Amino acid sequence and environment are the most important factors determining the structure, stability and dynamics of proteins. To evaluate their roles in the process of folding, we studied a retroversion of the well-described Trp-cage miniprotein in water and 2,2,2-trifluoroethanol (TFE) solution. We show, by circular dichroism spectroscopy and nuclear magnetic resonance (NMR) measurement, that the molecule has no stable structure under conditions in which the Trp-cage is folded. A detectable stable structure of the retro Trp-cage, with the architecture similar to that of the original Trp-cage, is established only upon addition of TFE to 30% of the total solvent volume. The retro Trp-cage structure shows a completely different pattern of stabilizing contacts between amino acid residues, involving the guanidinium group of arginine and the aromatic group of tryptophan. The commonly used online prediction methods for protein and peptide structures Robetta and PEP-FOLD failed to predict that the retro Trp-cage is unstructured under default prediction conditions. On the other hand, both methods provided structures with a fold similar to those of the experimentally determined NMR structure in water/TFE but with different contacts between amino acids.