State-dependent activity dynamics of hypothalamic stress effector neurons

Version of Record

Accepted for publication after peer review and revision.

Download
Cite
Share
CommentOpen annotations (there are currently 0 annotations on this page).

Version of Record published: July 13, 2022 (This version)
Accepted Manuscript published: June 30, 2022 (Go to version)
Accepted: June 17, 2022
Preprint posted: February 10, 2022 (Go to version)
Received: January 6, 2022

1. Of interest
Dynamic organization of cerebellar climbing fiber response and synchrony in multiple functional components reduces dimensions for reinforcement learning

Huu Hoang, Shinichiro Tsutsumi ... Keisuke Toyama

Research Article Sep 27, 2023
Further reading

Abstract
Editor's evaluation
Introduction
Results
Discussion
Materials and methods
Data availability
References
Decision letter
Author response
Article and author information
Metrics

Abstract

The stress response necessitates an immediate boost in vital physiological functions from their homeostatic operation to an elevated emergency response. However, the neural mechanisms underlying this state-dependent change remain largely unknown. Using a combination of in vivo and ex vivo electrophysiology with computational modeling, we report that corticotropin releasing hormone (CRH) neurons in the paraventricular nucleus of the hypothalamus (PVN), the effector neurons of hormonal stress response, rapidly transition between distinct activity states through recurrent inhibition. Specifically, in vivo optrode recording shows that under non-stress conditions, CRH_PVN neurons often fire with rhythmic brief bursts (RB), which, somewhat counterintuitively, constrains firing rate due to long (~2 s) interburst intervals. Stressful stimuli rapidly switch RB to continuous single spiking (SS), permitting a large increase in firing rate. A spiking network model shows that recurrent inhibition can control this activity-state switch, and more broadly the gain of spiking responses to excitatory inputs. In biological CRH_PVN neurons ex vivo, the injection of whole-cell currents derived from our computational model recreates the in vivo-like switch between RB and SS, providing direct evidence that physiologically relevant network inputs enable state-dependent computation in single neurons. Together, we present a novel mechanism for state-dependent activity dynamics in CRH_PVN neurons.

Editor's evaluation

This is the first electrophysiological study showing high-quality spike activity in vivo from identified CRH neurons in the hypothalamic paraventricular nucleus. The authors make the surprising observation that CRH neurons exhibit brief high-frequency rhythmic bursts in unstressed mice that are converted to a sustained, single spike firing mode by an acute stressor. Their findings are supported by a computational model that suggests that feedback inhibition may regulate the activity patterns of CRH neurons in distinct states. The work provides a framework for new studies exploring firing characteristics in discrete physiological and emotional states in the hypothalamus and perhaps other regions.

https://doi.org/10.7554/eLife.76832.sa0

Introduction

Mounting an optimal stress response is critical for survival. This requires neural mechanisms that effectively switches vital physiological functions between homeostatic operation and emergency response. For example, the activation of the hypothalamic-pituitary-adrenal (HPA) axis elevates the systemic levels of glucocorticoids, a hallmark of stress. Under non-stress conditions, glucocorticoids are also produced, albeit at lower levels, for their widespread physiological actions (de Kloet et al., 2005; Lightman, 2008). While HPA axis (patho)physiology has been extensively studied with relevance to stress-related disorders (Gold and Chrousos, 1999; McEwen, 2005), we know surprisingly little about the basic neural mechanisms that powerfully and bidirectionally transition HPA axis activity from baseline to levels of elevated stress.

The brain ultimately drives the HPA axis by exciting a group of neuroendocrine neurons that synthesize corticotropin-releasing hormone (CRH). The cell bodies of CRH neurons reside in the paraventricular nucleus of the hypothalamus (PVN) and project their axons to the median eminence. At the median eminence, axonally released CRH enters the hypophyseal portal circulation and initiates the downstream hormonal cascades (Swanson and Sawchenko, 1980). Numerous studies have shown that various acute stressors induce de novo expression of immediate early genes in CRH_PVN neurons, establishing their stress-induced activity increase on the time scale of minutes to hours (Cole and Sawchenko, 2002; Itoi et al., 2004; Matovic et al., 2020; Wamsteeker Cusulin et al., 2013). Recent in vivo fiber photometry imaging revealed highly dynamic activity changes where population Ca²⁺ signals in CRH_PVN neurons (a proxy for population activity) increase rapidly (within seconds) and reversibly in response to repetitive, short-aversive stimuli (Daviu et al., 2020; Kim et al., 2019b; Kim et al., 2019a; Yuan et al., 2019).

How do individual CRH_PVN neurons fire action potentials during the rapid transitions between baseline and stress conditions? Although the in vivo firing activity of single units in the PVN was first studied more than 30 years ago (Hamamura et al., 1986; Kannan et al., 1987; Saphier and Feldman, 1985; Saphier and Feldman, 1988; Saphier and Feldman, 1991), these pioneering experiments lacked neurochemical identification; i.e., it could not be confirmed whether the spiking activity was of CRH neurons or another sub-population of neurons within the PVN. This is because in vivo extracellular electrophysiology is inherently blinded to neuronal identities, and CRH neurons are intermingled with a mosaic of neuron-types deep in the hypothalamus (Simmons and Swanson, 2009). Thus, despite decades of research on CRH_PVN neurons, their firing activities in vivo have remained inaccessible. This is an important gap in our knowledge because single-neuron firing patterns result from dynamic interaction between intrinsic properties and network dynamics (Destexhe et al., 2001; Herz et al., 2006; Leng and MacGregor, 2018), and hence will provide essential clues to probe how CRH_PVN neurons—the gatekeepers of stress outputs—process the converging brain commands during baseline conditions and in response to stress.

Here we have implemented in vivo single-unit extracellular recording of optogenetically identified CRH_PVN neurons in mice, and subsequently developed a network model that explains in vivo state-dependent firing dynamics changes. For the first time, we report that CRH_PVN neurons fire in two distinct modes: rhythmic brief bursts (RB) and single spikes (SS) of random spike intervals. Somewhat counterintuitively, RB occurred under the baseline conditions and constrained the overall firing rate at low levels due to long (~2 s) interburst silence. On the other hand, sustained SS underlies stress-induced increase in firing. Using tight combinations of experimental and computational approaches, we show that a recurrent inhibitory network generates RB and constrains the overall firing rate due to negative feedback inhibition. A spiking network model showed that release from recurrent inhibition upon stress stimuli permits continuous SS that increases the overall firing rate, thereby revealing a novel mechanism that enables efficient bidirectional controls of stress output neurons. To validate the spiking network model, we used ex vivo patch-clamp electrophysiology and injected excitatory and inhibitory inputs received by a cell in the computational model into actual biological neurons. CRH_PVN neurons in these experiments exhibited in vivo-like bursting behavior when driven by the network current, but not when driven by simple step-current stimuli. Ultimately, our collective results identified a novel network mechanism that underlies state-dependent firing activity dynamics at the effector neurons of hormonal stress response.

Results

Identifying CRH neurons in vivo

To examine firing patterns of individual CRH_PVN neurons in vivo, we used optrode recording (Lima et al., 2009). Briefly, we first ‘tagged’ CRH neurons by expressing an excitatory opsin channelrhodopsin2 (ChR2) using adeno-associated virus (AAV) carrying a cre-dependent promoter construct injected into the PVN of CRH-Ires-Cre mice crossed with Ai14 td-tomato reporter line (Wamsteeker Cusulin et al., 2013). This resulted in ChR2-EYFP expression in CRH_PVN neurons that express td-tomato (Figure 1A), similar to recent studies using the same approach (Bittar et al., 2019; Füzesi et al., 2016; Kim et al., 2019a; Wamsteeker Cusulin et al., 2013; Yuan et al., 2019). Important for the desired PVN neuron identification, these ChR2-expressing neurons could now be detected on the basis of their light-induced spike firing. We isolated 36 single units in the PVN area from 10 adult male mice under urethane anaesthesia. Among these preisolated single units, 18 neurons (from 10 mice) were ‘light-responsive’; i.e., they fired action potentials in response to pulses of blue light (5 ms, λ=465 nm) with short latency (7.2 ± 2.6 ms, n=18; Figure 1B and C). This responsiveness to light had a binary effect (Figure 1D): the light-responsive neurons reliably increased their probability of firing (7.1 ± 6.3% preonset to 66.3 ± 21.3% postonset, n=18), whereas the light-non-responsive neurons did not (3.3 ± 3.1% preonset to 2.0 ± 2.8% postonset, n=18). Furthermore, in response to longer light pulses (50 ms), the light-responsive neurons fired a train of action potential with frequency adaptation (Figure 1E and F). Consequently, these light-responsive neurons were defined as CRH_PVN neurons.

Figure 1

Download asset Open asset

Optogenetic identification of CRH_PVN neuron single-unit.

(A) Electrode tract (purple), TdTomato-expressing CRH_PVN neurons (red) and ChR2-EYFP (green) expression. (B) An example for an isolated single-unit (blue) that also responded to light (cyan). (C) Raster plot (top) and peristimulus time histogram (PSTH, bottom) in response to 5 ms blue light from a representative unit. (D) A summary graph for the probability of firing before (−20–0 ms) and after blue light illumination (0–20 ms). Light-responsive units in black (n=18) and non-responsive units in gray (n=18). (E) Raster plot (top) and PSTH (bottom) for a representative single-unit responding to 50 ms blue light. (F) Summary of firing rate time course following 5 ms (tick) and 50 ms (open circle) blue light. n=18. SD is represented as error bars.

Figure 1—source data 1 Optogenetic identification of CRHPVN neuron single-unit.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig1-data1-v2.xlsx
Download elife-76832-fig1-data1-v2.xlsx

CRH_PVN neurons can fire in two distinct modes

We found that, under the baseline (non-stress) conditions, many CRH_PVN neurons occasionally fired in a distinct brief train of high-frequency (>100 Hz) bursts, in addition to spikes with variable interspike intervals (ISI, Figure 2A). Figure 2B shows an ISI histogram of a representative neuron that demonstrated a bimodal distribution with a sharp peak around 2–10 ms (burst) and another wider peak around 100 ms (SS). Burst trains typically showed the shortest ISI at the start and followed by a few spikes with slightly increased ISIs. Thus, to quantify these brief burst firing, we set a criterion for burst detection with ISI ≤ 6 ms for the start of a burst, and subsequent spikes were considered part of the burst train as long as ISI remained below 20 ms. Using this burst-detection criterion, we found that the majority of CRH_PVN neurons fired a brief burst at least once during the baseline (16 out of 18), but the burst rate was variable among CRH_PVN neurons (Figure 2C). Figure 2—figure supplement 1 shows ISI histograms for all units. For the analysis of burst properties below, we used neurons that showed at least 60 bursts during 10 min baseline recording (>0.1 bursts/s, purple circles in Figure 2C, n=10). Among these bursting CRH_PVN neurons, each individual burst episode was brief, on average 3.1 ± 0.5 spikes, ranging between 2 and 6 spikes (Figure 2D). These brief bursts occurred at slow rhythms (rhythmic brief burst, RB), intervened by long, mostly silent, interburst intervals (IBI, 1.8 ± 0.6 s, n=10, Figure 2E).

Figure 2 with 1 supplement see all

Download asset Open asset

CRH_PVN neurons fire in rhythmic bursts and single spikes.

(A) Two distinct spike patterns in a representative single-unit (top: single spikes; bottom: bursts indicated by arrow). (B) Interspike interval (ISI) distribution for the representative single-unit shown in A. (C) A summary plot for burst rates among CRH_PVN neurons during non-stress baseline recordings. The criterion for burst firing neurons (purple) was set as 0.1 burst/s. A horizontal bar indicates the average. (D) A summary graph for burst length distribution (n=10). (E) Summary interburst interval (IBI) distribution (n=10). Standard deviation is represented in the graphs as the error bars (D) and dotted line (E).

Figure 2—source data 1 CRHPVN neurons fire in rhythmic bursts and single spikes.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig2-data1-v2.xlsx
Download elife-76832-fig2-data1-v2.xlsx

CRH_PVN neurons are constrained to low activity during rhythmic bursting

In many neurons, burst firing is postulated to carry behaviorally relevant information, representing specific network states different from that represented by SS of variable frequencies (Krahe and Gabbiani, 2004; Steriade et al., 1993). The classic role of CRH_PVN neurons is to respond to stress stimuli with an increase in firing activity, driving the neuroendocrine release of CRH (Ulrich-Lai and Herman, 2009). Thus, we examined potential roles of RB in CRH_PVN neurons’ response to stress. To this end, we used electric stimulation of the sciatic nerve to model noxious sensory stimuli, as this approach has been shown to effectively elicit time-locked firing responses in median eminence-projecting parvocellular PVN neurons (Day et al., 1985; Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1985; Saphier and Feldman, 1991) as well as to increase the circulating ACTH levels, indicative of HPA axis activation (Feldman et al., 1981), in anesthetized rats. For this experiment, we recorded 13 CRH_PVN neurons (among 18 neurons recorded for the baseline firing analysis described above) that remained stable until the end of nerve stimulation session (see Experimental Design in Materials & Methods). Consistent with the earlier ‘blinded’ recordings (Day et al., 1985; Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1985; Saphier and Feldman, 1991), sciatic nerve stimulation (1.6 mA) increased the firing rate (time averaged spike number) of CRH_PVN neurons (BL 3.063 ± 1.646 Hz vs NS 4.900 ± 1.962 Hz, paired t-test, p=0.0108, n=13; Figure 3A, B and G). Somewhat counter-intuitively, however, this activity increase was paralleled by a striking loss of RB (BL 0.1406 ± 0.1375 Hz vs NS 0.0474 ± 0.0410 Hz, paired t-test, p=0.0167, n=13, Figure 3C, E and H), and the firing rate increased due to SS (BL 2.702 ± 1.428 Hz vs NS 4.782 ± 1.987 Hz, paired t-test, p=0.0025, n=13; Figure 3D, F). In a representative case, we found that the firing pattern change lasted longer with higher intensity stimulation. Figure 3—figure supplement 1 shows that, with the highest intensity stimulation (2 mA), the firing activity change did not completely return to the baseline within the trial duration of 15 s, resulting in continuous decrease of RB and increase of SS during the stimulation period. This finding suggests that RB paradoxically informs a certain ‘low activity state’ of CRH_PVN neurons, whereas the ‘high activity state’, induced by stress, is due to SS at high rates.

Figure 3 with 1 supplement see all

Download asset Open asset

CRH_PVN neurons are constrained to low activity during rhythmic bursting.

(A) Peristimulus time histogram (PSTH) for a representative single-unit responding to sciatic nerve stimulation (1.5 mA, 0.5 ms × 5 pulses at 20 Hz, red line). (**B–D**) Raster plots for the unit shown in A for all spikes (B), bursts (C) and single spikes (SS) (D). (**E, F**) Summary time course for bursts (E) and SS (F). n=13. (**G–I**) Summary changes in all spikes (G, paired t-test, n=13, p=0.0108), bursts (H, paired t-test, n=13, p=0.0167) and SS (I, paired t-test, n=13, p=0.0025) before (baseline, BL) and after sciatic nerve stimulation (NS). (**J, K**) Time course of firing rate (J) and interspike interval (ISI) (K) for a representative single-unit during baseline recording. Gray lines are the running average (30 s) of firing rate. (L) The relationship between burst rate and total firing rate for the representative single-unit shown in J and K. For each time bin (10 s), total spike rate was plotted against burst rate. (M) Pooled data for all units (n=18).

Figure 3—source data 1 CRHPVN neurons are constrained to low activity during rhythmic bursting.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig3-data1-v2.xlsx
Download elife-76832-fig3-data1-v2.xlsx

Under baseline conditions, CRH_PVN neurons showed spontaneous low-level firing activities; an observation in line with recent in vivo two-photon Ca²⁺ imaging of CRH neurons in zebra fish larvae (Vom Berg-Maurer et al., 2016). We also found that some of CRH_PVN neurons show time-dependent fluctuations in their firing rate with spontaneous and transient increase (Figure 3J). Thus, we next asked whether spontaneous ‘high activity state’ is also due to an increase in SS paralleled by a loss of RB. Figure 3J plots the running average firing rate (10 s bins) of one representative neuron showing spontaneous emergence of high activity states. We next overlaid the time course of individual spike’s ISI to visualize the temporal relationship between the overall activity states (i.e. firing rate) and specific firing patterns (i.e. RB vs SS). Similar to the stress-induced activity increase, a loss of RB temporally correlated with the emergence of the high activity state (Figure 3K). On the other hand, when CRH_PVN neurons fired mainly with RB, their overall firing rate remained at relatively low levels. Intuitively, this is because of the brevity of burst episodes with mostly silent IBIs (Figure 2A and E), precluding the high rate of SS. Figure 3L and M plot the relationship between burst and total spike rate for every 10 s time bin for this unit, as well as for all units pooled (n=18), respectively. This analysis revealed that high levels of total spike rate only emerged when burst rate was at or near zero. Thus, our data suggest that a loss of RB permits CRH_PVN neurons to increase firing rate with high rate of SS. In other words, the network activity state that drives RB constrains the activity of CRH_PVN neurons at low levels.

Prolonged silent periods precede burst firing

Given that the RB firing reflects a specific, low-activity state of CRH_PVN neurons, a question emerged: How do CRH_PVN neurons fire in these distinct, short burst trains? To address this, we examined the temporal properties of the burst spike trains by adopting the analysis used by Harris et al., 2001. Figure 4A illustrates the relationship between ISIs preceding and proceeding individual spikes of the same representative unit shown in Figure 2A and B. A cluster of spikes in the lower-left correspond with the spikes in the middle of bursts (i.e. both preceding and proceeding ISI are short). On the other hand, the lower-right cluster represents the initial spike of bursts (long preceding ISI and short proceeding ISI). Notably, the lack of spikes between these two clusters along the X-axis indicate that bursts preferentially start after a long silence period (i.e. preceding ISI >500 ms). By contrast, spikes with short preceding ISI on the left spread along the Y-axis, indicating that the end of bursts are followed by proceeding ISI of variable durations (i.e. not followed by an abrupt start of postburst silence). To express these features more explicitly, we plotted the probability of burst-range firing (i.e. proceeding ISI <6 ms) against the preceding ISI for individual units, and then averaged for all bursting units (Figure 4B). The graph shows that a burst train seldom initiates without a preceding silent period (i.e. ISI) of at least 200 ms, and that the likelihood of burst firing substantially increases when preceded by a silence period longer than 500 ms. In Figure 4C, which plots the probability of burst-range firing (i.e. preceding ISI <20ms) against proceeding ISI, there was a higher probability of bursting with longer proceeding silences, but the relationship was less prominent than the ‘preceding ISI analysis’: this likely reflects that burst trains end with gradual prolongation of ISI (i.e. frequency adaptation). We also found that burst firing, regardless of its length, was accompanied by longer preceding silences than SS (0.33 ± 0.16 s, 0.80 ± 0.31 s, 0.89 ± 0.32 s, 1.04 ± 0.43 s, for burst lengths of single, 2, 3, and 4+, respectively; One-way ANOVA, p<0.0001; Tukey’s multiple comparisons test, p<0.0001, p=0.0001, p=0.0004 for single vs 2, 3, 4+, respectively. Figure 4D). On the other hand, we did not observe these predictive trends in the ‘silence’ following SS or bursts. Furthermore, the bursts tended to have longer proceeding silences (0.59 ± 0.36 s, 0.57 ± 0.33 s, 0.51 ± 0.40 s, for burst lengths of 2, 3, and 4+, respectively) than SS (0.37 ± 0.18 s; One-way ANOVA, p=0.0131; Tukey’s multiple comparisons test, p=0.0573, P=0.0437, p=0.2885 for single vs 2, 3, 4+, respectively. Figure 4E), which resulted in significant difference only between SS and bursts with three spikes. Again, these data reflect the fact that bursts end with gradual prolongation of ISI and are not followed by an abrupt start of a silent period. These characteristics of burst generation suggest that a prolonged silent period is followed by a brief high-frequency spike trains, which in turn are followed by a prolonged silent period, leading to RB.

Figure 4

Download asset Open asset

Prolonged silent periods precede burst firing.

