Wireless Sensor Networks (WSNs) are indispensable for providing crucial sensing and data collection services to IoT systems. The diversity of IoT applications leads to data heterogeneity, which has become a prevalent characteristic of WSNs. Most existing work can only cope with data-homogeneous WSNs, where the same policy for clustering and data processing is applied to all network nodes. However, in practical IoT systems, different policies are desired to support the diverse types of IoT data. The neglect of data heterogeneity may result in poor data collection performance or even non-functionality. In this paper, we propose a cooperative data collection protocol (CO-LEACH) for data-heterogeneous WSNs. Specifically, 1) To mitigate the impact of data heterogeneity, CO-LEACH introduces an efficient cooperative clustering algorithm, in which heterogeneous nodes jointly form clusters, nodes with a higher proportion of intra-cluster type are more likely to be elected as cluster heads, thereby improving data collection efficiency. 2) To cope with the heterogeneous data, CO-LEACH presents an adaptive reprogramming strategy, making cluster heads capable of processing heterogeneous data. Simulation experiments show that CO-LEACH makes a better energy balance of all devices than state-of-the-art protocols, and prolongs the network lifetime by 3–5 times in data-heterogeneous WSNs.