In recent years, content-based publish/subscribe middleware has harnessed the power of Software-Defined Networking (SDN) to leverage performance gains in terms of throughput rates, end-to-end latency, etc. To this end, content filters are directly installed on the Ternary Content Addressable Memory (TCAM) of switches. Such a middleware assumes unlimited TCAM space to deploy content filters. However, in reality, TCAM is a scarce resource and the number of flow table entries available for publish/subscribe traffic is severely limited. While such a limitation poses severe problems for the deployment of publish/subscribe middleware in practice, it is yet to be addressed in literature. So, in this paper, we design a filter aggregation algorithm that merges content filters on individual switches to respect TCAM constraints while ensuring minimal increase in unnecessary network traffic. Our algorithm uses the knowledge of advertisements, subscriptions, and a global view of the network state to perform bandwidth-efficient aggregation decisions on necessary switches. We provide different flavors of this algorithm with varying degrees of accuracy and complexity and thoroughly evaluate their performances under realistic workload. Our evaluation results show that our designed aggregation algorithm successfully meets TCAM constraints on switches while also reducing unnecessary traffic introduced in the network due to aggregation as compared to a baseline approach by up to 99.9%.