This study focused on the evaluation of infant immune responses at 3 months post BCG vaccination, investigating potential factors that may affect cytokine/chemokine signatures in infants from different countries and vaccinated at different ages. It has been reported that variation in resistance to TB and efficacy of BCG may be influenced by latitude dependent factors such as climate, exposure to nontuberculous mycobacteria and genetics . In our comparisons of the immunogenicity of BCG the geographical location and environment of Malawi and The Gambia are much closer to each other than they are to the UK and this corresponds with the similarity of the cytokine/chemokine signatures observed. In addition, the difference in cytokine/chemokine signatures between Malawian and Gambian infants was partly due to the season of birth; the difference in median IFN-γ response in Gambian compared to Malawian infants born in the dry season was not statistically significant (P = 0.31). One potential confounder may be the strain of BCG used for vaccination. It is still unknown how the immune response can be affected by different vaccine strains as it may depend on stimulating antigen, population and assay protocol. In this study, the infants were not immunised with the same BCG vaccine strains; Malawian and UK infants were immunized with the Danish strain, SSI 1331 while Gambian infants received BCG-Russia. These two strains are genetically different and have been shown to induce a different immune response to crude CFP and Ag85 antigens in children in Uganda [20, 21]. Immunization with BCG-Denmark or BCG-Japan induced higher frequencies of mycobacterial-specific polyfunctional and cytotoxic T cells compared with BCG-Russia in Australia . However, statistical differences were shown only when BCG was used for in vitro stimulation, and not with PPD or MTB antigens . In addition, there were no significant differences in cytokine responses to BCG and PPD measured in culture supernatant between infants immunized with BCG-Denmark or BCG Russia . In our study, production of only 3 of 42 cytokines (IL-1, IL-1ra and IL-6) was significantly different in the infants from Malawi and the Gambia after adjusting for season of birth. Therefore, the differences in cytokine responses to M. tb PPD between Malawi and The Gambia may be due to season of birth rather than to the BCG vaccine strains.
The cytokine/chemokine signatures observed were influenced by the timing of BCG vaccination. In Malawi, when compared with infants vaccinated early, the infants vaccinated later (a median age of 5 weeks) showed higher responses to M. tb PPD for most cytokines and chemokines measured but not for IFN-γ. A similar finding of varying BCG immunogenicity by age at vaccination was reported in South African infants, showing that enhanced BCG-specific T cell responses were observed in infants when vaccination was delayed from birth to 10 weeks . However, two studies in The Gambia  and Uganda  showed contradicting results; the magnitude of the IFN-γ response was not significantly altered by the timing of vaccination when comparing infants vaccinated at birth or 2 months in The Gambia , whereas greater frequencies of BCG-specific IFN-γ positive CD4+ T cells were detected in the group vaccinated at birth compared with those vaccinated at 6 weeks of age in Uganda .
In this study, there was no evidence of a difference in median IFN-γ production between Malawian infants vaccinated early compared with those vaccinated late, but there was a trend of higher responses in those vaccinated late. The Malawian infants vaccinated late showed higher median (p < 0.001) Th2 (IL-4, IL-5 and IL-13) recall responses to M. tb PPD than infants vaccinated at birth. This observation suggests that different immune signatures develop according to timing of vaccination which might be an important factor in relation to the induction of protective immune responses following vaccination. The samples obtained from early and late BCG vaccinated infants in Malawi were derived from the same large cohort study . A total of 615 Infants were vaccinated with BCG between 2002 and 2004, and blood samples were obtained from 590 infants at 3 months post BCG vaccination between 2003 and 2005 . Blood samples were obtained from 590 of these infants at 3 months post BCG vaccination. Infants were vaccinated early or late throughout the whole recruiting period. This was as a result of whether the infants were born in hospital and vaccinated at birth or born elsewhere and brought to the clinic later for vaccination. There was no difference in the vaccination they received. Thus no other confounding factors that might be associated with delayed vaccination have been identified. The results of this study are also consistent with evidence that newborns can develop Th1 responses following BCG vaccination [24, 25] unlike the Th2 bias in mice .
In terms of the proportion of infants who showed greater than 10 mg/L of CRP in their plasma, the majority (7/8) were born in the rainy season. In the whole infant cohort no infants showed helminth infections at 3–6 months of age. No infants were bled if unwell or with symptoms of malaria or other infection. We excluded the infants whose mothers had HIV, malaria, or helminth infections but we did not test for other diseases which could induce inflammation in the infants and might affect the CRP concentration in the infants and their mothers. Therefore higher plasma CRP concentration, reflecting level of inflammation, most likely indicates a higher burden of infections, but there was no association between the IFN-γ responses to M. tb PPD and CRP concentration (data not shown). Most of the Malawian infants had a normal range of plasma zinc concentrations and none of the cytokines tested were associated with the concentration of zinc (data not shown), indicating that zinc status might not be a major factor affecting the differences in infant immune responses in Malawi, at least when measured at 3 months post BCG vaccination.
To examine if cytokine activity had been lost over time in storage, we tested culture supernatant samples which had been collected from 4 infants at 3 years post BCG vaccination and stored in the same place as the archived samples we used for this study, since 2006. There was no significant difference in IFN-γ concentrations of 19 culture supernatant samples from each of 4 subjects when measured in 2006 and again in 2010, and a strong correlation was found between IFN-γ values obtained in 2006 and 2010 (r = 0.9808, P < 0.0001) .
Differences in the infant immune responses to M. tb PPD following BCG vaccination are documented at different latitudes and in different populations. The extent to which these influence BCG efficacy remains to be evaluated alongside studies looking for correlates of protection against TB . Season of birth and timing of vaccination have been shown here to be critical factors affecting the cytokine/chemokine signatures induced by BCG vaccination in Malawian infants. In addition, this study indicates the potential importance of the balance between type 1 and 2 cytokines, or between pro- and anti-inflammatory cytokines rather than the absolute quantities of a particular cytokine. Although we were limited by the number of samples tested, the results of this study suggest the need for additional biomarkers to be measured in conjunction with IFN-γ to assess the immunogenicity of BCG, and that potential factors affecting immune responses following vaccination such as season of birth should be taken into account for future trials of new TB vaccines.