The tremendous increase of the traffic demand in cellular networks imposes a massive densification of the traditional cellular infrastructure. The network architecture becomes heterogenous, in particular 4G networks where LTE micro eNodeBs are deployed to strengthen the coverage of macro eNodeBs. This densification yields major issues related to the energy consumption of the infrastructure. Indeed, there is fixed and significant amount of energy required to run each additional node, whatever the traffic load of the network. Mitigating this fixed energy consumption is therefore a major challenge from a societal and economical viewpoint. Extensive researches about energy-saving highlight that to save energy the better strategy is to switch off the radio part of nodes. This is the heart of wireless sensor networks energy-saving strategies, even though the objective for WSN is to maximize the battery life of each individual nodes. In this paper, we develop a parallel between the principles of WSN protocols and the requirements of cellular infrastruc- tures. We then propose a distributed and localized algorithm to dynamically switch off and on the micro eNodeBs of an LTE heterogeneous network following the traffic demand evo- lution in time and analyze it in terms of energy savings. We show that one can expect energy savings of approximately 12% when implementing sleep modes whereas the energy cost for sending the traffic decreases by 24%.