This study was approved by the Ethics Committee ULB-Hôpital Erasme (aggregation number OMO21, study protocol P2018/515), and all the subjects gave their written informed consent.

Blood samples were collected from 12 wP-primed adults and two 5-6-years-old aP-primed children after the administration of an aP vaccine booster. Nine adults were recently boosted (between D13 and D37) whereas the three others received the boost some years before the blood sampling, with a maximum of 6 years (Boostrix^®, GSK Biologicals, Belgium). The two children were recently boosted (between 1-2 months; Tetravac™ -Sanofi Pasteur, Lyon, France). Mean age of the adults was 33 years (21-52 years in the range), whereas the children were 5-6 years-old. Venous blood samples were collected in sodium heparin tubes (BD Biosciences, Erembodegem, Belgium) and transported directly to the laboratory.

Pertussis toxin (PT) and filamentous hemagglutinin (FHA), the two Bp antigens present in all aP vaccines (36), were selected for in vitro stimulation of blood cells in the Bp-specific whole blood intracellular staining (BpWB-ICS) assay to expand vaccine-induced Bp-specific precursor T lymphocytes. A Bp lysate (BPL) was also used to detect cellular immune responses to Bp antigens that are not included in aP vaccines. A genetically detoxified PT mutant, with an inactivated S1 subunit (R9K, E129A), was chosen for the in vitro T cell assay (LIST Biological laboratories Inc., Campbell, CA, USA, #184) to avoid the toxic activity on target cells (37). PT was resuspended following the manufacturer’s instructions and stored in aliquots at 500 µg/mL at 4°C. The endotoxin content determined by a kinetic chromogenic LAL assay was 0.02 EU/µg. Prior to use, each aliquot of PT was heat-inactivated at 80°C during 10 minutes using a Polymerase Chain reaction block, as recommended to abolish in vitro mitogenic activity of the B oligomer on T lymphocytes (16, 31, 37). Heat-inactivated aliquots were kept at 4°C until use. FHA was kindly provided by Sanofi Pasteur (Marcy-l’Étoile, France) and stored at 700 µg/mL at 4°C. The endotoxin content was below the detection limit of 0.005 EU/µg FHA. The BPL of strain B1917 (38) was kindly provided by Q Biologicals (Ghent, Belgium) and stored at 3.4 mg/mL at -80°C. A Bp sonicate, initially used for assay optimization before the availability of the BPL, was kindly provided by K. Mills (TCD, Dublin, Ireland) and stored at 596 µg/mL at -80°C. BPL and BP sonicate were heat-inactivated as described above for PT. All Bp antigens were aliquoted in low-binding tubes (Maxymum recovery from Axygen, VWR, Leuven, Belgium) for long-term storage. TT used to assess the intra-assay reproducibility, was purchased from Calbiochem (Sigma-Aldrich, Bornem, Belgium) and stored at 200 µg/mL at -20°C. As a positive control, Staphylococcus aureus enterotoxin B (SEB) was purchased from Sigma (Sigma-Aldrich) and stored at 1 mg/mL at -20°C.

Blood samples were processed within 3 hrs maximum of their collection (median of 30 min, 25^th percentile - 75^th percentile (P25-P75): 18-60 min, range: 15-180 min), consistent with other studies that recommended a short delay between blood collection and processing (34, 39). Freshly drawn whole blood samples were diluted 1:1 with RPMI 1640 medium (Invitrogen/ThermoFisher, Merelbeke, Belgium), supplemented with 40 µg/ml gentamycin (Invitrogen/ThermoFisher). Costimulatory anti-CD28 and anti-CD49d antibodies (clones L293 and L25.3, respectively, BD Biosciences, Mountain View, CA) were added to the diluted blood, each at a final concentration of 1 µg/mL to increase cytokine expression in antigen-specific T cells (34). Two-fold diluted blood was distributed in 15 ml round-bottom polypropylene tubes, each of which containing 800 µl of diluted blood (corresponding to 400 µl original blood volume) and incubated in the absence (negative control) or presence of antigen, i.e. 5 µg/mL PT, 5 µg/mL FHA, 10 µg/mL BPL, 10 µg/mL TT, 10 µg/mL Bp sonicate or 1 µg/mL SEB. The optimal antigen concentrations were either based on the literature (BP sonicate, TT) or on preliminary experiments from the PERISCOPE Consortium (PT, FHA, BPL). To assess intra-assay variability, some stimulations were performed in duplicate. Tubes were loosely covered by caps and incubated for 24 hrs at 37°C and 5% CO₂. During the last 4 to 5 hrs of incubation, 10 µg/ml Brefeldin-A (Sigma-Aldrich) together with 1/1,000 Monensin (BD Biosciences, Erembodegem, Belgium) was added to each tube to inhibit cellular protein transport (26). Cells were subsequently pelleted at 500 x g for 10 min.

