Temperature and frequency dependence of conduction along sympathetic preganglionic axons

Abstract

Sympathetic preganglionic neurons (SPNs) distribute signals across paravertebral ganglia, yet the reliability of spike propagation along their predominantly unmyelinated axons remains poorly defined. We examined temperature- and activity-dependent modulation of SPN axonal conduction in an ex vivo adult mouse thoracic sympathetic-chain preparation. Population compound action potentials (CAP) responses were evoked by supramaximal T10 ventral-root stimulation and recorded from the interganglionic nerve (IGN), which contains pathways that branch within the paravertebral chain, and the splanchnic nerve, presumed to project unbranching axons toward prevertebral ganglia.

After scaling for temperature-dependent changes in conduction velocity and spike waveform, CAP magnitude at 36°C was approximately 50% lower than at 22°C, with preferential reduction of slower-conducting components. Reductions occurred in both pathways, indicating a shared vulnerability of small unmyelinated axons, whereas evidence for an additional branching-associated vulnerability was strongest during repetitive activation.

Activity dependence was assessed using 20-s trains at 5 and 20 Hz. At 22°C, slower-conducting components showed marked frequency-dependent depression that was greater in the IGN. Observed loss of individual units without change in spike amplitude support CAP depression as due to conduction block. At 36°C, depression of slow components was not statistically detectable, faster-conducting components facilitated at 20 Hz, and several responses showed post-train potentiation. These differences may reflect temperature-dependent filtering of vulnerable slow-conducting axons rather than uniform changes in frequency-following properties.

Thus, SPN axonal conduction is strongly modulated by temperature and activation history. Preferential vulnerability of slower-conducting, likely small-diameter axons identifies conduction reliability as a vulnerable site in sympathetic output.

INTRODUCTION

Sympathetic preganglionic neurons (SPNs) distribute signals widely across paravertebral ganglia, providing the primary drive to postganglionic sympathetic neurons that regulate cardiovascular, thermoregulatory, and metabolic functions. These projections are typically assumed to transmit action potentials reliably, such that preganglionic firing is faithfully translated into postganglionic output (McLachlan, 2003). However, this assumption has not been rigorously tested under physiological conditions. Given the distinctive structural features of SPN axons, including their predominantly unmyelinated nature and extensive branching within the sympathetic chain, it is unclear whether spike propagation remains reliable across their full anatomical extent.

Axonal geometry and diameter are key determinants of conduction reliability. SPN axons in mouse are overwhelmingly unmyelinated and extremely small in caliber, with mean diameters on the order of 0.4 μm and branch diameters as small as 0.1 μm (Lewis and Burton, 1977). Such small-diameter axons operate with a reduced safety factor for propagation and are therefore particularly sensitive to perturbations in membrane conductance and ionic currents. In other systems, axonal branch points are well established as sites of conduction failure, especially when the available depolarizing current is insufficient to support propagation into multiple daughter branches (Levy, 1980; Debanne et al., 2011). These considerations suggest that SPN axons, by virtue of their size and branching architecture, may be especially vulnerable to failures of spike propagation, with potential consequences for the effective distribution of sympathetic signals.

Temperature is a critical determinant of axonal excitability and conduction reliability, particularly in unmyelinated fibers. Increases in temperature accelerate channel kinetics but can reduce spike amplitude and duration, thereby lowering the safety factor for propagation (Hodgkin and Katz, 1949; Paintal, 1965; Pekala et al., 2016). Conversely, cooling can broaden action potentials and may enhance conduction reliability. Despite this, prior physiological studies of multisegmental SPN signaling have been performed at room temperature (Blackman and Purves, 1969; Nja and Purves, 1977; Lichtman et al., 1979, 1980; Purves and Lichtman, 1980), raising the possibility that the reliability and extent of SPN signal distribution under physiological conditions have been substantially mischaracterized. In addition to temperature, activation history is known to influence conduction in unmyelinated axons through activity-dependent hyperpolarization and sodium channel inactivation, which can lead to conduction slowing and failure during repetitive firing (Renganathan et al., 2002; De Col et al., 2008; Obreja et al., 2010; Rama et al., 2018).

Here, we tested the hypothesis that spike propagation in SPN axons is not uniformly reliable and is strongly modulated by temperature and activation history. Using an ex vivo adult mouse preparation that preserves the thoracic sympathetic chain and ventral roots, we recorded population compound action potentials (CAPs) evoked by supramaximal stimulation of SPN axons. By comparing responses at room and physiological temperatures and across different stimulation frequencies, we assessed how temperature and activation history influence effective axonal recruitment. We further examined whether slower-conducting axons, which likely correspond to smaller diameter and more extensively branching fibers, are preferentially vulnerable to conduction failure. These experiments were designed to determine whether axonal conduction reliability represents a previously unrecognized site of gain control in sympathetic output.

METHODS

Multisegmental ex vivo paravertebral preparation.

All animal procedures were approved by the Emory University Institutional Animal Care and Use Committee in accordance with regulations of the Guide for the Care and Use of Laboratory Animals. Adult (8 weeks+) C57BL/6 mice of both sexes were deeply anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) by intraperitoneal injection after brief isoflurane induction. The thoracic vertebral column and adjacent ribs were excised and transferred to a Sylgard dish containing ice-cold, oxygenated (95% O2/5% CO2) high-Mg2+/low-Ca2+ artificial cerebrospinal fluid (aCSF) containing (in mM): 128 NaCl, 1.9 KCl, 13.3 MgSO4, 1.1 CaCl2, 1.2 KH2PO4, 10 glucose, and 26 NaHCO3. Following complete laminectomy and vertebrectomy, the spinal cord and dorsal roots were exposed and removed. The remaining thoracic chain ganglia, in continuity with communicating rami, spinal nerves, and ventral roots, were cleaned of excess fat and muscle. The tissue was transferred to a Sylgard recording chamber superfused with oxygenated aCSF containing (in mM): 128 NaCl, 1.9 KCl, 1.3 MgSO4, 2.4 CaCl2, 1.2 KH2PO4, 10 glucose, and 26 NaHCO3 at approximately 40 ml/min and 28°C, then allowed to rest for 1 h.

