Actin plays crucial roles in the life of the cell while being notorious for its inability to reach a functional conformation without the help of assistant proteins. In eukaryotes, for example, the cytosolic chaperonin containing TCP-1 (CCT) and prefoldin (PFD) are required for actin folding assistance and prevention of protein aggregation in the crowded cellular environment. The folding of non-native actin is known to occur in a number of steps, but the reasons underlying its folding difficulty are unknown. Because a full, atomistic-level, investigation of the kinetics and thermodynamics of folding of such a large molecule is beyond computational reach, we focused our investigation on the role of topological frustration on the folding of actin. Namely, we studied the (re)folding of actin using simulations of a variant self-organized polymer model (SOP-DH) starting from a stretched state, leading to results that correlate well with experimentally driven conclusions and allowing us to make a number of testable predictions. Primarily, our simulations reveal that the successful refolding of the C-terminus end of actin occurs through a zipping process in which the alpha-helices wind up turn by turn upon formation of their native tertiary contacts. In turn, an early formation of the helical structure in this region of the chain has deleterious effects for actin's refolding fitness. Moreover, the C-terminus refolding is a very rare event in our simulations, in agreement with the large activation barrier predicted on the basis of experimental studies of actin unfolding in EDTA. We also discovered that subdomain 4 has a low refolding probability, which can help explain why many of the non-native actin target binding sites for CCT and PFD are located within this subdomain.