To better understand the biophysical basis of osmoprotection by glycine betaine (GB) and the roles of cytoplasmic osmolytes, water, and macromolecular crowding in the growth of osmotically stressed Escherichia coli, we have determined growth rates and amounts of GB, K(+), trehalose, biopolymers, and water in the cytoplasm of E. coli K-12 grown over a wide range of high external osmolalities (1.02-2.17 Osm) in MOPS-buffered minimal medium (MBM) containing 1 mM betaine (MBM+GB). As osmolality increases, we observe that the amount of cytoplasmic GB increases, the amounts of K(+) (the other major cytoplasmic solute) and of biopolymers remain relatively constant, and the growth rate and the amount of cytoplasmic water decrease strongly, so concentrations of biopolymers and all solutes increase with increasing osmolality. We observe the same correlation between the growth rate and the amount of cytoplasmic water for cells grown in MBM+GB as in MBM, supporting our proposal that the amount of cytoplasmic water is a primary determinant of the growth rate of osmotically stressed cells. We also observe the same correlation between cytoplasmic concentrations of biopolymers and K(+) for cells grown in MBM and MBM+GB, consistent with our hypothesis of compensation between the anticipated large perturbing effects on cytoplasmic protein-DNA interactions of increases in cytoplasmic concentrations of K(+) and biopolymers (crowding) with increasing osmolality. For growth conditions where the amount of cytoplasmic water is relatively large, we find that cytoplasmic osmolality is adequately predicted by assuming that contributions of individual solutes to osmolality are additive and using in vitro osmotic data on osmolytes and a local bulk domain model for cytoplasmic water. At moderate growth osmolalities (up to 1 Osm), we conclude that GB is an efficient osmoprotectant because it is almost as excluded from the biopolymer surface in the cytoplasm as it is from native protein surface in vitro. At very high growth osmolalities where cells contain little cytoplasmic water, predicted cytoplasmic osmolalities greatly exceed observed osmolalities, and the efficiency of GB as an osmolality booster decreases as the amount of cytoplasmic water decreases.