(A) Preceding and proceeding interspike interval (ISI) plotted for individual spikes recorded from a representative single-unit. Blue rectangle indicates spikes with proceeding ISI <6 ms. Red rectangle indicates spikes with preceding ISI <20 ms. Left. An example of a burst episode. (B) Summary of probability of burst firing relative to preceding silence (ISI). (C) Summary of probability of burst firing relative to proceeding silence (ISI). (D) Mean preceding silence as a function of event length. One-way ANOVA, p<0.0001; Tukey’s multiple comparisons test, single vs 2 (p<0.0001), 3 (p=0.0001), ≥ 4 (p=0.0004) (E) Mean proceeding silence as a function of event length. One-way ANOVA, p=0.0131; Tukey’s multiple comparisons test, single vs 2 (p=0.0573), 3 (p=0.0437), ≥ 4 (p=0.2885) SD is represented in the graphs as dotted lines (**B, C**) and the error bars (**D, E**).

Figure 4—source data 1 Prolonged silent periods precede burst firin.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig4-data1-v2.xlsx
Download elife-76832-fig4-data1-v2.xlsx

Recurrent inhibitory circuits underlie burst firing

What are the mechanisms generating RB in CRH_PVN neurons? Burst firing is known to rely on the intricate interactions between intrinsic neuronal properties, synaptic inputs and network characteristics (Krahe and Gabbiani, 2004). Indeed, when studied ex vivo in slice recordings, in which network inputs are substantially lost, CRH_PVN neurons do not show overt bursting properties (Bittar et al., 2019; Jiang et al., 2019; Khan et al., 2011; Luther et al., 2002; Matovic et al., 2020; Sarkar et al., 2011; Wamsteeker Cusulin et al., 2013). In the present study, we confirmed this lack of ex vivo burst firing phenotype using whole-cell patch-clamp recordings in acute slices. As shown in Figure 5A (red traces), in response to depolarizing (monotonic) current steps, CRH_PVN neurons fired with a gradual increase in frequency with a moderate adaptation and they did not generate in vivo-like high-frequency bursts. When considering more complex network inputs that dynamically interact with the intrinsic properties of CRH_PVN neurons and enable RB firing, our in vivo burst firing pattern predicts the following: (1) the intrinsic properties of CRH_PVN neurons favor high-frequency burst firing after a prolonged silence period, and (2) network inputs that cause the prolonged silence contribute to the slow ‘rhythms’' of bursting. Taken together, these features are indicative of an inhibitory-driven recurrent network burst mechanism (McCormick and Feeser, 1990; Sherman, 2001; Steriade et al., 1993). Supporting this idea, early in vivo recordings of PVN neurons predicted the existence of recurrent inhibitory circuits based on inhibition following antidromically elicited action potentials (Saphier and Feldman, 1985). More recent studies identified a group of local GABAergic interneurons that form recurrent inhibitory circuits (Jiang et al., 2018; Jiang et al., 2019; Ramot et al., 2017). Thus, we hypothesized that recurrent inhibitory circuits underlie the RB of CRH_PVN neurons. To test this, we computationally studied the dynamics of single CRH_PVN neurons within a network by using a spiking network of adaptive-exponential (AdEx) neuron models (Brette and Gerstner, 2005; Gerstner and Naud, 2009; Izhikevich, 2003; Markram et al., 2015). First, the single neuron AdEx models (Brette and Gerstner, 2005) were fit to slice patch-clamp recordings of CRH_PVN neurons (Figure 5A and Table 1). We evaluated the performance of the model by quantifying the mean squared difference in the subthreshold traces (Figure 5A) and the frequency-current (F-I) curve (Figure 5B) between the model and experimental data (Brette and Gerstner, 2005). Our simulations show that the single neuron models can capture a rapid spike adaptation that follows the first few spikes, timing of repetitive spike firing and the F-I relationship for a series of square pulse depolarizing current injections (Figure 5A and B).

Figure 5 with 1 supplement see all

Download asset Open asset

Recurrent inhibitory circuits generate burst firing and gate firing response to excitatory inputs in CRH_PVN neurons.

(A) CRH model neuron in silico (black) fitted to current-clamp recordings of CRH_PVN neurons ex vivo (red). (B) F-I curves for in silico (black) and ex vivo (red) neurons. (C) Network model diagram. (D) Time course of interspike interval (ISI) for a representative model CRH neuron. Gray line is the running average firing rate. (E) Summary burst length distribution for model simulations (n=283 for burst rate >0.1 Hz). (F) Summary IBI distribution for model simulations (n=283 for burst rate >0.1 Hz). (G) An example ISI time course before and after a drop of GABA release (Pr 0.8 → 0.1) from 200 s. Gray line is the running average firing rate. (**H–J**) Summary changes in all spikes (H, paired t-test, p<0.0001, n=500), bursts (I, paired t-test, p<0.0001, n=500) and single spikes (SS) (J, paired t-test, p<0.0001, n=500) before (BL) and after the GABA release drop. (K) An example ISI time course before and after a drop of GABA release (Pr 0.8 → 0.1) combined with an increase in spike-triggered adaptation current from 200 s. Gray line is the running average firing rate. (**L–N**) Summary changes in all spikes (L, paired t-test, p<0.0001, n=500), bursts (M, paired t-test, p<0.0001, n=500) and SS (N, paired t-test, p<0.0001, n=500) before (BL) and after combined removal of GABA release and increased spike-triggered adaptation current. (O) An example ISI time course before and after EPSP frequency increase (30 Hz → 60 Hz) from 200 s. Gray line is the running average firing rate. (P) An example ISI time course before and after the combined change in GABA release and spike-triggered adaptation current, followed by an increase in EPSP frequency (30 Hz → 60 Hz) from 400 s. (Q) Summary graph for changes in firing rate before (30 Hz) and after EPSP frequency increase (60 Hz) with (white) and without recurrent inhibition (pink). Two-way ANOVA (EPSP × Recurrent inhibition interaction, p<0.0001; Tukey’s multiple comparisons test, BL-30 Hz 2.942 ± 2.220 Hz vs BL-60 Hz 3.895 ± 3.252 Hz, p<0.0001; ↓GABA+↑Ad. Curr.-30 Hz 7.548 ± 2.068 Hz vs ↓GABA+↑Ad. Curr.-60 Hz 10.630 ± 3.387 Hz, p<0.0001). SD is represented in the graphs as the error bars.

Figure 5—source data 1 Recurrent inhibitory circuits generate burst firing and gate firing response to excitatory inputs in CRHPVN neurons.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig5-data1-v2.xlsx
Download elife-76832-fig5-data1-v2.xlsx

Table 1

Parameters for spiking network model.

Neuron parameters	Mean	S	Synapse parameters	Value
Number of neurons	1000		E_e (mV)	0
Number of CRH neurons	500		E_i (mV)	–80
Number of GABA neurons	500		τ_e (ms)	12.5
Capacitance (pF)	22.0	2.6	τ_CRH (ms)	232.6
τ_m (ms)	26.0	2.5	τ_i (ms)	20.4
g_L (nS)	0.9	0.19	w_e (nS)	3.9
∆_T (mV)	12.1	2.0	w_CRH (nS)	0.005
E_L (mV)	–67.9	2.9	w_i (nS)	3.3
V_T (mV)	–47.2	6.5	τ_br (s)	40
V_R (mV)	–58.8	2.5	τ_p (s)	80
τ_w (ms)	98.2	54.3
a (nS)	0.082	0.13
b (pA)	17.9	9.8

Next, the fitted single-neuron models were integrated into a spiking network model to test whether the CRH-GABA neuron network generates rhythmic bursting. We have constructed a network model of the PVN that contains 500 CRH neurons and 500 GABA neurons (Figure 5C). The excitatory CRH neurons send a sparse projection to the inhibitory GABA population (2% connection probability), and the GABA population projects back to the CRH neurons with a similar, sparse projection (2%) (Destexhe, 2009). CRH neurons also receive noisy random synaptic inputs (30 Hz fast EPSPs), approximated by an instantaneous rise and exponential decay (Rothman and Silver, 2014). GABAergic neurons, in turn, receive excitatory CRHergic inputs that have slow rise and decay time constant (~200 ms), modeled as single time constant alpha synapse (Destexhe et al., 1994), reflecting the slow CRHR1 mediated excitation (Jiang et al., 2018; Ramot et al., 2017). The putative network construction was then optimized to fit our in vivo observations. In brief, from a representative unit (shown in Figure 3K), temporal subsections were isolated as representative RB and SS firing patterns. Then, in the network model, several rounds of a genetic algorithm were used to optimize the synaptic network parameters; simulations were evaluated using the earth movers’ distance (EMD) between the log ISI distributions of the representative spike train and simulated spike trains (Materials and Methods). In this optimized network setting, we found that model CRH neurons generated RB (Figure 5D). In the model CRH neurons, the recurrent inhibitory inputs were critical for the long, mostly silent IBI, and the loss of inhibitory inputs (i.e. disinhibition) underlay the timing when excitatory inputs generated depolarization sufficient to trigger burst firing (Figure 5—figure supplement 1). In addition to the inhibitory and excitatory synaptic inputs, the spike-triggered adaptation current (Izhikevich, 2003) prevented the action potential firing of CRH neurons after the burst firing. However, its influence was less important. This was because its time constant (τ_w=98.2 ms, Table 1) was substantially shorter than the time constant of CRHergic excitation (τ_CRH = 232.6 ms, Table 1) and ensuing prolonged feedback inhibition. The model CRH neurons, which have heterogeneity in their intrinsic properties obtained in ex vivo recordings (Table 1), showed heterogeneity in their burst rates (0.1456 ± 0.1072 bursts/s, n=500). By further characterizing burst firing properties of these bursting model CRH neurons, we found that our network model recapitulated several in vivo features, including burst length and IBI (Figure 5E and F). Consistent with these features, RB in the model also constrained the overall firing rate at low levels (~2 Hz, Figure 5D gray line). These data show that CRH_PVN neurons (which are not intrinsically bursting) fire in RB within a recurrent inhibitory network.

Next, to test the causal roles of recurrent inhibition in constraining firing rate, we performed a series of tests with the network. First, a reduction of recurrent inhibition (by stochastically lowering the release probability [Pr] of GABA→CRH synapses from [0.9–1.0] to [0.01–0.2]) alone with no additional change in external excitatory inputs, robustly increased spiking, and consequently the overall firing rate (Figure 5G–J). This indicates that recurrent inhibition constrains the overall firing activity of CRH neurons. However, the removal of recurrent inhibition alone did not eliminate the burst-range high-frequency firing in CRH neurons (Figure 5G and L). Interestingly, we found that the in vivo-like switch from RB to SS can be achieved by an additional parameter change, an increase in the ‘spike-triggered adaptation current’ from (5–18) pA to (36–50) pA (Brette and Gerstner, 2005; Figure 5K–N). This result predicts potential intrinsic single-cell properties underlying the generation of RB (or its prevention). For example, the spike-triggered adaptation current can represent after-hyperpolarization, (a loss of) after-depolarization, or both (Izhikevich, 2003). Next, our modeling data also indicate that recurrent inhibition constrains the response (spike outputs) of CRH neurons to excitatory inputs. To test this idea directly, we increased external synaptic inputs under RB (i.e. in the presence of recurrent inhibition) and SS (i.e. in the absence of recurrent inhibition plus an increase in spike-triggered adaptation current) states (Figure 6O and P). In the presence of recurrent inhibition, an increase in excitatory synaptic inputs (from 30 Hz to 60 Hz) caused small change in the overall firing rate, namely CRH neurons’ response (Figure 5O and Q). By contrast, in the absence of recurrent inhibition, an identical increase in excitatory synaptic inputs robustly increased the firing rate of CRH neurons (Figure 5P and Q, Two-way repeated ANOVA, EPSP × recurrent inhibition, interaction p<0.0001), showing that recurrent inhibition works as a gain regulator at CRH neurons.

Figure 6

Download asset Open asset

Recurrent inhibitory circuits underlie burst firing.

(A) Diagram of ‘network clamp’ experiment. (**B–D**) Time course of interspike interval (ISI) for a representative model CRH neurons (B), a biological CRH_PVN neuron injected with network current (C) and overlay (D). (E) Correlation coefficient between the time series of model neuron spike train in silico, a biological CRH_PVN neurons’ spike train ex vivo, and a shuffled spike train consisting of the model spike train with ISI’s shuffled (n=5). (**F, G**) Spike firing patterns in response to network current (F) and depolarizing current step (G) in a biological CRH_PVN neuron. (H) ISI distribution of spikes generated by network current (red) and steps of current (purple). Note that spikes with ISI smaller than 30 ms were used for comparison. (**I, J**) Network currents of variable length (I) and corresponding membrane voltage changes in CRH_PVN neurons (J). (K) Summary graph for the shortest ISI triggered by network currents of variable lengths. SD is represented in the graph as the error bars. (L) CRH_PVN neuron’s firing response to network current with (blue) and without tetrodotoxin (TTX, 1 µM, red). (M) Firing of CRH_PVN neuron (orange) and magnocellular neuron (brown) in response to an identical network current (black). (N) Summary of ISI triggered by network current in CRH_PVN and magnocellular neurons. (**O–Q**) A representative ISI time course for the transition between rhythmic burst and single spiking in a model CRH neuron (O), a biological CRH_PVN neuron injected with the network current of the model neuron shown in O, and overlay (Q).

Figure 6—source data 1 Recurrent inhibitory circuits underlie burst firing.: https://cdn.elifesciences.org/articles/76832/elife-76832-fig6-data1-v2.xlsx
Download elife-76832-fig6-data1-v2.xlsx

Recurrent inhibition generates burst firing in single CRH_PVN neurons

Our model simulation predicted that recurrent inhibition is sufficient and necessary for RB and constrains the firing rate of CRH_PVN neurons. However, it remains unknown whether our single neuron model, while grounded in direct physiological measurements (Figure 5AB, Table 1), captures essential biophysical properties required for in vivo-like firing patterns. To directly test this, we went back to ex vivo slice patch clamp electrophysiology and asked if an injection of whole-cell (somatic) currents that mimic the recurrent network inputs of the spiking network model is sufficient to drive RB firing in biological CRH_PVN neurons (Figure 6A). To this end, we first selected a model CRH_PVN neuron that showed representative RB firing (Figure 6B), and then extracted the ‘network currents’ that this model neuron receives within the network model (see Materials and Methods). We then injected these ‘network currents’ into biological CRH_PVN neurons in slices using current-clamp electrophysiology (Figure 6C). We found a remarkable similarity between the RB firing patterns of a model neuron and a biological CRH_PVN neuron that received the ‘network currents’ (Figure 6D and E). Notably, in response to square pulse depolarization, biological CRH_PVN neurons in acute slices fire in regular spiking patterns (Figure 6F and H) with the lowest ISI > 10 ms. The same neurons, however, readily generated burst range high-frequency firing (ISI < 6 ms) in response to complex network current injections derived from the network model (Figure 6G and H). A closer examination of the network inputs indicated that high-frequency excitatory inputs drove burst firing (Figure 6G). Importantly, however, the high-frequency inputs alone were not sufficient to elicit burst firing, and preceding network inputs were also important. Figure 6I–K show that truncated current injection only containing the high-frequency inputs failed to elicit burst firing, and that preceding noisy network inputs (which caused subthreshold membrane potential fluctuations) were necessary to drive the burst firing. These results corroborate with in vivo data that burst firing almost exclusively occurred after a prolonged (>200 ms) silent period (Figure 4C), indicating the importance of dynamic interactions between network synaptic inputs (that generate subthreshold membrane potential changes) and intrinsic properties. The burst firing triggered by network synaptic currents were abolished in the presence of voltage-gated Na⁺ channel blocker tetrodotoxin (TTX, 1 µM, Figure 6L), confirming that they were indeed action potentials. Moreover, we found that in magnocellular (non-CRH) neuroendocrine neurons of the PVN, which are known to have ex vivo intrinsic properties (Luther et al., 2002) and in vivo firing patterns (Leng and MacGregor, 2018) different from CRH_PVN neurons, the identical network current failed to trigger the RB (Figure 6M and N). Lastly, Figure 6O shows an example of model neurons capturing in vivo-like alternations of RB and SS using a transient change in the recurrent inhibition and adaptation current (see in vivo data in Figure 4J). The injection of the network currents effectively elicited a similar transition of firing pattern in biological CRH_PVN neurons in ex vivo (Figure 6P and Q).

Discussion

Here, we report single-unit activities of identified CRH_PVN neurons in vivo for the first time. We show that CRH_PVN neurons, within intact circuits, can fire distinct brief, high-frequency (>100 Hz) bursts, a firing pattern that has not been observed in ex vivo slice electrophysiology (Bittar et al., 2019; Jiang et al., 2019; Khan et al., 2011; Luther et al., 2002; Matovic et al., 2020; Sarkar et al., 2011; Wamsteeker Cusulin et al., 2013) (but see Yuan et al., 2019 for slower [<100 Hz] bursts). Guided by the in vivo firing patterns, we have developed a computational model and showed that the RB firing mode reflects critical roles of recurrent inhibition in constraining the overall firing rate of CRH neurons. More generally, the recurrent inhibition controls the gain of CRH neurons’ response to excitatory inputs. In biological CRH_PVN neurons ex vivo, the injection of whole-cell currents derived from our network model effectively triggered the in vivo-like RB and recapitulated the transition from RB to SS, thus providing direct evidence that intrinsic properties of CRH_PVN neurons, which do not overtly fire in burst in slices, are capable of firing in vivo-like RB by interacting with complex inputs reflecting network characteristics. In summary, using a combination of experimental and computational approaches, we demonstrate a novel circuit mechanism that controls state-dependent activity switch of CRH_PVN neurons between the baseline and stress.

CRH_PVN neurons fire brief bursts with long interburst intervals

Pioneering studies in 1980s used antidromic activation of the median eminence and reported single-unit recordings from parvocellular neuroendocrine neurons of the PVN, a class of neurons that include CRH and several other hormone-releasing neurons, in anesthetized rats (Day et al., 1985; Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1985). Here, using the optrode technique (Lima et al., 2009), we report firing activities of identified CRH_PVN neurons in anesthetized mice. Overall, our results from CRH_PVN neurons agree with the early results from parvocellular neurons in that both populations show spontaneous firing at low rates, and that majority of them increased firing rates in response to stress-mimicking peripheral nerve stimulation.

A key new finding of our study is that CRH_PVN neurons fire distinct bursts characterized by RB, a brief train of high-frequency action potentials (~3 action potentials/bursts at >100 Hz) intervened between long periods of silence (constraining the overall firing rate), in addition to SS with more variable and lower spike-to-spike frequencies. This RB firing was not explicitly documented by the earlier studies in rats (Day et al., 1985; Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1985; Watanabe et al., 2004); likely because these studies analyzed the firing activities only as time-binned firing rates and did not examine spike-to-spike temporal patterns. In fact, representative traces in one of aforementioned studies showed spikes >100 Hz (see Figure 4 of Saphier and Feldman, 1985). It should be emphasized that the RB we report here is distinct from ‘phasic (burst) firing’ which is widely known as a signature firing patterns of vasopressinergic magnocellular neurons of the PVN and SON (Brown and Bourque, 2006). The magnocellular phasic burst consists of lower intraburst frequencies (2.5–20 Hz) and lasts much longer with continuous firing (>10 s) than the RB (<100 ms) we report here in CRH_PVN neurons (Poulain and Wakerley, 1982; Watanabe et al., 2004). Also, the phasic burst firing is followed by a prolonged (~10 s) silent period generated by intrinsic mechanisms (Brown and Bourque, 2006). We did not observe this type of phasic bursts in identified CRH_PVN neurons consistent with the fact magnocellular neurons do not express CRH (Biag et al., 2012; Simmons and Swanson, 2009).

The brief bursts were intervened with mostly silent IBIs of its median length around 2 s during the RB state. Consequently, despite the high frequency (>100 Hz) burst firing, the time-averaged firing rate of the RB state remained low (3–5 Hz). Furthermore, during non-stress conditions, the RB state dynamically shifted to a non-bursting, SS state where CRH_PVN neurons increased their firing rate (10–20 Hz) due to continuous single spiking. A similar shift from RB to continuous SS was evident when CRH_PVN neurons increase their firing rate in response to stress-mimicking sciatic nerve stimulations. When considering hormonal CRH levels, the drive for the hormone release is likely to be coded by firing rate over the time scale of seconds and longer, rather than millisecond precisions of firing patterns. Thus, we speculate that persistent SS firing and ensuing elevation of spike rate play a major role in the hormonal CRH release.

