Adenoviruses are used extensively as gene transfer agents, both experimentally and clinically. However, targeting of liver cells by adenoviruses compromises their potential efficacy. In cell culture, the adenovirus serotype 5 fiber protein engages the coxsackievirus and adenovirus receptor (CAR) to bind cells. Paradoxically, following intravascular delivery, CAR is not used for liver transduction, implicating alternate pathways. Recently, we demonstrated that coagulation factor (F)X directly binds adenovirus leading to liver infection. Here, we show that FX binds to the Ad5 hexon, not fiber, via an interaction between the FX Gla domain and hypervariable regions of the hexon surface. Binding occurs in multiple human adenovirus serotypes. Liver infection by the FX-Ad5 complex is mediated through a heparin-binding exosite in the FX serine protease domain. This study reveals an unanticipated function for hexon in mediating liver gene transfer in vivo.

Mutations in LMAN1 (also called ERGIC-53) result in combined deficiency of factor V and factor VIII (F5F8D), an autosomal recessive bleeding disorder characterized by coordinate reduction of both clotting proteins. LMAN1 is a mannose-binding type 1 transmembrane protein localized to the endoplasmic reticulum-Golgi intermediate compartment (ERGIC; refs. 2,3), suggesting that F5F8D could result from a defect in secretion of factor V and factor VIII (ref. 4). Correctly folded proteins destined for secretion are packaged in the ER into COPII-coated vesicles, which subsequently fuse to form the ERGIC. Secretion of certain abundant proteins suggests a default pathway requiring no export signals (bulk flow; refs. 6,7). An alternative mechanism involves selective packaging of secreted proteins with the help of specific cargo receptors. The latter model would be consistent with mutations in LMAN1 causing a selective block to export of factor V and factor VIII. But approximately 30% of individuals with F5F8D have normal levels of LMAN1, suggesting that mutations in another gene may also be associated with F5F8D. Here we show that inactivating mutations in MCFD2 cause F5F8D with a phenotype indistinguishable from that caused by mutations in LMAN1. MCFD2 is localized to the ERGIC through a direct, calcium-dependent interaction with LMAN1. These findings suggest that the MCFD2-LMAN1 complex forms a specific cargo receptor for the ER-to-Golgi transport of selected proteins.

No abstract