Recordings were made with Ag/AgCl pellet reference electrodes and suction electrodes containing chlorided silver wires. DC recordings were acquired with a DC-coupled amplifier at the indicated gain with no high-pass filtering. Preliminary recordings made through a 2 M NaCl agarose salt bridge produced results qualitatively similar to those obtained with the Ag/AgCl pellet placed directly in the chamber. Glass recording electrodes were placed on the interganglionic nerve at the caudal cut end of the T12 ganglion and on the adjacent splanchnic nerve using trumpet-shaped electrodes with progressively reduced internal diameters reaching approximately 100-125 μm (Halder et al., 2021). Glass suction stimulation electrodes were positioned on the T10 ventral root (200-250 μm tip diameter), and the distance between stimulation and recording sites was 5 mm (Fig. 1A). Dual IGN and splanchnic nerve recordings were chosen to compare responses of SPNs having axons presumed to undergo branching with the chain ganglia (IGN) and those presumed to be non-branching axons of SPNs projecting to prevertebral ganglia (splanchnic). While available evidence supports this view (Baron et al., 1985a; Baron et al., 1985b; Baron et al., 1985c; Janig and McLachlan, 1986; Jänig, 2022a), there is currently no direct evidence.

Figure 1. Experimental conditions.

[A] Experimental setup showing SPN electrical stimulation in the T10 ventral root (VR) while recording evoked CAPs from the caudal T12 IGN and the splanchnic nerve. [B] Synaptic blocking actions of hexamethonium (100 μM) were compared with high-Mg²⁺ (6.5 mM)/low-Ca²⁺ (0.85 mM) solution. Supramaximal T10 ventral-root stimulation evoked responses were recorded from T12 interganglionic and splanchnic nerves. [B₁] Example showing that hexamethonium and ionic block of synaptic transmission produced comparable reductions in evoked response; thick traces show averages overlaid on five raw responses. [B₂] Paired t tests were used to compare CAP reductions from baseline. IGN CAPs were reduced by 51.9 ± 0.8% in high-Mg²⁺/low-Ca²⁺ solution (p < 0.001) and by 51.0 ± 1.9% in hexamethonium (p < 0.001; n = 4). Splanchnic CAPs were reduced by 34.4 ± 18.1% and 41.6 ± 16.9%, respectively (p < 0.01; n = 3). Reductions produced by the two synaptic-blocking conditions were not detectably different. [C] Hexamethonium (100 μM) blocked spontaneous synaptic activity in thoracic-chain postganglionic neurons (n = 3). Synaptic block is shown here in two separate neurons as loss of EPSCs in voltage clamp [C₁] and EPSPs in current clamp [C₂]. [D] Strength-recruitment relations of SPN axons. Following synaptic isolation with hexamethonium, T10 ventral-root constant-current stimuli were delivered at increasing intensities and durations to assess recruitment patterns and identify supramaximal intensities (at ~30 sec intervals between different intensities). Responses at the stimulus intensity and duration used for subsequent experiments (200 μA, 500 μs) are shown in red, and these responses are also pasted as transparencies over the nearest lower stimulus intensity and the higher intensities tested.

Because population axonal responses could include synaptically mediated postganglionic components, synaptic activity was blocked with hexamethonium bromide (100 μM), a nicotinic acetylcholine receptor antagonist (Mason, 1962), and pancuronium (20 μM) was included to prevent possible intercostal muscle activity. Hexamethonium efficacy was previously demonstrated in this preparation (Halder et al., 2021; Li et al., 2026). Here, we further verified synaptic block by comparing hexamethonium with high-Mg²⁺/low-Ca²⁺ solution, which blocks synaptic transmission (Vyklicky et al., 1976; Shreckengost et al., 2010). Superfused aCSF was replaced with high-Mg²⁺/low-Ca²⁺ solution for at least 30 min, followed by wash-in of normal aCSF and then 100 μM hexamethonium for 30 min. Hexamethonium reduced the evoked CAP response to a degree comparable to high-Mg²⁺/low-Ca²⁺ solution (Fig. 1B). To further support use of hexamethonium to isolate the preganglionic volley, whole-cell recordings were obtained from postganglionic neurons within the chain (n = 3) using established methods (McKinnon et al., 2019; Halder et al., 2021). In all three neurons, hexamethonium (100 μM) abolished spontaneous postsynaptic responses (Fig. 1C).

Temperature and activity-dependence experiments: For temperature experiments, bath temperature was maintained at 36°C by a Peltier device (built in-house by Mr. Bill Goolsby) before the start of data acquisition. Each protocol consisted of three separate stimulus trains delivered at 120 s intervals; at this interval, evoked responses recovered to pre-train values. For each trial, baseline was recorded for 2 s before a pre-train stimulus delivered 1 s before the 20 s stimulus train (5 Hz or 20 Hz), with a post-train stimulus delivered 2 s after the train to assess short-term history-dependent changes (Figs. 4A, 5A). Quantified epochs were the first response in the train (A), the response 1 s into the train (B), the last response in the train (C), and the response evoked 2 s after the train (D). The bath was then cooled to 22°C to assess temperature-dependent changes in preganglionic conduction. Tissue was allowed to stabilize at 22°C for at least 15 min before data acquisition. Recorded signals were amplified (5000x), low-pass filtered at 3 kHz using in-house amplifiers, and digitized at 50 kHz (Digidata 1322A, Molecular Devices) using pClamp software (v. 10.7, Molecular Devices).

