Circular differential microphone arrays (CDMAs) provide compact geometry, full-azimuth coverage, and flexible beam steering, making them suitable for speech enhancement. However, practical acoustic environments often involve unknown desired-source directions and dynamic or multiple interferers. In such scenarios, fixed-null CDMA beamformers are sensitive to steering errors and null-placement mismatch, leading to degraded performance. This paper proposes a circular differential beamformer fusion framework with target DOA estimation. The desired-source direction is first estimated using a CDMA-based steered response power method, and multiple directional-derivative-constrained CDMA beamformers are then constructed toward the estimated direction with different null hypotheses. The candidate outputs are adaptively fused by minimizing the output variance, so that beamformers with lower residual interference are emphasized. Both hard-selection and soft-weighting strategies are investigated. Simulation results in reverberant environments with moving and multiple interferers show that the proposed method improves interference suppression and robustness over representative CDMA beamforming methods.