There has been rapid progress in our understanding of the generation of postsynaptic currents at excitatory synapses in the central nervous system. However, many important properties of synaptic transmission still remain unclear, such as: how do the kinetics and spatial location of transmitter release, the gating properties of receptors and location and properties of reuptake sites affect the quantal amplitude and variability. We have recently constructed a computational model that is highly constrained by recent data to study these questions. The model used Monte Carlo techniques to track molecular diffusion and chemical reactions in 3-d subcellular space and results in several surprising conclusions about the properties of central synapses. The model showed that the large variability in synaptic responses (at a single synapse) arose due to a combination of channel noise and number of vesicles released. We are currently interested in extending the model in a number of directions in order to accommodate additional physiological processes such as: 1) bimolecular reactions between diffusing species to model protein-protein interactions as well as interactions of second-messengers with proteins, 2) surface properties such as electric fields 3) mechanical properties of membranes to model intracellular trafficking of proteins.
It is now well documented that certain patterns of the activity of neurons can lead to long-lasting changes in the strength of synapses. Furthermore, there are strong reasons to believe that these synaptic changes contribute to the brain modifications that underlie learning and memory. Thus, in addition to being a communication channel, the synapse is also a biochemical computer, whose input is a change in calcium concentration and the output is a biochemical and/or structural change that leads to alterations in synaptic strength. We are now exploring how do brief changes in calcium concentration (milliseconds) trigger and regulate a multitude of signaling pathways that lead to plasticity (changes across multiple timescales, such as minutes, hours or days). One key point is that synapses are small chemical compartments (synapse volumes are typically measured in femtoliters). Thus the number of participating molecules is small (tens), implying that the reactions are dominated by noise and spatial separation between proteins plays a major role. We use mathematical techniques borrowed from dynamical systems theory, chemical kinetics and statistical mechanics to model synaptic plasticity. Such extended models when combined with the expanding experimental database of biochemical and biophysical pathways will allow us to provide testable quantitative predictions for many questions in synaptic physiology.
Figure 1: Electron micrograph of a hippocampal excitatory synapse (Kristen Harris, Georgia Medical College)
Figure 2: Close-up view of the synapse, showing the post-synaptic density (an electron-dense region that is highly enriched in receptors and other proteins). Ionotropic glutamate receptors (AMPA, red and NMDA, blue) are also shown along with CaMKII (pink), a protein kinase that is integral to synaptic plasticity (image by Mary Kennedy, Caltech).
Figure 3: Monte-Carlo simulation of the spontaneous release of glutamate leading to the generation of an excitatory post-synaptic response (mEPSC). The presynaptic vesicle opens releasing glutamate (pink spheres) that diffuse across a 20 nm synaptic cleft and activate glutamate receptors of the AMPA type. The receptors change color when activated, with different colors corresponding to different conductance levels. The total time is 150 microseconds. (Click
here to view).
Figure 4: a) Histogram of simulated synaptic currents showing the variability due to release of single vesicles (spread in each of the peaks in the histogram) and due to the release of multiple vesicles.b) Probability of receptor activation in response to the release of 2 vesicles 100 nm apart. The two vesicles activate pools of receptors in a non-overlapping manner.
