Expectant mothers, presenting at the Sukuta Health Centre (a peri-urban setting in The Gambia) for antenatal care were sensitized about the study. Their children were recruited into the study upon birth or when they presented for BCG vaccination (within 48 h of birth). BCG vaccination was withheld for those that agreed to participate (Table 1). Exclusion criteria included low birth weight (<2.5 kg), congenital defects, and multiple births.

The Scientific Coordinating Committee (SCC) at the Medical Research Council (MRC) Unit, The Gambia and the MRC-Gambian Government Joint Ethics Committee approved this study (study number SCC 1233). Written informed consent was provided by the parent/guardian of participating infants in accordance with the Declaration of Helsinki.

Infants were randomized at 6 weeks of age into one of two groups: 0.1 mL intradermal vaccination with BCG vaccine (Russian strain, SSI India) (BCG group) or no vaccination (Control group) to be given at 6 weeks of age. All infants received hepatitis B vaccine (HepB) (Serum Institute of India) and oral polio vaccine (OPV) (Sanofi Pasteur, France) at birth in accordance with the EPI schedule of The Gambia. At 8, 12, and 16 weeks, all infants received the pentavalent [diphtheria–tetanus–whole cell pertussis, Haemophilus influenzae B (Hib), HepB] vaccine (Panacea biotech, India), OPV, and the 13-valent pneumococcal conjugate vaccine (PCV-13) (Table 1). At 18 weeks of age, all infants were tested by standard tuberculin skin test (TST) for mycobacterial reactivity by injecting 0.1 mL containing 2 U of tuberculin (Serum Staten Institut, Denmark) intradermally. Induration was read longitudinally and transversely by a trained nurse at 48–72 h to determine the average induration. The infants in the BCG naïve control group were then vaccinated with BCG at 18 weeks of age. A maximum of 5 mL of venous whole blood was collected into EDTA tubes at 6 weeks (prior to BCG vaccination of group 1) [Visit 1 (V1)], 7 weeks (1 week post-BCG vaccination) [Visit 2 (V2)], and 18 weeks (12 weeks post-BCG vaccination) [Visit 3 (V3)] (Figure 1).

EDTA blood was spun at 1,500 rpm for 10 min and 1 mL of plasma collected and stored at −20°C until use. The remaining whole blood was diluted 1:1 with RPMI (Sigma, UK) and PBMC separation was performed via density gradient centrifugation on a Lymphoprep (Fresenius Kabi Norge AS, Norway) cushion. Harvested PBMCs were washed twice in RPMI and cells counted using Trypan blue exclusion. For every 10 million live cells, 500 µL of fetal calf serum (FCS) (Gibco, Life technologies, USA) was added, cells were left on ice for 30 min and 500 μL of 20% dimethyl sulfoxide (Sigma-Aldrich, UK) in FCS was added before cells were transferred into a Nalgene^® Mr Frosty container (Sigma-Aldrich) and stored at −70°C then transferred to liquid nitrogen within 24 h.

Peripheral Blood Mononuclear Cells were thawed rapidly at 37°C in a water bath. Cells were then resuspended slowly in warm 10% FCS in RPMI (R10) prior to washing twice. Viability counts were done on the samples before being resuspended in R10 and 25 U/mL of benzonuclease (Sigma, UK). PBMCs were then rested at 37°C 5% CO₂ for 6 h.

0.1 × 10⁶ PBMCs in 100 µL of culture medium (2% penicillin/streptomycin, 1% l-glutamine, and 10% FCS in RPMI) were stimulated in 96-well round bottom plates for 16 h with the following antigens: purified protein derivative (PPD) (10 µg/mL, Serum Staten Institute, Denmark), heat-killed Listeria monocytogenes (HKLM) (TLR2 agonist; 10⁹/mL, Invivogen), lipopolysaccharide (LPS) (TLR4 ligand; 1 µg/mL, Invivogen), CLO-75 (TLR7/8 agonist; 10 µg/mL, Invivogen); and heat-killed Streptococcus pneumonia (SP) (10⁵ cells/mL), CA (10⁵ cells/mL), or Escherichia coli (EC) (10⁶ cells/mL). PMA/ionomycin (0.1 μg/1 μg/mL, Sigma, UK) was used as a positive control, and RPMI as a negative control. Supernatants were collected after the addition of 100 µL of RPMI and centrifugation at 1,500 rpm for 10 min. Supernatants were stored at −20°C until needed for cytokine analysis.

