There are plenty of so called "derivations of Maxwell equations", however, I would not claim that they were based on U(1) necessarily. IMO ME are just the integrability conditions for any conserved quantity (continuity equation): take a generic three-form in R^4, then dJ=0 => J=dF. F has six components corresponding to two vectofields in R^3 which depend on the fourth coordinate and satisfy the div and curl relations as in ME ;)
U(1) symmetry (R also works in a classical setting: it's the Lie algebra that matters) is exactly how you get a "generic three-form" to drop out of the action.
