A Brake on Disinhibition: Non‐Hebbian Learning at <scp>VIP</scp> Interneuron Outputs

In the February issue of Acta Physiologica, Jabłońska and colleagues report a previously undocumented form of inhibitory synaptic plasticity that operates within hippocampal disinhibitory circuits [1]. Inhibitory plasticity is increasingly recognized as a central mechanism shaping neuronal computation, yet most studies have focused on inhibitory synapses onto excitatory neurons. In contrast, how inhibitory synapses between interneurons adapt and how such plasticity regulates disinhibitory circuits has remained largely unexplored. Jabłońska and colleagues address this knowledge gap by showing that inhibitory synapses formed by vasoactive intestinal peptide (VIP)-expressing interneurons onto stratum oriens interneurons undergo a long-lasting depression triggered solely by postsynaptic burst firing. This plasticity is non-Hebbian, independent of presynaptic activity, and selectively expressed at interneuron—interneuron (I→ I) connections. The study, therefore, reveals that disinhibitory circuits are not static relays but adaptive elements whose gain is dynamically tuned by the circuit's activity (Figure 1). Disinhibition mediated by VIP interneurons has emerged as a canonical circuit motif contributing to learning and memory [2]. By preferentially targeting other interneurons, often dendrite-targeting somatostatin-positive cells, VIP interneurons can transiently relieve pyramidal neurons from inhibition and thereby facilitate context-dependent computation. In the hippocampus, VIP interneurons contribute to goal-directed spatial learning and the reorganization of CA1 population dynamics [3]. This framework has largely portrayed VIP signaling as a fast, state-dependent gate. Jabłońska et al. now add a crucial new dimension by showing that the gate itself can be reshaped by experience [1]. The key finding is that burst firing of stratum oriens interneurons induces a persistent weakening of VIP interneuron-mediated inhibition, even in the complete absence of presynaptic activity of VIP interneurons [1]. Pairing presynaptic activation with postsynaptic bursts does not modify the magnitude of depression, confirming that classical spike-timing-dependent rules do not apply. Mechanistically, such inhibitory long-term depression (LTD) requires postsynaptic calcium influx through L- and T-type voltage-gated calcium channels but is independent of endocannabinoid signaling and presynaptic release probability, pointing to a postsynaptic locus of expression [1]. In fact, hippocampal inhibitory circuits express a diversity of learning rules that are distinct from classical Hebbian excitatory plasticity. Whereas correlated pre- and postsynaptic activity drives Hebbian long-term potentiation (LTP) and LTD at excitatory synapses, inhibitory synapses can develop anti-Hebbian or nonassociative forms of plasticity. Notably, many inhibitory neurons in hippocampal feedback circuits express anti-Hebbian LTP [4], demonstrating that inhibitory circuit plasticity can be governed by mechanisms fundamentally different from Hebbian timing-based rules. Against this background, the I→ I LTD described by Jabłońska et al. represents a postsynaptic, activity-dependent learning rule that further expands the repertoire of inhibitory plasticity in hippocampal networks. An important implication of these findings is that anatomically distinct inhibitory circuit types are shaped by different plasticity rules [5]. Equally important is the demonstration that this I→ I plasticity is recruited by physiologically relevant network activity. Theta-burst stimulation of excitatory inputs to stratum oriens interneurons induces their burst firing, potentiates excitation, and simultaneously depresses VIP interneuron-mediated inhibition, shifting the excitation-inhibition balance toward stronger recruitment of stratum oriens interneurons [1]. This coordinated adjustment echoes a broader principle: synaptic plasticity often serves to reset circuits via coordinated (including heterosynaptic) changes that calibrate excitation and inhibition [6]. From a neuronal network perspective, non-Hebbian I→ I LTD provides an activity-dependent brake on disinhibition. Repeated recruitment of stratum oriens interneurons progressively weakens VIP interneuron input, enhances stratum oriens interneuron output and prevents excessive disinhibition while preserving the capacity for rapid state-dependent gating [1]. Crucially, the functional impact of inhibitory plasticity depends not only on changes in synaptic conductance but also on the GABAA receptor reversal potential (EGABA). GABAergic inputs act through a fast, spatially confined shunting conductance and a slower voltage component (polarization) that spreads along dendrites and persists beyond receptor activation. The balance between these components is determined by EGABA, dendritic location, and membrane properties, in addition to synaptic weight [7]. In this context, I→ I LTD is expected to change the relative contribution of shunting and polarization, thereby altering dendritic integration in stratum oriens interneurons. This aspect of I→ I LTD needs to be further explored. At the network level, plasticity of VIP interneuron-mediated disinhibition may influence memory allocation and ensemble formation in the hippocampal CA1 region. Local inhibitory feedback constrains rapid place-field formation and biases recruitment into emerging representations [8]. By regulating disinhibition, I→ I plasticity links interneuron firing history to mesoscale assembly dynamics [8, 9]. In summary, Jabłońska et al. (2026) identify a novel form of non-Hebbian inhibitory plasticity that selectively targets disinhibitory synapses in the hippocampus [1]. Their work demonstrates that interneuron—interneuron connections are adaptive elements governed by distinct learning rules and highlights the importance of incorporating I→ I LTD into models of hippocampal computation [10]. The authors have nothing to report. The authors declare no conflicts of interest. The authors have nothing to report.