Fake news detection has received great attention in recent years due to the proliferation of disinformation. Existing researches focus on modeling the news content and user comments on social media. However, these researches ignore the cascades of news propagation. Accordingly, we propose a Social Content and Cascade Neural Network (SCCN) that integrates cascade information into the model. Specifically, we adopt Bi-LSTM, graph attention mechanism and average pooling to represent the social content as a vector, encoding the cascade structures by graph neural network. We compare our model with representative baselines to validate the performance and perform an ablation analysis to verify the effectiveness of social context and cascade structures. The results show that SCCN model performs better and yields improvement over the baselines according to precision, recall, and F1-score based on two real-world datasets.