The nicotinamide cofactor specificity of enzymes plays a key role in regulating metabolic processes and attaining cellular homeostasis. Multiple studies have used enzyme engineering tools or a directed evolution approach to switch the cofactor preference of specific oxidoreductases. However, whole-cell adaptation toward the emergence of novel cofactor regeneration routes has not been previously explored. To address this challenge, we used an Escherichia coli NADPH-auxotrophic strain. We continuously cultivated this strain under selective conditions. After 500 to 1,100 generations of adaptive evolution using different carbon sources, we isolated several strains capable of growing without an external NADPH source. Most isolated strains were found to harbor a mutated NAD+-dependent malic enzyme (MaeA). A single mutation in MaeA was found to switch cofactor specificity while lowering enzyme activity. Most mutated MaeA variants also harbored a second mutation that restored the catalytic efficiency of the enzyme. Remarkably, the best MaeA variants identified this way displayed overall superior kinetics relative to the wild-type variant with NAD+. In other evolved strains, the dihydrolipoamide dehydrogenase (Lpd) was mutated to accept NADP+, thus enabling the pyruvate dehydrogenase and 2-ketoglutarate dehydrogenase complexes to regenerate NADPH. Interestingly, no other central metabolism oxidoreductase seems to evolve toward reducing NADP+, which we attribute to several biochemical constraints, including unfavorable thermodynamics. This study demonstrates the potential and biochemical limits of evolving oxidoreductases within the cellular context toward changing cofactor specificity, further showing that long-term adaptive evolution can optimize enzyme activity beyond what is achievable via rational design or directed evolution using small libraries. IMPORTANCE In the cell, NAD(H) and NADP(H) cofactors have different functions. The former mainly accepts electrons from catabolic reactions and carries them to respiration, while the latter provides reducing power for anabolism. Correspondingly, the ratio of the reduced to the oxidized form differs for NAD+ (low) and NADP+ (high), reflecting their distinct roles. We challenged the flexibility of E. coli's central metabolism in multiple adaptive evolution experiments using an NADPH-auxotrophic strain. We found several mutations in two enzymes, changing the cofactor preference of malic enzyme and dihydrolipoamide dehydrogenase. Upon deletion of their corresponding genes we performed additional evolution experiments which did not lead to the emergence of any additional mutants. We attribute this restricted number of mutational targets to intrinsic thermodynamic barriers; the high ratio of NADPH to NADP+ limits metabolic redox reactions that can regenerate NADPH, mainly by mass action constraints.