Figure 4.

Effect of 5Hz and 20 Hz trains on evoked response at 22°C.

[A] Experimental design. [A₁] 5Hz, 20 second stimulus train. [A₂] 20Hz, 20 second stimulus train. Quantified epochs are expressed as % of the pre-train test response. Epoch (A), the first response in the train occurs 1 second (1 Hz) after a pre-train evoked response. Epoch (B) is the response amplitude 1 second into the train, epoch (C) is the last response in the train, and epoch (D) is the response evoked 2 seconds after the train. Responses for each epoch were divided into faster- and slower-conducting components and expressed as percentages of the rectified-integral value of the pre-train response. [B] Frequency-dependent effects are shown as box-and-whisker plots indicating median, interquartile range, and minimum-to-maximum whiskers. [B₁] At 5 Hz, faster axon CAPs did not reach α = 0.05 relative to baseline in either IGN or splanchnic nerves. Slower axons exhibited transient depression during the train followed by recovery after 2 s. Relative to the test response, slow IGN axons were reduced to 40 ± 5% (p < 0.05) and 29 ± 8% (p < 0.001) in epochs B and C, respectively, whereas splanchnic slow axons were reduced to 66 ± 7% (p < 0.05) and 66 ± 10% (p < 0.05) in the same epochs. [B₂] At 20 Hz, faster axons in both nerves did not reach α = 0.05 relative to baseline, whereas slower axons were more variable and were reduced in the IGN at epoch C (26 ± 10%; (p < 0.001). Slow-conducting units from epochs A-B were excluded from the 20 Hz analysis because CAP arrival overlapped with subsequent stimulus artifacts. Statistical testing used a baseline-normalized ratio design where test pulse amplitude is set to 1 as the reference point, with statistical testing undertaken as one-sample t-tests of A-D against test pulse amplitude. Separately, repeated-measures ANOVA with post hoc tests evaluate differences between A-D. (n = 7; *, p < 0.05; **, p < 0.01; ***, p < 0.001).

Figure 5.

Effect of 5Hz and 20Hz trains on evoked response at 36°C.

[A] Experimental design. [A₁] 5Hz, 20 second stimulus train. [A₂] 20Hz, 20 second stimulus train. Quantified epochs are expressed as % of the pre-train test response. Epoch (A), the first response in the train occurs 1 second (1 Hz) after a pre-train evoked response. Epoch (B) is the response amplitude 1 second into the train, epoch (C) is the last response in the train, and epoch (D) is the response evoked 2 seconds after the train. Responses for each epoch were divided into faster- and slower-conducting components and expressed as percentages of the rectified-integral value of the pre-train response. [B] Frequency-dependent effects are shown as box-and-whisker plots indicating median, interquartile range, and minimum-to-maximum whiskers. [B₁] At 5 Hz, fast axon CAPs in both IGN and splanchnic nerves did not reach α = 0.05 relative to baseline, but responses at the end of the train after conditioning (D) were facilitated in slower-conducting axons in both IGN and splanchnic nerves (138 ± 11% [p < 0.01] and 152 ± 28% [p < 0.05], respectively). [B₂] At 20 Hz, fast axons in both nerves exhibited increased responses at the end of the train (C), and this increase was sustained 2 s after the train (D). For fast IGN axons, facilitation was 128 ± 6% (C) and 127 ± 4% (D) (p < 0.01 and p < 0.001, respectively); for splanchnic axons, facilitation was 119 ± 4% [p < 0.01] (C) and 116 ± 5% (p < 0.05) (D). Slow-conducting responses were more variable. Statistical testing used a baseline-normalized ratio design where test pulse amplitude is set to 1 as the reference point, with statistical testing undertaken as one-sample t-tests of A-D against test pulse amplitude. Separately, repeated-measures ANOVA with post hoc tests evaluate differences between A-D. (n = 8; *, p < 0.05; **, p < 0.01; ***, p < 0.001).

Whole-cell patch clamp recordings.

Details of the dissection and approach have been described previously (McKinnon et al., 2019; Halder et al., 2021). Briefly, the intact sympathetic chain was then separated by severing rami in dissection solution and then transferred and pinned down into a clear Sylgard recording dish. Recordings were conducted at room temperature in oxygenated (aCSF) and postganglionic neurons were identified using an upright microscope (Olympus, BX51WI) equipped with a low-light camera (Olympus, OLY-150). Intracellular electrode solution contained [in mM]: K-gluconate [140], EGTA [11], HEPES [10], CaCl₂ [1.32], ATP [4] and GTP [1]. Osmolarity ranged from 284 to 290 mOsm and pH was adjusted to 7.3 using KOH. Signals were amplified using a MultiClamp 700A and digitized at 10 kHz using a Digidata 1322A and Clampex software (Molecular Devices). The measured liquid junction potential of ~10mV was subtracted in membrane potential measurements.

Analyses.

Evoked SPN CAPs were recorded after T10 ventral-root stimulation. In experiments assessing individual responses, stimuli were evoked once every 60 s. Experiments involving trains of stimuli had intervals of 120 s. Trains were delivered at 5 and 20 Hz for 20 seconds and included a pre-train event delivered one second prior and a post-train event delivered two seconds afterwards. Supramaximal recruitment of slow-conducting axons was targeted with constant-current electrical stimuli delivered at 200 μA for 500 μs, because longer pulse durations improve recruitment of unmyelinated axons (Thompson et al., 1990; Brocker and Grill, 2013; Idlett et al., 2019). This stimulus intensity and duration were chosen after testing strength-duration recruitment relationships (Fig. 1D); aside from a tendency for lower intensities to recruit faster components first, stimulus intensity did not reliably separate fast from slow components. Stimulus artifact amplitude remained stable during trains, providing an internal control against declining stimulus delivery during repetitive stimulation.

