Hyperexcitable Neurons Enable Precise and Persistent Information Encoding in the Superficial Retrosplenial Cortex

Abstract

Highlights d Two distinct subtypes of excitatory neurons in superficial retrosplenial cortex (RSC) d Most common neuron in layer 2/3 of RSC is excitatory lowrheobase (LR) neuron d LR intrinsic properties enable precise, sustained encoding of information d Layer 2/3 of RSC is dominated by feedforward, not feedback, inhibition

Thalamus and claustrum control parallel layer 1 circuits in retrosplenial cortex

“…LR neurons are morphologically distinct compared to their neighboring RS neurons (Brennan et al, 2020; Kurotani et al, 2013; Yousuf et al, 2020). Given the precise laminar pattern of anterior thalamic inputs to RSG (Fig.…”

“…Given the somewhat surprising depressing nature of anterior thalamic synapses onto LR neurons (compared to the RSG and ACC, we next sought to understand its computational implication with regard to the processing of incoming head direction inputs, particularly as it contrasts with the previously-reported facilitating input from the subiculum (Yamawaki et al, 2019a), another known source of spatial input to LR cells (Figure S3; Nitzan et al, 2020; Yamawaki et al, 2019a). Using a morphologically realistic model of retrosplenial LR cells (Brennan et al, 2020), we investigated how the short-term synaptic dynamics impact the processing of depressing versus facilitating inputs by LR neurons. Utilizing the short-term-plasticity framework of Tsodyks et al (Tsodyks et al, 1998; Tsodyks and Wu, 2013), we input Poisson spike trains at a range of input frequencies to the model LR cells (Figure 4A).…”

“…We normalized the net charge across inputs, such that the inputs only differed in their short-term-plasticity (see Methods & Figure S4A&B). Since the LR neuron has a late-spiking current attributed to Kv1 expression (Brennan et al, 2020), we provided the synaptic input either early or late to simultaneously study the impact of the late-spiking current on depressing versus facilitating synaptic integration (Figure 4B). We found that over a broad range of input frequencies, depressing synapses reliably led to faster responses (reduced response latency) in comparison to facilitating synapses (Figure 4C, E).…”

“…These results suggest that the presence of distinct pyramidal cell populations within RSG facilitates simultaneous processing of converging multimodal inputs by parallel circuits. Specifically, LR neurons are intrinsically (Brennan et al, 2020) and morphologically optimally suited to receive and integrate head direction inputs from the anterior thalamus, while RS cells are better positioned to receive higher-order inputs from ClaC and corticocortical projections.…”

“…The retrosplenial cortex (RSC) is a densely connected cortical region essential for spatial learning, memory, and navigation (Alexander et al, 2020; Alexander and Nitz, 2017; Chrastil, 2018; Hinman et al, 2018; Spiers and Barry, 2015; Todd and Bucci, 2015; Valenstein et al, 1987; Vann et al, 2009), as well as contextual fear conditioning (Corcoran et al, 2011; Keene and Bucci, 2008a, 2008b; Kwapis et al, 2015; Yamawaki et al, 2019b, 2019a). The superficial layers of the granular RSC (RSG) contain two subtypes of principal excitatory pyramidal neurons that have strikingly distinct physiology, morphology, and computational capabilities: the low-rheobase (LR) neuron and the regular-spiking (RS) neuron (Brennan et al, 2020; Kurotani et al, 2013; Yousuf et al, 2020). The RSG receives inputs from many regions, including the anterior thalamus (Ichinohe et al, 2008; O‘Mara, 2013; Shibata, 1993; van Groen and Wyss, 1995; Yamawaki et al, 2019b), subiculum (Yamawaki et al, 2019a, 2019b, Opalka et al, 2020; Nitzan et al, 2020), hippocampus (Alexander et al, 2018; Wyss and van Groen, 1992; Yamawaki et al, 2019b), claustrum (van Groen and Wyss, 2003, 1990; Wang et al, 2017), and cingulate cortex (van Groen and Wyss, 1990, 2003), among others.…”