John F. Cryan* Neuromodulation Unit, Nervous System Research, Novartis Pharma AG, WSJ 386.344, Basel, CH-4002, Switzerland. *e-mail: john_f.cryan@ pharma.novartis.com Athina M arkou Dept of Neuropharmacology, CVN-7, The Scripps Research Institute, 10550 North Torrey Pines Rd, La Jolla, CA 92037, USA. Irwin Lucki Depts of Psychiatry and Pharmacology, University of Pennsylvania, 538A Clinical Research Building, Philadelphia, PA 19104, USA. Depression is a serious disorder in today’s society, with estimates of lifetime prevalence as high as 21% of the general population in some developed countries [1]. As defined by the American Psychiatric Association [2], depression is a heterogeneous disorder often manifested with symptoms at the psychological, behavioral and physiological levels (Box 1). As with all diseases, approximations of both the disorder and the actions of corrective medications in laboratory animals are essential for the development of effective therapies. The wide spectrum of disruptions that characterize depression highlight the difficulty posing researchers to mimic the disorder in the laboratory. Indeed, two human symptoms, recurring thoughts of death or suicide, or excessive thoughts of guilt, are impossible to model in laboratory animals. The question remains impenetrable as to whether we can ever know whether a laboratory animal is ‘depressed’ (Fig. 1). Nonetheless, numerous attempts have been made to create animal models of depression, or at least of the symptoms of depression, and criteria for their evaluation have been established. Some of the most widely cited criteria were developed by McKinney and Bunney >30 years ago [3]. They proposed that the minimum requirements for an animal model of depression are: (1) it is ‘reasonably analogous’ to the human disorder in its manifestations or symptomatology; (2) there is a behavioral change that can be monitored objectively; (3) the behavioral changes observed should be reversed by the same treatment modalities that are effective in humans; and (4) it should be reproducible between investigators. Most models of depression in use at that time were based on primate separation experiments that attempted to model the entire syndrome of depression. However, subsequent efforts to delineate validity criteria for animal models often do not take into account the reliability and usability of the paradigm in the everyday rodent laboratory setting and are often based on esoteric, theoretical principals rooted in comparing the etiological basis between the human condition and the syndrome in the animal model ([4], but also see [5]). Unlike other medical disorders where the pathology is well defined, such as diabetes or Parkinson’s disease, the underlying pathophysiology of depression is still unresolved, thus making it virtually impossible to fulfill criteria solely based on etiology. Most recently, it has become clear that a more useful strategy might be to model single endophenotypic differences (i.e. one clear-cut behavioral output) relevant to the disease state as opposed to a syndrome [4]. Geyer and Markou [4] have proposed that the only criteria that are necessary and sufficient for initial use are that the paradigm has strong predictive validity and that the behavioral readout be reliable and robust in the same laboratory and between laboratories. The satisfaction of other criteria such as construct or discriminant validity might have heuristic value and are desirable but not essential for the model to provide important initial uses in both basic neurobiological research and drug discovery. Various paradigms have been developed and are instrumental in detecting the antidepressant-like potential of novel compounds in preclinical settings. The models commonly used are diverse and were developed originally based on the behavioral consequences of stress, drug, lesion or genetic manipulations (Table 1). Many of these models have undergone iterative improvements to keep pace with continuing advances in the development of drugs with an increasingly wide array of pharmacological actions. Moreover, such improvements to models continue to be necessary to detect antidepressant effects more precisely in genetically engineered animals and after modifications of cellular and molecular targets. In this review, some of the recent improvements to models that enhance their utility to detect antidepressant effects are highlighted. Animal models are indispensable tools in the search to identify new antidepressant drugs and to provide insights into the neuropathology that underlies the idiopathic disease state of depression. As new targets are developed, both serendipitously and through hypothesis-driven research, existing animal paradigms are being modified and new tests are being developed to detect antidepressant actions of compounds acting on a broad range of neural and genetic targets. This review focuses on recent findings regarding some of the most w idely employed animal models used currently to predict antidepressant potential. Emphasis is placed on recent modifications to such paradigms that have increased their utility and reliability. Furthermore, some key issues that need to be addressed for future discovery of novel antidepressant agents are examined, and the available data on genetically altered mice that might lead to the discovery of novel targets for antidepressant action are collated. Assessing antidepressant activity in rodents: recent developments and future needs John F. Cryan, Athina Markou and Irwin Lucki TRENDS in Pharmacological Sciences Vol.23 No.5 May 2002 http://tips.trends.com 0165-6147/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0165-6147(02)02017-5 238 Review