Fibro-adipogenic progenitors (FAPs) are muscle-resident mesenchymal progenitors that can contribute to muscle tissue homeostasis and regeneration, as well as postnatal maturation and lifelong maintenance of the neuromuscular system. Recently, traumatic injury to the peripheral nerve was shown to activate FAPs, suggesting that FAPs can respond to nerve injury. However, questions of how FAPs can sense the anatomically distant peripheral nerve injury and whether FAPs can directly contribute to nerve regeneration remained unanswered. Here, utilizing single-cell transcriptomics and mouse models, we discovered that a subset of FAPs expressing GDNF receptors Ret and Gfra1 can respond to peripheral nerve injury by sensing GDNF secreted by Schwann cells. Upon GDNF sensing, this subset becomes activated and expresses Bdnf. FAP-specific inactivation of Bdnf (Prrx1^Cre; Bdnf^fl/fl) resulted in delayed nerve regeneration owing to defective remyelination, indicating that GDNF-sensing FAPs play an important role in the remyelination process during peripheral nerve regeneration. In aged mice, significantly reduced Bdnf expression in FAPs was observed upon nerve injury, suggesting the clinical relevance of FAP-derived BDNF in the age-related delays in nerve regeneration. Collectively, our study revealed the previously unidentified role of FAPs in peripheral nerve regeneration, and the molecular mechanism behind FAPs’ response to peripheral nerve injury.

What determines when neural representations of memories move together (integrate) or apart (differentiate)? Classic supervised learning models posit that, when two stimuli predict similar outcomes, their representations should integrate. However, these models have recently been challenged by studies showing that pairing two stimuli with a shared associate can sometimes cause differentiation, depending on the parameters of the study and the brain region being examined. Here, we provide a purely unsupervised neural network model that can explain these and other related findings. The model can exhibit integration or differentiation depending on the amount of activity allowed to spread to competitors — inactive memories are not modified, connections to moderately active competitors are weakened (leading to differentiation), and connections to highly active competitors are strengthened (leading to integration). The model also makes several novel predictions — most importantly, that when differentiation occurs as a result of this unsupervised learning mechanism, it will be rapid and asymmetric, and it will give rise to anticorrelated representations in the region of the brain that is the source of the differentiation. Overall, these modeling results provide a computational explanation for a diverse set of seemingly contradictory empirical findings in the memory literature, as well as new insights into the dynamics at play during learning.