It emerges that a transcription program differentially regulates inhibitory inputs in distinct neuronal compartments — an unexpected coordinated switch for achieving experience-dependent 'plasticity' in neural circuits. See Letter p.121
Experiences trigger long-lasting changes in memory and behaviour. Although manifested at the organismal level, such changes arise from modifications of individual synaptic connections between neurons. Specifically, new sensory experiences alter a neuron's activity (its output of action potentials) by modulating the strength of the synaptic connections that it receives1. On page 121 of this issue, Bloodgood et al.2 demonstrate that a transcriptional program drives highly localized changes in synaptic connections between neurons. The authors' results thus offer a refined molecular explanation for experience-dependent learning processes.
When an animal explores an uncharted environment, neural activity originating from its sensory organs propagates into the brain, where neurons receive a barrage of concurrent excitatory and inhibitory inputs. The net activity signal that each neuron receives is determined by a delicate balance of these two competing forces. In some neurons, excitation prevails and the cells produce an action-potential spike; in others, inhibition overrides excitation and the cells remain silent.
The cellular location of an inhibitory input — that is, whether it impinges on a neuron's cell body (soma) or dendritic projections — as well as its strength govern how effectively excitatory inputs sum up. Somatic inhibitory synapses, which lie close to the spike-initiation machinery, can effectively cancel excitation3. Farther away along the dendrite, an inhibitory input can merely dampen an excitatory signal. Separate regulation of somatic and dendritic synapses allows a neuron to both 'gate' its spiking activity and control the dendritic integration of excitatory inputs. Nonetheless, little is known about the molecular mechanisms that selectively regulate these inhibitory inputs at a subcellular level.
A leading hypothesis has been that neuronal activity triggers specific gene-expression programs downstream, with the resulting gene products modulating transmission at synapses4. Because these gene products would be synthesized in the soma, they have been implicated in rather non-selective, cell-wide changes in total synapse number or strength5. The transcription factor NPAS4 is one such activity-regulated gene product, which itself initiates a gene-expression program that modulates the total number of inhibitory synapses in vitro6. But whether activity-regulated gene products can act on a spatially restricted set of synapses to provide more-refined control of neuronal function has remained unclear.
To address this question, Bloodgood and colleagues housed mice either in standard cages or in enriched environments containing running wheels and novel objects to stimulate activity in the hippocampus — a brain structure that is crucial for spatial navigation and memory. Exposure to the enriched environment resulted in a marked increase in NPAS4 levels in hippocampal neurons. Moreover, this transcription factor changed the number of inhibitory synaptic connections made with pyramidal cells, the main excitatory cell type in the hippocampus. Most notably, NPAS4 triggered an increase in somatic inhibitory synapses, with a simultaneous reduction in dendritic inhibitory synapses (Fig. 1).
NPAS4 can control the expression of many genes, one or more of which might mediate the observed differential effects on the number of somatic versus dendritic inhibitory synapses. To identify the genes responsible, Bloodgood et al. performed a comprehensive screen involving three search criteria: for genes that are directly regulated by NPAS4; those that are strongly upregulated by neuronal activity; and those that control the number of inhibitory synapses. The gene encoding brain-derived neurotrophic factor (BDNF) satisfied all three requirements. (Intriguingly, BDNF has also been shown7,8 to be regulated independently in different cellular compartments.)
The authors' further analyses confirmed that NPAS4 promotes BDNF production during bouts of increased neuronal activity. Furthermore, using mice lacking BDNF, they showed that this factor is necessary for the increase in the number of somatic inhibitory synapses. However, the reduction in dendritic-synapse number persisted in these mutants, indicating that different NPAS4 target genes regulate this aspect of NPAS4-dependent synapse remodelling.
Bloodgood and colleagues' characterization of the NPAS4-mediated control of specific sets of inhibitory synapses is a significant step forward in our understanding of how activity can exert long-lasting changes in synapse number. Nonetheless, some questions about the implications of these findings remain.
The hippocampus contains several types of inhibitory interneuron cell, and these vary in their activity, synaptic function and subcellular targeting of pyramidal cells9. Whereas Bloodgood and co-workers' experimental set-up was designed to separate somatic and dendritic inputs through spatially restricted stimulation, it would be beneficial to extend this study to investigate the specific cell types at play. Various mouse lines that allow distinct interneuron subpopulations to be accessed10 could be useful for this purpose. Knowing precisely which cell groups are involved will be instructive in further refinement of hypotheses for how activity-induced changes in inhibition affect hippocampal processing and behaviour.
Another question is how the effects of BDNF become confined to specific subcellular compartments. Neurons transcribe several different messenger RNAs of BDNF, and it could be that activity-induced BDNF mRNAs are preferentially trafficked to somatic synapses, where they are translated locally. Another possibility is that newly synthesized BDNF protein is selectively released near somatic synapses. Alternatively, specificity might arise from the axonal processes of other neurons that make contact, such that inhibitory axons targeting the soma are more responsive to BDNF than their dendrite-targeting counterparts.
More broadly, the ability of NPAS4 to coordinately and bi-directionally modulate somatic versus dendritic inhibition has implications for synaptic plasticity (the processes by which synapses grow stronger or weaker depending on their activity level), which relates to learning. To induce plasticity at excitatory synapses, the net excitatory–inhibitory input must trigger the entry of calcium ions into the dendrite11. Nearby dendritic inhibition normally dampens the efficacy of synaptic input. NPAS4 activation would relieve this inhibition, making individual synapses more likely to undergo the lasting changes associated with learning and memory. At the same time, increased somatic inhibition would tighten the time window in which combined synaptic input can produce a spike. The consequences of both actions are that fewer excitatory synapses would be needed to cause the cell to spike, but that those inputs must be more temporally coincident.
Bloodgood and colleagues' data unify several previous observations concerning the activity dependence and synapse-forming potential of NPAS4 and BDNF. Moreover, they characterize synapse-specific rules governing the effects of NPAS4 activation, and thereby highlight an extra layer of adaptability of neural circuits and a mechanism by which sensory input can produce long-lasting changes in synaptic connectivity.
Turrigiano, G. G. Cell 135, 422–435 (2008).
Bloodgood, B. L., Sharma, N., Browne, H. A., Trepman, A. Z. & Greenberg, M. E. Nature 503, 121–125 (2013).
Pouille, F. & Scanziani, M. Science 293, 1159–1163 (2001).
West, A. E., Griffith, E. C. & Greenberg, M. E. Nature Rev. Neurosci. 3, 921–931 (2002).
Turrigiano, G. Cold Spring Harb. Perspect. Biol. 4, a005736 (2012).
Lin, Y. et al. Nature 455, 1198–1204 (2008).
Dieni, S. et al. J. Cell Biol. 196, 775–788 (2012).
Matsuda, N. et al. J. Neurosci. 29, 14185–14198 (2009).
Freund, T. F. & Buzsáki, G. Hippocampus 6, 347–470 (1996).
Taniguchi, H. et al. Neuron 71, 995–1013 (2011).
Bear, M. F. & Malenka, R. C. Curr. Opin. Neurobiol. 4, 389–399 (1994).