Recorded CAPs were divided into fast and slow conduction-latency windows (Li and Horn, 2006). Because temperature and repetitive stimulation alter conduction velocity, non-adjacent latency windows were selected to capture the fastest and slowest components while minimizing overlap between components (Table 2; Fig. 2B,C). All analysis was undertaken following high pass filtering (25Hz single pole RC) to remove any possible contribution from subthreshold postganglionic population synaptic potentials (Li et al., 2026). For each condition, episodes were rectified and integrated in Clampfit. Background noise was subtracted using an equal-duration baseline window, and the episode rectified integrals were averaged before statistical analysis. Unless otherwise stated, for 20 Hz trains, the slower response overlapped with subsequent stimulus artifact during early train epochs; therefore, slow-component analysis at 20 Hz was restricted to the final stimulus-evoked train response and the 2 s post-train response. For comparisons of baseline-normalized responses with the reference value of 1.0, one-sample t tests were used. Differences among epochs A–D were assessed using one-way repeated-measures ANOVA followed by Bonferroni-adjusted comparisons. Paired t tests were used for direct within-preparation comparisons between two conditions. Responses from repeated trains were averaged within each preparation before statistical testing. The criterion for rejecting the null hypothesis was α = 0.05; P values are reported using the following thresholds: P < 0.05, P < 0.01, and P < 0.001.

Table 2.

Conduction time ranges from start of stimulus used to quantify magnitude changes in fast and slow conducting axons at 22° and 36°C.

T° & category	Conduction velocity range	Epoch duration
22°C fast	2.2-0.31 m/s	2.0-14 ms from stim. onset
22°C slow	0.23-0.06 m/s	19-70 ms from stim. onset
36°C fast	3.4-0.58 m/s	1.3-7.6 ms from stim. onset
36°C slow	0.44-0.11 m/s	10-40 ms from stim. onset

Figure 2.

Increased temperature leads to reduced axonal recruitment.

[A] Graph derived from Table 1 in Pinto et al (2008), estimating the changes in conduction velocity for Aδ and C-fibers, which exhibit conduction velocities comparable to SPN fibers observed in mice. Using trend line formulas and experimental data collected at 22°C (fast: 0.31-2.2m/s [2-14ms from stim onset]; slow: 0.06-0.23m/s [19-70ms from onset]), predicted changes to conduction velocity ranges were extrapolated for fast (0.58-3.4m/s) [1.3-7.6ms from onset] and slow populations (0.11-0.44 m/s) [10-40ms from onset] at 36°C as shown in temp and frequencies comparison studies summarized in Table 2. [B] Example T° differences in CAP appearance and separation into fast and slow groups. Fast and slow SPNs within conduction velocity ranges above are highlighted with bars and shaded boxes at 36°C (red) and 22°C (blue). [C] Scaling approach shown in another example recording. [C₁] shows difference in responses at 36°C and 22°C with colored bars identifying conduction ranges for fast and slow CAPs as in [B]. [C₂] shows overlaid volleys with 36°C response scaled to 22°C using pre-determined scaling factors (shown). Note preferential loss of slower responses. [D] Evoked responses from 36°C and 22°C were rectified, integrated, scaled, and compared. The percent reduction in magnitude of evoked response was determined for fast and slow-conducting epochs of the CAP. The IGN slower-conducting axons exhibited a substantially higher % reduction in evoked response (48.1 ± 6.4%; p<0.001) compared to that seen in faster-conducting axons (75.3 ± 5.1%; p<0.05) [n=6]. For the splanchnic nerve these values were reduced by 50.0 ± 9.1% (p<0.01) and 63.7 ± 8.0%, (p<0.05), respectively [n=5]. (*, p<0.05; **,p<0.01; ***, p<0.001; paired t-tests).

RESULTS

Temperature-dependent reduction in population recruitment of SPN axons.

We first examined the effect of temperature on SPN axonal conduction by comparing evoked compound action potentials (CAPs) recorded at room temperature (22°C) and physiological temperature (36°C). Conduction velocities increased with temperature, approximately doubling between 22°C and 32°C (Table 1), consistent with established temperature dependence of axonal conduction.

Table 1.

Conduction velocity range (in m/s) of thoracic SPN axons at 22° and 36°C.

	22°C	36°C
IGN (n=6)	0.10±0.001 - 0.40±0.04	0.19±0.02 - 0.81±0.06
Splanchnic (n=5)	0.12±0.02 - 3.1±1.5	0.21±0.03 - 7.0±3.9

To compare responses across temperatures, CAPs recorded at 36°C were scaled to account for temperature-dependent changes in conduction velocity, spike amplitude, and duration. Using the framework provided by Pinto et al 2008 (Pinto et al., 2008), which quantified temperature-dependent changes in Aδ- and C-fiber conduction velocity, we adjusted the expected latency windows from 22°C to 36°C (Fig. 2A–C). Then, we scaled the amplitude and width by factors of 1.53 and 1.87, respectively, based on prior measurements of temperature-dependent changes in spike waveform (Hodgkin and Katz, 1949; Stegeman and De Weerd, 1982) (Fig. 2C). In preliminary fitting, the width factor matched the observed change well, whereas the amplitude factor tended to overestimate the 36°C response. We nevertheless retained this scaling because it biases against detecting a reduction at 36°C. Following scaling, CAP magnitude at 36°C was statistically reduced relative to 22°C in the interganglionic (IGN) and in splanchnic nerve recordings. In the IGN, overall responses were reduced to 48.8 ± 6.4% of scaled 22°C values (p < 0.001, n = 6), and in the splanchnic nerve to 52.8 ± 7.9% (p < 0.01, n = 5). These data indicate a substantial temperature-dependent decrease in effective axonal recruitment, consistent with reduced conduction reliability rather than a simple temperature-dependent change in spike waveform.

