A critical step in understanding the mode of action of insecticidal crystal toxins from Bacillus thuringiensis is their partitioning into membranes and, in particular, the insertion of the toxin into insect brush border membranes. The Umbrella and Penknife models predict that only α-helix 5 of domain I along with adjacent helices α-4 or α-6 insert into the brush border membranes because of their hydrophobic nature. By employing fluorescent-labeled cysteine mutations, we observe that all three domains of the toxin insert into the insect membrane. Using proteinase K protection assays, steady state fluorescence quenching measurements, and blue shift analysis of acrylodan-labeled cysteine mutants, we show that regions beyond those proposed by the two models insert into the membrane. Based on our studies, the only extended region that does not partition into the membrane is that of α-helix 1. Bioassays and voltage clamping studies show that all mutations examined, except certain domain II mutations in loop 2 (e.g. F371C and G374C), which disrupt membrane partitioning, retain their ability to form ion channels and toxicity in Manduca sexta larvae. This study confirms our earlier hypothesis that insertion of crystal toxin does not occur as separate helices alone, but virtually the entire molecule inserts as one or more units of the whole molecule.