A well-known case of evolutionary adaptation is that of ribulose-1,5-bisphosphate carboxylase (RubisCO), the enzyme responsible for fixation of CO 2 during photosynthesis. Although the majority of plants use the ancestral C 3 photosynthetic pathway, many flowering plants have evolved a derived pathway named C 4 photosynthesis. The latter concentrates CO 2 , and C 4 RubisCOs consequently have lower specificity for, and faster turnover of, CO 2 . The C 4 forms result from convergent evolution in multiple clades, with substitutions at a small number of sites under positive selection. To understand the physical constraints on these evolutionary changes, we reconstructed in silico ancestral sequences and 3D structures of RubisCO from a large group of related C 3 and C 4 species. We were able to precisely track their past evolutionary trajectories, identify mutations on each branch of the phylogeny, and evaluate their stability effect. We show that RubisCO evolution has been constrained by stability-activity tradeoffs similar in character to those previously identified in laboratory-based experiments. The C 4 properties require a subset of several ancestral destabilizing mutations, which from their location in the structure are inferred to mainly be involved in enhancing conformational flexibility of the open-closed transition in the catalytic cycle. These mutations are near, but not in, the active site or at intersubunit interfaces. The C 3 to C 4 transition is preceded by a sustained period in which stability of the enzyme is increased, creating the capacity to accept the functionally necessary destabilizing mutations, and is immediately followed by compensatory mutations that restore global stability.T he adaptive diversification of organisms often requires the evolution of novel enzymatic properties. The evolutionary shift from one enzymatic function to another involves crossing an energetic barrier in a fitness landscape (1). The number of mutations that confer advantageous function during such a shift is consequently limited. Some residues are critical for maintaining the stability of the protein fold, others are important for the catalytic activity itself. Due to the multiple roles of amino acids in proteins, the adaptation of one physical parameter of an enzyme is likely to affect other properties (2). As proteins usually form thermodynamically stable structures, their evolutionary trajectories are constrained to a narrow range of stability (3). Stability and activity are likely to be negatively correlated. Most possible amino acid changes in native proteins are destabilizing and, consequently, mutations that lead to a more favorable enzyme activity are likely to decrease the stability of the protein (2, 4). Compensatory mutations are then needed to restore global stability. These processes are referred to as stability-activity tradeoffs (5-7). Furthermore, proteins with higher stability confer greater evolvability, because there is more scope to accept destabilizing yet functionally beneficial changes (8). Wh...