Working memory requires efficient excitatory drive to parvalbuminpositive (PV) interneurons in the primate dorsolateral prefrontal cortex (DLPFC). Developmental pruning eliminates superfluous excitatory inputs, suggesting that working memory maturation during adolescence requires pruning of excitatory inputs to PV interneurons. Therefore, we tested the hypothesis that excitatory synapses on PV interneurons are pruned during adolescence. The density of excitatory synapses, defined by overlapping vesicular glutamate transporter 1-positive (VGlut1+) and postsynaptic density 95-positive (PSD95+) puncta, on PV interneurons was lower in postpubertal relative to prepubertal monkeys. In contrast, puncta levels of VGlut1 and PSD95 proteins were higher in postpubertal monkeys and positively predicted activity-dependent PV levels, suggesting a greater strength of the remaining synapses after pruning. Because excitatory synapse number on PV interneurons is regulated by erb-b2 receptor tyrosine kinase 4 (ErbB4), whose function is influenced by alternative splicing, we tested the hypothesis that pruning of excitatory synapses on PV interneurons is associated with developmental shifts in ErbB4 expression and/or splicing. Pan-ErbB4 expression did not change, whereas the minor-to-major splice variant ratios increased with age. In cell culture, the major, but not the minor, variant increased excitatory synapse number on PV interneurons and displayed greater kinase activity than the minor variant, suggesting that the effect of ErbB4 signaling in PV interneurons is mediated by alternative splicing. Supporting this interpretation, in monkey DLPFC, higher minor-to-major variant ratios predicted lower PSD95+ puncta density on PV interneurons. Together, our findings suggest that ErbB4 splicing may regulate the pruning of excitatory synapses on PV interneurons during adolescence.synaptic pruning | parvalbumin interneuron | ErbB4 | alternative splicing | schizophrenia I n primates, certain complex cognitive processes, such as working memory, depend in part on the proper activation of specific neural circuits in the dorsolateral prefrontal cortex (DLPFC) (1). In both monkeys and humans, recruitment of DLPFC activity and performance during working memory tasks continue to improve through adolescence (2-4). During this period, excitatory synapses and their principal targets, pyramidal neuron dendritic spines, massively decline in number in the primate DLPFC (5-8). Synaptic pruning is thought to eliminate unwanted or imprecise connections and to strengthen the remaining connections (9, 10), suggesting that this process could contribute to the maturation of working memory function during adolescence.Working memory is thought to emerge from oscillatory activity of DLPFC neurons at gamma frequency (30-80 Hz) (11), which requires the activity of local GABAergic interneurons that express the calcium binding protein parvalbumin (PV) (12, 13). During working memory tasks, excitation from pyramidal neurons recruits PV interneurons, which in turn pro...