<p>Empathetic Question Generation (EQG), an essential component of emotional support dialogue systems. Existing EQG models face two significant issues. First, the perception of long-distance emotional clues suffers from attenuation. Second, semantic focus tends to be diluted across multi-perspective knowledge representations. To address these challenges, the <b>Emp</b>athetic Question Generation via Syntactic-guided <b>M</b>ulti-perspective <b>H</b>eterogeneous <b>K</b>nowledge Adaptive <b>C</b>oupling framework (Emp-MHKC) is proposed. Specifically, syntactic dependency representations are constructed using an Empathy-Oriented Global Hierarchical Syntactic Dependency Representation method. A global syntactic dependency tree is constructed. Tree-structured positional encoding is then employed to capture hierarchical syntactic dependency features and represent long-distance dependencies across sentences. A Context-aware Multi-perspective Heterogeneous Knowledge Coupling mechanism is designed. It constructs novel multi-perspective heterogeneous knowledge for emotional cognition, contextual inference, and psychological motivation. Through context-aware semantic modulation factors, the mechanism adaptively couples heterogeneous knowledge from multiple perspectives. The coupling promotes focused semantic representation in EQG tasks. Experimental results demonstrate that Emp-MHKC achieves strong performance on both automatic and human evaluations, demonstrating robust contextual alignment and practical empathy guidance.</p>