A computational model predicts a novel disinhibition mechanism of CRH_PVN neurons

Decades of research has established that, under no-stress conditions, the excitability of CRH_PVN neurons is tonically constrained by powerful GABAergic synaptic inhibition, and that a release from this tonic inhibition, or disinhibition, is one dominant mechanism for the activation of these neurons, and consequently the HPA axis (Bains et al., 2015; Cole and Sawchenko, 2002; Cullinan et al., 2008; Inoue and Bains, 2014; Levy and Tasker, 2012; Roland and Sawchenko, 1993; Ulrich-Lai and Herman, 2009). Multiple disinhibitory mechanisms have been identified, likely reflecting diverse and overlapping neural and hormonal pathways controlling the activities of the HPA axis (Joëls and Baram, 2009; Ulrich-Lai and Herman, 2009). These include the inhibition of the presynaptic GABAergic neurons (Anthony et al., 2014; Johnson et al., 2019), depression of GABAergic synaptic terminals (Ferri and Ferguson, 2005; Han et al., 2002; Hewitt and Bains, 2006; Khazaeipool et al., 2018) and a depolarizing shift in the equilibrium potential of GABA_A receptor in the postsynaptic CRH_PVN neurons (Hewitt et al., 2009; Sarkar et al., 2011). However, it remained unknown how GABAergic inhibition controls the firing activities of CRH_PVN neurons in vivo. In this study, we developed a simple computational model, which was guided by distinct brief burst firing in vivo, and revealed previously underappreciated contribution of recurrent inhibition to the tonic inhibition of CRH_PVN neurons. Specifically, the feedback inhibition effectively prevented continuous firing and created a lasting (~2 s) silent periods in between brief bursts, and consequently constrained the time-averaged firing rate low (~3 Hz). Strikingly, a decrease of recurrent inhibition, with no additional change in the excitatory inputs, profoundly increased the overall firing rate of CRH_PVN neurons. In other words, (a loss of) the recurrent inhibition alone partially recapitulates the disinhibitory mechanisms of CRH_PVN neurons. That said, there is more to consider regarding the activity of CRH_PVN neurons, as the removal of recurrent inhibition alone results in continuous burst-range high frequency firing, which does not typically occur during stress-induced firing increase in vivo (Figure 3A–F). Thus, our model predicts that the in vivo-like transition from RB to SS requires an increase in spike-triggered adaptation current (Brette and Gerstner, 2005), pointing to single-neuron intrinsic properties underlying burst generation. For example, the spike-triggered adaptation current can represent after-hyperpolarization, (a loss of) after-depolarization, or both (Brette and Gerstner, 2005).

It should be noted that our model for the recurrent excitatory-inhibitory circuit in the PVN is simplified and does not incorporate external inhibitory inputs to CRH_PVN neurons or external excitatory and inhibitory inputs to the GABAergic neurons (Ulrich-Lai and Herman, 2009). Furthermore, transmission from CRH neurons to GABAergic neurons is exclusively modeled by slow CRHergic transmission based on recent experimental data (Jiang et al., 2019; Ramot et al., 2017), but CRH_PVN neurons have been shown to release glutamate at some synapses (Füzesi et al., 2016). Despite these simplification, our model parsimoniously captures the key features of firing behavior in vivo, including the brevity of burst episodes and long IBI (Figure 5E and F). Thus, our simplified model, which is the first network model for CRH_PVN neurons to our best knowledge, is a useful first step to understand the spiking behaviors of CRH_PVN neuron in a network, and to generate experimentally testable hypotheses for future research.

So, what are the identities of the GABAergic neurons that form recurrent connectivity with CRH_PVN neurons? Indeed, the recurrent inhibitory circuits for CRH_PVN neurons has been first suggested by an early in vivo electrophysiology recordings of parvocellular neurons in rats based on an inhibition following action potentials antidromically activated from the median eminence (Saphier and Feldman, 1985). Recent studies have revealed local recurrent circuits within the PVN, where CRH_PVN neurons excite CRHR1-expressing GABAergic interneurons, which in turn send GABAergic inputs to CRH_PVN neurons (Jiang et al., 2018). CRH_PVN neurons excite CRHR1-expressing neurons via slow metabotropic actions with little, if any, fast (glutamatergic) excitatory synaptic transmission (Jiang et al., 2018; Ramot et al., 2017). Notably, our network model found that the long IBI (~2 s) of brief bursts require slow excitatory EPSPs consistent with metabotropic signaling (τ_CRH >200 ms) rather than fast ionotropic EPSPs (τ_e <20 ms). In addition to this intra-PVN microcircuit, it is possible that CRH_PVN neurons form additional recurrent inhibitory circuits. Indeed, there is growing appreciation that CRH_PVN neurons send projections within the brain besides their well established projections to the median eminence (for hormonal CRH release) (Füzesi et al., 2016; Jiang et al., 2018; Jiang et al., 2019; Kim et al., 2019a; Li et al., 2020; Ono et al., 2020; Ramot et al., 2017; Rho and Swanson, 1989; Yuan et al., 2019). For example, the perifornical area (Füzesi et al., 2016; Rho and Swanson, 1989) and the lateral hypothalamus (Li et al., 2020; Ono et al., 2020) receive direct synaptic inputs from CRH_PVN neurons: these brain areas in turn send direct GABAergic inputs to CRH_PVN neurons (Boudaba et al., 1996; Cullinan et al., 2008; Roland and Sawchenko, 1993; Ulrich-Lai and Herman, 2009).

Synaptic activity underlies flexible firing patterns of CRH_PVN neurons in vivo

In acute slices ex vivo, CRH_PVN neurons typically show a SS electrophysiological phenotype, or ‘regular’ spiking, with up to around 50 Hz of spike-to-spike frequency in response to steps of intracellular current injections in mice and rats (Bittar et al., 2019; Jiang et al., 2019; Khan et al., 2011; Luther et al., 2002; Matovic et al., 2020; Sarkar et al., 2011; Wamsteeker Cusulin et al., 2013). Thus, our in vivo recordings of the RB (>100 Hz spike-to-spike frequency) revealed a previously underappreciated repertoire of the firing patterns of CRH_PVN neurons. In this regard, CRH_PVN neurons are different from thalamic relay neurons that have been extensively studied for their properties to switch between bursting and single (tonic) spiking modes both in vivo and ex vivo (McCormick and Pape, 1990; Sherman, 2001; Steriade et al., 1993). The difference can be partly explained by the fact that CRH_PVN neurons express only modest levels of T-type currents (Luther et al., 2002) and little, if any, H-currents (see Figure 5A): these intrinsic properties play crucial roles in the generation of rhythmic burst firing, and switch between burst and single spiking modes in thalamic relay neurons (McCormick and Pape, 1990; Sherman, 2001; Steriade et al., 1993).

In other words, our data (i.e. the absence of overt burst-generating intrinsic properties) suggest that the intrinsic properties of CRH_PVN neurons conduct flexible input/output signal processing by interacting with the physiological network dynamics which is substantially lost in ex vivo slice preparations. Supporting this idea, our modeling showed that model CRH neurons, which is fitted to ‘regular’ spiking phenotype characterized ex vivo, readily fire in RB when they receive the bombardment of excitatory inputs (30 Hz EPSPs) combined with the recurrent inhibition. Thus, our modeling results demonstrate that as simple as two types of network-like inputs are sufficient to generate a firing pattern that is qualitatively different from and not readily evident in conventional (e.g. square pulse) electrophysiological characterizations ex vivo.

Interestingly, Yuan et al. recently reported spontaneous ‘burst firing’ of CRH_PVN neurons ex vivo using cell-attached recordings in slices prepared from repeatedly stressed mice (Yuan et al., 2019). It should be noted, however, that the intraburst frequency of this ex vivo ‘burst firing’ was <100 Hz (averaging around 50 Hz), which was different from the in vivo RB (>100 Hz intraburst frequency) we report here in non-stressed mice. Furthermore, the ex vivo ‘burst firing’ involved stress-induced plasticity that developed over days because it only became evident after three daily foot shock, but was absent when mice were not previously stressed, or even after two daily foot shock. Accumulating evidence show that CRH_PVN neurons undergoes diverse forms of synaptic and intrinsic plasticity in response to acute as well as during chronic stress (Bains et al., 2015; Herman and Tasker, 2016; Matovic et al., 2020; Salter et al., 2018; Yuan et al., 2019). Thus, it is an important direction of future studies to examine the plasticity of CRH_PVN neurons’ firing in vivo and their contributions to (mal)adaptations of the stress response.

Our computational modeling addressed the potential interplay between in vivo-like network inputs and the intrinsic properties of CRH_PVN neurons. The flip side of our approach, however, is the oversimplification of the network inputs and questions about their physiological relevance. To directly address the latter concern, we applied the modeling results to biological neurons ex vivo. Specifically, we extracted network inputs from representative model neurons that fire RB, and then injected the network current waveform into biological neurons under whole-cell current-clamp configuration ex vivo. We term this new approach as ‘network clamp’. Network clamp injects a predetermined current waveform which can be readily available from the leaky-integrate-and-fire model, and therefore much simpler, operationally, and computationally, than dynamic clamp. One limitation is that the amplitude of the current fluctuations can be larger than physiologically realistic range. Despite this limitation, however, we found that the network current generated RB in CRH_PVN neurons and the transition to SS upon changes in network parameters (i.e. removal of recurrent inhibition and increase in adaptation current), providing direct evidence for the capability of CRH_PVN neurons to fire with high frequency bursts (i.e. RB pattern) and also bidirectionally switch to regular spiking (i.e. SS pattern). Importantly, the burst firing does not simply occur as a results of current injection patterns but requires unique interaction between the intrinsic properties of CRH_PVN neurons and the network inputs. This is supported by the following two controls. First, network inputs, which drove subthreshold membrane potential changes preceding (>200 ms) the burst firing, were necessary because truncated (<200 ms) network currents, which only contain high-frequency synaptic inputs component, failed to elicit the brief bursts in CRH_PVN neurons (Fig. 6I-K). Notably, this result corroborates well with in vivo data that RB almost exclusively occur after long (>200 ms) silent period (Figure 4B). Second, magnocellular neurons, a separate class of neuroendocrine neurons of the PVN with different intrinsic properties (Luther et al., 2002), did not fire in brief bursts in response to the same network current injection that elicited brief bursts in CRH_PVN neurons (Figure 6M and N). The intrinsic properties key to this burst firing are currently unknown and warrants future investigation.

These results point to future applications of the ‘network clamp’ in order to examine specific biophysical mechanisms that interact with network inputs to generate in vivo-like firing patterns in other brain areas. It is well known that neurons in slice differ in their response properties from intact in vivo preparations (Destexhe and Paré, 1999; Destexhe et al., 2003) and that injection of noise approximating the inputs from the large number of synaptic contacts in intact preparations can recreate firing conditions observed in vivo (Destexhe and Rudolph-Lilith, 2012; Destexhe et al., 2003; Rudolph and Destexhe, 2003; Rudolph and Destexhe, 2006). Advanced techniques have been developed to measure neuronal firing rate responses under injection of theoretically motivated noise processes (Zerlaut et al., 2016). In this work, however, we have focused on injecting the time-varying inputs from a single cell in the network model to a biological neuron, so that we can directly compare the activity patterns evoked in the biological and simulated cells. We find that computationally generated inputs can recreate the specific burst firing mode we have reported in this work. In this way, this injection protocol bears some similarity to the ‘iteratively constructed network (ICN)’ developed by Reyes, 2003 to provide a direct test of whether biological cells can propagate synchrony without some of the caveats that can be imposed by computational simulations (Rudolph and Destexhe, 2007). We suggest that this ‘network clamp’ protocol has more general applicability; however, for validating spiking network models of neural circuits. In this work, the spiking network model provided a testable prediction: initially, biological neurons recorded ex vivo did not exhibit bursting behavior when stimulated with step current pulses, but would they burst when driven with network-generated input? In this case, the biological neuron in this ‘network clamp’ paradigm exhibited behavior consistent with this prediction, strengthening confidence that our spiking network model captures a meaningful underlying mechanism in this system. Because this paradigm allows to test multiple model predictions in a single setup (for conventional patch-clamp electrophysiology), and because it also opens the possibilities for testing how neuronal firing and response patterns may change with neuromodulation applied in the slice, we suggest that the model-experiment protocol developed and tested in this work may be able to distinguish between multiple models that can account for a single dataset in isolation (Marder and Goaillard, 2006).

Limitations of the study

Our in vivo recordings were performed under urethane anesthesia, as was the case for the majority of previous studies using in vivo recordings from the PVN in anesthetized rats (Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1985; Saphier and Feldman, 1988; Saphier and Feldman, 1990). Urethane produces a long-lasting steady level of surgical anesthesia, and preserves the subcortical and peripheral neural functions (Maggi and Meli, 1986), making it suitable for long electrophysiological recordings from the hypothalamus. For example, under urethane anesthesia in rabbits, hypothalamic neurons have been shown to readily respond to varieties of stress stimuli including hypoxia, hypercapnia, loud noise, and pain (Cross and Silver, 1963). Of particular relevance to our study, sciatic nerve stimulation (under urethane anesthesia) has been shown to effectively elevate blood ACTH levels (Hamamura et al., 1986), validating that the stress-responsiveness of the HPA axis is preserved. However, it should be noted that urethane has also been shown to enhance autonomic (Shimokawa et al., 1998a) and HPA axis activity (Hamstra et al., 1984), likely by enhancing adrenergic inputs to the PVN (Shimokawa et al., 1998b). Considering that (nor)adrenaline has both excitatory and inhibitory effects on PVN neurons (Han et al., 2002; Itoi et al., 1994; Saphier and Feldman, 1991) and that it can potently influence state-switch between bursting and single spiking (Pape and McCormick, 1989), urethane may have affected the baseline firing activities and burst firing properties of CRH_PVN neurons in general. Future studies using awake animal recordings are warranted.

Ideas and speculations

Our study found brief, high-frequency burst firing of CRH_PVN neurons in vivo. This unexpected finding led us to propose, using a tight combination of experimental and computational approaches, that recurrent inhibition can generate the brief bursts and constrain prolonged firing activities of CRH_PVN neurons leading to hormonal release. Then, what is the advantage of a RB mechanism during periods when CRH neuron activity is not required for the hormone release? Does RB firing encode and convey specific information? One possibility is that burst firing facilitates the self-maintenance of the RB firing state. That is, a brief burst enables high-fidelity synaptic transmission (Lisman, 1997) and efficiently excites GABAergic neurons that return recurrent inhibition to CRH_PVN neurons: this feedback inhibition, in turn, facilitates subsequent burst firing. Notably, our spiking network model found that recurrent inhibition works as a gain regulator at CRH neurons (Figure 5). Thus, our model prediction offers a new circuit mechanism for the experimental data that disinhibition, more so than increasing excitatory inputs, is crucial for the activation of CRH_PVN neurons for hormonal release (Cole and Sawchenko, 2002; Hewitt et al., 2009; Kovács et al., 2004; Sarkar et al., 2011). Beyond constraining hormonal release of CRH, the idea that CRH_PVN neurons use brief bursts for high-fidelity synaptic transmission, without triggering massive HPA axis activation, raises an intriguing possibility in light of emerging new roles of CRH_PVN neurons that are independent of hormone release. Recent studies showed that CRH_PVN neurons are involved in controlling wakefulness (Ono et al., 2020), valence encoding (Kim et al., 2019a), reward processing (Kim et al., 2019a; Yuan et al., 2019), and defensive behavior control (Daviu et al., 2020; Füzesi et al., 2016) under non-stress and stress conditions. We propose that brief intermittent bursts of activity are ideally suited to drive rapid behavioral and emotional changes independently of systemic and long-lasting hormonal response.

Materials and methods

Animals

All experimental procedures were performed in accordance with the Canadian Council on Animal Care guidelines and approved by the University of Western Ontario Animal Use Subcommittee (AUP: 2018–130). Homozygous Crh-IRES-Cre (B6(Cg)-Crhtm1(cre)Zjh/J) mice (Stock No: 012704, the Jackson Laboratory) were crossed with homozygous Cre-reporter Ai14 (B6.Cg-Gt(ROSA)26Sortm14(CAG-TdTomato)Hze/J) mice (Stock No: 007908, the Jackson Laboratory) to produce CRH-TdTomato reporter offspring. The specificity of cre-expression in the PVN in these mice has been characterized in previous studies (Chen et al., 2015; Wamsteeker Cusulin et al., 2013). 10 adult male mice (>60 days old) were used for viral injections and recordings. Animals were group-housed (2–4 per cage) in standard shoebox mouse cages with ad libitum access to food and water. Animals were housed on a 12 hr dark/12 hr light cycle (lights on from 07:00 to 19:00) in a temperature-controlled room (23 ± 1°C).

Viral injection

Request a detailed protocol

To express channelrhodopsin-2 (ChR2) in CRH_PVN neurons, we used an AAV carrying the EF1α promoter and the double-floxed inverted ChR2(H134R)-EYFP coding sequence that is inverted and turned on by cre-recombination (AAV2/5- EF1α-DIO-ChR2(H134R)-EYFP). The plasmid was a gift from Dr. Karl Deisseroth http://n2t.net/addgene:20298; RRID:Addgene_20298. The AAV preparation was obtained from the Neurophotonics Centre (5 × 10¹³ GC/ml, Laval University, Canada).

For AAV injection, the animal was anesthetized under isoflurane (2%) using a low flow gas anesthesia system, (Kent Scientific Corporation) and placed in a stereotaxic apparatus on a heating pad. A finely pulled glass capillary was loaded with the virus and slowly lowered into the brain of animals, targeting the PVN on each side of the brain (A/P: –0.70 mm, M/L:±0.25 mm, and D/V:−4.75 mm from bregma). A total of ~240 nL (41.4 nL × 6 at 23 nL/s) was pressure-injected on each side using the Nanoject II (Drummond Scientific Company). The pipette was held in place for 5 min after injection to allow diffusion before being slowly retracted from the brain. The incision was closed by suturing and the animals were injected with analgesic (buphrenorphone 0.1 mg/kg, s.c.) at the end of surgery. Animals were allowed to recover for at least 6 weeks for optimal ChR2 expression before electrophysiological recordings.

Electrophysiology

Request a detailed protocol

Animals were initially anesthetized under isoflurane (1–2%) and urethane (1.5 g/kg in 0.9% saline, intraperitoneal) to perform surgery for sciatic nerve isolation and a craniotomy/durotomy for insertion of the recording probe into the brain. We first isolated the sciatic nerve (see details in Sciatic nerve stimulation), and then the probe was lowered into the brain. Thereafter, the animal was taken off isoflurane. The probe was vertically inserted above the PVN (A/P: –0.70 mm from bregma, M/L:±0.25 mm from bregma) and slowly lowered ventrally to the PVN (target D/V: –4.5 mm from cortical surface, photo-tagged units (see details in Optogenetic identification of CRH neurons) were found between –4.20 mm and –4.80 mm from cortical surface).

Extracellular neural signals were recorded using a single shank, 32-channel (8 rows × 4 columns) silicon probe, with recording sites spanning 60 µm in depth (Cambridge NeuroTech). An optic fiber (100 µm core diameter, 0.37 NA, A45; Doric Lenses) was attached by the manufacturer parallel to the electrode shank with a vertical offset of 250 µm, constituting an optrode. The electrode was connected to a digital headstage (ZD ZIF-Clip; Tucker-Davis Technologies) with an internal Intan amplifier chip. The digitized signals were sent to the amplifier (PZ5; Tucker-Davis Technologies), which in turn connects out to the processing unit (RZ5D, Tucker-Davis Technologies) via a fiber optic cable. Broadband signals were sampled at 25 kHz and bandpass filtered between 300 and 3000 Hz. The threshold for spike detection was a minimum three SDs above the noise floor. The waveforms for detected spikes were sorted offline (Offline Sorter, Plexon) using manual and automatic clustering (T-Distribution E-M). To ensure the quality of single-unit isolation, only discrete clusters with L-ratios <0.05 (in 2D or 3D spaces) was included for single-unit analysis.

Optogenetic identification of CRH neurons

Request a detailed protocol

Transistor-transistor logic (TTL) signals to trigger optogenetics light stimulation were sent from the signal processing and acquisition software Synapse (Tucker-Davis Technologies) to a LED driver (PlexBright LD-1 Single Channel LED Driver, Plexon). The LED driver controlled the intensity of light emitted from the LED module (PlexBright Table-Top LED Module Blue 465 nm, Plexon) which was coupled to an optic patch cable (200 µm core, PlexBright Optical Patch Cable, Plexon). The patch cable connected via a plastic sleeve to the optic fiber of the optrode.

For each electrode location, 5 ms and 50 ms single light pulses were delivered at 2 mW. Following offline sorting of waveforms, ChR2-expressing CRH single units were identified by their time-locked responses to light (see details in Data analyses and statistics: Identification of light-evoked spikes). Specifically, we clustered all waveforms (both spontaneous and light-evoked in a blinded manner) for single unit isolation (L-ratios <0.05). Thereafter, single units clustered with the light-evoked waveforms were considered as light responsive and thus CRH neurons, and other single units simultaneously recorded with light responsive neurons but did not respond to the light were considered as light non-responsive, which includes both non-CRH neurons and CRH neurons with insufficient ChR2 expression and/or light exposure.

