We have studied binding and block of sodium channels by 12 derivatives of the 22-residue peptide mu-conotoxin GIIIA (mu-CTX) in which single amino acids were substituted as follows: Arg or Lys by Gln, Gln-18 by Lys, Asp by Asn, and HO-Pro by Pro. Derivatives were synthesized as described by Becker et al. [(1989) Eur. J. Biochem. 185, 79]. Binding was measured by displacement of labeled saxitoxin from eel electroplax membranes (100 mM choline chloride, 10 mM HEPES-NaOH, pH 7.4). Blocking kinetics were evaluated from steady-state, single-channel recordings from rat skeletal muscle sodium channels incorporated into planar, neutral phospholipid/decane bilayers (200 mM NaCl, 10 mM HEPES-NaOH, pH 7.0). Blocking events generally appeared as periods of seconds to minutes in which current through the single channel was completely eliminated. A notable exception was seen for the substitution Arg-13-Gln for which the "blocked" events showed measurable conductances of about 20-40% of the open state. The substitution of Arg-13 reduced binding to electroplax membranes to undetectable levels and increased the apparent dissociation constant determined for skeletal muscle channels by greater than 80-fold compared with the native peptide. Other substitutions caused smaller decreases in affinity. The decreased potency of the toxin derivatives resulted both from increases in the rates of dissociation from the channel, and from decreases in association rates. Our data support the suggestion by Sato et al. [(1991) J. Biol. Chem. 265, 16989] that Arg-13 associates intimately with the binding site on the channel. In addition, our results suggest that certain residues affect almost exclusively the approach and docking of the toxin with its binding site, others appear to be important only to the strength of the association once binding has taken place, and yet others affect both.