Cytokines in culture supernatants were assessed using Bioplex Pro custom made kits according to the manufacturer’s instructions using a 1:2 dilution of reagents (Bio-Rad, Belgium). The cytokines analyzed were IFN-γ, TNF-α, IL-2, IL-4, IL-10, IL-17, and IL-12(p70). Defrosted culture supernatants were centrifuged for 5 min to pellet any precipitate and left on ice until use. Coupled magnetic beads were diluted 1:2 with assay buffer, and added to pre-wet 96-well filter plates. Plates were washed using a vacuum manifold station. 50 µL of samples and standard were added and the plate incubated on a shaker at 300 rpm for 30 min at room temperature. Detection antibodies were diluted 1:2 with detection antibody diluents, 25 µL added to the plate after three washes, and incubated for 30 min on a shaker at room temperature. The plate was washed three times using the vacuum manifold before addition of 50 µL of streptavidin, and then incubated for 10 min on a shaker. 125 µL of assay buffer was added and plates left on a shaker until ready to be read using the Bioplex 200 system, Luminex x-map technology, and Bioplex pro software version 4.0 (Bio-Rad, Belgium).

All values less than the lowest value of the standard within the standard curve were given the value of the lowest standard and all values higher than the highest standard were given the value of the highest standard. The range in picograms per milliliter for each analyte is as follows: IL-2 (0.88–14,426), IL-4 (0.35–5,693), IL-10 (1.45–23,718), IL-12p70 (2.65–43,449), IL-17 (2.33–38,209), IFN-γ (2.05–33,560), and TNF-α (6.69–109,604). Background (unstimulated) values were subtracted from antigen-stimulated values to determine the antigen-specific response.

0.5 × 10⁶ PBMCs were stimulated with RPMI, PMA/ionomycin, PPD, a peptide pool of the TB-specific antigens ESAT-6/CFP-10 (2.5 µg/mL, ProImmune, UK), heat-killed SP (10⁵ cells/mL), C. albicans (10⁵ cells/mL), or E. coli (10⁶ cells/mL) per test. Cells were incubated at 37°C 5% CO₂ for 18 h with Brefeldin A (10 µg/mL) added after the first 2 h of incubation. Cells were stained with a live/dead aqua yellow stain (Invitrogen, USA), incubated for 10 min at room temperature before staining with a surface marker cocktail consisting of either CD3-PerCP Cy5.5 or CD3 APC efluor 750 and CD8 APC efluor 780. Cells were incubated for 30 min at room temperature and washed with 1 mL of PBS/1% FCS/0.2% Sodium Azide (FACS buffer). Supernatant was removed and cells washed in perm/wash solution (Ebioscience, UK). Permeabilization was then performed using fix/perm solution (Ebioscience) and cells incubated for 20 min at 4°C before centrifugation at 1,800 rpm for 5 min. A cocktail of intracellular cytokines IL-2 FITC, IL-17 PE, TNF-α PE-Cy7, IL-10 efluor-450, and IFN-γ APC (all from ebioscience) was added, and cells incubated for 30 min at room temperature. Cells were washed with perm/wash solution and 300 µL FACS buffer was added before acquisition. At least 100,000 lymphocytes were acquired for each sample on a CyanADP (Beckman Coulter, USA) flow cytometer using Summit v4 software and analyzed using FlowJo v10.0.2 (Tree Star, USA) according to the gating strategy shown (Figures S1 and S2 in Supplementary Material).

Stored plasma samples were used to measure antibody levels to polio 1 and 3 using the antibody neutralizing assay as described previously (39). Briefly, Hep-2 Cincinnati cells were cultured with different dilutions of the virus and subject plasma, cytopathic effect of the virus on the plasma was assessed after 5 days incubation. Hepatitis B antibody levels were measured using a commercial ELISA kit (Diasorin, Italy) as per the manufacturer’s instructions. Diphtheria, tetanus, and pertussis antibodies were assessed in house using a multiplex immunoassay developed by the RIVM in the Netherlands as described previously (40), and antibody levels were measured using the Bioplex 200 system (Bio-Rad, Belgium).

Cytokine multiplex, intracellular staining (ICS), and antibody data were analyzed by fitting a linear mixed model using restricted maximum likelihood. Bleed, vaccine group and sex, and their interactions were fitted as fixed effects, and infant and sample within infant, were fitted as random effects. The F-test was used to test for interactions between the fixed effects, in particular the two-way group-by-bleed and three-way sex-by-group-by-bleed interactions. Comparisons between treatment groups, and sexes within treatment groups, at each time point, were based on t-tests utilizing the predicted means and SEs of differences recovered from the fitted mixed model. Comparisons were conducted at the 5% significance level with no adjustments for multiplicity of outcome variables or comparisons. Ratios of cytokines were intrinsically skewed and required logarithmic transformation prior to analysis, whereas diagnostic plots of residuals indicated that the cytokine data did not require log transformation. Logarithmic transformations were also required to overcome heterogeneity of variance in the antibody data. Data were analyzed using GenStat 17 (VSN International), R 3.1.2 (www.r-project.org), and Stata version 12.1 (StataCorp LP, USA).