Sciatic nerve stimulation

Request a detailed protocol

Under isoflurane anaesthesia, a small incision was made into the hind limb of the animal and the sciatic nerve was isolated from surrounding tissue. A bipolar tungsten stimulating electrode was gently placed on the nerve and the nerve was kept hydrated with periodic applications of sterile saline. The contralateral nerve to the recording site was stimulated with 1.6 mA negative pulses at 20 Hz (5 × 0.5 ms) using a pulse stimulation unit (S88, GRASS Instrument Co) connected to a stimulus isolation unit (Model PSIU6, GRASS Instrument Co) triggered via TTL signals.

Histology

Request a detailed protocol

At the end of recording, animals were euthanized with an overdose of pentobarbital sodium (150 mg/kg, i.p.) and transcardially perfused with 0.9% saline and 4% paraformaldehyde (PFA). The brain was collected and left to postfix in 4% PFA at 4°C overnight. The brain was washed in phosphate buffer saline (PBS) solution and sliced into 40 µm coronal sections and counter stained with 4,6-diamidino-2-phenylindole (DAPI). The slices were mounted and verified for CRH expression (tdTomato), ChR2 expression (eYFP), and the dye-painted electrode tract (Vybrant DiD, Thermo Fisher V22889).

Experimental design

Request a detailed protocol

Each recording started with the baseline recordings of spontaneous firing without any intentional sensory stimulation for at least 10 min. This was followed by light stimulation for CRH neuron identification, then subsequently sciatic nerve stimulation (minimum 3 min between light stimulation and sciatic nerve stimulation).

Data analyses and statistics

Data analyses were carried out using built-in and custom-built software in MATLAB (MathWorks). Data analysis codes are available on GitHub, copy archived at swh:1:rev:d4d153e44882a0c2a7d0a1ef11193347aa5d6c96 (Ichiyama and Mestern, 2022). Graphs were made and statistical analysis were conducted using Prism 8 (GraphPad).

Identification of light-evoked spikes

Request a detailed protocol

Time-locked response to light stimulation was tested by an increase in spiking activity to a 5 ms pulse of blue light (25 trials). The response was plotted on a peristimulus time histogram (PSTH) aligned to the onset of light. Peristimulus frequency before and after light onset (20 ms windows) was compared by a trial-by-trial paired t-test: a significant increase was defined as light-responsive. The event probability before and after light (20 ms windows) was calculated as the number of trials that had at least 1 event within the response window, divided by the total number of trials.

Spike pattern analysis

Request a detailed protocol

For each single-unit, spike activity was classified into bursts and SS. Bursts were detected as series of two or more spikes starting with the initial ISI less than 6 ms and subsequent ISIs less than 20 ms. Although rare, to prevent contamination of bursts by high-firing SS which gradually speed up (i.e. shortening ISIs) into the burst range, we adopted a secondary criterion for bursts to be preceded by an ISI >25 ms. Spikes not associated with bursts were defined as SS. For analyses involving burst rates (i.e. burst rate, burst index, IBI), each burst was counted as a single event regardless of burst length (the number of spikes per burst). IBI distribution was first calculated for each cell (100 ms bins, plotted by upper limit of bin), and then the group distribution was generated by plotting the mean and standard deviation. Similarly, the distribution of burst length (the number of spikes per burst) was first calculated for each cell, and then the group distribution was generated by plotting the mean and SD.

To examine spiking dynamics that may influence burst activity, the probability of burst initiation (ISI_(i)<6 ms) as a function of preceding silence (ISI_(i-1)) was plotted for each cell. To do this, spikes were binned by ISI_(i-1) (10 ms and 500 ms bins for periods <500 ms and ≥500 ms, respectively), and the probability of ISI_(i)<6 ms was calculated for each bin for each cell. Then, the group average was plotted with the mean and SD. To examine the influence of preceding silence on burst length, spikes were binned by event length, and the mean preceding silence was calculated for each cell. Then, the group average was plotted with the mean and SD. The silent period analysis was repeated for proceeding silent periods. In this case, the probability of burst (ISI_(i)<6 ms) as a function of the proceeding silence (ISI_(i+1)) was plotted. Finally, the mean proceeding silence was compared by event lengths.

Firing rate analysis

Request a detailed protocol

The average baseline firing rates were calculated as the total number of spikes divided by total time during the spontaneous baseline recording (10 min). The firing rate time course was calculated as the number of spikes in every 10 s bins. The running average firing rate represents a moving average of five consecutive bins (bin[i-2]+bin[i-1]+bin[i]+bin[i+1]+bin[i+2]/5) and was used to visualize spontaneous fluctuations of firing rate. These fluctuations in firing rate were then overlaid onto ISIs plotted across time to visualize the relationship between firing rate to firing patterns. To examine the relationship between burst rate and total spike rate, for every 10 s bin in individual units, total spike rate was plotted against burst rates of 0.1, 0.2, 0.3, 0.4, 0.5, or ≥ 0.6 bursts/s. The relationship was plotted for individual unit (Figure 3L), as well as pooled across units (Figure 3M, n=18).

Sciatic nerve stimulation

Request a detailed protocol

The response to sciatic nerve stimulation was plotted on PSTHs aligned to the onset of stimulation (the first of the 5-pulse train). A paired t-test was performed between the average post-onset firing rates (2 s across 30–60 trials) and the baseline firing for each cell. Responses to sciatic nerve stimulation was analyzed in three categories: all spikes regardless of spiking pattern, SS only, and bursts only (each burst train counted as a single event regardless of the burst length).

Recurrent inhibitory network model

Single neuron model fit

Request a detailed protocol

Single neurons were modeled using the Adaptive Exponential Integrate and Fire (AdEx) model (Brette and Gerstner, 2005).

C \frac{d v}{d t} = g_{L} (E_{L} - v) + g_{L} Δ_{T} \exp (\frac{v - V_{t h}}{Δ_{T}}) + g_{e} (E_{e} - v) + g_{i} (E_{i} - v) - w + I

\begin{array}{ll} v \to v_{r} \\ w \to w + b \end{array}

Here, $C$ is the membrane capacitance, $w$ is an adaptation variable, $I$ is the applied current, $g_{L}$ is the leak conductance, $E_{L}$ is the resting membrane potential, $Δ_{T}$ is the slope factor, and $V_{t h}$ is the threshold potential. The slope factor determines the sharpness of the threshold. The membrane potential $v$ is modeled as a sum of these conductance parameters (Equation 1). When the membrane potential $v$ exceeds the threshold $V_{t h}$ the neuron emits a spike, and the neuron is reset according to Equation 2. w is computed as a function of the subthreshold, and spike-triggered adaptations, respectively (Equation 3)

τ_{w} \frac{d w}{d t} = a (v - E_{L}) - w

where $τ_{w}$ is the time constant and $a$ is subthreshold adaptation conductance.

To refine the model to capture the behavior of CRH_PVN neurons, we fit the model with whole-cell patch-clamp recordings of CRH_PVN neurons in acute brain slices (n=33, 5 mice) using methods detailed in Matovic et al., 2020. Recordings consisted of 14 trials containing a hyperpolarizing (–20 pA for 200 ms) pulse followed by a stepped depolarizing current pulse (10 pA steps for 500 ms). The same stimulus was used in subsequent model simulations. First, the firing rate-current (F-I) curve for the model was simulated across a large parameter space. Then, the posterior distribution for the parameter space was fit using the Sequential Neural Posterior Estimation (SNPE) method as previously described (Gonçalves et al., 2020). Second, for each reference cell, the F-I curve was used to draw a restricted parameter space from the previously fit posterior. Then, 200 rounds of evolutionary optimization were used to refine the model (Lynch and Houghton, 2015). The penalty (or evaluation of fit) was calculated as the sum of the mean squared error (MSE) between the subthreshold voltage of the model and reference, and the log MSE between the F-I curve of the model and reference. Models that produced an MSE above 5 mV (for the subthreshold voltage) and above 10 Hz (for the F-I curve) were considered poor fits, and were discarded. To produce a robust general model, the average of the parameter fit for all reference traces was used in the network model (Table 1).

Network model

Request a detailed protocol

To construct the network model, we utilized a simplified recurrent network template with two distinct populations of AdEx neurons (Destexhe, 2009). These neurons consisted of a modified AdEx model with added excitatory and inhibitory synaptic parameters (Equation 1).

\begin{array}{ll} g_{e} \to g_{e} + w_{e} \\ g_{i} \to g_{i} + w_{i} \end{array}

\begin{array}{ll} τ_{e} \frac{d g_{e}}{d t} = - g_{e} \\ τ_{i} \frac{d g_{i}}{d t} = - g_{i} \end{array}

Here, $g_{e}$ & $τ_{e}$ and $g_{i}$ & $τ_{i}$ represent the inhibitory and excitatory conductance and time constant, respectively. Following a presynaptic spike, the inhibitory and excitatory conductance were incremented according to Equation 4, followed by a loss of conductance via a simple exponential decay (Equation 5).

Both populations were interconnected, with one population presenting as excitatory and one as inhibitory. For simplicity, only the excitatory neurons received external input. First, we constrained the excitatory portion (n=500) of the network to our biologically realistic CRH_PVN neurons. To do so, we used previously fit parameters from the single neuron model as described in Single neuron model fit. Next, the remaining inhibitory population was constrained with parameters, reflecting the tonic spiking GABAergic neurons of the thalamus (Destexhe, 2009). Changes in adaptation current $b$ were modeled with instantaneous increment followed by an exponential decay function (Equation 6), reflecting the slow change in intracellular parameters.

\begin{array}{ll} τ_{b} b_{i n c} = - b_{i n c} \\ b \to b_{i n t i a l} + b_{i n c} \end{array}

For each synapse, the probability of release $p$ was modelled by the logistic differential equation

\begin{array}{ll} \frac{d p}{d t} = \frac{p}{τ_{p}} (1 - p) \\ p \to 0.1 \end{array}

Based on previous literature, the excitatory input of CRH_PVN neurons to the GABAergic population was modeled to be mediated by the CRHR1 receptor (Jiang et al., 2018; Jiang et al., 2019; Ramot et al., 2017). Therefore, this synapse was modeled by a slow-growing alpha function, a common model of neuromodulator function (Destexhe et al., 1994):

\begin{array}{ll} \frac{d [C R H]}{d t} = \frac{y - w_{C R H}}{τ_{C R H}} \\ \frac{d y}{d t} = \frac{- y}{τ_{C R H}} \\ y \to [C R H] + w_{C R H} \end{array}

The remaining free parameters $w_{e}$ , $w_{i}$ , $τ_{e}$ , $τ_{i}$ , $τ_{C R H}$ , and the EPSP input frequency were tuned with respect to the observed extracellular spike trains in vivo. To begin with, we selected a representative opto-tagged CRH neuron baseline recording as reference (Figure 2B). Next, we selected two characteristic (50 s) sections of the reference spike train that reflected bursting and single spiking independently. Then, several rounds of evolutionary algorithms were used to tune the parameters. For each round, the network was simulated with the given parameters with an increased adaptation current (2 × 50 s). Then, the distance between simulated and reference data was computed as the earth mover’s distance (Wasserstein metric) between the log-scaled ISI distributions (Olkin and Pukelsheim, 1982). The total distance was taken as the sum of the smallest 100 distances between each simulated unit and the reference. The standard adaptation current and the increased adaptation current simulations were compared to the bursting and tonic references, respectively.

Generation of network clamp input for ex vivo experiments

Request a detailed protocol

To investigate the ability of our model to induce burst in biological neurons ex vivo, we extracted the input currents from a model neuron in the network. First, a model neuron, isolated within the network, had its initial resting membrane potential V_mØ and leak reversal E_L constrained to the mean resting membrane potential of the CRH_PVN neurons in our recording conditions ex vivo (–67.9 mV, Table 1). Then, the model was simulated for 50 s, and the model neuron was allowed to respond to the incoming input freely. Every time-step (100 µs), the sum of the incoming excitatory, inhibitory, and adaptation currents were extracted following Equation 9:

I (t) = g_{e} (t) (E_{e} - v (t)) + g_{i} (t) (E_{i} - v (t)) - w (t)

where $g_{e}$ , $g_{i}$ , $v$ , and $w$ are extracted from the values of the individual neuron found in Equation 1 at given time step $t$ . In essence, this formulation allowed observation of the model neuron state without intervention. We theorized that inclusion of the computed leak current: $g_{L} (E_{L} - v)$ , was unnecessary as the ex vivo neurons express this intrinsically. Our network model generated very large synaptic currents (~500 pA), which is unlikely to reflect realistic inputs to CRH_PVN neurons observed in ex vivo preparation (Salter et al., 2018; Sterley et al., 2018; Wamsteeker Cusulin et al., 2013). Thus, we capped the maximum amplitude of PSCs to ±200 pA without changing the timings of PSPs.

Software

Simulations and modeling were completed using software packages within the Python 3.5 ecosystem. Single neuron and network simulations were completed using the Brian 2 (2.4) package (Stimberg et al., 2019). MSE, EMD, and other mathematical operations were computed using modified implementations of the functions available in NumPy (1.20.1) and SciPy (1.6.0) software packages (Harris et al., 2020; Virtanen et al., 2020). Backend probability inference was completed using the sbi (0.16.0) package (Gonçalves et al., 2020). Optimization of fits was performed using the nevergrad (0.4.3) package (Liu et al., 2020). Specifically, we utilized a custom portfolio consisting of two-points differential evolution (two-points DE), covariance matrix adaptation evolutionary strategy (CMA-ES), and a random search using scrambled Hammersley sequence (SCR-Hammersly). Axon binary file (ABF) recordings were imported to python using the pyABF package (Harden, 2020). Extraction of single-neuron action potentials and other electrophysiological features from intracellular recordings was performed using the ipfx (0.1.1) package (Gouwens et al., 2019). A generalized formulation of the fitting code, including the interface between Brian 2, nevergrad, and sbi are publicly available on GitHub (Ichiyama and Mestern, 2022). This includes the methodology for comparing EMD between spike trains in 1D and 2D using a sliced Wasserstein implementation.

Data availability

All data analyzed in this study are included in the manuscript, figures, and figure-supplement. Data analysis code and source code for figures is available at https://github.com/smestern/ichiyama_2022_code, (copy archived at swh:1:rev:d4d153e44882a0c2a7d0a1ef11193347aa5d6c96). The full list of model parameters are listed in Table 1. Figure source data files contain the numerical data used to generate figures.

References

1. Anthony TE
2. Dee N
3. Bernard A
4. Lerchner W
5. Heintz N
6. Anderson DJ
(2014) Control of stress-induced persistent anxiety by an extra-amygdala septohypothalamic circuit
Cell 156:522–536.
https://doi.org/10.1016/j.cell.2013.12.040
- PubMed
- Google Scholar
(2015) Stress-related synaptic plasticity in the hypothalamus
Nature Reviews. Neuroscience 16:377–388.
https://doi.org/10.1038/nrn3881
- PubMed
- Google Scholar
1. Biag J
2. Huang Y
3. Gou L
4. Hintiryan H
5. Askarinam A
6. Hahn JD
7. Toga AW
8. Dong HW
(2012) Cyto- and chemoarchitecture of the hypothalamic paraventricular nucleus in the C57BL/6J male mouse: a study of immunostaining and multiple fluorescent tract tracing
The Journal of Comparative Neurology 520:6–33.
https://doi.org/10.1002/cne.22698
- PubMed
- Google Scholar
(2019) Corticosterone mediated functional and structural plasticity in corticotropin-releasing hormone neurons
Neuropharmacology 154:79–86.
https://doi.org/10.1016/j.neuropharm.2019.02.017
- PubMed
- Google Scholar
(1996) Physiological mapping of local inhibitory inputs to the hypothalamic paraventricular nucleus
The Journal of Neuroscience 16:7151–7160.
https://doi.org/10.1523/JNEUROSCI.16-22-07151.1996
- PubMed
- Google Scholar
1. Brette R
2. Gerstner W
(2005) Adaptive exponential integrate-and-fire model as an effective description of neuronal activity
Journal of Neurophysiology 94:3637–3642.
https://doi.org/10.1152/jn.00686.2005
- PubMed
- Google Scholar
1. Brown CH
2. Bourque CW
(2006) Mechanisms of rhythmogenesis: insights from hypothalamic vasopressin neurons
Trends in Neurosciences 29:108–115.
https://doi.org/10.1016/j.tins.2005.11.005
- PubMed
- Google Scholar
1. Chen Y
2. Molet J
3. Gunn BG
4. Ressler K
5. Baram TZ
(2015) Diversity of Reporter Expression Patterns in Transgenic Mouse Lines Targeting Corticotropin-Releasing Hormone-Expressing Neurons
Endocrinology 156:4769–4780.
https://doi.org/10.1210/en.2015-1673
- PubMed
- Google Scholar
1. Cole RL
2. Sawchenko PE
(2002) Neurotransmitter Regulation of Cellular Activation and Neuropeptide Gene Expression in the Paraventricular Nucleus of the Hypothalamus
The Journal of Neuroscience 22:959–969.
https://doi.org/10.1523/JNEUROSCI.22-03-00959.2002
- PubMed
- Google Scholar
1. Cross BA
2. Silver IA
(1963) Unit activity in the hypothalamus and the sympathetic response to hypoxia and hypercapnia
Experimental Neurology 7:375–393.
https://doi.org/10.1016/0014-4886(63)90019-0
- PubMed
- Google Scholar
(2008) Functional role of local GABAergic influences on the HPA axis
Brain Structure and Function 213:63–72.
https://doi.org/10.1007/s00429-008-0192-2
- PubMed
- Google Scholar
1. Daviu N
2. Füzesi T
3. Rosenegger DG
4. Rasiah NP
5. Sterley TL
6. Peringod G
7. Bains JS
(2020) Paraventricular nucleus CRH neurons encode stress controllability and regulate defensive behavior selection
Nature Neuroscience 23:398–410.
https://doi.org/10.1038/s41593-020-0591-0
- PubMed
- Google Scholar
(1985) Noradrenergic Afferents Facilitate the Activity of Tuberoinfundibular Neurons of the Hypothalamic Paraventricular Nucleus
Neuroendocrinology 41:17–22.
https://doi.org/10.1159/000124148
- PubMed
- Google Scholar
(2005) Stress and the brain: from adaptation to disease
Nature Reviews Neuroscience 6:463–475.
https://doi.org/10.1038/nrn1683
- PubMed
- Google Scholar
(1994) Synthesis of models for excitable membranes, synaptic transmission and neuromodulation using a common kinetic formalism
Journal of Computational Neuroscience 1:195–230.
https://doi.org/10.1007/BF00961734
- PubMed
- Google Scholar
1. Destexhe A
2. Paré D
(1999) Impact of Network Activity on the Integrative Properties of Neocortical Pyramidal Neurons In Vivo
Journal of Neurophysiology 81:1531–1547.
https://doi.org/10.1152/jn.1999.81.4.1531
- PubMed
- Google Scholar
(2001) Fluctuating synaptic conductances recreate in vivo-like activity in neocortical neurons
Neuroscience 107:13–24.
https://doi.org/10.1016/S0306-4522(01)00344-X
- PubMed
- Google Scholar
(2003) The high-conductance state of neocortical neurons in vivo
Nature Reviews Neuroscience 4:739–751.
https://doi.org/10.1038/nrn1198
- PubMed
- Google Scholar
1. Destexhe A
(2009) Self-sustained asynchronous irregular states and Up–Down states in thalamic, cortical and thalamocortical networks of nonlinear integrate-and-fire neurons
Journal of Computational Neuroscience 27:493–506.
https://doi.org/10.1007/s10827-009-0164-4
- PubMed
- Google Scholar
Book
1. Destexhe A
2. Rudolph-Lilith M
(2012) Neuronal Noise
Boston, MA: Springer US.
https://doi.org/10.1007/978-0-387-79020-6
- Google Scholar
(1981) Effects of sciatic nerve stimulation on ACTH secretion, in intact and in variously hypothalamic deafferentated male rats
Experimental Brain Research 42–42:486–488.
https://doi.org/10.1007/BF00237514
- PubMed
- Google Scholar
1. Ferri CC
2. Ferguson AV
(2005) Prostaglandin E2 Mediates Cellular Effects of Interleukin-1beta on Parvocellular Neurones in the Paraventricular Nucleus of the Hypothalamus
Journal of Neuroendocrinology 17:498–508.
https://doi.org/10.1111/j.1365-2826.2005.01336.x
- PubMed
- Google Scholar
(2016) Hypothalamic CRH neurons orchestrate complex behaviours after stress
Nature Communications 7:11937.
https://doi.org/10.1038/ncomms11937
- PubMed
- Google Scholar
1. Gerstner W
2. Naud R
(2009) How Good Are Neuron Models?
Science 326:379–380.
https://doi.org/10.1126/science.1181936
- PubMed
- Google Scholar
1. Gold PW
2. Chrousos GP
(1999) The endocrinology of melancholic and atypical depression: relation to neurocircuitry and somatic consequences
Proceedings of the Association of American Physicians 111:22–34.
https://doi.org/10.1046/j.1525-1381.1999.09423.x
- PubMed
- Google Scholar
(2020) Training deep neural density estimators to identify mechanistic models of neural dynamics
eLife 9:e56261.
https://doi.org/10.7554/eLife.56261
- PubMed
- Google Scholar
1. Gouwens NW
2. Sorensen SA
3. Berg J
4. Lee C
5. Jarsky T
6. Ting J
7. Sunkin SM
8. Feng D
9. Anastassiou CA
10. Barkan E
11. Bickley K
12. Blesie N
13. Braun T
14. Brouner K
15. Budzillo A
16. Caldejon S
17. Casper T
18. Castelli D
19. Chong P
20. Crichton K
21. Cuhaciyan C
22. Daigle TL
23. Dalley R
24. Dee N
25. Desta T
26. Ding S-L
27. Dingman S
28. Doperalski A
29. Dotson N
30. Egdorf T
31. Fisher M
32. de Frates RA
33. Garren E
34. Garwood M
35. Gary A
36. Gaudreault N
37. Godfrey K
38. Gorham M
39. Gu H
40. Habel C
41. Hadley K
42. Harrington J
43. Harris JA
44. Henry A
45. Hill D
46. Josephsen S
47. Kebede S
48. Kim L
49. Kroll M
50. Lee B
51. Lemon T
52. Link KE
53. Liu X
54. Long B
55. Mann R
56. McGraw M
57. Mihalas S
58. Mukora A
59. Murphy GJ
60. Ng L
61. Ngo K
62. Nguyen TN
63. Nicovich PR
64. Oldre A
65. Park D
66. Parry S
67. Perkins J
68. Potekhina L
69. Reid D
70. Robertson M
71. Sandman D
72. Schroedter M
73. Slaughterbeck C
74. Soler-Llavina G
75. Sulc J
76. Szafer A
77. Tasic B
78. Taskin N
79. Teeter C
80. Thatra N
81. Tung H
82. Wakeman W
83. Williams G
84. Young R
85. Zhou Z
86. Farrell C
87. Peng H
88. Hawrylycz MJ
89. Lein E
90. Ng L
91. Arkhipov A
92. Bernard A
93. Phillips JW
94. Zeng H
95. Koch C
(2019) Classification of electrophysiological and morphological neuron types in the mouse visual cortex
Nature Neuroscience 22:1182–1195.
https://doi.org/10.1038/s41593-019-0417-0
- PubMed
- Google Scholar
(1986) Parvocellular neurosecretory neurons : Converging inputs after saphenous nerve and hypovolemic stimulations in the rat
The Japanese Journal of Physiology 36:921–933.
https://doi.org/10.2170/jjphysiol.36.921
- Google Scholar
(1984) The effect of urethane on pituitary-adrenal function of female rats
Acta Endocrinologica 106:362–367.
https://doi.org/10.1530/acta.0.1060362
- Google Scholar
1. Han SK
2. Chong W
3. Li LH
4. Lee IS
5. Murase K
6. Ryu PD
(2002) Noradrenaline excites and inhibits GABAergic transmission in parvocellular neurons of rat hypothalamic paraventricular nucleus
Journal of Neurophysiology 87:2287–2296.
https://doi.org/10.1152/jn.2002.87.5.2287
- PubMed
- Google Scholar
Software
1. Harden SW
(2020) pyabf: Python library for reading files in Axon Binary Format (ABF
PyPI.

