for ground state ATP analogues use AMP-PNP (or AMP-PCP) or ADP in complex with beryllium fluoride.
In some circumstances ATP-gamma-S will work too. It's less resistant to hydrolysis, so it's a bad idea for many enzymes, but the resulting active site geometry can be different depending on which analogue is used. I know of at least one case where independent structures of the same protein had different nucleotide conformations due to the choices of analogue (ATPgS vs. AMPPNP).
-Nat
