It is well-established that in vitro measurements do not reflect protein behaviors in-cell, where macromolecular crowding and chemical interactions modulate protein stability and kinetics. Recent work suggests that peptides and small proteins experience the cellular environment differently from larger proteins, as their small sizes leave them primarily susceptible to chemical interactions. Here, we investigate this principle in diverse cellular environments, different intracellular compartments and host organisms. Our small protein folding model is barnase, a bacterial ribonuclease that has been extensively characterized in vitro. Using fast relaxation imaging, we find that FRET-labeled barnase is stabilized in the cytoplasm and destabilized in the nucleus of U2-OS cells. These trends could not be reproduced in vitro by Ficoll and M-PER™, which mimic macromolecular crowding and non-specific chemical interactions, respectively. Instead, in-cell trends were best replicated by cytoplasmic and nuclear lysates, indicating that weak specific interactions with proteins in either compartment are responsible for the in-cell observations. Interestingly, in the cytoplasm barnase's unfolded state is unstable and prone to aggregation, while in the nucleus a stable unfolded state exists prior to aggregation. In the more biologically relevant environment of bacterial cells, barnase folding resembled that in the nucleus, but with no aggregation at higher temperatures. These findings show that protein interactions are evolved for their native environment, which highlights the importance of studying and designing proteins in situ.