Traffic responsive signal control systems bear high potential in reducing delays in congested networks due to their ability of dynamically adjusting right-of-way assignment among conflicting movements, based on real-time traffic measurements. In this work, we focus on distributed traffic signal control for large-scale networks based on the existing Max-Pressure controller, which has been shown to stabilize queues and maximize throughput in congested conditions. Max-Pressure constitutes a feedback control law that tries to balance queues around an intersection by updating green times between signal stages as a function of current queue measurements. Nevertheless, its increased infrastructure requirements impose high implementation costs. Our objective is to investigate how network performance changes when controller is installed only in subsets, (instead of all) of network nodes, while exploring strategies of identifying the most critical nodes. A modified version of Store-and-Forward traffic model is used to emulate spatio-temporal traffic evolution in a large-scale network and evaluate system performance for different controller layouts. Firstly, we observe significant improvement in terms of total delay and network MFD production when Max-Pressure control is applied. More than 85% of the improvement observed when controlling all network nodes can be achieved by controlling only 25% of properly selected nodes, thus reducing implementation costs to one fourth. Further research is needed in order to optimize node selection for the Max-Pressure layout, through evaluation of node impact to network performance. Moreover, investigating the potential of further gains via combining Max-Pressure with centralized control strategies, e.g. perimeter control, is a promising research direction.