https://swharden.com/pyabf/
(2001) Temporal interaction between single spikes and complex spike bursts in hippocampal pyramidal cells
Neuron 32:141–149.
https://doi.org/10.1016/s0896-6273(01)00447-0
- PubMed
- Google Scholar
1. Harris CR
2. Millman KJ
3. van der Walt SJ
4. Gommers R
5. Virtanen P
6. Cournapeau D
7. Wieser E
8. Taylor J
9. Berg S
10. Smith NJ
11. Kern R
12. Picus M
13. Hoyer S
14. van Kerkwijk MH
15. Brett M
16. Haldane A
17. Del Río JF
18. Wiebe M
19. Peterson P
20. Gérard-Marchant P
21. Sheppard K
22. Reddy T
23. Weckesser W
24. Abbasi H
25. Gohlke C
26. Oliphant TE
(2020) Array programming with NumPy
Nature 585:357–362.
https://doi.org/10.1038/s41586-020-2649-2
- PubMed
- Google Scholar
1. Herman JP
2. Tasker JG
(2016) Paraventricular Hypothalamic Mechanisms of Chronic Stress Adaptation
Frontiers in Endocrinology 7:137.
https://doi.org/10.3389/fendo.2016.00137
- PubMed
- Google Scholar
(2006) Modeling single-neuron dynamics and computations: A balance of detail and abstraction
Science 314:80–85.
https://doi.org/10.1126/science.1127240
- PubMed
- Google Scholar
1. Hewitt SA
2. Bains JS
(2006) Brain-derived neurotrophic factor silences GABA synapses onto hypothalamic neuroendocrine cells through a postsynaptic dynamin-mediated mechanism
Journal of Neurophysiology 95:2193–2198.
https://doi.org/10.1152/jn.01135.2005
- PubMed
- Google Scholar
(2009) Altered chloride homeostasis removes synaptic inhibitory constraint of the stress axis
Nature Neuroscience 12:438–443.
https://doi.org/10.1038/nn.2274
- PubMed
- Google Scholar
Software
1. Ichiyama A
2. Mestern S
(2022) ichiyama_2022_code, version swh:1:rev:d4d153e44882a0c2a7d0a1ef11193347aa5d6c96
Software Heritage.

https://archive.softwareheritage.org/swh:1:dir:0ecc3d15e1b05b9ffa1667cbf7d10af56369abd6;origin=https://github.com/smestern/ichiyama_2022_code;visit=swh:1:snp:c7fa3278248382328e2ea694f874b4360132595a;anchor=swh:1:rev:d4d153e44882a0c2a7d0a1ef11193347aa5d6c96
1. Inoue W
2. Bains JS
(2014) Beyond inhibition: GABA synapses tune the neuroendocrine stress axis
BioEssays 36:561–569.
https://doi.org/10.1002/bies.201300178
- PubMed
- Google Scholar
1. Itoi K
2. Suda T
3. Tozawa F
4. Dobashi I
5. Ohmori N
6. Sakai Y
7. Abe K
8. Demura H
(1994) Microinjection of norepinephrine into the paraventricular nucleus of the hypothalamus stimulates corticotropin-releasing factor gene expression in conscious rats
Endocrinology 135:2177–2182.
https://doi.org/10.1210/endo.135.5.7956940
- PubMed
- Google Scholar
1. Itoi K
2. Jiang Y-Q
3. Iwasaki Y
4. Watson SJ
(2004) Regulatory mechanisms of corticotropin-releasing hormone and vasopressin gene expression in the hypothalamus
Journal of Neuroendocrinology 16:348–355.
https://doi.org/10.1111/j.0953-8194.2004.01172.x
- PubMed
- Google Scholar
1. Izhikevich EM
(2003) Simple model of spiking neurons
IEEE Transactions on Neural Networks 14:1569–1572.
https://doi.org/10.1109/TNN.2003.820440
- PubMed
- Google Scholar
(2018) Local Corticotropin-Releasing Factor Signaling in the Hypothalamic Paraventricular Nucleus
The Journal of Neuroscience 38:1874–1890.
https://doi.org/10.1523/JNEUROSCI.1492-17.2017
- PubMed
- Google Scholar
(2019) CRF signaling between neurons in the paraventricular nucleus of the hypothalamus (PVN) coordinates stress responses
Neurobiology of Stress 11:100192.
https://doi.org/10.1016/j.ynstr.2019.100192
- PubMed
- Google Scholar
1. Joëls M
2. Baram TZ
(2009) The neuro-symphony of stress
Nature Reviews. Neuroscience 10:459–466.
https://doi.org/10.1038/nrn2632
- PubMed
- Google Scholar
(2019) Prefrontal-Bed Nucleus Circuit Modulation of a Passive Coping Response Set
The Journal of Neuroscience 39:1405–1419.
https://doi.org/10.1523/JNEUROSCI.1421-18.2018
- PubMed
- Google Scholar
1. Kannan H
2. Kasai M
3. Osaka T
4. Yamashita H
(1987) Neurons in the paraventricular nucleus projecting to the median eminence: a study of their afferent connections from peripheral baroreceptors, and from the A1-catecholaminergic area in the ventrolateral medulla
Brain Research 409:358–363.
https://doi.org/10.1016/0006-8993(87)90722-0
- PubMed
- Google Scholar
(2011) MAP kinases couple hindbrain-derived catecholamine signals to hypothalamic adrenocortical control mechanisms during glycemia-related challenges
The Journal of Neuroscience 31:18479–18491.
https://doi.org/10.1523/JNEUROSCI.4785-11.2011
- PubMed
- Google Scholar
(2018) Prostaglandin E₂ depresses GABA release onto parvocellular neuroendocrine neurones in the paraventricular nucleus of the hypothalamus via presynaptic receptors
Journal of Neuroendocrinology 30:e12638.
https://doi.org/10.1111/jne.12638
- PubMed
- Google Scholar
(2019a) Stress experience and hormone feedback tune distinct components of hypothalamic CRH neuron activity
Nature Communications 10:5696.
https://doi.org/10.1038/s41467-019-13639-8
- PubMed
- Google Scholar
1. Kim J
2. Lee S
3. Fang YY
4. Shin A
5. Park S
6. Hashikawa K
7. Bhat S
8. Kim D
9. Sohn JW
10. Lin D
11. Suh GSB
(2019b) Rapid, biphasic CRF neuronal responses encode positive and negative valence
Nature Neuroscience 22:576–585.
https://doi.org/10.1038/s41593-019-0342-2
- PubMed
- Google Scholar
(2004) GABAergic Mechanisms Constraining the Activity of the Hypothalamo-Pituitary-Adrenocortical Axis
Annals of the New York Academy of Sciences 1018:466–476.
https://doi.org/10.1196/annals.1296.057
- PubMed
- Google Scholar
1. Krahe R
2. Gabbiani F
(2004) Burst firing in sensory systems
Nature Reviews Neuroscience 5:13–23.
https://doi.org/10.1038/nrn1296
- PubMed
- Google Scholar
1. Leng G
2. MacGregor DJ
(2018) Models in neuroendocrinology
Mathematical Biosciences 305:29–41.
https://doi.org/10.1016/j.mbs.2018.07.008
- PubMed
- Google Scholar
1. Levy BH
2. Tasker JG
(2012) Synaptic regulation of the hypothalamic-pituitary-adrenal axis and its modulation by glucocorticoids and stress
Frontiers in Cellular Neuroscience 6:24.
https://doi.org/10.3389/fncel.2012.00024
- PubMed
- Google Scholar
(2020) Hypothalamic circuitry underlying stress-induced insomnia and peripheral immunosuppression
Science Advances 6:eabc2590.
https://doi.org/10.1126/sciadv.abc2590
- PubMed
- Google Scholar
1. Lightman SL
(2008) The neuroendocrinology of stress: A never ending story
Journal of Neuroendocrinology 20:880–884.
https://doi.org/10.1111/j.1365-2826.2008.01711.x
- PubMed
- Google Scholar
(2009) PINP: A new method of tagging neuronal populations for identification during in vivo electrophysiological recording
PLOS ONE 4:e6099.
https://doi.org/10.1371/journal.pone.0006099
- PubMed
- Google Scholar
1. Lisman JE
(1997) Bursts as a unit of neural information: making unreliable synapses reliable
Trends in Neurosciences 20:38–43.
https://doi.org/10.1016/S0166-2236(96)10070-9
- PubMed
- Google Scholar
Preprint
1. Liu J
2. Moreau A
3. Preuss M
4. Rapin J
5. Roziere B
6. Teytaud F
7. Teytaud O
(2020) Versatile Black-Box Optimization
arXiv.

https://arxiv.org/abs/2004.14014
- Google Scholar
1. Luther JA
2. Daftary SS
3. Boudaba C
4. Gould GC
5. Halmos KC
6. Tasker JG
(2002) Neurosecretory and non-neurosecretory parvocellular neurones of the hypothalamic paraventricular nucleus express distinct electrophysiological properties
Journal of Neuroendocrinology 14:929–932.
https://doi.org/10.1046/j.1365-2826.2002.00867.x
- PubMed
- Google Scholar
1. Lynch EP
2. Houghton CJ
(2015) Parameter estimation of neuron models using in-vitro and in-vivo electrophysiological data
Frontiers in Neuroinformatics 9:10.
https://doi.org/10.3389/fninf.2015.00010
- PubMed
- Google Scholar
1. Maggi CA
2. Meli A
(1986) Suitability of urethane anesthesia for physiopharmacological investigations in various systems: Part 1: General considerations
Experientia 42:109–114.
https://doi.org/10.1007/BF01952426
- PubMed
- Google Scholar
1. Marder E
2. Goaillard JM
(2006) Variability, compensation and homeostasis in neuron and network function
Nature Reviews Neuroscience 7:563–574.
https://doi.org/10.1038/nrn1949
- PubMed
- Google Scholar
1. Markram H
2. Muller E
3. Ramaswamy S
4. Reimann MW
5. Abdellah M
6. Sanchez CA
7. Ailamaki A
8. Alonso-Nanclares L
9. Antille N
10. Arsever S
11. Kahou GAA
12. Berger TK
13. Bilgili A
14. Buncic N
15. Chalimourda A
16. Chindemi G
17. Courcol J-D
18. Delalondre F
19. Delattre V
20. Druckmann S
21. Dumusc R
22. Dynes J
23. Eilemann S
24. Gal E
25. Gevaert ME
26. Ghobril J-P
27. Gidon A
28. Graham JW
29. Gupta A
30. Haenel V
31. Hay E
32. Heinis T
33. Hernando JB
34. Hines M
35. Kanari L
36. Keller D
37. Kenyon J
38. Khazen G
39. Kim Y
40. King JG
41. Kisvarday Z
42. Kumbhar P
43. Lasserre S
44. Le Bé J-V
45. Magalhães BRC
46. Merchán-Pérez A
47. Meystre J
48. Morrice BR
49. Muller J
50. Muñoz-Céspedes A
51. Muralidhar S
52. Muthurasa K
53. Nachbaur D
54. Newton TH
55. Nolte M
56. Ovcharenko A
57. Palacios J
58. Pastor L
59. Perin R
60. Ranjan R
61. Riachi I
62. Rodríguez J-R
63. Riquelme JL
64. Rössert C
65. Sfyrakis K
66. Shi Y
67. Shillcock JC
68. Silberberg G
69. Silva R
70. Tauheed F
71. Telefont M
72. Toledo-Rodriguez M
73. Tränkler T
74. Van Geit W
75. Díaz JV
76. Walker R
77. Wang Y
78. Zaninetta SM
79. DeFelipe J
80. Hill SL
81. Segev I
82. Schürmann F
(2015) Reconstruction and Simulation of Neocortical Microcircuitry
Cell 163:456–492.
https://doi.org/10.1016/j.cell.2015.09.029
- PubMed
- Google Scholar
1. Matovic S
2. Ichiyama A
3. Igarashi H
4. Salter EW
5. Sunstrum JK
6. Wang XF
7. Henry M
8. Kuebler ES
9. Vernoux N
10. Martinez‐Trujillo J
11. Tremblay M
12. Inoue W
(2020) Neuronal hypertrophy dampens neuronal intrinsic excitability and stress responsiveness during chronic stress
The Journal of Physiology 598:2757–2773.
https://doi.org/10.1113/JP279666
- PubMed
- Google Scholar
1. McCormick DA
2. Feeser HR
(1990) Functional implications of burst firing and single spike activity in lateral geniculate relay neurons
Neuroscience 39:103–113.
https://doi.org/10.1016/0306-4522(90)90225-S
- PubMed
- Google Scholar
1. McCormick DA
2. Pape HC
(1990) Properties of a hyperpolarization-activated cation current and its role in rhythmic oscillation in thalamic relay neurones
The Journal of Physiology 431:291–318.
https://doi.org/10.1113/jphysiol.1990.sp018331
- PubMed
- Google Scholar
1. McEwen BS
(2005) Glucocorticoids, depression, and mood disorders: structural remodeling in the brain
Metabolism 54:20–23.
https://doi.org/10.1016/j.metabol.2005.01.008
- PubMed
- Google Scholar
1. Olkin I
2. Pukelsheim F
(1982) The distance between two random vectors with given dispersion matrices
Linear Algebra and Its Applications 48:257–263.
https://doi.org/10.1016/0024-3795(82)90112-4
- Google Scholar
1. Ono D
2. Mukai Y
3. Hung CJ
4. Chowdhury S
5. Sugiyama T
6. Yamanaka A
(2020) The mammalian circadian pacemaker regulates wakefulness via CRF neurons in the paraventricular nucleus of the hypothalamus
Science Advances 6:eabd0384.
https://doi.org/10.1126/sciadv.abd0384
- PubMed
- Google Scholar
1. Pape HC
2. McCormick DA
(1989) Noradrenaline and serotonin selectively modulate thalamic burst firing by enhancing a hyperpolarization-activated cation current
Nature 340:715–718.
https://doi.org/10.1038/340715a0
- PubMed
- Google Scholar
1. Poulain DA
2. Wakerley JB
(1982) Electrophysiology of hypothalamic magnocellular neurones secreting oxytocin and vasopressin
Neuroscience 7:773–808.
https://doi.org/10.1016/0306-4522(82)90044-6
- PubMed
- Google Scholar
1. Ramot A
2. Jiang Z
3. Tian JB
4. Nahum T
5. Kuperman Y
6. Justice N
7. Chen A
(2017) Hypothalamic CRFR1 is essential for HPA axis regulation following chronic stress
Nature Neuroscience 20:385–388.
https://doi.org/10.1038/nn.4491
- PubMed
- Google Scholar
1. Reyes AD
(2003) Synchrony-dependent propagation of firing rate in iteratively constructed networks in vitro
Nature Neuroscience 6:593–599.
https://doi.org/10.1038/nn1056
- PubMed
- Google Scholar
1. Rho JH
2. Swanson LW
(1989)
A morphometric analysis of functionally defined subpopulations of neurons in the paraventricular nucleus of the rat with observations on the effects of colchicine

The Journal of Neuroscience 9:1375–1388.
- PubMed
- Google Scholar
1. Roland BL
2. Sawchenko PE
(1993) Local origins of some GABAergic projections to the paraventricular and supraoptic nuclei of the hypothalamus in the rat
The Journal of Comparative Neurology 332:123–143.
https://doi.org/10.1002/cne.903320109
- PubMed
- Google Scholar
1. Rothman JS
2. Silver RA
(2014) Data-driven modeling of synaptic transmission and integration
Progress in Molecular Biology and Translational Science 123:305–350.
https://doi.org/10.1016/B978-0-12-397897-4.00004-8
- PubMed
- Google Scholar
1. Rudolph M
2. Destexhe A
(2003)
A fast-conducting, stochastic integrative mode for neocortical neurons in vivo

