Dendritic spikes consist of ABT 263 an initial fast, followed by a slower component (Figure 7G, upper traces), as described previously (Losonczy and Magee, 2006 and Remy et al., 2009). The initial fast component of the dendritic spike was particularly apparent as a marked increase

in the first derivative of the somatic voltage trace (Figure 7G inset, Figure 7I, δV/δt). Dendritic spikes could never be elicited by synchronous uncaging in dentate granule cells (n = 47 dendrites). Thus, CA1 dendrites are capable both of linear and supralinear integration via dendritic spikes. Dentate granule cells, in contrast, invariably exhibit linear integration, but with a variable gain. The observation that the relationship of measured versus expected gluEPSPs was linear over a wide range of input strengths was surprising, since we expected that the loss of local driving force at the dendritic stimulation site would lead to a saturation of the local EPSP size with increasing stimulation

strength (for estimation of the magnitude of this effect see Experimental Procedures). These data suggested the presence of a voltage-dependent selleck boosting mechanism that normalizes EPSPs for the loss of driving force and causes a linear gain. Because synaptically elicited perforant path EPSCs had a substantial NMDA component (NMDA/AMPA peak current ratio 1.08 ± 0.12, n = 9, Figure S3, see also Keller et al., 1991), we explored how these receptors impact processing of synchronous

input. In the presence of the NMDA receptor blocker D-APV, the ratio of measured versus expected gluEPSPs declined when the number of synchronously stimulated spines was increased (Figures 8A and 8B, n = 14 branches). Application of TTX (1 μM, n = 23 branches, isothipendyl Figures 8C and 8D) or Ni2+ (1 mM, n = 17 branches, Figures 8E and 8F) also decreased the ratio of measured versus expected gluEPSPs (see Figure 8G for summary, Dunnett test, p < 0.0001, p = 0.004, p = 0.02, respectively), but not as strongly as the application of D-APV. These data indicate that linear integration in granule cells requires NMDA receptors, and—to a lesser extent—voltage-gated Na+ and Ca2+ channels. It should be noted that the NMDA/AMPA ratio could be enhanced in uncaging experiments, because it cannot be excluded that photoliberated glutamate gains access to perisynaptic NMDA receptors. We explored this effect in the computational model. We stimulated up to 13 synapses on the dendritic tree of a model granule cell, with synapses exhibiting the experimentally determined NMDA/AMPA ratio of 1.08, while recording voltage from the dendritic stimulation site and the soma. We first stimulated an individual synaptic spine alone (•), then 12 further spines (∑○), and finally all 13 spines (∑○+ •, Figure 9A, upper traces) in a comparable manner as during uncaging experiments.