To determine whether temperature-dependent reductions in CAP magnitude were distributed uniformly across axonal populations, responses were separated into fast- and slow-conducting components based on conduction latency (Table 2).

At 36°C, reductions in CAP magnitude were greater in slower-conducting components than in faster-conducting components. In the IGN, slower-conducting responses were reduced to 48.1 ± 6.4% (p < 0.001) of scaled 22°C values, whereas faster-conducting responses were reduced to 75.3 ± 5.1% (p < 0.05; Fig. 2D). Similar trends were observed in the splanchnic nerve (slow: 50.0 ± 9.1% [ p < 0.01]; fast: 63.7 ± 8.0%). These findings indicate preferential temperature-sensitive loss of slower-conducting axons.

Maximal recruitment does not change with temperature.

We next assessed whether the temperature-dependent reduction in CAP magnitude could be explained by failure to recruit axons at the stimulation site. In related strength-duration experiments (Halder et al., 2026), 200 μA, 500 μs stimulation produced suprathreshold or near-maximal axonal recruitment in 8/10 experiments at 22°C and 12/13 experiments at 32-36°C. An example of recruitment testing at several intensities below and above the chosen value of 200 μA, 500 μs is shown at 32°C (Fig. 1D). Responses evoked at intensities and durations below and above the selected stimulus parameters did not reveal appreciable additional recruitment.

Frequency-dependent modulation of CAP magnitude over the first 9 events.

We examined the effects of activation history on SPN conduction by delivering a pre-pulse stimulus 1 second prior to stimulus trains of 5 Hz and 20 Hz for 20 seconds. Inspection of DC-coupled recordings during 5-Hz (Fig. 3A₁) and 20-Hz trains (Fig. 3A₂) revealed progressive loss of distinguishable population events without a corresponding uniform decline in the amplitudes. Stimulus trains displayed little sustained polarization and (Fig. 3A₂).

Figure 3.

Effect of stimulus frequency on early changes in evoked CAP response amplitude during stimulation at 22° and 36°C.

(A) Examples of DC recordings of evoked responses in the IGN during 5 and 20 Hz trains at 22°C. (A₁) Rasters of successive responses evoked in two mice. The first evoked event was delivered one second prior to delivery of a stimulus train at 5Hz (the first 30 events are shown). Note that frequency-dependent depression of the population response reflects progressive loss of individual spiking events (unit dropout) rather than a decline in the amplitude of individual surviving units. (A₂) Top. Two-20 second trains delivered in the same mouse. Note that the stimulus trains do not lead to pronounced changes in DC. Bottom. Raw and high-pass filtered (40Hz RC) magnification of the first 27 events in a stimulus train. Stimulus artifact was reduced (top panel) or masked (bottom panel) but identified by blue vertical lines to aid identification individual spikes for amplitude comparison. Reference horizontal dotted line aids comparison spike amplitudes over time. (B) Effect of stimulus frequency at 5 and 20 Hz on early evoked responses in stimulus trains at 22°C (B₁) and 36°C (B₂). CAP responses were normalized to the first evoked response in the train and compared across temperature and frequency in IGN and splanchnic nerves. Activation history affected responses during the train (events 1-9) in all conditions in both nerves (n = 4; repeated-measures ANOVA; *, p < 0.05; **, p < 0.01; ***, p < 0.001). Pairwise differences are shown between individual events and from the first response in the train (Bonferroni t test). Note that responses tended to stabilize by the 5th-6th pulse. Relative changes were comparable within nerves across temperatures and frequencies, but IGN axons consistently underwent greater frequency dependent reduction in CAP regardless of temperature. Post-hoc one-tailed paired t-tests found significant differences between IGN and splanchnic at 5Hz at 22°C events 8 and 9 and 20Hz at 36°C event 8 (all p < 0.05). (*, p < 0.05; **, p < 0.01; ***, p < 0.001).

At 22°C, CAP magnitude exhibited overall frequency-dependent depressions when examined during the first nine events during stimulus trains in both IGN and splanchnic recordings (Fig. 3B₁). Depression generally increased with stimulus frequency and stabilized by the 5th–6th pulse. Overall depression was larger in IGN compared to splanchnic and reached statistical significance at 5 Hz (events 8 and 9).

At 36°C, the overall trajectory of CAP changes during stimulus trains was similar to that observed at 22°C, although response magnitudes differed across temperature and frequency (Fig. 3B₂). Depression generally increased with stimulus frequency and stabilized by the 5th–6th pulse. Overall depression was larger in IGN compared to splanchnic and reached statistical significance at 20 Hz (event 8).

Frequency-dependent modulation of fast and slower conducting axons at 22°C and 36°C.

Axonal subpopulations were then analyzed at defined epochs during the train (Fig. 4A). Epoch (A) occurs 1 second (1 Hz) after a pre-train evoked response. Epoch (B) is the response amplitude 1 second into the train, epoch (C) is the last response in the train, and epoch (D) is the conditioned response evoked 2 seconds after the train.

For faster conducting SPNs, there was no effect of the pre-train response on the amplitude of the first response in the train delivered 1 second later (1 Hz). However, slower conducting responses at 22°C were depressed at 1 Hz in both IGN and splanchnic nerves to 71.7 ± 4.6% (p<0.001) and 84.5 ± 5.0% (p<0.05) of pre-train values, respectively (Table 3). Conversely, at 36°C, the slower conducting splanchnic response was facilitated at 1 Hz to 116.2 ± 5.5% (p<0.05) of pre-pulse amplitude (Table 4).

