Local axonal conduction delays underlie precise timing of a neural sequence

Abstract

SUMMARYSequential activation of neurons has been observed during various behavioral and cognitive processes and is thought to play a critical role in their generation. Here, we studied a circuit in the songbird forebrain that drives the performance of adult courtship song. In this region, known as HVC, neurons are sequentially active with millisecond precision in relation to behavior. Using large-scale network models, we found that HVC sequences could only be accurately produced if sequentially active neurons were linked with long and heterogeneous axonal conduction delays. Although such latencies are often thought to be negligible in local microcircuits, we empirically determined that HVC interconnections were surprisingly slow, generating delays up to 22 ms. An analysis of anatomical reconstructions suggests that similar processes may also occur in rat neocortex, supporting the notion that axonal conduction delays can sculpt the dynamical repertoire of a range of local circuits.

Addition of new neurons and the emergence of a local neural circuit for precise timing

“…We also show that our model creates the observed spatial distribution profile for the connections between HVC RA neurons [14]. With a wide delay distribution between these connections, as observed by experiments [29], our model produces a robust polychronous chain network with continuous and precise time representation, which is recently proposed to be the structure of the synaptic chain network in HVC [29]. Our model also predicts that HVC RA neurons in the growing chain network receive less foreward inhibition from the HVC RA neurons that drive them.…”

“…We also created axonal time delays between all neurons by setting the conduction velocity to 100 μ m/ms and using distances between neurons on the sphere. The conduction velocity was chosen such that the computed axonal delays in the model approximately match the measured axonal delays in zebra finch HVC (1 to 7.5 ms) [29].…”

“…In our model, the axonal conduction velocity controlled the axonal time delays between neurons. With the conduction velocity set to 100 μ m/ms, which creates the realistic axonal time delays observed in HVC [29], the emerged network showed continuous dynamics and nearly uniform temporal distribution of burst onset times (Fig. 4a).…”

“…4d). These are the characteristics of polychronous chain network proposed as the connectivity of HVC RA neurons within HVC in a recent study [29].…”

“…We investigate this hypothesis through a computational model that builds on the previous models of network growth in HVC [25, 26]. Unlike these earlier models, our model incorporates more biologically realistic features, including explicit incorporation of HVC INT neurons rather than simplifying inhibitory actions as idealized global inhibition between HVC RA neurons; implementation of axonal delays between HVC RA neurons, which has shown to be substantial and is important for determining the connectivity structure of the synaptic chain network [29]; and spatial structure of HVC RA connectivity, which has been recently measured in zebra finch [14]. Most importantly, the maturation dynamics of HVC RA neurons is modeled.…”