137 infants were recruited into the study at birth with males and females randomized separately (Figure 1). The groups were of equivalent weight-for-age z scores at recruitment (data not shown). At 6 weeks (range 5–7 weeks), 131 babies remained in the study and were randomized to one of the two study groups: 67 infants (34 males and 33 females) received BCG and 64 (35 males and 29 females) were randomized to the control group and received BCG at 18 weeks upon study completion (Figure 1).

Tuberculin skin test reactivity and in vitro PPD assays were used to assess mycobacterial responses post-vaccination. The BCG vaccinated group had greater TST reactivity than the naïve control group (Table 2). Thus, while only 7 of 52 (13.5%) BCG naïve infants had a positive TST response (any induration > 1 mm), 32 of 53 (60.4%) BCG vaccinated infants were TST positive. The median induration in the unvaccinated infants was 0 (IQ range 0, 0) further indicating low level TST activity in this group, as compared to the BCG vaccinated group who had a median induration of 4.5 mm (IQ range 0, 10.5) (Control group vs BCG group p < 0.0001). The BCG vaccinated males had greater induration than females (M 8.5 mm, F 1.25 mm, M vs F p = 0.014) and responded more frequently than females (M = 69%, F = 50%).

Surprisingly, there was no evidence of a boosting of PPD reactivity in the BCG vaccinated group at V2 or V3 for any of the cytokines tested in culture supernatants (Table S1 in Supplementary Material). On the contrary, in BCG vaccinated females but not males the PPD-specific IL-12(p70) (p = 0.0025) (Figure 2A), IFN-γ (p = 0.0003) (Figure 2B), IL-4 (p = 0.0018) (Figure 2C), IL-10 (p = 0.004) (Figure 2D), and IL-17 (p = 0.0003) (Figure 2E) had all declined by V3 compared to V2. By contrast, the BCG naïve group had an increase in IFN-γ (p = 0.0164) (Figure 2A), IL-4 (p = 0.0018) (Figure 2C), IL-10 (p = 0.004) (Figure 2D), and IL-17 (p = 0.0003) (Figure 2E) at V2 compared to baseline.

Intracellular staining by flow cytometry was performed to characterize CD4 and CD8 T cells, and innate cells (CD3⁻ cells) expressing IL-2, IL-17, TNF-α, IL-10, IFN-γ, or combinations of these cytokines intracellularly (Figures S1 and S2 in Supplementary Material). This failed to show any significant sex, group, or time interactions for PPD-specific CD4⁺, CD8⁺, or CD3⁻ ICS responses for any of the cytokines in BCG vaccinated or control infants (data not shown).

Levels of the innate pro-inflammatory cytokine TNF-α were analyzed in the TLR ligand and pathogen culture supernatants after overnight stimulation. While robust TNF-α responses were detected to the TLR ligands (Table S2 in Supplementary Material) and heat-killed pathogens (Table S3 in Supplementary Material), there was no difference in TNF-α responses between BCG vaccinated and naïve infants nor between males and females. In keeping with the soluble cytokine data, there was no effect of vaccine group, bleeding time point or sex on TNF-α responses by ICS of CD4⁺, CD8⁺, or CD3⁻ cells (data not shown).

IL-12(p70) (V1 vs V3 p = 0.0355) and IFN-γ (V1 vs V3 p = 0.0018; V2 vs V3 p = 0.0232) responses declined to the TLR4 ligand LPS in BCG vaccinated females, and IL-12(p70) (V1 vs V3 p = 0.0129) but not IFN-γ declined in the BCG naïve control females too (Figures 3A,B). BCG naïve females showed an increase in IFN-γ production to PMA/ionomycin by V3 (V1 to V3: p = 0.0019; and V2 to V3: p = 0.0334), while the BCG vaccinated females did not (Figure 3C). BCG naïve, but not vaccinated, males had increased IL-2 production to both the TLR2 ligand HKLM (V1 vs V2 p = 0.0057) (Figure 3D) and CA (V1 vs V2 p = 0.0207) (Figure 3E) at V2, which declined by V3 (HKLM V2 vs V3 p = 0.0005; CA V2 vs V3 p = 0.026). The BCG vaccinated males also had an increase in IL-2 to CA at V2, which did not quite reach significance (p = 0.0512), but declined significantly by V3 (V2 vs V3 p = 0.0498) (Figure 3E). The only effect on intracellular cytokine production by flow cytometry after stimulating with the TLR ligands and heat-killed pathogens was an increase in IFN-γ from CD8⁺ T cells after CA stimulation in the BCG vaccinated females at V2 compared to V1 (p = 0.0316) (Figure 3F). The similar increase in CA-specific CD8⁺IFN-γ⁺ responses from V1 to V2 was not significant in the males, but the level at V2 was significantly higher than at V3 (p = 0.0151) (Figure 3G).