Table 3.

22°C results of stimuli delivered at 1 Hz and at 5 & 20 Hz trains for 20 seconds (% of test response). Quantified epochs in IGN and splanchnic nerve (Spl.), expressed as mean % of the test response (± SEM), where the first response in the train delivered after a 1 second test pulse (A), the response 1 s into the train (B), the last response in the train (C), and the conditioned response evoked 2 s post-train (D).

N=7	Fast SPNs (0.31-2.2 m/s)		Slow SPNs (0.06-0.23 m/s)
IGN	1Hz		1Hz
A	98.9 ± 2.5		71.7 ± 4.6***
	5Hz	20Hz	5Hz	20Hz
B	94.8 ± 6.8	103.3 ± 18.6	40.2 ± 4.7***‡
C	92.2 ± 13.6	102.5 ± 3.3	29.1 ± 8.0***†	25.7 ± 10.4***
D	94.0 ± 11.1	101.4± 3.6	102.8± 13.6	56.2 ± 19.1
Spl.	1Hz		1Hz
A	101 ± 1.1		84.5 ± 5.0*
	5Hz	20Hz	5Hz	20Hz
B	99.2 ± 2.2	102.2 ± 2.2	65.7 ± 6.6**‡
C	96.1 ± 8.4	92.4 ± 9.2	65.5 ± 10.4*†	109.7 ± 65.0
D	97.5 ± 7.0	89.5 ± 11.5	108.2± 12.1	90.6 ± 23.8

^*(, p<0.05;

^**, p<0.01,

^***; and p<0.001, respectively)

^{‡ and †}, p<0.05 differences between IGN and Spl. 5Hz slow SPNs at epoch B and C, respectively). 1 Hz values were captured from the pre-pulse and first response of the trains at 5Hz.

Table 4.

36°C results of stimuli delivered at 1 Hz and at 5 & 20 Hz trains for 20 seconds (% of test response). Quantified epochs in IGN and splanchnic nerve (Spl.), expressed as mean % of the test response (± SEM), where the first response in the train delivered after a 1 second test pulse (A), the response 1 s into the train (B), the last response in the train (C), and the conditioned response evoked 2 s post-train (D).

N=8	Fast SPNs (0.58-3.4 m/s)		Slow SPNs (0.11-0.44 m/s)
IGN	1Hz		1Hz
A	93.9 ± 4.6		103.2 ± 4.9
	5Hz	20Hz	5Hz	20Hz
B	92.9 ± 4.3	95.3 ± 4.0	89.4 ± 6.2‡‡‡
C	104.5 ± 5.0	127.7 ± 6.1**	92.2 ± 9.9	53.9 ± 19.7
D	106.0 ± 5.6	127.0 ± 4.3***	138.4 ± 10.5**	149.7 ± 24.8*
Spl.	1Hz		1Hz
A	95.7 ± 3.8		116.2 ± 5.5*
	5Hz	20Hz	5Hz	20Hz
B	92.5 ± 3.8	93.3 ± 3.8	118.0 ± 7.5*‡‡‡
C	101.8 ± 3.8	118.6 ± 4.4**	111.0 ± 7.3	182.1± 93.4
D	103.0 ± 4.3	116.1 ± 4.9*	151.6 ± 28.3	385.6 ± 231.3*

^*(, p<0.05;

^**, p<0.01,

^***; and p<0.001, respectively)

^‡‡‡(, p<0.001 differences between IGN and Spl. 5Hz slow SPNs at epoch B). 1 Hz values were captured from the pre-pulse and first response of the trains at 5Hz.

At 22°C, slower-conducting components exhibited pronounced reductions in CAP magnitude during trains, whereas faster-conducting components were largely unaffected (Fig. 4B). For example, in the IGN at 5 Hz, slower-conducting responses were reduced to 40.2 ± 4.7% and 29.1 ± 8.0% of baseline at 1 second and at the end of the train (epochs B and C, respectively; p < 0.05 and p < 0.001), whereas faster-conducting responses did not reach α = 0.05 relative to baseline (Fig. 4B₁). Similar patterns were observed at 20 Hz, where slower-conducting IGN responses were reduced to 25.7 ± 10.4% at the end of the train (epoch C; p < 0.001; Fig. 4B₂).

As we were unable to selectively quantify the slower-conducting volley components during the 20 Hz train due to overlap with stimulus artifact, comparison of ‘pre-train’ evoked responses are only compared to the final response in the pulse train (epoch C). A summary of results obtained at 22°C is provided in Table 3.

Analysis of fast- and slow-conducting components revealed additional temperature-dependent differences at 36°C (Fig. 5). At 5 Hz (Fig. 5 B₁), faster-conducting responses in both IGN and splanchnic nerves did not reach α = 0.05 relative to baseline during the 20 s train. Slower-conducting responses also did not show statistically detectable depression during the train but exhibited statistical post-train increases 2 s after stimulation (epoch D).

At 20 Hz (Fig. 5B₂), faster-conducting responses statistically increased during trains and remained elevated 2 s after the stimulus train. In the IGN, faster-conducting responses reached 127.7 ± 6.1% of baseline at the end of the train (p < 0.001) and remained statistically elevated at 127.0 ± 4.3% 2 s post-train (p < 0.001). Similar increases were observed in the splanchnic nerve during (p < 0.01) and in the 2 s post-train conditioned response (p < 0.05). Slower-conducting responses were more variable and did not show consistent changes except for the IGN conditioned response 2 s post-train (p < 0.01). A summary of results obtained at 36°C is provided in Table 4.

Interaction between temperature and activation history.

