4 blank wells. Individual samples are tested in triplicates at up to three dilutions on different plates. EU/mL of each sample are obtained by interpolating the OD values against the standard curve run on each plate. By using this method, up to 70 different sera can be assayed on a set composed of nine 96-well plates ( Figure 1 ).

To be valid, an assay must pass multiple non-mutually exclusive quality control criteria (R-square value for the 5 PL curve fit of the standard dilution series > 0.98, maximum background OD < 0.15, the minimum value of OD maximum >2.6, with 0-5-2 range of acceptance in terms of OD for 1 EU/mL, deviation to the expected EU/mL both for high and low controls <40%).

When we tested the accuracy of the standard curve according to this ELISA setup, we measured an accuracy range between 0.113 EU/mL and 4.456 EU/mL. To assess linearity and precision, the anti-Vi reference serum, obtained by pooling high responder vaccine sera, was tested undiluted and after six sequential dilution steps (2-fold apart) in 4 independent replicates, run on 3 days and by 2 operators ( Figure 2A ). The assay proved to be linear, either in terms of linearity graphs, which fell within 0.7-1.3 predefined range, and as per regression analysis (R^{2 =} 0.9917). The quantification of the seven samples in the different replicates provided an intra-assay precision with CV% ≤ 20% (7.32 to 17.05%) and inter-assay precision with a CV% ≤ 25% (7.11 to 15.88%). These values are within the limits of acceptability that were predetermined for this parameter.

To assess the specificity, an initial set-up experiment was performed, by pre-incubating homologous purified Vi antigen at different concentrations with test samples to determine the lowest concentration of Vi able to cause a reduction of the ELISA Units of ≥ 80%. This was shown to be 20 µg/mL of Vi. Such concentration was used to assess the homologous and heterologous specificity, with Vi used as homologous competitor and <i>Shigella sonnei and S. Typhimurium OAg used as heterologous competitors. The assay resulted to be specific: a signal’s inhibition >80% (82.37%) was observed with Vi, while no inhibition was found with the heterologous antigens ( Figure 2B ).

We calibrated the assay against the international standard to establish a conversion factor between EU/mL and International Unit/mL (IU/mL). To this end, anti-Vi IgG international standard (16/138), was assayed for a total of 32 independent tests, run in four replicates during four different days, by two operators ( Figure 3 ). The Intermediate Precision and Intra-Assay Precision of these tests were 4.09% and 4.82%, respectively.

The conversion factor between EU/mL and IU/mL was therefore calculated as the ratio between the average EU/mL obtained considering all the replicates reported in Figure 3 (equal to 148.04 EU/mL, range 132.46-163.62 CI95%) and the 100 IU/mL assigned to 16/138, and thus resulted to be 0.676 ( Table 1 ). The IgG values, expressed as EU/mL, can therefore be converted to µg/ml based on the conversion factor between EU/mL and IU/mL (0.676) and the one previously established between IU/mL and µg/mL (0.01775 (10)), thus resulting in 1 EU/mL equivalent to 0.137 µg/mL of anti-Vi IgG ( Table 1 ).

The qualified ELISA was compared with an anti-Vi ELISA commercial kit used in the field (9, 10, 20, 25, 31–33). This was done by testing five different pooled serum samples with different anti-Vi IgG content (1037, 548, 236, 123 and 67 EU/mL for S1, S2, S3, S4 and S5 respectively) and the 16/138 standard serum, in three different days, by two operators ( Figure 4 ).

Both assays resulted to be precise, with CV% reproducibility and CV% repeatability, respectively, between 4.85-7.40 and 3.30-6.56 for the commercial ELISA, and between 6.48-12.76 and 6.35-7.05 for the qualified ELISA. Hence, both assays met the acceptance criteria for repeatability (20%) and reproducibility (25%), confirming their precision. Furthermore, neither the day nor the operator influenced the overall variability in both assays (p-values >0.05 for all the samples).