The Journal of Neuroscience 23:2466–2476.
- PubMed
- Google Scholar
1. Rudolph M
2. Destexhe A
(2006) A multichannel shot noise approach to describe synaptic background activity in neurons
The European Physical Journal B 52:125–132.
https://doi.org/10.1140/epjb/e2006-00261-3
- Google Scholar
1. Rudolph M
2. Destexhe A
(2007) How much can we trust neural simulation strategies?
Neurocomputing 70:1966–1969.
https://doi.org/10.1016/j.neucom.2006.10.138
- Google Scholar
(2018) Chronic stress dampens excitatory synaptic gain in the paraventricular nucleus of the hypothalamus
The Journal of Physiology 596:4157–4172.
https://doi.org/10.1113/JP275669
- PubMed
- Google Scholar
1. Saphier D
2. Feldman S
(1985) Effects of neural stimuli on paraventricular nucleus neurones
Brain Research Bulletin 14:401–407.
https://doi.org/10.1016/0361-9230(85)90016-4
- PubMed
- Google Scholar
1. Saphier D
2. Feldman S
(1988) Iontophoretic application of glucocorticoids inhibits identified neurones in the rat paraventricular nucleus
Brain Research 453:183–190.
https://doi.org/10.1016/0006-8993(88)90157-6
- PubMed
- Google Scholar
1. Saphier D
(1989) Catecholaminergic projections to tuberoinfundibular neurones of the paraventricular nucleus: I. Effects of stimulation of A1, A2, A6 and C2 cell groups
Brain Research Bulletin 23:389–395.
https://doi.org/10.1016/0361-9230(89)90179-2
- PubMed
- Google Scholar
1. Saphier D
2. Feldman S
(1990) Iontophoresis of cortisol inhibits responses of identified paraventricular nucleus neurones to sciatic nerve stimulation
Brain Research 535:159–162.
https://doi.org/10.1016/0006-8993(90)91836-6
- PubMed
- Google Scholar
1. Saphier D
2. Feldman S
(1991) Catecholaminergic projections to tuberoinfundibular neurones of the paraventricular nucleus: III. Effects of adrenoceptor agonists and antagonists
Brain Research Bulletin 26:863–870.
https://doi.org/10.1016/0361-9230(91)90250-n
- PubMed
- Google Scholar
(2011) Neurosteroidogenesis is required for the physiological response to stress: role of neurosteroid-sensitive GABAA receptors
The Journal of Neuroscience 31:18198–18210.
https://doi.org/10.1523/JNEUROSCI.2560-11.2011
- PubMed
- Google Scholar
1. Sherman SM
(2001) Tonic and burst firing: dual modes of thalamocortical relay
Trends in Neurosciences 24:122–126.
https://doi.org/10.1016/s0166-2236(00)01714-8
- PubMed
- Google Scholar
1. Shimokawa A
2. Jin QH
3. Ishizuka Y
4. Kunitake T
5. Takasaki M
6. Kannan H
(1998a) Effects of anesthetics on norepinephrine release in the hypothalamic paraventricular nucleus region of awake rats
Neuroscience Letters 244:21–24.
https://doi.org/10.1016/s0304-3940(98)00119-0
- PubMed
- Google Scholar
(1998b) Differential effects of anesthetics on sympathetic nerve activity and arterial baroreceptor reflex in chronically instrumented rats
Journal of the Autonomic Nervous System 72:46–54.
https://doi.org/10.1016/s0165-1838(98)00084-8
- PubMed
- Google Scholar
1. Simmons DM
2. Swanson LW
(2009) Comparison of the spatial distribution of seven types of neuroendocrine neurons in the rat paraventricular nucleus: toward a global 3D model
The Journal of Comparative Neurology 516:423–441.
https://doi.org/10.1002/cne.22126
- PubMed
- Google Scholar
(1993) Thalamocortical oscillations in the sleeping and aroused brain
Science 262:679–685.
https://doi.org/10.1126/science.8235588
- PubMed
- Google Scholar
(2018) Social transmission and buffering of synaptic changes after stress
Nature Neuroscience 21:393–403.
https://doi.org/10.1038/s41593-017-0044-6
- PubMed
- Google Scholar
(2019) Brian 2, an intuitive and efficient neural simulator
eLife 8:e47314.
https://doi.org/10.7554/eLife.47314
- PubMed
- Google Scholar
1. Swanson LW
2. Sawchenko PE
(1980) Paraventricular Nucleus:A Site for the Integration of Neuroendocrine and Autonomic Mechanisms
Neuroendocrinology 31:410–417.
https://doi.org/10.1159/000123111
- PubMed
- Google Scholar
1. Ulrich-Lai YM
2. Herman JP
(2009) Neural regulation of endocrine and autonomic stress responses
Nature Reviews Neuroscience 10:397–409.
https://doi.org/10.1038/nrn2647
- PubMed
- Google Scholar
1. Virtanen P
2. Gommers R
3. Oliphant TE
4. Haberland M
5. Reddy T
6. Cournapeau D
7. Burovski E
8. Peterson P
9. Weckesser W
10. Bright J
11. van der Walt SJ
12. Brett M
13. Wilson J
14. Millman KJ
15. Mayorov N
16. Nelson ARJ
17. Jones E
18. Kern R
19. Larson E
20. Carey CJ
21. Polat İ
22. Feng Y
23. Moore EW
24. VanderPlas J
25. Laxalde D
26. Perktold J
27. Cimrman R
28. Henriksen I
29. Quintero EA
30. Harris CR
31. Archibald AM
32. Ribeiro AH
33. Pedregosa F
34. van Mulbregt P
35. SciPy 1.0 Contributors
36. Vijaykumar A
37. Bardelli AP
38. Rothberg A
39. Hilboll A
40. Kloeckner A
41. Scopatz A
42. Lee A
43. Rokem A
44. Woods CN
45. Fulton C
46. Masson C
47. Häggström C
48. Fitzgerald C
49. Nicholson DA
50. Hagen DR
51. Pasechnik DV
52. Olivetti E
53. Martin E
54. Wieser E
55. Silva F
56. Lenders F
57. Wilhelm F
58. Young G
59. Price GA
60. Ingold G-L
61. Allen GE
62. Lee GR
63. Audren H
64. Probst I
65. Dietrich JP
66. Silterra J
67. Webber JT
68. Slavič J
69. Nothman J
70. Buchner J
71. Kulick J
72. Schönberger JL
73. de Miranda Cardoso JV
74. Reimer J
75. Harrington J
76. Rodríguez JLC
77. Nunez-Iglesias J
78. Kuczynski J
79. Tritz K
80. Thoma M
81. Newville M
82. Kümmerer M
83. Bolingbroke M
84. Tartre M
85. Pak M
86. Smith NJ
87. Nowaczyk N
88. Shebanov N
89. Pavlyk O
90. Brodtkorb PA
91. Lee P
92. McGibbon RT
93. Feldbauer R
94. Lewis S
95. Tygier S
96. Sievert S
97. Vigna S
98. Peterson S
99. More S
100. Pudlik T
101. Oshima T
102. Pingel TJ
103. Robitaille TP
104. Spura T
105. Jones TR
106. Cera T
107. Leslie T
108. Zito T
109. Krauss T
110. Upadhyay U
111. Halchenko YO
112. Vázquez-Baeza Y
(2020) SciPy 1.0: fundamental algorithms for scientific computing in Python
Nature Methods 17:261–272.
https://doi.org/10.1038/s41592-019-0686-2
- PubMed
- Google Scholar
(2016) The Severity of Acute Stress Is Represented by Increased Synchronous Activity and Recruitment of Hypothalamic CRH Neurons
The Journal of Neuroscience 36:3350–3362.
https://doi.org/10.1523/JNEUROSCI.3390-15.2016
- PubMed
- Google Scholar
(2013) Characterization of corticotropin-releasing hormone neurons in the paraventricular nucleus of the hypothalamus of Crh-IRES-Cre mutant mice
PLOS ONE 8:e64943.
https://doi.org/10.1371/journal.pone.0064943
- PubMed
- Google Scholar
1. Watanabe S
2. Kunitake T
3. Kato K
4. Chu CP
5. Nakao H
6. Qiu DL
7. Kannan H
(2004) Single-unit activity of paraventricular nucleus neurons in response to intero- and exteroceptive stressors in conscious, freely moving rats
Brain Research 995:97–108.
https://doi.org/10.1016/j.brainres.2003.09.058
- PubMed
- Google Scholar
1. Yuan Y
2. Wu W
3. Chen M
4. Cai F
5. Fan C
6. Shen W
7. Sun W
8. Hu J
(2019) Reward Inhibits Paraventricular CRH Neurons to Relieve Stress
Current Biology 29:1243–1251.
https://doi.org/10.1016/j.cub.2019.02.048
- PubMed
- Google Scholar
1. Zerlaut Y
2. Teleńczuk B
3. Deleuze C
4. Bal T
5. Ouanounou G
6. Destexhe A
(2016) Heterogeneous firing rate response of mouse layer V pyramidal neurons in the fluctuation-driven regime
The Journal of Physiology 594:3791–3808.
https://doi.org/10.1113/JP272317
- PubMed
- Google Scholar

Decision letter

Julie A Kauer

Reviewing Editor; Stanford University, United States
John R Huguenard

Senior Editor; Stanford University School of Medicine, United States
Akihiro Yamanaka

Reviewer; Research Institute of Environmental Medicine, Nagoya University, Japan
Jaideep Singh Bains

Reviewer; University of Calgary, Canada

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "State-dependent activity dynamics of hypothalamic stress effector neurons" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by John Huguenard as the Senior Editor. The following individuals involved in the review of your submission have agreed to reveal their identity: Akihiro Yamanaka (Reviewer #2); Jaideep Singh Bains (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) If the authors have data on what happens either in vivo or in silico if they stimulate the sciatic nerve for long periods, it would improve the manuscript significantly.

2) Please add a discussion of the Yuan et al. 2019 paper which has opposite results in slices after a stressful experience.

3) Please add a thoughtful discussion of the likely effects of the anesthetic on cell firing properties and of bursting more generally.

4) Please discuss the possible mechanism requiring the long silent period between bursts and the role of H-currents, as well as likely contributions of peptide neuromodulators vs. fast neurotransmitters likely contained in the same neurons.

Reviewer #1 (Recommendations for the authors):

1. Previous work from thalamic neurons has demonstrated how burst firing can be increased with hyperpolarization, and this can be accompanied by lower firing rates than in the "up-state" in which the cell fires tonically. Do the authors think that a similar modulation may be occurring in the CRH neurons of PVN as well, as many PVN cells express these channels (e.g. Luther et al., 2002, Luther and Tasker, 2000)? Does their model provide any insights into this question? While the decrease in GABAergic inhibition and increased spike attenuation can lead to "stress-like" firing outputs, might a variety of different mechanisms that lead to hyperpolarization also underlie the actual change in vivo?

2. Work by Yuan et al. 2019, cited here, is not really discussed in the context of stressors. The first page of the discussion argues that burst firing in this cell population has not been characterized, but Yuan et al. have indeed reported it. This earlier work indicates that state changes in CRH cells have been previously identified, but muddies the waters as to how these in vitro data can be reconciled with the present work. Yuan et al. used an ex vivo preparation from control or stressed animals and report that the CRH cells also can show burst behaviors after specific stressful stimuli in vivo. However, the neurons are tonically firing from control animals and bursting when in stressed animals. How do the authors think this work fits into their observations in vivo?

3. Why do the graphs for Figure 3G and Figure 3I look nearly identical? This figure may suggest that the burst behavior hardly contributes to the overall firing rate at all, even under basal conditions.

Reviewer #2 (Recommendations for the authors):

1) Although computational remodeling well recreates in vitro-like switch between RB and SS, it is not clear as remodeling parameters the amount of contribution of fast neurotransmitters such as GABA and glutamate and slow neurotransmitter such as neuropeptide, CRH. CRH neurons would release both fast neurotransmitter glutamate and slow and sustained neurotransmitter CRH. Does the computational remodeling consider these parameters?

2) CRH release probability from synaptic vesicle might be different between RB and SS. It is not clear if the authors consider this difference in their computational remodeling.

3) Sciatic nerve stimulation-induced transient firing pattern change from RB and SS around 2-3 sec. However, intrinsic state change representatively shown in Figure 3J looks much slower and lasted longer around ~100 sec. Could the authors reproduce a continuous increase of firing by changing the firing pattern from RB to SS if the sciatic nerve stimulation is repetitively applied for a longer period around 100 sec in vivo or in silico?

4) The authors stated that high-frequency burst firing requires a preceding prolonged silence period. It is not clear what neural mechanism is involved in prolonged silence periods in vivo or in silico.

5) H-current through hyperpolarization-activated cyclic nucleotide-gated (HCN) channels play an important role in the generation of rhythmic burst firing. Do the authors record or consider changing of H-current in the activity state switch between RB and SS.

Reviewer #3 (Recommendations for the authors):

1) The neurons that were sensitive to blue light ("light-responsive") had a latency of approximately 7 ms. This is sufficient time for action potential in one cell, release of glutamate and an action potential in another cell. Can the authors provide assurances that blue light is directly activating the cell they are recording? In brain slices, light pulses elicit almost simultaneous action potentials, so the delay here needs to be addressed.

2) What is the advantage of a rhythmic burst mechanism during periods when CRH neuron activity is not required for hormone release? The authors offer one sentence in the discussion, but more is warranted here.

3) There needs to be some discussion on the possible contributions of the anesthetic, urethane on the properties described here. This is an important limitation of the study, as it may increase inhibitory tone in the system and needs to be considered.

https://doi.org/10.7554/eLife.76832.sa1

Author response

Essential revisions:

1) If the authors have data on what happens either in vivo or in silico if they stimulate the sciatic nerve for long periods, it would improve the manuscript significantly.

We thank the reviewer for this comment, which helps us to strengthen our proposal that the firing change from RB to SS underlies both transient and more sustained firing rate increase of CRH_PVN neurons.

In a pilot in vivo recording, we applied four different stimulation intensities (0.5, 1, 1.5, and 2 mA) of a brief nerve stimulation that was repeated over multiple trials (5 pulses at 20 Hz, every 15 s): the stimulation pattern is identical to what we reported in Figure 3. Notably, at the highest intensity (2 mA), the firing activity did not completely return to the baseline within the trial duration of 15 s, resulting in a continuous decrease of RB and an increase of SS during the stimulation period. These results demonstrate that nerve stimulation (with high intensity and/or short trial intervals) can result in a continuous change from RB to SS, suggesting that common mechanisms underlie both transient and long-lasting change from RB to SS in vivo. In the current study, we focused to study the transient changes triggered by a lower intensity sciatic nerve stimulation, and multiple trials were averaged to analyse firing activity changes.

We added these results as Figure 3—figure supplement in the revised manuscript. We also added the following sentences in the corresponding Results section (Line 185-189).

“In a representative case, we found that the firing pattern change lasted longer with higher intensity stimulation. Figure 3—figure supplement shows that, with the highest intensity stimulation (2 mA), the firing activity change did not completely return to the baseline within the trial duration, resulting in continuous decrease of RB and increase of SS during the stimulation period.”

In silico, our model can produce continuous firing pattern changes from RB to SS by persistently lowering GABAergic inputs to CRH neurons and increasing spike-triggered adaptation current in CRH neurons (Figure 5K). In Figure 6O, we also recreated in vivo-like (smooth) transition between RB and SS by transiently changing the same two parameters. Thus, our model results indicate that the sciatic nerve stimulation in vivo triggered transient and reversible changes in neuronal mechanisms which were modelled by these parameters. Our model is consistent with the observation that repetitive (and/or high intensity) sciatic nerve stimulation in vivo produced long-lasting changes in the firing pattern from RB to SS.

2) Please add a discussion of the Yuan et al. 2019 paper which has opposite results in slices after a stressful experience.

We thank the reviewers for pointing out the importance of discussing “burst firing” in slices reported by Yuan et al. (Yuan et al., 2019). To clarify the relationship between our new data and the observation made by Yuan et al. (2019), we added the following section in the Discussion of the revised manuscript (Line 609-619).

“Interestingly, Yuan et al. recently reported spontaneous “burst firing” of CRH_PVN neurons ex vivo using cell-attached recordings in slices prepared from repeatedly stressed mice (Yuan et al., 2019). It should be noted, however, that the intra-burst frequency of this ex vivo “burst firing” was <100 Hz (averaging around 50 Hz), which was different from the in vivo RB (>100 Hz intra-burst frequency) we report here in non-stressed mice. Further, the ex vivo “burst firing” involved stress-induced plasticity that developed over days because it only became evident after 3 daily foot shock, but was absent when mice were not previously stressed, or even after 2 daily foot shock. Thus, our data and those by Yuan et al., (2019) are consistent in that CRH_PVN neurons do not generate burst firing (and primarily fire in single spiking (SS)) in slices from previously unstressed mice. Accumulating evidence show that CRH_PVN neurons undergo diverse forms of synaptic and intrinsic plasticity in response to acute as well as during chronic stress (Bains et al., 2015; Herman and Tasker, 2016; Matovic et al., 2020; Salter et al., 2018; Yuan et al., 2019). Thus, it is an important direction of future studies to examine the plasticity of CRH_PVN neurons’ firing in vivo and their contributions to (mal)adaptations of the stress response.”

3) Please add a thoughtful discussion of the likely effects of the anesthetic on cell firing properties and of bursting more generally.

We agree with the reviewers about the importance of discussing the likely effects of anesthetics. We added a new section “Limitations of the study” in our Discussion and added the following paragraph. (Line 675-692)

“Our in vivo recordings were performed under urethane anesthesia, as was the case for the majority of previous studies using in vivo recordings from the PVN in anesthetized rats (Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1990, 1988, 1985). Urethane produces a long-lasting steady level of surgical anesthesia while preserving the subcortical and peripheral neural functions (Maggi and Meli, 1986), making it suitable for long electrophysiological recordings from the hypothalamus. For example, under urethane anesthesia in rabbits, hypothalamic neurons have been shown to readily respond to varieties of stress stimuli including hypoxia, hypercapnia, laud noise and pain (Cross and Silver, 1963). Of particular relevance to our study, sciatic nerve stimulation (under urethane anesthesia) has been shown to effectively elevate blood ACTH levels (Hamamura et al., 1986), validating that the stress-responsiveness of the HPA axis is preserved. However, it should be noted that urethane has also been shown to enhance autonomic (Shimokawa et al., 1998b) and HPA axis activity (Hamstra et al., 1984), likely by enhancing adrenergic inputs to the PVN (Shimokawa et al., 1998a). Considering that (nor)adrenaline has both excitatory and inhibitory effects on PVN neurons (Han et al., 2002; Itoi et al., 1994; Saphier and Feldman, 1991) and that it can potently influence state-switch between bursting and single spiking (Pape and McCormick, 1989), urethane may have affected the baseline firing activities and burst firing properties of CRH_PVN neurons in general. Future studies using awake animal recordings are warranted.”

4) Please discuss the possible mechanism requiring the long silent period between bursts and the role of H-currents, as well as likely contributions of peptide neuromodulators vs. fast neurotransmitters likely contained in the same neurons.

Regarding the potential roles of H-currents and the possible mechanisms requiring the long silent period for burst firing, we have revised the following sections of the Discussion.

(Line 585-598)

“In acute slices ex vivo, CRH_PVN neurons typically show a SS electrophysiological phenotype, or “regular” spiking, with up to around 50 Hz of spike-to-spike frequency in response to steps of intracellular current injections in mice and rats (Bittar et al., 2019; Jiang et al., 2019; Khan et al., 2011; Luther et al., 2002; Matovic et al., 2020; Sarkar et al., 2011; Wamsteeker Cusulin et al., 2013). Thus, our in vivo recordings of the RB (>100 Hz spike-to-spike frequency) revealed a previously underappreciated repertoire of the firing patterns of CRH_PVN neurons. In this regard, CRH_PVN neurons are different from thalamic relay neurons that have been extensively studied for their properties to switch between bursting and single (tonic) spiking modes both in vivo and ex vivo (McCormick and Pape, 1990a; Sherman, 2001; Steriade et al., 1993). The difference can be partly explained by the fact that CRH_PVN neurons express only modest levels of T-type currents (Luther et al., 2002) and little, if any, H-currents (see Figure 5A): these intrinsic properties play crucial roles in the generation of rhythmic burst firing, and switch between burst and single spiking modes in thalamic relay neurons (McCormick and Pape, 1990a; Sherman, 2001; Steriade et al., 1993).”

(Line 634-645)

“Importantly, the burst firing does not simply occur as a results of current injection patterns but requires unique interaction between the intrinsic properties of CRH_PVN neurons and the network inputs. This is supported by the following two controls. First, network inputs, which drove subthreshold membrane potential changes preceding (>200 ms) the burst firing, were necessary because truncated (<200 ms) network currents, which only contain high-frequency synaptic inputs component, failed to elicit the brief bursts in CRH_PVN neurons (Figure I-K). Notably, this result corroborates well with in vivo data that RB almost exclusively occur after long (>200 ms) silent period (Figure 4B). Second, magnocellular neurons, a separate class of neuroendocrine neurons of the PVN with different intrinsic properties (Luther et al., 2002), did not fire in brief bursts in response to the same network current injection that elicited brief bursts in CRH_PVN neurons (Figure 6M, N). The intrinsic properties key to this burst firing are currently unknown and warrants future investigation.”

For more detailed considerations about H-current and other intrinsic mechanisms that may be involved in the burst generation, please see our Responses to Reviewer 1-Point #1 and Reviewer 2-Point #5 below.

