Learning accurate users and news representations is critical for news recommendation. Despite great progress, existing methods seem to have a strong bias towards content representation or just capture collaborative filtering relationship. However, these approaches may suffer from the data sparsity problem (user-news interactive behavior sparsity problem) or maybe affected more by news (or user) with high popularity. In this paper, to address such limitations, we propose content filtering enriched GNN framework for news recommendation, ConFRec in short. It is compatible with existing GNN-based approaches for news recommendation and can capture both collaborative and content filtering information simultaneously. Comprehensive experiments are conducted to demonstrate the effectiveness of ConFRec over the state-of-the-art baseline models for news recommendation on real-world datasets for news recommendation.