Wiring, computation and plasticity of cortical neurons
How does the brain process and store information? Although the brain functions in many ways different from a computer, the picture of an electrical circuit provides a useful analogy to describe the elements of neural circuits and their function. In the brain, neurons are the basic functional units. To form the neural circuit they are wired up by synaptic connections for signal transmission between their axonal and dendritic processes. Individual segments of the dendritic processes are partially independent integrative units. Therefore, the computation performed by a neuron on particular inputs is determined by addressing specific dendritic segments in relation to other inputs. The particular function of the circuit arises from the specific layout of connections, the way the neurons integrate their inputs and compute their output, and the strength and dynamic properties of the synapses.
The synapses between excitatory neurons, which represent 80% of cortical neurons and form 85% of cortical synapses, are located on small protrusions of the target cell’s dendrite, called dendritic spines (see Figures 1, 2). The stronger a synapse the larger is the spine and vice versa. Spines contain molecular signaling complexes, which control synaptic strength in response to synaptic activity. Furthermore, spines serve as substrate for synapse formation by bridging the distance between axons and dendrites of synaptic partners. Spine synapses appear to assume different states with respect to plasticity. In vivo imaging shows a large fraction of stable persistent spines and a small fraction of spines, which turn over within days. Similarly, in vitro studies on single spines identified populations, which were either highly responsive or resistant to plasticity inducing stimuli.
Learning and memory storage in a neuronal circuit is achieved in two ways. First, particular synapses are strengthened or weakened, i.e. in terms of neuronal network theory their weights are modified. Second, formation of new and elimination of old synapses rewires and thereby reprograms the circuit.
Our studies address the following open questions regarding the synaptic basis of neuronal circuit function and plasticity:
- What are the determinants of spine synapse stability and the processes underlying plasticity?
- Do spine synapses connecting particular types of neurons display particular properties and differences in their stability?
- Which dendritic segments are addressed by inputs from particular types of neurons?
- How do inputs from different types of neurons combine on the dendritic segments of the target cell?
To study the properties of single spine synapses and the processes underlying their plasticity, we use a combination of morphological and functional 2-photon imaging, 2-photon glutamate uncaging and electrophysiology as well as correlative 2-photon and electron microscopy. In particular we focus on spine synapses in acute or cultured slices from rat or mouse hippocampus, a brain area with a unique function in learning and memory. We probe the functional properties of spines by selectively stimulating single visualized spines with local 2-photon glutamate uncaging (see Figure 1). Furthermore, we used 2-photon glutamate uncaging to induce and study plasticity at single identified spine synapses (see Figure 2).
To explore neuronal circuit function and plasticity in terms of the properties and locations of spine synapses connecting particular types of neurons, we implemented a system combining optogenetic stimulation of axons and calcium imaging in dendritic spines. Because it is particularly suited for circuit analysis, we focus on mouse primary visual cortex, which has a retinotopic organization and performs complex processing of visual stimuli, such that neurons display e.g. orientation and direction selectivity. We express labeled channelrhodopsin, a light-gated cation channel, in a set of presynaptic cells to render their axons visible and sensitive to photostimulation. By monitoring calcium signals in spines of a target neuron in response to axon stimulation, we can identify and localize spine synapses belonging to a particular projection (see Figure 3).
We showed that unconventional dendritic spines on spiny GABAergic interneurons share key properties of pyramidal cell spines suggesting that spines serve as universal synaptic connection elements rather than as defining feature of pyramidal neurons (Scheuss and Bonhoeffer, Cereb Cortex, 2013).
Recently, we demonstrated that structural synaptic potentiation is only stabilized in the case of balanced pre- and postsynaptic enlargement, suggesting that structural correlations might serve synaptic stability despite continuous protein turnover (see Figure 2; Meyer, Bonhoeffer and Scheuss, Neuron, 2014).