Note on anomalies in the AdS-CFT correspondence