The protocol was adapted from the BD protocol (Alternative Protocol, Activation and Intracellular Staining of Whole Blood, BD Biosciences). Briefly, erythrocytes were lysed by adding 10 mL BD PharmLyse (BD Biosciences) to the pelleted cells for 10 min at room temperature (RT) in the dark. After washing with phosphate buffered saline (PBS, Westburg – Lonza, The Netherlands), cells were centrifuged at 500 x g for 10 min, the supernatants were discarded, and the white blood cells were fixed by incubation with 1 mL of the Fixation/Permeabilisation solution (from the BD Cytofix/Cytoperm Fixation/Permeabilisation kit, BD biosciences) during 20 min at RT in the dark. After centrifugation at 500 x g for 5 min, the white blood cells were cryopreserved in 500 µl Recovery Cell Culture Freezing medium (ThermoFisher), a formulation based on Dulbecco’s Modified Eagle Medium (High Glucose) with optimized levels of foetal bovine serum, bovine serum and 10% DMSO per condition. Cells were transferred into cryovials transferred into cryovial, before storage at -80°C for delayed analysis by batches. In case of direct staining and flow cytometry acquisition, the white blood cells were directly permeabilised by incubation with 2 mL of the Perm/Wash solution (from the BD Cytofix/Cytoperm Fixation/Permeabilisation kit, BD biosciences) for 10 min at RT in the dark.

Cryovials containing the stimulated, fixed and frozen white blood cells were retrieved from the -80°C freezer and thawed in a water bath at 37°C for 2 min. Thawed cells were transferred from cryovials to 15 mL conical tubes containing 5 mL PBS. After centrifugation at 500 x g for 5 min, the cells were permeabilized by adding 2 mL of the Perm/Wash solution (BD Biosciences) and incubated at RT for 10 min.

After permeabilization, either directly on freshly stimulated cells or after cryopreservation of the stimulated and fixed cells, surface and intracellular antibody staining of the cells was combined and performed in the presence of Perm/Wash buffer during 30 min at RT in the dark. Cells were stained with one of the two monoclonal antibody panels; panel 1: CD3-PcP, CD4-APC H7, CD8-PE, CD14/CD19-V500, IFN-γ-BV421, IL-5/IL-13-APC, IL-17A-FITC (Details see Table 1 , panel 1); or panel 2: CD3-APC-H7, CD4-BB515, CD8-PerCP-Cy5.5, IFN-γ-BV421, IL-4-APC, IL-5-APC, IL-13-APC, IL-17A-PE, IL-17F-PE, IL-22-PE-Cy7 (Details see Table 1 , panel 2). The antibody panel 1 was optimised in order to improve the detection of Th1 and Th17-type responses. After staining, cells were washed with 2 mL Perm/Wash buffer, transferred in FACS tubes (BD Biosciences), and washed again with PBS before they were recovered in FACS Flow buffer (BD Biosciences) for acquisition.

Acquisition was initially performed on a FACSCanto II flow cytometer (BD Biosciences) to assess the intra-assay reproducibility and a BD-LSR Fortessa flow cytometer (BD Biosciences) was used subsequently for the next experiments. The content of the tube was completely acquired. Compensations were performed using Comp Beads (BD Bioscience) tubes individually stained with each fluorophore, and compensation matrices were calculated with FACSDiva. Cytometer Setting and Tracking (CST) beads (BD Biosciences) were acquired before each experiment to ensure that cytometer parameters remained consistent across all experiments.

All flow cytometry data analyses were performed with FlowJo software (version 9.5.3, Tree Star, Ashland, OR USA) by using sequential gating. Preliminary experiments using a live/dead dye demonstrated that most cells were alive after the stimulation, and successive gates were applied to exclude dead cells. The gating strategy shown in Supplementary Figure 1 is an example of cryopreserved cells stained with antibody panel 2. Briefly, after selection of a time of homogenous acquisition, lymphocytes were determined according to their size and granularity. After doublet exclusions and CD3⁺ T cells selection, CD4⁺ CD8^- T cells were analysed for their content in cytokines. The numbers of acquired events and percentages of CD4⁺ T cells that produce IFN-γ, IL-4/IL-5/IL-13, IL-17A/IL-17F or IL-22 were reported. The numbers of events recorded in the CD3 gate were 264,500 (median with range: 97,396-533,000) and 344,000 (median with range: 96,432-577,000) for the samples stained freshly and after cryopreservation, respectively. For CD4⁺ T lymphocytes, these numbers were 119,000 (median with range: 43,536-346,000) and 144,500 (median with range: 53,800-377,000), respectively. CD8⁺ T lymphocytes were analysed for their content in IFN-γ, and the numbers of CD8 events were 72,973 (median, range: 17,647-149,000) and 89,034 (median, range: 23,288-164,000), respectively. The numbers of acquired events were thus higher after cryopreservation within the CD3, CD4 and CD8 gates (p=0.0001, p=0.0003, p=0.0004, respectively).

