Acidic fibroblast growth factor (FGF-1), keratinocyte growth factor (FGF-7), and FGF-10 are homologues with distinct specificity. In the presence of heparin, FGF-1 binds and activates in vitro all FGFR subtypes, while FGF-7 exhibits absolute specificity for the IIIb splice variant of FGFR2. FGF-10 exhibits a similar specificity but also binds the FGFR1IIIb isoform. Neither FGF-7 nor FGF-10 will bind to IIIc isoforms of FGFR. Molecular models of FGF, heparin, and the FGFR ectodomain suggested that sequences between beta-strands 10 and 12 of FGF may be important for the interaction of FGF with the heparin-FGFR ectodomain duplex. Site-directed mutants of FGF-7 and FGF-10 were prepared to test whether this domain might underlie failure of FGF-7 and FGF-10 to bind to the FGFRIIIc isoforms. Constructions with substitution of FGF-1 sequences spanning the entire C-terminus encoded in exon 3 or only C-terminal sequences spanning beta-strands 10 through 12 conferred ability on FGF-7 to bind to and activate FGFRIIIc without a significant loss in binding to or activation of FGFR2IIIb. A series of twelve different substitutions of shorter segments of FGF-1 sequences into the C-terminal portion of FGF-7 or FGF-10 revealed that substitution of GSCKRG for GIPVRG or the tri-peptide sequence KKN for NQK just N-terminal to it conferred dual activities on both the FGF-7 and FGF-10 backbones. The results suggest that the combined sequence domain, which we call the FGF glycine box (G-box), is a major determinant for the specificity of the binding of FGF to heparan sulfate-FGFR duplexes.