A novel theory on the pathophysiology of depression would be expected to resolve a contradiction between therapeutic time lag and monoamine hypothesis. On the basis of the fact that a subgroup of depression appears during or after stress, we exposed rats to a long-term (2 weeks) forced walking stress and produced depression-model rats in one group and spontaneous recovery rats in another. The density of axon terminals of the locus coeruleus (LC) neurons in the frontal cortex stained by dopamine β-hydroxylase antiserum was lower in the depression-model rats than in the spontaneous recovery rats and in the control rats without stress. The density was higher in the model rats daily treated with imipramine than in those treated with saline. Morphological projection (MP) index (a percentage of horseradish peroxidase-positive LC cells in total number of LC cells) and electrophysiological projection index (a percentage of LC neurons activated antidromically by electrical stimulation of the cerebral cortex) were lower in the depression-model rats than in the recovery and control rats. MP index was higher in the imipramine-treated rats than that in the saline-treated rats. Electron microscopic examination of the LC disclosed such degenerative changes as low-dense areas without structure, aggregation of intracellular organs, destroyed membranes around the rough endoplasmic reticulum (rER), a decreased number of deformed subsurface cisterns, glia invaginated into the LC neurons and prominent appearance of microglia containing increased number of lipofustin or lysosome in the model rats, but not in the spontaneous recovery rats. These findings suggest that the terminals and cell bodies of the LC noradrenergic neurons degenerate in the stress-induced depression-model rats and regenerate in the imipramine-treated model rats. This degenerative change may possibly contribute to the decrease in synthesis and metabolism of noradrenaline (NA), the slowing of axonal flow, the accumulation of NA in the neurons, the decrease in discharge rate of LC neurons without stress and the increase in release of NA in response to an additional stress. It may also explain the therapeutic time lag that is required to repair the noradrenergic neurons.