Transcriptomic studies in infants suggested that the relative proportions of myeloid and lymphoid cells in infants have predictive value for TB risk (HF, unpublished). Concordant animal studies support this notion , as do recent studies in adults with AIDS . We found that the ML ratio in peripheral blood at around three months of age was associated with TB disease or MTB infection-free survival in South African children, notwithstanding conventional risk factors. Per 0.1 unit increase in ML ratio at three months of age, the hazard of probable or definite TB disease before two years old increased by roughly 4% (approximately 1.50.1). While significant, the modest effect size suggests that the ML ratio plays a modest role in predicting TB disease-free survival; its utility may, therefore, be limited to combination with other tools to stratify TB risk or to study pathophysiologic determinants of TB disease.
Several factors support the ML ratio being on the causal pathway for TB development. First, monocytes are target cells for mycobacterial growth and lymphocytes are the major effectors for mycobacterial clearance. Second, there appears to be a dose or gradient effect with higher ratios being more predictive than lower ratios across the ML ratio gradient. Third, as we demonstrated, altered ML ratios precede active disease; hence, reverse causality is unlikely. Fourth, our data in children are consistent with in vitro findings. Fifth, there is overall coherence of this finding with experimental animal and observational adult studies. Finally, the association between the ML ratio and subsequent TB disease has partial specificity. The association is stronger for the ML ratio rather than monocyte counts and definite/probable TB rather than possible TB. We have also recently reported that the ML ratio may be associated with childhood malaria incidence ; hence, the ML ratio may actually have pleiotropic associations with some childhood infectious diseases. However, the effect size seems modest. Further studies should endeavor to assess more detailed subsets of monocytes and lymphocytes to identify precise cellular players in this association.
The mechanistic basis for the association is not addressed in this work. It is plausible that the ML ratio reflects relative frequency of myeloid or lymphoid biased hematopoietic stem cells, and that the ratio may, therefore, reflect ontogeny driven differences in function [20,21]. This argument remains speculative. Alternatively, the ML ratio may reflect the relative frequency of monocytes as target cells and lymphocytes as effectors against TB. The likelihood of the latter explanation is reduced by the lack of association between monocyte counts and/or lymphocyte counts alone with TB outcomes. A third possible explanation is that an altered ML ratio is due to a specific monocyte subset defect. Since HIV infects and alters monocyte function in a subset specific manner , it is possible that HIV exposure alters monocyte functions and ratios. The similarity in effect between HIV-infected and HEU infants, however, mitigates the likelihood of the association being driven by HIV infection.
Our analysis has several limitations. Firstly, the definition of TB outcomes is challenging in infants. This analysis benefits from using data from a trial where outcomes were stringently defined and prospectively recorded, without knowledge of this hypothesis. Clearly specified definitions, including those delineating certainty of endpoints, allowed evaluation of the sensitivity of our finding to alternative outcome definitions, including more stringent definitions of TB disease. Secondly, this observational study cannot conclusively demonstrate a causal relationship between the ML ratio and TB disease. Mendelian randomization approaches, leveraging genetic correlates of the ratio, may facilitate more stringent evaluation of the association. Thirdly, although we observed association between the ML ratio and incident TB disease in all HIV-exposed infants regardless of HIV infection status, further studies are required to establish whether the association is present in infants not exposed to HIV. Use of an a priori statistical analysis plan ameliorated the likelihood of the significant association between the ML ratios and TB risk being a false discovery due to multiple hypothesis testing.