The non-specific backgrounds were defined as the percentage of cytokine-producing CD4⁺ or CD8⁺ T lymphocytes among cells incubated in the absence of antigen. Percentages of antigen-specific cytokine-producing CD4⁺ and IFN-γ-producing CD8⁺ T cells were determined after the non-specific background values were subtracted.

To determine whether a Bp antigen-specific T cell response was relevant, we first compared the results obtained for the stimulated conditions to those from unstimulated condition by checking carefully the Flowjo analyses. The percentages of cytokine-containing cells in a diluted whole blood sample after Bp antigen stimulation were arbitrarily defined as significant i.e. different from the non-specific background, based on two criteria: a stimulation index (SI, percentage of cytokine positive cells in Bp antigen-stimulated condition/percentage of cytokine positive cells in the absence of antigen) ≥ 2, and a percentage of Bp antigen-specific cells (i.e. after subtraction of the non-specific background) ≥ 0.030% ( Table 2 ). Both a high number of events acquired in the parent gate and a low background in the absence of antigen allowed us to consider low percentages of positive cells in stimulated conditions to be antigen-specific. Results were considered doubtful in two cases (1): a SI > 2 with a percentage of positive cells between 0.010–0.030%, and (2) a SI between 1.5 and 2, with a percentage of positive cells ≥ 0.010%. Results were always considered negative in case of a percentage of positive cells lower than 0.010%. These criteria were validated by checking back the Flowjo analyses.

Subjects with clearly detectable or doubtful percentages of Bp antigen-specific T cells were identified as responders or doubtful responders to that antigen, respectively. Subjects were considered as non-responders when a negative result was recorded.

Graphpad Prism 7.03 for Windows (Graphpad software, La Jolla, CA, USA) was used for statistical analysis. Correlations were evaluated by a non-parametric Spearman test. Wilcoxon matched-pairs signed rank test was applied to compare the results obtained for paired samples. Friedman’s test was used to compared three or more paired groups, with Dunn’s test as post hoc analysis to compare two conditions. A value of p<0.05 was considered significant. *, p<0.05; **, p<0.01.

In order to examine the intra-assay reproducibility of the BpWB-ICS assay, blood samples from five adults (n°1 to n°5) was divided into two aliquots and processed separately from the incubation step until the direct staining of the cells, to detect intracellular cytokines within CD4⁺ T cells ( Figure 1 , Table 1 , panel 1). The percentages of IFN-γ- or IL-17A-producing CD4⁺ T cells obtained for the duplicates were measured under unstimulated and stimulated conditions with antigens (PT, FHA, Bp lysate, Bp sonicate, TT or SEB).

No Th2-containing CD4⁺ T cells were detected whereas the percentages of antigen-induced IFN-γ- or IL-17A-producing CD4⁺ T reached 0.604% and 0.034% in the median, respectively (P25-P75: 0.097%-3.18% for IFN-γ and 0.010%-0.118% for IL17-A). Results obtained in the absence of antigen were very low (0.011% and 0.004% in the median, for IFN-γ and IL-17A, respectively). The results obtained for the two different aliquots of blood were very well correlated (r=0.980, p<0.0001 for IFN-γ and r=0.958, p<0.0001 for IL-17A ( Figure 2 ). The coefficients of variation (CVs) of the results obtained for the duplicates were 16% for IFN-γ and 22% for IL-17A, indicating that the precision of the assay may be considered as good (40, 41).

To assess a potential operator effect on the flow cytometry analyses, we compared results of acquired data obtained for the same samples and analysed with the FlowJo software by two different operators, one experienced and one novice operator.

Blood samples from four adults (n°6 to n°9) and from two children (n°1, n°2), were included here and they were processed with an optimized antibody panel (panel 2, Table 1 ) to further expand the detection of Th2 and Th17-type CD4⁺ T cells ( Figure 1 ). As a flow cytometer LSR Fortessa became available, datasets were analysed on this cytometer, and we extent the analysis to CD8⁺ T lymphocytes in addition to CD4⁺ T lymphocytes.

