FixL is a multiple-domain bacterial O2-sensing protein that modulates the activity of its kinase domain in response to O2 concentration. The kinase activity is coupled, via phosphoryl transfer, to transcriptional activation by a response-regulating protein, FixJ. Heme ligation resulting in a transition from high to low spin inhibits the kinase through an, as yet, ill-defined mechanism. This report presents spectroscopic, kinetic, and thermodynamic data on various complexes of two deletion derivatives of Rhizobium meliloti FixL, FixLN (the heme domain) and a functional heme kinase, FixL*. Resonance Raman characterization of metFixLN and metFixL* indicates that the heme core is smaller than that observed in metmyoglobin and is indicative of a five-coordinate high-spin heme in metFixLs. Resonance Raman spectra of FixL-CO adducts reveal that the Fe-C = O unit and/or its electrostatic environment in FixL*-CO is distorted relative to that in FixLN-CO. The 1H NMR spectra of the met forms further support the model of an asymmetric perturbation of the heme pocket structure associated with the presence of the kinase domain in FixL*. Observation of equivalent Fe-imidazole stretching vibrations for deoxyFixLN and deoxyFixL* (212 cm-1) indicates that the source of this perturbation in the heme pocket of FixL* does not lie on the proximal side of the heme. The equivalent Fe-imidazole stretching frequencies for deoxyFixLN and FixL* indicate that the presence of the kinase domain does not alter the relative strength of the proximal Fe-imidazole bond and that the proximal imidazole ligand is weakly H-bonded, probably to a backbone carbonyl group. Kinetic and thermodynamic data for the reactions of cyanide and fluoride ions with FixL are consistent with shape selectivity due to steric and/or an anisotropic electrostatic field in the distal heme pocket being responsible for the unique reactivities (or lack thereof) of FixL with ligands, i.e., O2, CO, CN-, F-, N3-, and SCN-. While the rate constants for binding of CN- to metFixLN and metFixL* are an order of magnitude slower than that for metMb, the stabilities of these complexes and metMb-CN are nearly the same. Neither N3- nor SCN- binds to the heme with measurable affinity. Since other ferric heme proteins form stable adducts with these ligands, the inability of FixL to form analogous complexes suggests that the ligand selectivity of this protein is rooted in insurmountable activation barriers to the binding of ligands containing more than two atoms and for ligands whose lowest-energy coordination geometries are linear. This allows the natural O2 ligand to compete kinetically with other naturally occurring ligands that form stable complexes with unencumbered hemes. Moreover, the rate constant for binding of CN- to the functional heme-kinase (metFixL*) is smaller than its metFixLN counterpart and the stability of metFixL*-CN is measurably lower than that of metFixLN-CN. This indicates that the contacts between the heme and kinase domains of FixL* impose more stringent geometric constraints on ligand binding than FixLN. The kinase is thus implicated in a possible mechanism for phosphate-dependent feedback control over ligand affinity of the heme.