Dysregulated neural coding in the vagus nerve during long sepsis

Figures

• 

  Fig. 1. Neurophysiological recordings in the cervical vagus nerve. (A) The cervical surgery involves a neck incision, isolation of the vagus nerve and removal of the surrounding sheath (top left). The desheathed nerve is placed over the recording electrode (bottom left) and the ground wire is placed under the left salivary gland. The electrode's leads and ground wire are connected to an electronic interface board (EIB, Omnetics) that carries the signal to the data acquisition system (top right), which records compound action potentials (bottom right). (B) Raw signals showing a 10-s interval (top) and an expanded 2-s view (bottom), which contains both neural and non-neural components. An adaptive threshold (in green) is applied for isolating the CAPs, accounting for the effect of respiration that leads to respiratory bursts (“Resp. Burst” in yellow). These are recurring periods with higher background noise and robust vagus nerve's activity. Cardiac signals (in red) are also recorded. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

• 

  Fig. 2. Clustering and waveform analysis of compound action potentials. (A)Top, representative t-SNE plot with CAP clusters (left) and waveforms of sorted units (right) for the vagus nerve of a single CON mouse. Bottom, same as above for a single LS mouse. Each color corresponds to an isolated vagal unit seen as a distinct cluster in the t-SNE plot (at left) and as overlaid waveforms (at right). Clusters corresponding to cardiac signals (waveforms shown in grey boxes) are discarded from further analysis. (B) Representative 1-s recording showing simultaneous traces from the vagus nerve cuff electrode (top) and the recording from the electrocardiogram electrode (bottom). The vagus nerve recording contains two distinct CAPs (blue and green), and a cardiac component (violet) that was excluded from all further analysis. The electrocardiogram trace shows cardiac signals that temporally align with the cardiac component in the vagus recording but contains no CAPs, confirming the neural origin of the CAP signals. (C) CAP waveforms (n = 15) from a single CON unit and a single LS unit, with their average waveform (mean ± SEM) shown to the right of each set. (C) Violin plots for the width of the CAP…

• 

  Fig. 3. Baseline activity of the vagus nerve. (A) Raster plots showing 3 representative recordings per group; insets (at left) display the associated CAP waveforms. (B) Time series for CAP frequency (mean ± SEM, 1-min bins) showing that LS mice have significantly higher values than CON mice for the entire baseline period (∗∗∗P < 0.0001, 2-way ANOVA). (C) Violin plots of CAP frequency confirm that the LS group has much higher firing rate, calculated per unit (left, ∗∗∗P < 0.0001, KS test) and per mouse (right, ∗∗∗P < 0.0001, KS test). (D) Histogram showing the mean CAP frequency for each unit and cumulative probability plots for CON and LS groups (note the different scales for the X axes for the two groups); μ,mean, σ, standard deviation.

• 

  Fig. 4. Respiratory modulation of the activity of the vagus nerve. (A) Representative recording (1-s epoch) showing respiratory bursts in purple and inter-burst intervals in black. (B) Box-and-whisker plots showing that LS and CON groups have respiratory bursts of similar duration (left, P = 0.271, t-test) and period (right, P = 0.342, MW test). Each plot displays the mean (square), median (line), inter-quartile range (box), and 10–90 range (whiskers), each dot represents a mouse. (C) Peri-event histogram showing the summed number of CAPs across CON or LS mice aligned from the start of each respiratory burst (“Resp. burst”). The highlighted portion depicts the time corresponding to the average duration of respiratory events. Inset, cumulative probability plots showing that LS mice have significantly higher number of CAPs during respiratory bursts (top, ∗∗∗P < 0.0001, KS test) and inter-burst intervals (bottom, ∗∗∗P < 0.0001, KS test). (D) Violin plot for the CAP frequency during respiratory bursts for CON and LS units, each dot represents one CAP unit (∗∗∗P < 0.0001, KS test). (E) Same as (D) but for CAP frequency during inter-burst intervals (∗P < 0.022, KS test). (For interpretation of…

• 

  Fig. 5. Vagal activity in response to cytokine injections. (A) Time series for CAP frequency (mean ± SEM, 1-min bins) of all units recorded in CON and LS mice. A baseline (BL) period is followed by injection of TNF (at 20 min) and injection of IL-1β (at 40 min). (B) Same as (A) but IL-1β is injected at 20 min and TNF at 40 min. (C) Plots showing the CAP frequency (mean ± SEM) for the last 10 min of BL, TNF, and IL-1β periods; left, TNF followed by IL-1β (CON: BL v. TNF, P = 0.06; BL v. IL-1β, ∗P = 0.02; LS: BL v. TNF, P = 0.929; BL v. IL-1β, P = 0.826; MW test). Right, IL-1β followed by TNF (CON: BL v. IL-1β, ∗P = 0.041; BL v. TNF, ∗P = 0.034; LS: BL v. IL-1β, P = 0.967; BL v. TNF, P = 0.941; MW test). (D)Left, violin plots showing the summed frequency of all CAPs from each mouse for the last 10 min of BL (∗∗∗P < 0.0001, MW test), TNF (∗∗P = 0.001, MW test), and IL-1β (∗∗P = 0.008, t-test). Right, violin plots showing the cytokine ratios (defined in Materials and Methods) for TNF (∗∗P = 0.001, MW test), and IL-1β (∗∗P = 0.008, MW test). Each dot represents a mouse.

• 

  Fig. 6. Firing rate dynamics for the vagal units. (A) Heatmaps of normalized firing rates for single vagal units in CON (left) and LS (right) mice. Each row of a heatmap represents a vagal unit and each column a 10-s bin. The units are ordered from top to bottom according to the time they exhibit their peak firing rate (PFR). Each heatmap is separated into baseline (BL), TNF and IL-1β periods (each lasting 20 min) by vertical white bars. (B) Each vagal unit is categorized into one of 3 categories: units that increase their firing after each cytokine injection, units that decrease their firing after each cytokine injection, or units which have a mixed response, defined as an increase to one cytokine and a decrease to the other cytokine regardless of the order in which each is injected. Left, donut plot showing the percent of units which fall into each category. Middle, line-series plots showing the frequency of firing for each vagal unit during the BL period, following injection of the first cytokine (C1), and the second cytokine (C2), regardless of which cytokine is actually administered. Right, stacked histograms showing the percent of vagal units which fall into each category for each…

• 

  Fig. 7. Bayesian decoding based on vagal activity. (A) Schematic of the decoding procedure. (B) Examples from CON and LS mice showing each decoded timepoint and the probability of prediction for each timepoint to be categorized as one of the 3 possible classes: baseline (BL, black), TNF (purple), or IL-1β (blue). (C) Examples of decoded recordings shown as dot plots in CON and LS mice. Each decoded timepoint is represented as a dot assigned to the class with the highest probability (among the 3 possible classes): BL (bottom of each plot), TNF (middle of each plot), or IL-1β (top of each plot). (D) Violin plots showing the probability of correct class assignment in each of the 3 classes for CON and LS mice: BL (left, ∗∗P = 0.008, MW test), TNF (middle, ∗P = 0.046, t-test), and IL-1β (right, P = 0.183, t-test) periods; each dot represents a mouse; ns, non-significant. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)