The percentages of total CD3⁺, CD4⁺ and CD8⁺ T lymphocytes obtained by the two operators were strongly correlated ( Figure 3A , r=0.992, p<0.0001). Similarly, the percentages of IFN-γ-producing-CD4⁺ T cells, IL-4/IL-5/IL-13 and IL-17A/IL-17F were highly correlated ( Figures 3B–D , p<0.0001). The percentages of IL-22-producing CD4⁺ T cells were low but also well correlated Supplementary Figure 2A ). Similar to CD4⁺ T cells, the percentages of IFN-γ-producing CD8⁺ T cells were also well correlated Supplementary Figure 2B ). The percentages of other cytokines produced by CD8⁺ T cells were insignificant.

To evaluate the effect of cryopreservation of stimulated and fixed cells, blood samples from four adults (n°6 to n°9) and two 5-6 years old-children (n°1, n°2), were divided into two series of aliquots. The first series was stained directly, while the second one was stained after cryopreservation of the stimulated and fixed cells at -80°C during a median of 12 months (range: 1 month – 22 months), for a head to head comparison ( Figure 1 ).

The percentages of total CD3⁺, CD4⁺ and CD8⁺ T lymphocytes obtained from fresh versus frozen samples were highly correlated ( Figure 4A , r=0.956, p<0.0001), as were the percentages of IFN-γ-producing CD4⁺ T cells ( Figure 4B , r=0.951 and p<0.0001). The correlation between the percentages of CD4⁺ T cells-producing IL-4/IL-5/IL-13 obtained by the two procedures was slightly lower albeit still satisfactory ( Figure 4C , r=0.766, p<0.0001). Eight discrepancies out of 30 results were noticed between results of Th2-producing cells obtained by the two procedures, with slightly higher percentages noticed for the direct staining than for the procedure performed after cryopreservation ( Figure 4C ). These discrepancies originated from two adults for several stimulating conditions, and from one child for the unstimulated condition. The percentages of IL-17A/IL-17F-producing CD4⁺ T cells obtained by the two procedures were perfectly correlated ( Figure 4D , r=0.974 and p<0.0001). Good correlations were also found for the very low percentages of IL-22-producing CD4⁺ T cells (r=0.860, p<0.0001, Supplementary Figure 3A ), and for IFN-γ-producing CD8⁺ T cells (r=0.883, p<0.0001, Supplementary Figure 3B ). The non-specific backgrounds were similar for direct staining compared to staining after cryopreservation ( Supplementary Figure 4 ). Figure 5 summarized the final workflow of the BpWB-ICS assay.

To determine the assay performance and the capacity of the assay to detect Bp antigen-specific Th1, Th2, and Th17-type responses, we compared the percentages of cytokine-producing CD4⁺ T cells of cryopreserved samples stimulated with PT, FHA, BPL or SEB, to the non-specific background, after staining with antibody panel 2 ( Figure 1 ). Samples from seven adults (n°6 to n° 12) and from two children (n°1, n°2), all recently aP vaccine boosted, were analysed ( Figure 1 ).

Except for Th2-producing cells, background percentages were remarkably low, with median values of 0.012%, 0.007%, and 0.009% for IFN-γ, IL-17A/IL-17F, and IL-22-producing CD4⁺ T cells, respectively ( Supplementary Figure 5 ). This background was higher for the percentage of IL-4/IL-5/IL-13-producing CD4⁺ T lymphocytes with a median of 0.031%, although they were still in an acceptable range ( Supplementary Figure 5 ).

In contrast, the percentages of IFN-γ-producing CD4⁺ T cells obtained after stimulation with PT, FHA, BPL or SEB were higher and significantly different from the background percentages ( Figure 6A ). Overall, the percentages of antigen-induced-IL-4/IL-5/IL-13-producing CD4⁺ T cells were not significantly different from the background percentages, due to one subject with a very high IL-4/IL-5/IL-13 background ( Figure 6B ). Excluding this outlier subject resulted in significant differences between PT, FHA or SEB-stimulated compared to unstimulated cells (p=0.016 for PT and FHA and 0.008 for SEB). The percentages of IL-17A/IL-17F-producing CD4⁺ T cells obtained after stimulation were higher and significantly different than the background percentages ( Figure 6C ). Significant percentages of IL-22-producing CD4⁺ T cells, different from the background, were only noticed after stimulation with BPL or SEB ( Figure 6D ).

