An important research area in sensor networks is the design and analysis of distributed estimation algorithms for dynamic information fusion in the presence of heterogeneity resulting from (i) nonidentical information roles of nodes and (ii) nonidentical modalities of nodes. In particular, (i) implies that both active (i.e., subject to observations of a process of interest) and passive (i.e., subject to no observations) nodes can be present in the sensor network. Furthermore, (ii) implies that active nodes can observe different measurements from a process (e.g., a subset of active nodes can observe position measurements and the rest can observe velocity measurements for a target tracking problem). In this paper, we focus on heterogeneous sensor networks, sensor networks with (i) and (ii), and present a new distributed input and state estimation approach. In addition to the presented theoretical contribution including the stability and performance of the proposed estimation approach, an illustrative numerical example is also given to demonstrate its efficacy.