The development of genetically modified adenoviral vectors capable of specifically transducing a given cell population requires the addition and functional presentation of particular tropism determinants within the virus capsid, together with the abrogation of the molecular determinants that dictate their natural tropism in vivo. The human adenovirus serotype 5 (Ad5) first attaches to the cell surface following high-affinity binding of the C-terminal knob of the fiber capsid protein to the coxsackie and adenovirus receptor (CAR). Here we have assessed whether genetic shortening of the fiber shaft (virus BS1), or replacing the Ad5 fiber shaft and knob with their Ad3 counterparts (virus DB6), could cripple this interaction in vitro and in vivo. A 10-fold decrease in the binding of the modified capsids to soluble CAR was evidenced, which correlated with a similar reduction of their ability to transduce CAR-positive cells in vitro. The ability of BS1 to interact with cellular integrins was also impaired, suggesting that the penton base and the short-shafted fiber when embedded in the capsid preclude each other from efficiently interacting with their cognate cell surface receptors (CAR and integrins respectively). BS1 and DB6 intravenous injections in mice further supported a profound impairment of the ability of the capsid-modified viruses to transduce the liver as demonstrated by a 10-fold reduction of intracellular viral DNA and transgene expression. Interestingly enough, the host humoral response was also specifically weakened in BS1- and DB6-inoculated animals. Taken together, these observations indicate that (i) fiber shortening and (ii) pseudo-typing of Ad5-based vectors with the shaft and knob from non-CAR-binding serotypes constitute two promising strategies to successfully attenuate their native tropism in vitro and most importantly in vivo.