The sensitivity of the assay was assessed after staining cryopreserved samples previously stimulated with PT, FHA, BPL, with antibody panel 2. Samples were from seven adults (n°6 to n°12) and from two children (n°1, n°2), all recently aP vaccine boosted ( Figure 1 ).

Based on defined criteria of positivity ( Table 2 ), PT-specific IFN-γ-producing CD4⁺ T cells were detected for 6/7 adults and 1/2 children ( Figure 7A ), with 0.047% of positive cells (median) and a SI of 8 (median) ( Supplementary Table 1 ). The second child had a doubtful response. All the adults (7/7) and one child had FHA- and BPL-specific IFN-γ-producing CD4⁺ T cells, with both high percentages of positive cells and SI ( Supplementary Table 1 ). The second child had a doubtful percentage of positive cells ( Figure 7A and Supplementary Table 2 ). Overall, IFN-γ-producing CD4⁺ T cells were detected to at least one Bp antigen in all subjects.

Bp antigen induced-Th2-type CD4⁺ T cells were clearly detected in the children only. Both children had FHA-specific Th2-type CD4⁺ T cells, one of them had also a PT-specific response, whereas the other one had also a BPL-specific response ( Figure 7B and Supplementary Table 1 ). The apparent Th2-type CD4⁺ T cell response detected for one adult was non-specific, as it resulted from very high background ( Figure 6 ). Some additional Th2-type responses were detected but considered doubtful, and induced by FHA (2/7 adults), PT (the second child), and BPL (1/7 adult) ( Figure 7B and supplementary Table 2 ).

As for IFN-γ, most subjects had PT-specific IL-17A/IL-17F-producing CD4⁺ T cells (5/7 adults and 2/2 children) ( Figure 7C ), with 0.064% of positive cells (median) and a SI of 41 (median) ( Supplementary Table 1 ). Almost all the subjects were responders to BPL (6/7 adults and 2/2 children) ( Figure 7C and Supplementary Table 1 ). In contrast, only three subjects had detectable FHA-specific IL-17A/IL-17F-producing CD4⁺ T cells (1/7 adults and 2/2 children) ( Figure 7C and supplementary Table 1 ). In addition, doubtful Th17-type responses were noticed for three adults (one induced by both PT and FHA, a second by FHA, and a third one by BPL) ( Figure 7C and Supplementary Table 2 ). Of note, even if most subjects had both PT-specific IFN-γ- and IL-17A/IL-17F CD4⁺ T cells, these cytokines were rarely co-expressed by the same cells ( Supplementary Figure 6 ), so that Boolean analysis was not further performed.

Finally, PT-specific IL-22-producing CD4⁺ T cells were detected only in three subjects (2/7 adults and 1/2 children), while no FHA-specific IL-22-producing CD4⁺ T cells were noticed ( Figure 7D and supplementary Table 1 ). In contrast, seven subjects had BPL-induced IL-22-producing CD4⁺ T cells (6/7 adults and 1/2 children) ( Figure 7D and supplementary Table 1 ). In addition, doubtful IL-22 responses were noticed for four subjects, and induced by PT (1/7 adults and 1/2 children) and BPL (1/7 adults and 1/2 children) ( Figure 7D and supplementary Table 2 ). Subjects with specific IL-22-producing CD4⁺ T lymphocytes also had IL-17-producing cells, but the cells were different ( Supplementary Figure 6 ).

The presence of IFN-γ-producing CD8⁺ T cells was investigated in adults n°6 to n°12 and in the two children, following the same procedure as described in Figure 5 .

The median background for IFN-γ-producing CD8⁺ T cells was 0.014% (P25-P75:0.010%-0.049%, Figure 8A ), which was significantly higher than the background of IFN-γ-producing CD4⁺ T lymphocytes (median: 0.012%, P25-P75:0.004%-0.020% ( Supplementary Figure 5 ), and p=0.039, data not shown). Overall, the percentages of IFN-γ-producing CD8⁺ T cells after stimulation were significantly higher than the background for PT, BPL and SEB, but not for FHA ( Figure 8B ). The same criteria used to determine Bp antigen-specific CD4⁺ T cell responses were applied for IFN-γ-producing CD8⁺ T cells. Even if less frequent than for the CD4⁺ T cells, PT- and FHA-specific IFN-γ-producing CD8⁺ T lymphocytes were detected in 3 and 2 adults, respectively, while BPL induced IFN-γ-producing CD8⁺ T cells in all subjects. In addition, one adult and 2 children had a doubtful response to PT, and three adults to FHA ( Figure 8C ).