In this paper, we consider a hybrid vehicular network, in which vehicles transmit data via the cellular network and dispose of a V2V interface. In this context, we propose an auto-adaptive multi-hop clustering algorithm, which optimizes the cellular radio resource under the constraint of a maximum packet loss rate (PLR) in the V2V network. The larger V2V based clusters are, the higher the data compression ratio at the cluster head is and the smaller the amount of required resource on the cellular link is. However, PLR becomes higher due to the collision on the V2V channel when increasing the number of hops. The proposed algorithm thus adapts dynamically the maximum number of hops in clusters according to the vehicular traffic density. By simulations, we show that it performs better in terms of aggregated cellular data and packet loss rate than any fixed-hop clustering algorithm in a dynamic scenario.