The lymphocyte pore-forming protein perforin is essential for maintaining immune homeostasis and for effective defense against intracellular pathogens. To date, there have been no reported structure-function studies to substantiate the function of any putative domains of perforin, which have been postulated totally on primary sequence similarities with domains in other proteins. In this report, we have used recently developed modalities for expressing full-length perforin and robust functional assays to investigate one of the hallmarks of perforin function: its absolute dependence on calcium for lipid binding and cell lysis. We provide, for the first time, experimental evidence that the predicted C-terminal C2 motif constitutes a functional domain that is responsible for membrane binding of perforin. Whereas conserved aspartate residues at positions 429, 435, 483, and 485 were essential for calcium-dependent plasma membrane binding and cell lysis, the contribution of Asp-491 was limited. Finally, after experimentally verifying an optimized three-dimensional model, we have made predictions on the impact of two inherited perforin mutations of the C2 domain on calcium-dependent lipid binding and cell lysis.

Up to 60% of cases of the autosomal recessive immunodeficiency hemophagocytic lymphohistiocytosis (HLH) are associated with mutations in the perforin (PRF1) gene. In this study, we expressed wild-type and mutated perforin in rat basophil leukemia cells to study the effect on lytic function of the substitutions A91V and N252S (commonly considered to be neutral polymorphisms) and 22 perforin missense substitutions first identified in HLH patients. Surprisingly, we found that A91V perforin was expressed at reduced levels compared with wild-type perforin, resulting in partial loss of lytic capacity. In contrast, expression and function of N252S-substituted perforin were normal. Most HLH-associated mutations resulted in protein degradation (probably due to misfolding) and complete loss of perforin activity, the exception being R232H, which retained approximately 30% wild-type activity. Several other mutated proteins (H222Q, C73R, F157V, and D313V) had no detectable lytic activity but were expressed at normal levels, suggesting that