Regarding the potential contributions of CRHergic vs. glutamatergic transmission to the generation of RB, we have made the following changes in the discussion. (Line 556-563)

“Furthermore, transmission from CRH neurons to GABAergic neurons is exclusively modelled by slow CRHergic transmission based on recent experimental data (Jiang et al., 2019; Ramot et al., 2017), but CRH_PVN neurons have been shown to release glutamate at some synapses (Füzesi et al., 2016). Despite these simplifications, our model parsimoniously captures the key features of firing behavior in vivo, including the brevity of burst episodes and long IBI (Figure 5E, F). Thus, our simplified model, which is the first network model for CRH_PVN neurons to our best knowledge, is a useful first step to understand the spiking behaviors of CRH_PVN neuron in a network, and to generate experimentally testable hypotheses for future research.”

Reviewer #1 (Recommendations for the authors):

1. Previous work from thalamic neurons has demonstrated how burst firing can be increased with hyperpolarization, and this can be accompanied by lower firing rates than in the "up-state" in which the cell fires tonically. Do the authors think that a similar modulation may be occurring in the CRH neurons of PVN as well, as many PVN cells express these channels (e.g. Luther et al., 2002, Luther and Tasker, 2000)? Does their model provide any insights into this question? While the decrease in GABAergic inhibition and increased spike attenuation can lead to "stress-like" firing outputs, might a variety of different mechanisms that lead to hyperpolarization also underlie the actual change in vivo?

We thank the reviewer for this comment. We agree that changes in membrane potential likely underlie the state change between RB and SS spiking modes in CRH_PVN neurons, and there are some similarity with thalamic neurons (McCormick and Pape, 1990a; Sherman, 2001; Steriade et al., 1993). Specifically, we consider that hyperpolarized membrane potential may underly the prolonged silence period preceding the burst firing of CRH_PVN neurons in vivo (Figure 4), and hence the maintenance of the RB spiking mode and overall low firing rate. On the other hand, we consider that depolarization of membrane potential likely modulates the intrinsic properties of CRH_PVN neurons and diminishes their ability to fire the high-frequency burst. As the reviewer pointed out, the models proposed for thalamic neurons can guide us towards further mechanistic understanding. For example, T-type ca²⁺ channels potentially contribute to the generation of burst firing and their inactivation by membrane depolarization may prevent burst firing during the SS mode. Our consideration is in line with the earlier studies reporting that neuroendocrine parvocellular neurons (putative CRH neurons) express moderate levels of T-type currents (Luther et al., 2002; Luther and Tasker, 2000). However, compared to thalamic neurons, the amount of T-type currents in CRH_PVN neurons is moderate, and consequently CRH_PVN neurons do not readily fire low-threshold burst spike upon release from hyperpolarization (Luther et al., 2002) (also see Figure 5A of our manuscript confirming the finding by Luther et al. 2002). Along the similar line, we found that square pulse depolarization (from hyperpolarized membrane potential below -80 mV) could not trigger high-frequency (>100 Hz) burst firing ex vivo, which was the key characteristic of in vivo RB (Figure 6G, H). Thus, T-type currents alone is not sufficient to explain the in vivo RB. Our data point to additional intrinsic mechanisms (and their interplay with synaptic inputs) in the generation of the high-frequency burst firing as well as the switch between RB and SS modes.

It should also be noted that there are other important differences between the intrinsic properties of CRH_PVN neurons and thalamic neurons. First, unlike thalamic relay neurons (McCormick and Pape, 1990a), CRH_PVN neurons do not readily generate recurring rhythmic burst firing upon hyperpolarization in acute slices ex vivo (note that the burst firing reported by Yuan et al. 2019 will be discussed in the next section). Second, this is partly because CRH_PVN neurons express little, if any, H-currents (see Figure 5A for the absence of Sag current in response to a hyperpolarizing step): our observation is consistent with previous publications from other groups studying intrinsic properties of CRH_PVN neurons in mice (see Figure 6G of (J. S. Kim et al., 2019)) and neuroendocrine parvocellular neurons (putative CRH neurons) in rats (see Figure 1 of (Luther and Tasker, 2000)). In the thalamic neurons, H-currents, together with T-type currents, play key roles in the generation of recurring rhythmic burst firing (McCormick and Pape, 1990a).

Our spiking network model indicated that recurrent GABAergic inputs and spike-triggered adaptation were “sufficient” to drive the spiking state change in silico as well as ex vivo. However, this does not exclude the possibility that different mechanisms, which were not incorporated into our current model, may additionally contribute to, or can independently drive, the high-frequency burst firing and the switch between RB and SS modes. For example, neuromodulators, such as noradrenaline and 5-HT, modulate intrinsic and synaptic properties of CRH_PVN neurons (Sunstrum and Inoue, 2018) and excite/inhibit CRH_PVN neurons (Mukai et al., 2020). Also, in the thalamic relay neurons, both NA and 5-HT causes depolarization of the thalamic neurons and drive the change from burst firing to tonic firing mode (McCormick and Pape, 1990b; Pape and McCormick, 1989). Thus, neuromodulators may play roles in the firing activity switch partly by changing the membrane potentials of CRH_PVN neurons.

We have revised the following section of Discussion related to our response.

(Line 585-598)

(Line 634-645)

2. Work by Yuan et al. 2019, cited here, is not really discussed in the context of stressors. The first page of the discussion argues that burst firing in this cell population has not been characterized, but Yuan et al. have indeed reported it. This earlier work indicates that state changes in CRH cells have been previously identified, but muddies the waters as to how these in vitro data can be reconciled with the present work. Yuan et al. used an ex vivo preparation from control or stressed animals and report that the CRH cells also can show burst behaviors after specific stressful stimuli in vivo. However, the neurons are tonically firing from control animals and bursting when in stressed animals. How do the authors think this work fits into their observations in vivo?

We thank the reviewer for pointing out the importance of discussing “burst firing” in slices reported by Yuan et al. (Yuan et al., 2019). We agree with the reviewer that our original manuscript lacked clear discussion about the relationship between our observations and those by Yuan et al.: this can generate unnecessary confusion. Our considerations are as follows:

1) Our in vivo rhythmic brief bursts (RB) and the ex vivo “burst firing” by Yuan et al. are two distinct modes of firing/bursting. Specifically, the intra-burst frequency is >100 Hz for our in vivo RB and <100 Hz (averaging around 50 Hz) for Yuan et al. 2019 (see Figure 4E of Yuan et al. 2019).

2) The ex vivo “burst firing” by Yuan et al. resulted from stress-induced plasticity because it became evident only when slices were prepared from mice that were repeatedly stressed (3 daily foot shocks), but it was not evident in slices from previously unstressed mice. Interestingly, a single day or 2 daily foot shocks were not sufficient to induce the plasticity required for the “burst firing” indicating its slow development over the course of a few days of repeated stress (see Figure S3 of Yuan et al. 2019). Yuan et al. showed that the burst firing was diminished by bath application of NMDA receptor antagonist, indicating roles of synaptic plasticity.

3) As the reviewer noted, Yuan et al. observed that CRH_PVN neurons fire tonically (using cell-attached recording) in slices from previously unstressed mice. Similar observation (using cell-attached recording) was made by an independent group (Sarkar et al., 2011). In this regard, our observation that CRH_PVN neurons do not generate burst firing (and primarily fire in single spiking (SS)) in slices from previously unstressed mice is consistent with Yuan et al.

We added the following paragraph in Discussion to clarify the differences between our in vivo RB and ex vivo burst firing by Yuan et al. (Line 608-619)

“Interestingly, Yuan et al. recently reported spontaneous “burst firing” of CRH_PVN neurons ex vivo using cell-attached recordings in slices prepared from repeatedly stressed mice (Yuan et al., 2019). It should be noted, however, that the intra-burst frequency of this ex vivo “burst firing” was <100 Hz (averaging around 50 Hz), which was different from the in vivo RB (>100 Hz intra-burst frequency) we report here in non-stressed mice. Further, the ex vivo “burst firing” involved stress-induced plasticity that developed over days because it only became evident after 3 daily foot shocks, but was absent when mice were not previously stressed, or even after 2 daily foot shocks. Accumulating evidence show that CRH_PVN neurons undergo diverse forms of synaptic and intrinsic plasticity in response to acute as well as during chronic stress (Bains et al., 2015; Herman and Tasker, 2016; Matovic et al., 2020; Salter et al., 2018; Yuan et al., 2019). Thus, it is an important direction of future studies to examine the plasticity of CRH_PVN neurons’ firing in vivo and their contributions to (mal)adaptations of the stress response.”

We also revised the first paragraph of the discussion as follows (Line 471-475):

“We show that CRH_PVN neurons, within intact circuits, can fire distinct brief, high-frequency (>100 Hz) bursts, a firing pattern that has not been observed in ex vivo slice electrophysiology (Bittar et al., 2019; Jiang et al., 2019; Khan et al., 2011; Luther et al., 2002; Matovic et al., 2020; Sarkar et al., 2011; Wamsteeker Cusulin et al., 2013)”.

3. Why do the graphs for Figure 3G and Figure 3I look nearly identical? This figure may suggest that the burst behavior hardly contributes to the overall firing rate at all, even under basal conditions.

As the reviewer pointed out, actions potentials associated with burst firing account for a very small proportion of total spikes. Our explanation is as follows: the burst is very brief (on average 3.1 ± 0.5 spikes/burst) and the burst rate is low even during the baseline period (0.1406 ± 0.1375 Hz), this means that the action potential associated with burst firing is approximately 3.1 x 0.1406 = 0.4359 Hz. Please note that during the baseline period neurons do not always fire in the burst mode and occasionally generate single spiking at relatively high rate (Figure 3K). On the other hand, the average spike rate (including both burst and single spikes) during the baseline period is 3.063 ± 1.646 Hz (Results, line181). Taken together, the proportion of action potentials associated with burst firing is approximately 14% of the total spikes. Accordingly, our paper shows that single spiking primarily contributes to the overall the firing rate of CRH_PVN neurons, and that burst firing reflects a state of CRH_PVN neurons (which is not evident from overall firing rate) where their firing activity is constrained (Figure 3L, M).

Reviewer #2 (Recommendations for the authors):

1) Although computational remodeling well recreates in vitro-like switch between RB and SS, it is not clear as remodeling parameters the amount of contribution of fast neurotransmitters such as GABA and glutamate and slow neurotransmitter such as neuropeptide, CRH. CRH neurons would release both fast neurotransmitter glutamate and slow and sustained neurotransmitter CRH. Does the computational remodeling consider these parameters?

The current computational model exclusively used slow (i.e. CRHergic) excitatory transmission for CRH neuron→GABA neurons (τCRH (ms) = 232.6) modelled as single time constant α synapse (Destexhe et al., 1994), reflecting the slow CRHR1 mediated excitation (Jiang et al., 2019; Ramot et al., 2017). For GABA neurons→CRH neurons, our model uses fast GABAergic transmission (τi (ms) = 20.4). In addition, CRH neurons receive fast glutamatergic transmission from external sources (τe (ms) = 12.5) (Materials and methods and Table 1). The rationale for the parameter choice (i.e. CRH instead of glutamate for CRH neuron→GABA neurons synapse) is that recent studies showed that transmission at CRH neuron→GABA(CRHR1) neurons in the PVN (which form recurrent inhibitory circuit with CRH_PVN neurons) is almost exclusively mediated by CRH with little contribution by glutamate (Jiang et al., 2019; Ramot et al., 2017). However, as the reviewer pointed out, CRH neurons are known to release glutamate onto other neuron types (Füzesi et al., 2016): these neurons can potentially form recurrent inhibitory circuit with CRH neurons. Thus, our model results provide a testable hypothesis for future experiments to address the roles of specific neurotransmitters and/or co-release for the firing patterns of CRH_PVN neurons.

For the clarification of modelling parameters, we have added the following sentences in the Discussion (Line 556-559).

2) CRH release probability from synaptic vesicle might be different between RB and SS. It is not clear if the authors consider this difference in their computational remodeling.

We agree with the reviewer’s comment that different spiking modes (RB and SS) may change the release probability of CRH release. Our current computational model, however, does not incorporate this interesting possibility. This is because experimental data for the relationship between spiking patterns and CRH release is currently not well characterized. Our in vivo data revealing RB and SS spiking modes offers future studies to apply physiologically relevant spiking patterns to examine how spiking patterns affect CRH release. The experimental data, in turn, will further improve the computational model.

3) Sciatic nerve stimulation-induced transient firing pattern change from RB and SS around 2-3 sec. However, intrinsic state change representatively shown in Figure 3J looks much slower and lasted longer around ~100 sec. Could the authors reproduce a continuous increase of firing by changing the firing pattern from RB to SS if the sciatic nerve stimulation is repetitively applied for a longer period around 100 sec in vivo or in silico?

In a pilot in vivo recording, we applied four different stimulation intensities (0.5, 1, 1.5, and 2 mA) of a brief nerve stimulation that was repeated over multiple trials (5 pulses at 20 Hz, every 15 s): the stimulation pattern is identical to what we reported in Figure. 3. Notably, at the highest intensity (2 mA), the firing activity did not completely return to the baseline within the trial duration of 15 s, resulting in continuous decrease of RB and increase of SS during the stimulation period. These results demonstrate that nerve stimulation (with high intensities and/or short trial intervals) can result in continuous change from RB to SS, suggesting that common mechanisms underlies both transient and long-lasting change from RB to SS in vivo. In the current study, we focused to study the transient changes triggered by a lower intensity (1.6 mA) sciatic nerve stimulation, and multiple trials were averaged to analyse firing activity changes.

We added these results as Figure 3—figure supplement. We also added the following sentences in corresponding Result section (Line 185-189).

“In a representative case, we found that the firing pattern change lasted longer with higher intensity stimulation. Figure 3—figure supplement 1 shows that, with the highest intensity stimulation (2 mA), the firing activity change did not completely return to the baseline within the trial duration of 15 s, resulting in continuous decrease of RB and increase of SS during the stimulation period.”

We do not exclude the possibility that long stimulation (with higher intensity) could lead to lasting plasticity and irreversible change in the firing pattern after the cessation of stimulation. However, mechanisms that involved lasting plasticity are beyond the scope of the current study which focuses on the neural mechanisms for rapid, and reversible firing changes.

In silico, our model can produce continuous firing pattern change from RB to SS by persistently lowering GABAergic inputs to CRH neurons and increasing spike-triggered adaptation current in CRH neurons (Figure 5K). In Figure 6O, we also recreated in vivo-like (smooth) transition between RB and SS by transiently changing the same two parameters. Thus, our model results indicate that the sciatic nerve stimulation in vivo triggered transient and reversible changes in nerve mechanisms which were modelled by these parameters. Our model is also consistent with the observation that repetitive (and/or high intensity) sciatic nerve stimulation in vivo produced long-lasting changes in the firing pattern from RB to SS.

4) The authors stated that high-frequency burst firing requires a preceding prolonged silence period. It is not clear what neural mechanism is involved in prolonged silence periods in vivo or in silico.

We thank the reviewer for this comment that helps us to clarify the neural mechanisms involved in prolonged silence.

Our spiking network model shown in Figure 5C recapitulated in vivo-like rhythmic brief bursting (RB) and prolonged silence preceding the high frequency burst firing in model CRH neurons (Figure 5D). Our model indicates that recurrent inhibitory inputs almost exclusively drive the prolonged silence. More specifically, the model CRH neurons receive noisy excitatory inputs at an average frequency of 30 Hz. On the other hand, the recurrent inhibitory inputs are dynamic, driven by the burst firing of CRH neurons → prolonged CRHergic excitation (described in point #1) → prolonged excitation of GABA neurons. As the feedback inhibition decreases (i.e. disinhibition), the constant external excitatory inputs can generate membrane depolarization sufficient to trigger the next high-frequency burst firing. In addition to the inhibitory and excitatory synaptic inputs, the spike-triggered adaptation current, which represents after-hyperpolarization, (a loss of) after-depolarization, or both (Izhikevich, 2003), prevents the action potential firing of CRH neurons after the burst firing. However, its influence is less important. This is because its time constant (τ_w = 98.2 ms, Table 1) is substantially shorter than the time constant of CRHergic excitation (τ_CRH = 232.6 ms, Table 1) and ensuing prolonged feedback inhibition.

To clarify the contribution of inhibitory inputs to the prolonged silence, we added Figure 5—figure supplement in the revised manuscript that shows the temporal relationship between the burst firing patterns of representative model CRH neurons and the breakdown of three model parameters (i.e. excitatory conductance, inhibitory conductance and adaptation current). We also added the following description in the Results of the revised manuscript (Line 374-382).

“In the model CRH neurons, the recurrent inhibitory inputs were critical for the long, mostly silent IBI, and the loss of inhibitory inputs (i.e. disinhibition) underlay the timing when excitatory inputs generated depolarization sufficient to trigger burst firing (Figure 5—figure supplement). In addition to the inhibitory and excitatory synaptic inputs, the spike-triggered adaptation current (Izhikevich, 2003), prevented the action potential firing of CRH neurons after the burst firing. However, its influence was less important. This was because its time constant (τw = 98.2 ms, Table 1) was substantially shorter than the time constant of CRHergic excitation (τCRH = 232.6 ms, Table 1) and ensuing prolonged feedback inhibition.”

In ex vivo patch-clamp recordings, CRH_PVN neurons do not intrinsically generate RB unlike thalamic relay neurons that have been shown to switch between burst and single spiking modes (McCormick and Pape, 1990a; Sherman, 2001; Steriade et al., 1993). Thus, we reasoned that in vivo-like synaptic inputs (as modelled above) play important roles in the rhythmic burst generation in CRH_PVN neurons. To this end, we used the current injection protocol derived from our spiking network model. Our data showed that the network-derived current protocol was sufficient to generate the high-frequency burst firing, together with the prolonged silence preceding the burst firing in biological CRH_PVN neurons.

Interestingly, the prolonged silence had an important influence on the intrinsic properties of the biological CRH_PVN neurons (Figure 6I-K) because injections of truncated network-derived current (which temporary correspond with the timing of burst firing) did not reliably elicit burst firing, and the preceding noisy network inputs (which caused subthreshold membrane potential fluctuations) were necessary to efficiently drive the burst firing (Figure 6I-K). Based on these observations, we speculate that inhibitory feedback inputs play key roles in the prolonged silence/subthreshold membrane potential fluctuations, and consequently the ensuing generation of high-frequency burst firing.

The neural mechanisms for the prolonged silence required for the high-frequency burst firing in vivo would be most directly examined by in vivo patch clamp recordings, which is beyond the scope of this paper and an important direction of future research.

5) H-current through hyperpolarization-activated cyclic nucleotide-gated (HCN) channels play an important role in the generation of rhythmic burst firing. Do the authors record or consider changing of H-current in the activity state switch between RB and SS.

In ex vivo patch-clamp recordings shown in Figure 5A, we observed little, if any, Sag current in CRH_PVN neurons in response to a hyperpolarizing current step (from −70 mV to near −100 mV). While the holding potential of −70 mV is not ideal for studying Sag current as it partially opens HCN channels, it is still well above its maximum activation voltage (~−100 mV) (McCormick and Pape, 1990a). Therefore, if CRH_PVN neurons possess prominent HCN channels, Sag current should be readily detectable in our recordings. Our observation is consistent with previous publications from other groups that reported similar current clamp trances from CRH_PVN neurons in mice (Figure 6G, (J. S. Kim et al., 2019)) and parvocellular neuroendocrine neurons (putative CRH neurons) in rats (Figure 1, (Luther and Tasker, 2000)). Thus, we conclude that HCN channels makes little, if any, contributions to the intrinsic membrane properties of CRH_PVN neurons involved in the burst firing.

However, our data do not fully exclude the possibility that HCN channels may become facilitated in CRH_PVN neurons, for example after stress, by neuromodulators that increase intracellular cAMP. As the reviewer pointed out, HCN channels play important roles in burst firing, with both facilitatory and inhibitory effects, depending on the magnitude of activity (McCormick and Pape, 1990a). With relevance to the firing mode switch, it has been shown that facilitation of HCN channels by noradrenaline and 5-HT contribute to the switch from rhythmic burst firing to single spike firing in thalamic relay neurons (McCormick and Pape, 1990b). In this case, activation of HCN channels prevented hyperpolarization required for subsequent burst generation. The same study also showed that HCN channels increased after-hyperpolarization and prevented burst firing. In this regard, HCN channels are potential intrinsic properties reflected by the spike-triggered adaptation current in our working model.

