The G-loop is a 10-residue glycine-rich loop that protrudes from the surface of the mature bacteriophage HK97 capsid at the C-terminal end of the long backbone helix of major capsid protein subunits. The G-loop is essential for assembly, is conserved in related capsid and encapsulin proteins, and plays its role during HK97 capsid assembly by making crucial contacts between the hill-like hexamers and pentamers in precursor proheads. These contacts are not preserved in the flattened capsomers of the mature capsid. Aspartate 231 in each of the ~400 G-loops interacts with Lysine 178 of the E-loop (Extended Loop) of a subunit on an adjacent capsomer. Mutations disrupting this interaction prevented correct assembly, and in some cases induced abnormal assembly into tubes, or small, incomplete capsids. Assembly remained defective when D231 and K178 were replaced with larger charged residues or when their positions were exchanged. Second-site suppressors of lethal mutants containing substitution D231L replaced the ionic interaction with new interactions between neutral or hydrophobic residues of about the same size: D231L/K178V, D231L/K178I, and D231L/K178N. We conclude that it is not the charge, but the size and shape of the side chains of residues 178 and 231 that are important. These two residues control the geometry of contacts between the E-loop and the G-loop, which apparently must be precisely spaced and oriented for correct assembly to occur. We present a model for how the G-loop could control HK97 assembly and identify G-loop-like protrusions in other capsid proteins that may play analogous roles.