A salt bridge is a non-covalent interaction between two ionized sites. It has two components: a hydrogen bond and an electrostatic interaction. In a salt bridge, a proton migrates from a carboxylic acid group to a primary amine or to the guanidine group in Arg. Typical salt bridges involve Lys or Arg as the bases and Asp or Glu as the acids. Of all the non-covalent interactions, salt bridges are among the strongest.
When hydrogen atoms are added to a PDB file using the MOPAC keyword ADD-H, the default is for all ionizable sites, including the type that allow salt bridges to form, to be neutralized. In other words, salt bridges cannot be created by ADD-H; in fact, if salt bridges existed before ADD-H was run, then they would be destroyed by the ADD-H operation. A legitimate question at this point is: "Should ionizable residues be ionized?" Some residues, e.g., Asp, Glu, Arg, Lys, His, and Tyr as well as the first and last residues of a protein are potentially ionizable, so why not ionize them by default? Conventional wisdom suggests that most ionizable residues should be ionized, but if that is done some unreasonable structures might be created. Thus in 1QGK the sequence Gln333-Asp334-Glu335-Asn336-Asp-337-Asp-338-Asp339-Asp340-Asp341-Trp342, or (QDENDDDDDW), is present. If the Glu and all the Asp were ionized, then this set of connected residues would have a net charge of -7. This is unlikely to happen because of the large buildup of charge; a more likely structure would involve one or more salt bridges and possibly one or two of the other acid residues also being ionized. When the sequence Asp334-Asp341 was extracted from 1QGK and optimized, on its own, as the 7+ anion, the system expanded by about 12%, when solvated the system contracted by about 3%. Without more information, deciding which residues should be ionized is very difficult, if not impossible.
The alternative to selecting which residues to ionize is to define all residues as being initially neutral; this is the default in MOPAC. Although this might seem to be extreme, two simple operations are available to allow ionized residues to form. First, when the positions of hydrogen atoms are optimized, some salt bridges might form spontaneously as a result of protons migrating from acid to basic residues. Second, after all strain in the structure has been relieved, the state of ionization can be inspected, and if a potentially salt bridge is identified, the proton involved can be migrated from the acid to the basic residue (use SITE for this operation) and the structure re-optimized. If the heat of formation decreases, then the new structure would be used in subsequent operations.
After ADD-H has been run, i.e., after hydrogen atoms have been added to a biomolecule, the positions of the hydrogen atoms must be optimized before work can be started on modeling protein chemistry. During this operation some hydrogen atoms might migrate from acid to basic residues to give rise to salt bridges, but is unlikely that all of the ones that should be present would form. Therefore, carefully review the structure of the protein, in particular, look at the ionization state of the Lys and Arg residues. Examine the environment of each neutral Lys and Arg residue, if there is a nearby unionized acidic residue, carefully move the proton from the acid to the base residue.
At this point, a useful operation would be to run the data set with SITE=(COO,NH3,ARG(+),His(+)). This generates a list of potential ionizable sites. Use this list to identify possible salt bridges.
Once all changes have been made, re-run the partial optimization to allow the hydrogen atoms that have been modified to move into their correct positions. If there is any doubt regarding whether a salt bridge should or should not be present, optimize the positions of the hydrogen atoms in the two candidate systems - the salt bridge and the neutral form - separately. Of course, in order to allow heats of formation to be compared, the set of keywords used should be the same, in particular if EPS=78.4 is used in one job it should be used in the other. Whichever one has the lower energy is the preferred structure. The resulting structure will then be a good starting point for subsequent operations.
If solvent is considered to be important, then add keyword EPS=78.4; this will cause Andreas Klamt's COSMO implicit solvation model to be used. When a readily ionizable residue is on or near the surface of a protein, implicit solvation will lower the energy required for ionization. If two ionizable residues are near to each other, but are still too far apart to allow a salt bridge to be the more stable form, then solvation might provide enough extra stabilization to make the salt bridge the stable form.
Sometimes a salt bridge can be stabilized by the presence of a single bridging water or hydronium ion. This situation can occur inside, i.e., not on the surface of, a protein when the two ionizable residues are further apart than normal.