Simulations reveal how M-currents and memory-based inputs from CA3 enable single neuron mismatch detection for EC3 inputs to the CA1 subfield of hippocampus.

Significant evidence has accumulated to support the hypothesis that hippocampal region CA1 operates as an associative mismatch detector (e.g., Hasselmo ME, Schnell E, Barkai E. 15: 5249-5262, 1995; Duncan K, Curtis C, Davachi L. 29: 131-139, 2009; Kumaran D, Maguire EA. 27: 8517-8524, 2007; Lisman JE, Grace AA. 46: 703-713, 2005; Lisman JE, Otmakhova NA. 11: 551-568 2001; Lörincz A, Buzsáki G. 911: 83-111, 2000; Meeter M, Murre JMJ, Talamini LM. 14: 722-741, 2004; Schiffer AM, Ahlheim C, Wurm MF, Schubotz RI. 7: e36445, 2012; Vinogradova OS. 11: 578-598 2001). CA1 compares predictive synaptic signals from CA3 with synaptic signals from EC3, which reflect actual sensory inputs. The new CA1 pyramidal model presented here shows that the distal-proximal segregation of synaptic inputs from EC3 versus CA3, along with other biophysical features, enable such pyramids to serve as comparators that switch output encoding from a brief burst, for a match, to prolonged tonic spiking, for a mismatch. By including often-overlooked features of CA1 pyramidal neurons, this new model allows simulation of pharmacological effects that can eliminate either the match (phasic mode) response or the mismatch (tonic mode) response. These simulations reveal that dysfunctions can arise from either too much or too little ACh stimulation of the muscarinic receptors that control KCNQ channels. Additionally, a dysfunction caused by administration of an -methyl-d-aspartate antagonist could be rescued by simultaneous administration of a KCNQ channel agonist, such as retigabine. Hippocampal region CA1 operates as an associative mismatch detector, comparing predictive signals from CA3 with signals from EC3 reflecting sensory inputs. This new CA1 pyramidal model shows that biophysical features enable these comparators to switch output between brief bursts for matches and tonic spiking for mismatches. This suggests that cognitive learning models (e.g., predictive coding) may require much less match/mismatch circuitry than commonly assumed. Additional simulations illuminate deficits seen in psychiatric disorders and drug-induced states.