Comparisons across temperatures indicate that frequency-dependent effects on CAP magnitude differ between axonal populations. At 22°C, slower-conducting components exhibited pronounced depression during repetitive stimulation, whereas at 36°C these components were already reduced in magnitude under baseline conditions compared to 22°C (Fig. 2D) and did not show statistically detectable depression during trains. In contrast, faster-conducting components were better preserved at 36°C (Fig. 2D) and exhibited stable or increased responses during high-frequency stimulation. Regardless of whether post-train conditioned responses measured 2 s later differed from baseline, all responses returned to baseline within the 120 s inter-episode interval, indicating short-term and reversible activity-dependent modulation.

Branching adds a secondary vulnerability that becomes most apparent during repetitive activation. At 5 Hz, overall depression was greater in IGN compared to splanchnic both at 22°C and 36°C, (Fig 3B). Moreover, slow IGN axons depressed more than slow splanchnic axons during 5 Hz trains at 22°C after 1 second during (epoch B; p <0.05) and at the end of the train (epoch C; p <0.05 [Table 3; Fig. 4B₁).

DISCUSSION

Classical views assume SPN axons reliably relay centrally generated activity to postganglionic targets. In a separate study, we reported that there is a high recruitment variability at 22°C indicating this assumption does not hold (Halder et al., 2026). The present study shows that recruitment is further compromised at normal body T°. At 36°C, the population compound action potential (CAP) was reduced by approximately 50% relative to room temperature, with preferential loss of slower-conducting axonal components. These observations indicate a substantial reduction in effective axonal conduction at physiological temperature and suggest that spike propagation along SPN axons is highly vulnerable to temperature-dependent disruption. Given that prior studies examining multisegmental sympathetic transmission were conducted at room temperature (Blackman and Purves, 1969; Nja and Purves, 1977; Lichtman et al., 1979, 1980; Purves and Lichtman, 1980), these findings raise the possibility that the extent of SPN influence across paravertebral ganglia has been overestimated.

Earlier recordings on the effects of temperature and repetitive stimulation from rabbit cervical sympathetic trunk provide important context for the present findings (Ritchie and Straub, 1956a, b; Ritchie and Straub, 1957). These studies showed that cooling increased unmyelinated axon CAP amplitude and duration and that repetitive stimulation can produce an afterdepolarization followed by a long-lasting afterhyperpolarization that is sensitive to ouabain. The cervical sympathetic trunk is predominantly preganglionic but also contains postganglionic axons from the middle and inferior cervical ganglia (Wang et al., 2018), which complicates direct comparison with the present synaptically isolated preganglionic axonal population. Nonetheless, their observations support the central premise that temperature and recent activity can alter the waveform and excitability state of sympathetic axons. In contrast to their findings, we did not observe long-lasting post-stimulation hyperpolarization shifts in our SPN CAP recordings, but the analogous event has been seen in postganglionic neurons in thoracic (McKinnon et al., 2019) and in superior cervical ganglia (Lees and Wallis, 1974) where it was shown to be ouabain-sensitive.

The magnitude of the reduction in CAP amplitude at physiological temperature is most consistent with a loss of propagating spikes across a substantial fraction of SPN axons. After temperature-dependent changes in conduction velocity and spike waveform were accounted for by scaling, the remaining decrease in response magnitude exceeded that expected from these biophysical effects alone. This suggests that a substantial subset of axons fails to sustain action potential propagation at higher temperatures. This interpretation must be considered alongside alternative explanations, including altered axonal excitability, elevated recruitment thresholds, or activation of temperature-sensitive conductances that reduce propagation safety factor. However, the absence of major temperature-dependent changes in maximal stimulus recruitment argues against reduced stimulation recruitment efficacy as the primary explanation. Thus, the data supports temperature-dependent reductions in conduction reliability while acknowledging that CAP measurements cannot resolve individual axonal failure directly (but see (Halder et al., 2026)).

Temperature- and branching-dependent vulnerabilities appear to be separable.

At 36°C, both fast- and slow-conducting SPN axon populations showed temperature-sensitive reductions, indicating that warming compromises conduction reliability across both IGN and splanchnic pathways. If branching were the primary determinant of temperature sensitivity, IGN axons would be expected to show substantially greater loss than splanchnic axons. Instead, the data indicates that both populations contain small-diameter unmyelinated axons that share vulnerability to warming-dependent mechanisms, including reduced spike duration and reduced electrotonic current spread. Branching adds a secondary vulnerability that becomes most apparent during repetitive activation.

Temperature-sensitive leak channels and spike fidelity.

RNA sequencing shows that the two-pore-domain potassium (K2P) channel TREK-1 (KCNK2), is present in thoracic SPNs (Alkaslasi et al., 2021). TREK-1 channel activity increases with increasing T°, with peak thermal responsiveness from 32-37°C, with ~7-fold increase in current amplitude for every 10°C, indicating that physiological temperature variations can substantially influence its activity (Maingret et al., 2000). Increased TREK-1 channel activity may compromise spike conduction through two distinct mechanisms. (i) Hyperpolarization towards E_K would necessitate a larger depolarization for spike recruitment. (ii) An increased membrane conductance would reduce passive voltage amplitude and propagation space constant for regenerative spiking by downstream Na_V channels. Both hyperpolarization and increased membrane conductance would be expected to reduce the safety factor for spike propagation. In small-diameter axons, where electrotonic spread is already limited, such changes could readily precipitate conduction failure. While the present data are consistent with this mechanism, direct experimental manipulation of these channels will be required to establish their specific contribution.

Preferential vulnerability of small, slow-conducting axons.

