M. tuberculosis (Mtb) senses and responds to changes in its environment primar-ily through two-component signalling systems (TCSs). Each TCS contains a trans-membrane histidine kinase (HK ) protein and a cytoplasmic response regulator (RR) protein. HK detects a stimulus and gets phosphorylated. It then binds and transfers the phosphoryl group to the RR of the same TCS. Activated RR then triggers gene ex-pression, including upregulation of the HK and RR involved, eliciting responses that are essential for the bacterium to adapt. Though di erent TCSs detect distinct stimuli, the binding regions of the HK s and RRs share signi cant similarity. This raises the possibil-ity of crosstalk, where HK s dissipate signals to RRs that do not belong to the same TCS. Studies have argued that such dissipation of signals impairs the fitness of the organism, as it decreases the output levels as well as triggers unwanted responses. In contrast, a recent experimental study has discovered that TCSs of Mtb share extensive crosstalk, violating the widely accepted specificity paradigm. In this study, we have attempted to unravel the evolutionary underpinnings of this extensive crosstalk observed in Mtb. We hypothesised that such crosstalk may be advantageous in programmed environments, where there are well-defined sequences of stimuli. In such situations, crosstalk can up-regulate HK s and RRs of non-cognate TCSs. This up-regulation primes the latter TCSs for upcoming signals, increasing their sensitivity. We constructed a mechanistic model of the functioning of TCSs and a fitness variable to qualitatively measure the response of a TCS to a signal, to test the hypothesis. We performed population genetics simulations of the evolution of phenotypes of different crosstalk patterns. We found that in a random environment, the phenotype without any crosstalk is selected over time, which is in agreement with prevalent arguments in favour of specificity of TCSs. But when the environment is programmed, the phenotype with a crosstalk pattern mirroring the pattern of stimuli dominates the population. Finally, we found evidence for the evolutionary preference to preserve crosstalk in gene sequences of HK s and RRs encoded in Mtb. We found that the binding domains of HK s and RRs, which were predicted to share crosstalk, are under greater pressure to be similar than those domains which do not crosstalk. Our study thus provides a plausible explanation of the unexpected presence of crosstalk in Mtb. Since these cross-interactions aid the pathogen to adapt in the host, inhibitors of such interactions are likely to have therapeutic potential.