Maladaptive fears are often treated using therapies based on extinction—re-exposure to a feared stimulus in a safe environment. Although these treatments can be effective, relapse is common [1]. Since the time of Pavlov, relapse has been thought to indicate that extinction is new learning, not unlearning [2]. From this perspective, relapse is a mnemonic phenomenon in which two opposing memory traces vie for expression. Recent work suggests that the hippocampus is an arena in which this competition plays out.

Experiments in mice demonstrate that the hippocampal dentate gyrus (DG) generates a contextual fear “engram.” Immediate-early gene-based tagging of neurons active during acquisition of contextual fear—fear of the chamber in which a footshock was given—shows that these neurons are reactivated during recall of contextual fear [3]. Furthermore, optogenetic experiments demonstrate that this reactivation is necessary and sufficient for expression of contextual fear [3, 4]. Two recent studies investigated what happens to hippocampal fear ensembles when contextual fear is extinguished.

The first of these studies, by Lacagnina et al. [5], used a transgenic mouse line to tag neurons active during acquisition or extinction of contextual fear. During a test session shortly after extinction training, fear acquisition neurons were suppressed and extinction neurons were reactivated. A month after extinction training mice displayed spontaneous recovery (relapse) of fear, and the pattern reversed: fear acquisition neurons were reactivated while extinction neurons were suppressed. The results suggest that the DG generates distinct fear acquisition and extinction representations, and competition between these representations determines whether fear is suppressed or recovers after extinction. Consistent with this interpretation, optogenetic manipulations demonstrated that reactivation of extinction neurons is necessary for suppression of fear after extinction, whereas reactivation of fear acquisition neurons is necessary for spontaneous recovery [5]

The other recent study, by Khalaf et al. [6], highlights that fear acquisition neurons also play an important role in extinction learning. In this study, neurons active during recall of a remote fear memory (acquired a month before extinction training) were tagged. Reactivation of these neurons was necessary for effective extinction, and artificial stimulation of these neurons improved extinction. For extinction to be effective, the fear acquisition memory must be reactivated during extinction training. Whether these findings apply to recent fear memories, like those studied in the Lacagnina et al. [5] experiments, is not yet known.

These studies reveal new ensemble mechanisms of fear extinction and relapse and raise some interesting questions. For instance, what aspects of extinction training stimulate creation of hippocampal extinction representations, and how do these representations suppress fear? Do they do so by activating extra-hippocampal fear-suppressive networks? In the Lacagnina study, why did fear acquisition neurons become more active over time and extinction neurons less active? Is it because of intrinsic differences between the two populations, changes in upstream input pathways, or different plasticity mechanisms involved in fear acquisition and extinction? Addressing these and other questions raised by the Lacagnina and Khalaf papers may provide keys to making extinction more resistant to relapse.

Funding and disclosure

M.R.D. was supported by NIH grants R01 MH102595 and R01 MH117426. E.T.B. was supported by NIH grant T32MH106454. The authors declare no competing interests.