The presence of a mixed common neighbor resulted in a significantly lower probability of an electrical connection and a significantly higher probability of a chemical connection compared to other pairs see more (χ2 test, p = 0.016 and 0.003,

respectively) and to the nonuniform random predictions (Monte Carlo, p = 0.0012 and 0.0296, respectively). This provides the first indication that the overlap between electrical and chemical connectivity is more structured than predicted by the random connectivity model. We then examined the effect of a chemical common neighbor, first disregarding the direction of the chemical connections (Figure 6C, n = 92). We observed an excess of selleck compound chemical connections in these pairs compared to the other pairs and to the random model prediction (χ2 test, p = 1.39 × 10−5 and Monte Carlo, p = 0.0020), confirming the preference for fully connected chemical triplets,

including the transitive ones seen in Figure 5A. Finally, we investigated the particular case of a common chemical neighbor in a chain configuration (Figure 6D, n = 11). This resulted in an underrepresentation of electrical connections compared to other pairs (χ2 test, p = 0.030; compared to the nonuniform random prediction p = 0.061). This result provides a second indication that the overlap between electrical and chemical networks is structured at the level of triplets of MLIs. We next devised an independent way to obtain connectivity information from cells that were not directly recorded by measuring common synaptic inputs to a pair. This allows us to examine the configuration of diverging chemical connections made onto a pair of recorded neurons. The level of synchrony of IPSCs has been used previously as a measure for the likelihood of two neurons sharing Cell press a presynaptic

partner (Sippy and Yuste, 2013 and Vincent and Marty, 1993). We recorded spontaneous inhibitory input in simultaneously recorded pairs of MLIs in voltage clamp and estimated the level of synchrony using the normalized cross-correlogram of their IPSC trains (Figure 7A; Supplemental Experimental Procedures). We found no difference in the level of synchrony between pairs of neurons sharing an electrical connection and those that did not (t test, p = 0.95, n = 36 and 50, respectively; Figure 7B). However, we found a significantly higher level of synchrony between pairs that were connected by a chemical synapse (t test, p = 0.00054, n = 18 and 68, respectively; Figure 7B). This result provides independent confirmation of the presence of transitive patterns (10, 14) in the chemical network. Although transitive connections are a signature of the chemical network (Figures 5B and 7B), it appears that the feedforward pattern (10), in particular, is a preferred motif of this network (Figures 5A and 8A; n = 13 cases).