Reviewer #3 (Recommendations for the authors):

1) The neurons that were sensitive to blue light ("light-responsive") had a latency of approximately 7 ms. This is sufficient time for action potential in one cell, release of glutamate and an action potential in another cell. Can the authors provide assurances that blue light is directly activating the cell they are recording? In brain slices, light pulses elicit almost simultaneous action potentials, so the delay here needs to be addressed.

We appreciate the reviewer’s concern that 7 ms could be sufficient for the excitation of the postsynaptic neurons via fast glutamatergic transmission. For the considerations listed below, we believe that the light-responsive neurons are directly activated by blue light.

1) In the PVN, CRH neurons are unlikely to excite other neurons via fast glutamatergic transmission. Jiang et al. showed that optogenetic stimulation of CRH_PVN neurons excites CRHR1 neurons, but this indirect excitation required trains of light stimulation, slow onset and almost exclusively mediated by CRH. Indeed, the optogenetic stimulation of CRH neurons evoked fast EPSCs in only 2% of CRHR1 neurons (Jiang et al., 2018). We validate that our recording electrode hit the PVN with post-mortem histology. Therefore, although we cannot fully exclude the possibility of secondary activation, it is highly unlikely for our optogenetic stimulation to evoke indirect activation after a single light stimulus with a latency of approximately 7 ms.

2) Outside of the PVN, CRH neurons have been shown to release glutamate onto neurons in the lateral hypothalamus (Füzesi et al., 2016). They can be excited indirectly by blue light. However, as described above, our recordings were validated to be made in the PVN: we always found our electrode tract between the 3^rd ventricle and the fornix.

3) One key factor for the latency of direct light-triggered action potential is the intensity of the light-induced current in the target neurons, which will be determined by the combination of light intensity and the density of ChR2 expression. Indeed, even in ex vivo, the latency can vary (2-10 ms) within the same neurons depending of the light intensity [Figure 6C of (Wamsteeker Cusulin et al., 2013)]. in vivo, the light intensity (at the target neuron) is difficult to control. We tried to avoid increasing light intensity (kept it just enough to reliably trigger action potentials) because high intensity light tends to generate large light-induced artifacts which prevent the detection of short-latency action potentials.

2) What is the advantage of a rhythmic burst mechanism during periods when CRH neuron activity is not required for hormone release? The authors offer one sentence in the discussion, but more is warranted here.

We thank the reviewer for this comment and for encouraging us to think more about the significance of burst firing. We took the opportunity that eLife offers, that is authors are allowed include “Ideas and Speculation” in the Discussion.

We added the following paragraph at the end of Discussion as “Ideas and Speculations”. (Line 694-716)

“Our study found brief, high-frequency burst firing of CRH_PVN neurons in vivo. This unexpected finding led us to propose, using a tight combination of experimental and computational approaches, that recurrent inhibition constrains the firing rate of CRH_PVN neurons and ensuing hormonal release. Then, what is the advantage of a RB mechanism during periods when CRH neuron activity is not required for hormone release? Does RB firing encode and convey specific information? One possibility is that burst firing facilitates the self-maintenance of the RB firing state. That is, a brief burst enables high fidelity synaptic transmission (Lisman, 1997) and efficiently excites GABAergic neurons that return recurrent inhibition to CRH_PVN neurons: this feedback inhibition, in turn, facilitates subsequent burst firing. Notably, our spiking network model found that recurrent inhibition works as a gain regulator at CRH neurons (Figure 5). Thus, our model prediction offers a new circuit mechanism for seminal experimental studies that revealed disinhibition, more so than increasing excitatory inputs, activates CRH_PVN neurons for hormonal release (Cole and Sawchenko, 2002; Hewitt et al., 2009; Kovács et al., 2004; Sarkar et al., 2011). Beyond constraining hormonal release of CRH, the idea that CRH_PVN neurons use brief bursts for high fidelity synaptic transmission, without triggering massive HPA axis activation, raises an intriguing possibility in light of emerging new roles of CRH_PVN neurons. Recent studies showed that CRH_PVN neurons are involved in controlling wakefulness (Ono et al., 2020), valence encoding (J. Kim et al., 2019), reward processing (Yuan et al., 2019), and defensive behavior control (Daviu et al., 2020; Füzesi et al., 2016) under non-stress and stress conditions. We propose that burst firing is an ideal mode of signal transmission for CRH_PVN neurons to drive rapid behavioral and emotional changes without triggering systemic and long-lasting hormonal response.”

3) There needs to be some discussion on the possible contributions of the anesthetic, urethane on the properties described here. This is an important limitation of the study, as it may increase inhibitory tone in the system and needs to be considered.

We added a new section “Limitations of the study” in our Discussion and added the following paragraph specifically for this topic. (Line 675-692)

“Our in vivo recordings were performed under urethane anesthesia, as was the case for the majority of previous studies using in vivo recordings from the PVN in anesthetized rats (Hamamura et al., 1986; Kannan et al., 1987; Saphier, 1989; Saphier and Feldman, 1990, 1988, 1985). Urethane produces a long-lasting steady level of surgical anesthesia, and preserves the subcortical and peripheral neural functions (Maggi and Meli, 1986), making it suitable for long electrophysiological recordings from the hypothalamus. For example, under urethane anesthesia in rabbits, hypothalamic neurons have been shown to readily respond to varieties of stress stimuli including hypoxia, hypercapnia, laud noise and pain (Cross and Silver, 1963). Of particular relevance to our study, sciatic nerve stimulation (under urethane anesthesia) has been shown to effectively elevate blood ACTH levels (Hamamura et al., 1986), validating that the stress-responsiveness of the HPA axis is preserved. However, it should be noted that urethane has also been shown to enhance autonomic (Shimokawa et al., 1998b) and HPA axis activity (Hamstra et al., 1984), likely by enhancing adrenergic inputs to the PVN (Shimokawa et al., 1998a). Considering that (nor)adrenaline has both excitatory and inhibitory effects on PVN neurons (Han et al., 2002; Itoi et al., 1994; Saphier and Feldman, 1991) and that it can potently influence state-switch between bursting and single spiking (Pape and McCormick, 1989), urethane may have affected the baseline firing activities and burst firing properties of CRH_PVN neurons in general. Future studies using awake animal recordings are warranted.”

References:

Bittar TP, Nair BB, Kim JS, Chandrasekera D, Sherrington A, Iremonger KJ. 2019. Corticosterone mediated functional and structural plasticity in corticotropin-releasing hormone neurons. Neuropharmacology. doi:10.1016/j.neuropharm.2019.02.017

Cole RL, Sawchenko PE. 2002. Neurotransmitter Regulation of Cellular Activation and Neuropeptide Gene Expression in the Paraventricular Nucleus of the Hypothalamus. J Neurosci 22:959–969.

Cross BA, Silver IA. 1963. Unit activity in the hypothalamus and the sympathetic response to hypoxia and hypercapnia. Experimental Neurology 7:375–393. doi:10.1016/0014-4886(63)90019-0

Daviu N, Füzesi T, Rosenegger DG, Rasiah NP, Sterley T-L, Peringod G, Bains JS. 2020. Paraventricular nucleus CRH neurons encode stress controllability and regulate defensive behavior selection. Nat Neurosci 23:398–410. doi:10.1038/s41593-020-0591-0

Destexhe A, Mainen ZF, Sejnowski TJ. 1994. Synthesis of models for excitable membranes, synaptic transmission and neuromodulation using a common kinetic formalism. J Comput Neurosci 1:195–230. doi:10.1007/BF00961734

Füzesi T, Daviu N, Wamsteeker Cusulin JI, Bonin RP, Bains JS. 2016. Hypothalamic CRH neurons orchestrate complex behaviours after stress. Nature Communications 7:11937. doi:10.1038/ncomms11937

Hamamura M, Onaka T, Yagi K. 1986. Parvocellular neurosecretory neurons: converging inputs after saphenous nerve and hypovolemic stimulations in the rat. The Japanese journal of physiology 36:921–933. doi:10.2170/jjphysiol.36.921

Hamstra WN, Doray D, Dunn JD. 1984. The effect of urethane on pituitary-adrenal function of female rats. Acta Endocrinol (Copenh) 106:362–367. doi:10.1530/acta.0.1060362

Han SK, Chong W, Li LH, Lee IS, Murase K, Ryu PD. 2002. Noradrenaline excites and inhibits GABAergic transmission in parvocellular neurons of rat hypothalamic paraventricular nucleus. J Neurophysiol 87:2287–2296.

Hewitt SA, Wamsteeker JI, Kurz EU, Bains JS. 2009. Altered chloride homeostasis removes synaptic inhibitory constraint of the stress axis. Nat Neurosci 12:438–443. doi:10.1038/nn.2274

Itoi K, Suda T, Tozawa F, Dobashi I, Ohmori N, Sakai Y, Abe K, Demura H. 1994. Microinjection of norepinephrine into the paraventricular nucleus of the hypothalamus stimulates corticotropin-releasing factor gene expression in conscious rats. Endocrinology 135:2177–2182. doi:10.1210/en.135.5.2177

Izhikevich EM. 2003. Simple model of spiking neurons. IEEE Transactions on Neural Networks 14:1569–1572. doi:10.1109/TNN.2003.820440

Jiang Z, Rajamanickam S, Justice NJ. 2019. CRF signaling between neurons in the paraventricular nucleus of the hypothalamus (PVN) coordinates stress responses. Neurobiol Stress 11. doi:10.1016/j.ynstr.2019.100192

Kannan H, Kasai M, Osaka T, Yamashita H. 1987. Neurons in the paraventricular nucleus projecting to the median eminence: a study of their afferent connections from peripheral baroreceptors, and from the A1-catecholaminergic area in the ventrolateral medulla. Brain Res 409:358–363. doi:10.1016/0006-8993(87)90722-0

Khan AM, Kaminski KL, Sanchez-Watts G, Ponzio TA, Kuzmiski JB, Bains JS, Watts AG. 2011. MAP kinases couple hindbrain-derived catecholamine signals to hypothalamic adrenocortical control mechanisms during glycemia-related challenges. J Neurosci 31:18479–18491. doi:10.1523/JNEUROSCI.4785-11.2011

Kim J, Lee S, Fang Y-Y, Shin A, Park S, Hashikawa K, Bhat S, Kim D, Sohn J-W, Lin D, Suh GSB. 2019. Rapid, biphasic CRF neuronal responses encode positive and negative valence. Nat Neurosci 22:576–585. doi:10.1038/s41593-019-0342-2

Kim JS, Han SY, Iremonger KJ. 2019. Stress experience and hormone feedback tune distinct components of hypothalamic CRH neuron activity. Nature Communications 10:5696. doi:10.1038/s41467-019-13639-8

Kovács KJ, Miklós IH, Bali B. 2004. GABAergic mechanisms constraining the activity of the hypothalamo-pituitary-adrenocortical axis. Ann N Y Acad Sci 1018:466–476. doi:10.1196/annals.1296.057

Lisman JE. 1997. Bursts as a unit of neural information: making unreliable synapses reliable. Trends Neurosci 20:38–43. doi:10.1016/S0166-2236(96)10070-9

Luther JA, Daftary SS, Boudaba C, Gould GC, Halmos KC, Tasker JG. 2002. Neurosecretory and non-neurosecretory parvocellular neurones of the hypothalamic paraventricular nucleus express distinct electrophysiological properties. Journal of Neuroendocrinology 14:929–932. doi:10.1046/j.1365-2826.2002.00867.x

Luther JA, Tasker JG. 2000. Voltage-gated currents distinguish parvocellular from magnocellular neurones in the rat hypothalamic paraventricular nucleus. J Physiol 523:193–209. doi:10.1111/j.1469-7793.2000.t01-1-00193.x

Maggi CA, Meli A. 1986. Suitability of urethane anesthesia for physiopharmacological investigations in various systems Part 1: General considerations. Experientia 42:109–114. doi:10.1007/BF01952426

Matovic S, Ichiyama A, Igarashi H, Salter EW, Sunstrum JK, Wang XF, Henry M, Kuebler ES, Vernoux N, Martinez‐Trujillo J, Tremblay M-E, Inoue W. 2020. Neuronal hypertrophy dampens neuronal intrinsic excitability and stress responsiveness during chronic stress. The Journal of Physiology 598:2757–2773. doi:10.1113/JP279666

McCormick DA, Pape HC. 1990a. Properties of a hyperpolarization-activated cation current and its role in rhythmic oscillation in thalamic relay neurones. J Physiol 431:291–318. doi:10.1113/jphysiol.1990.sp018331

McCormick DA, Pape HC. 1990b. Noradrenergic and serotonergic modulation of a hyperpolarization-activated cation current in thalamic relay neurones. J Physiol 431:319–342. doi:10.1113/jphysiol.1990.sp018332

Mukai Y, Nagayama A, Itoi K, Yamanaka A. 2020. Identification of substances which regulate activity of corticotropin-releasing factor-producing neurons in the paraventricular nucleus of the hypothalamus. Scientific Reports 10:13639. doi:10.1038/s41598-020-70481-5

Ono D, Mukai Y, Hung CJ, Chowdhury S, Sugiyama T, Yamanaka A. 2020. The mammalian circadian pacemaker regulates wakefulness via CRF neurons in the paraventricular nucleus of the hypothalamus. Science Advances 6:eabd0384. doi:10.1126/sciadv.abd0384

Pape HC, McCormick DA. 1989. Noradrenaline and serotonin selectively modulate thalamic burst firing by enhancing a hyperpolarization-activated cation current. Nature 340:715–718. doi:10.1038/340715a0

Ramot A, Jiang Z, Tian J-B, Nahum T, Kuperman Y, Justice N, Chen A. 2017. Hypothalamic CRFR1 is essential for HPA axis regulation following chronic stress. Nat Neurosci 20:385–388. doi:10.1038/nn.4491

Saphier D. 1989. Catecholaminergic projections to tuberoinfundibular neurones of the paraventricular nucleus: I. Effects of stimulation of A1, A2, A6 and C2 cell groups. Brain Research Bulletin 23:389–395. doi:10.1016/0361-9230(89)90179-2

Saphier D, Feldman S. 1991. Catecholaminergic projections to tuberoinfundibular neurones of the paraventricular nucleus: III. Effects of adrenoceptor agonists and antagonists. Brain Res Bull 26:863–870.

Saphier D, Feldman S. 1990. Iontophoresis of cortisol inhibits responses of identified paraventricular nucleus neurones to sciatic nerve stimulation. Brain Res 535:159–162.

Saphier D, Feldman S. 1988. Iontophoretic application of glucocorticoids inhibits identified neurones in the rat paraventricular nucleus. Brain Res 453:183–190.

Saphier D, Feldman S. 1985. Effects of neural stimuli on paraventricular nucleus neurones. Brain Res Bull 14:401–407. doi:10.1016/0361-9230(85)90016-4

Sarkar J, Wakefield S, MacKenzie G, Moss SJ, Maguire J. 2011. Neurosteroidogenesis is required for the physiological response to stress: role of neurosteroid-sensitive GABAA receptors. J Neurosci 31:18198–18210. doi:10.1523/JNEUROSCI.2560-11.2011

Sherman SM. 2001. Tonic and burst firing: dual modes of thalamocortical relay. Trends Neurosci 24:122–126. doi:10.1016/s0166-2236(00)01714-8

Shimokawa, Jin QH, Ishizuka Y, Kunitake T, Takasaki M, Kannan H. 1998a. Effects of anesthetics on norepinephrine release in the hypothalamic paraventricular nucleus region of awake rats. Neurosci Lett 244:21–24. doi:10.1016/s0304-3940(98)00119-0

Shimokawa, Kunitake T, Takasaki M, Kannan H. 1998b. Differential effects of anesthetics on sympathetic nerve activity and arterial baroreceptor reflex in chronically instrumented rats. J Auton Nerv Syst 72:46–54. doi:10.1016/s0165-1838(98)00084-8

Steriade M, McCormick DA, Sejnowski TJ. 1993. Thalamocortical oscillations in the sleeping and aroused brain. Science 262:679–685. doi:10.1126/science.8235588

Sunstrum JK, Inoue W. 2018. Heterosynaptic modulation in the paraventricular nucleus of the hypothalamus. Neuropharmacology. doi:10.1016/j.neuropharm.2018.11.004

Wamsteeker Cusulin JI, Füzesi T, Watts AG, Bains JS. 2013. Characterization of corticotropin-releasing hormone neurons in the paraventricular nucleus of the hypothalamus of Crh-IRES-Cre mutant mice. PLoS ONE 8:e64943. doi:10.1371/journal.pone.0064943

Yuan Y, Wu W, Chen M, Cai F, Fan C, Shen W, Sun W, Hu J. 2019. Reward Inhibits Paraventricular CRH Neurons to Relieve Stress. Curr Biol 29:1243-1251.e4. doi:10.1016/j.cub.2019.02.048

https://doi.org/10.7554/eLife.76832.sa2

Article and author information

Author details

Aoi Ichiyama

Graduate Program in Neuroscience, Western University, London, Canada

Contribution
Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review and editing

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-0981-2369
Samuel Mestern

Graduate Program in Neuroscience, Western University, London, Canada

Contribution
Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-5062-6712
Gabriel B Benigno
1. Department of Mathematics, Western University, London, Canada
2. Brain and Mind Institute, Western University, London, Canada
Contribution
Formal analysis, Software, Writing – original draft

Competing interests
No competing interests declared
Kaela E Scott
1. Graduate Program in Neuroscience, Western University, London, Canada
2. Department of Anatomy and Cell Biology, Schulich School of Medicine and Dentistry, Western University, London, Canada
Contribution
Methodology

Competing interests
No competing interests declared
Brian L Allman

Department of Anatomy and Cell Biology, Schulich School of Medicine and Dentistry, Western University, London, Canada

Contribution
Methodology, Supervision, Writing – review and editing

Competing interests
No competing interests declared
Lyle Muller
1. Department of Mathematics, Western University, London, Canada
2. Brain and Mind Institute, Western University, London, Canada
3. Robarts Research Institute, Western University, London, Canada
Contribution
Conceptualization, Funding acquisition, Investigation, Methodology, Software, Supervision, Writing – review and editing

For correspondence
lmuller2@uwo.ca

Competing interests
No competing interests declared

Additional information
Co-senior authors.

"This ORCID iD identifies the author of this article:" 0000-0001-5165-9890
Wataru Inoue
1. Robarts Research Institute, Western University, London, Canada
2. Department of Physiology and Pharmacology, Schulich School of Medicine and Dentistry, Western University, London, Canada
Contribution
Conceptualization, Funding acquisition, Investigation, Supervision, Validation, Writing – original draft, Writing – review and editing

For correspondence
winoue@robarts.ca

Competing interests
No competing interests declared

Additional information
Co-senior authors.

"This ORCID iD identifies the author of this article:" 0000-0002-2438-5123

Funding

Natural Sciences and Engineering Research Council of Canada (RGPIN-2015-06106)

Wataru Inoue

Canadian Institutes of Health Research (PJT-148707)

Wataru Inoue

Canada First Research Excellence Fund (BrainsCAN Accelerator)

Wataru Inoue
Lyle Muller

Compute Canada

Lyle Muller

Canadian Open Neuroscience Platform

Samuel Mestern
Gabriel B Benigno

Vector Institute (Postgraduate Affiliate Program)

Gabriel B Benigno

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Karl Deisseroth (Stanford University) for providing pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA was a gift from (Addgene plasmid # 20298; http://n2t.net/addgene:20298; RRID:Addgene_20298).

This work was funded through a Discovery Grant (RGPIN-2015–06106) from the Natural Sciences and Engineering Research Council of Canada (NSERC; to W I), a Project Grant (PJT-148707) from the Canadian Institutes of Health Research (CIHR; to W I), BrainsCAN at Western University through the Canada First Research Excellence Fund (CFREF; to W I and L M), and the Compute Canada (to L M). A I was a recipient of an Ontario Graduate Scholarship, SM is a recipient of the Canadian Open Neuroscience Platform studentship. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication

Ethics

All experimental procedures were performed in accordance with the Canadian Council on Animal Care guidelines and approved by the University of Western Ontario Animal Use Subcommittee (AUP: 2018-130).

Senior Editor

John R Huguenard, Stanford University School of Medicine, United States

Reviewing Editor

Julie A Kauer, Stanford University, United States

Reviewers

Akihiro Yamanaka, Research Institute of Environmental Medicine, Nagoya University, Japan
Jaideep Singh Bains, University of Calgary, Canada

Version history

Received: January 6, 2022
Preprint posted: February 10, 2022 (view preprint)
Accepted: June 17, 2022
Accepted Manuscript published: June 30, 2022 (version 1)
Version of Record published: July 13, 2022 (version 2)

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.