While the Th2 cytokine IL-4 was not affected by vaccine group or sex in culture supernatants, the ratio of IFN-γ to IL-4, indicative of the Th1:Th2 cytokine balance, was affected. Thus, BCG naïve males had decreased IFN-γ:IL-4 at V3 to TLR4 (LPS) (V1 vs V3 p = 0.039) (Figure 4A) and TLR7/8 (CLO-75) stimulation (V2 vs V3 p = 0.0194) (Figure 4B); and BCG vaccinated females had decreased IFN-γ:IL-4 to SP at V2 compared to baseline (V1 vs V2 p = 0.0128) which increased again by V3 (V2 vs V3 p = 0.0227) (Figure 4C; Table 3; Tables S2 and S3 in Supplementary Material).

IL-10 is an anti-inflammatory cytokine that can be produced by T helper 1 (Th1), Th2, and regulatory T cells (41). IL-10 responses to TLR4 (LPS) declined in the BCG vaccinated females by V3 (V2 vs V3 p = 0.0482) (Figure 5A), while IL-10 responses to CA increased in the BCG naïve but not vaccinated females at V2 with a further increase at V3 (V1 vs V3 p = 0.0001, V2 vs V3 p = 0.0217) (Figure 5B; Tables S2 and S3 in Supplementary Material). There was no effect of vaccine group, bleeding time point, or sex on ICS IL-10 responses to any antigen (data not shown).

Bacillus Calmette–Guérin naïve females had a decrease in IFN-γ:IL-10 to SP at V2 and V3 (V1 vs V2 p = 0.0191; V1 vs V3 p = 0.0141) (Figure 5C), and BCG naïve males likewise had a decline in IFN-γ:IL-10 to SP at V3 (V2 vs V3 p = 0.0472) (Figure 5D). BCG naïve males also had a decline in IFN-γ:IL-10 to TLR2 (HKLM) (V2 vs V3 p = 0.017) (Figure 5E) and TLR4 (LPS) in vitro at V3 (V1 vs V3 p = 0.0342; V2 vs V3 p = 0.0163) (Figure 5F). Furthermore, the ratio of TNF-α to IL-10 in EC (V1 vs V3 p = 0.005; V2 vs V3 p = 0.0457) (Figure 5G) and SP cultures (V1 vs V3 p = 0.0028) (Figure 5H) decreased by V3 in the naïve females, but remained constant in those who received BCG. This effect was not observed in male infants (Table 3).

IL-17 production following TLR2 ligand (HKLM) stimulation increased at V2 in the BCG naïve (V1 vs V2 p = 0.0407), but not vaccinated, males (Figure 6A); and HKLM-specific IL-17 declined at V3 in BCG vaccinated (V1 vs V3 p = 0.0165), but not naïve, females (Figure 6B). There was no effect of vaccine group or sex on IL-17 responses for any of the other stimuli (Tables S2 and S3 in Supplementary Material). IL-17 reactivity in CD4⁺, CD8⁺, or CD3⁻ cells by ICS was not affected by BCG vaccination or sex at any time point (Table 3).

Since BCG has been shown to boost antibody responses to other EPI vaccines, even when given before the vaccine in question (35), we measured antibody levels to hepatitis B vaccine, poliovirus serotypes 1 and 3, diphtheria toxoid, tetanus toxoid (TT), and pertussis toxoid. There was significant induction of antibody responses to all vaccines by V3, but not V2, compared to V1 (Table 4). All infants except for one had protective antibody levels to TT at V1, confirming a benefit of maternal vaccination, as expectant mothers in The Gambia receive two doses of TT vaccine during the second and third trimesters. No additional boosting effect was observed in the BCG vaccinated group compared to the naïve group for any of the vaccines. The only sex difference was higher HepB IgG levels in control males compared to control females at V3 (p = 0.040), while no such difference was observed in the BCG vaccinated group.