The preferential vulnerability of slower-conducting axons provides important insight into the structural determinants of this phenomenon. SPN axons are predominantly unmyelinated and extremely small in diameter, with many projecting across multiple paravertebral ganglia via branching pathways (Lewis and Burton, 1977; Halder et al., 2026). Smaller diameter axons are expected to operate with a reduced safety factor for propagation and should be particularly sensitive to perturbations in membrane conductance and spike waveform. Consistent with this, the greatest temperature-dependent reductions in CAP magnitude were observed in slower-conducting populations, which are likely to include the smallest-caliber axons. During repetitive stimulation, slow-conducting IGN responses were more strongly depressed than slow splanchnic responses, consistent with an additional vulnerability associated with axonal geometry or branching (Levy, 1980; Debanne et al., 2011). Together, these results identify axonal diameter and branching architecture as key determinants of temperature- and frequency-sensitive conduction reliability in SPNs.

Temperature-dependent selective filtering of axonal populations and apparently silent SPNs.

The interaction between temperature and activation history further suggests that changes in population response reflect selective filtering of axonal subpopulations rather than uniform modulation of all fibers. At room temperature, slower-conducting axons were present and exhibited pronounced frequency-dependent depression during stimulus trains, consistent with a reduced excitability described in unmyelinated fibers (Renganathan et al., 2002; De Col et al., 2008; Obreja et al., 2010; Rama et al., 2018). At physiological temperature, many vulnerable slow components were already reduced, and the remaining responses were dominated by faster-conducting axons that could sustain or even facilitate responses during higher-frequency stimulation. Thus, the apparent shift from depression at 22°C to facilitation at 36°C may reflect a change in the composition of the active axonal population rather than a reversal of intrinsic frequency-following properties. This mechanism could contribute to the observation that many anatomically identified SPNs show little ongoing or reflex activity under some in vivo conditions (Jänig, 2022b), although direct in vivo testing will be required.

Changes in core body temperature may influence SPN recruitment.

That SPN conduction undergoes more failures at 36°C than at 22°C may have ethological relevance in mice. While core body temperature range for humans and mice are similar (Reitman, 2018), mice are heterothermic and can enter hypometabolic torpor where core body temperature can reach 20°C (Geiser, 2004; Hrvatin et al., 2020). Given that mice exhibit large physiological fluctuations in body temperature during torpor, temperature-dependent modulation of SPN conduction could potentially contribute to adaptive autonomic regulation. For example, as reduced temperatures promote conduction, this may adaptively strengthen sympathetic drive to cutaneous vasoconstrictors for heat retention and to brown adipose tissue for heat generation through non-shivering thermogenesis (Enerbäck, 2009; Madden and Morrison, 2016; Francois et al., 2019). Additional studies would benefit from more detailed analysis of the effects of smaller temperature changes on the spike conduction in functional subpopulations of thoracic SPNs.

Implications for sympathetic output regulation.

Rather than functioning as a purely reliable transmission system, SPN axons appear to operate with limited conduction reserve, such that changes in temperature or activation history can alter the number and identity of axons that successfully propagate spikes. This introduces a mechanism of presynaptic gain control based on conduction reliability, whereby the effective size of the recruited postganglionic population can be dynamically regulated. Because depression was greatest in slow-conducting axons during trains, these axons may be preferentially suited to low-frequency population-encoded output rather than sustained high-frequency signaling. Given this, one target population may be the smaller NPY-expressing vasomotor postganglionic neurons (Furlan et al., 2016) as observations from human microneurography show that individual postganglionic sympathetic neurons are recruited probabilistically across cardiac cycles (Macefield and Wallin, 1999, 2018), suggesting that variability in sympathetic postganglionic output may arise, in part, from upstream constraints on preganglionic conduction.

Experimental considerations and limitations.

Several factors should be considered when interpreting results obtained with this experimental approach. (1) Use of hexamethonium to block synaptic transmission. These studies observed comparable effects on CAP using hexamethonium or a high-Mg²⁺/low-Ca²⁺ solution used to block synaptic transmission (Vyklicky et al., 1976; Shreckengost et al., 2010). Together, these findings and the whole-cell verification of synaptic block with hexamethonium strongly support isolation of the preganglionic volley. Still, a minor residual contribution from postganglionic activity in IGN recordings cannot be completely excluded. (2) Several factors can impact use of change in CAP magnitude as an estimate of conduction failures. (a) Temperature dependent waveform cancellation and scaling. Conclusions regarding temperature dependent conduction failure are inferred from population CAP measurements rather than direct recordings from individual axons. Scaling procedures were used to account for temperature-dependent changes in spike waveform, but the scaling framework was developed for simpler waveforms. Positive/negative phase cancellation in multiphasic CAPs introduces uncertainty into rectified-integral measurements. In the current work, this would make estimated changes conservative: cooling broadens individual fiber spikes, which should increase overlap cancellation at 22°C and thereby underestimate the difference between 22°C and 36°C. We also note that the amplitude and width scaling factors applied here were derived from squid giant axon (Hodgkin and Katz, 1949)) and predominantly fast-conducting human sural nerve fibers (Stegeman and De Weerd, 1982) and were applied uniformly to fast- and slow-conducting SPN components. If the true temperature dependence of spike amplitude and duration differs between these populations, the absolute magnitude of the fast-versus-slow difference in Fig. 2D should be interpreted cautiously. This limitation does not affect the comparison between IGN and splanchnic pathways within a matched conduction-velocity class, since identical scaling was applied to both; the similarity between IGN and splanchnic slow-component reductions is accordingly the more robust finding. (b) Frequency and temperature dependent loss of axonal recruitment at the stimulation site. The 200 μA, 500 μs stimulation intensity was supramaximal or near maximal in 8/10 experiments at 22°C and 12/13 experiments at 32-36°C. A long stimulus pulse duration of 500μs was intentionally chosen to ensure it extends beyond the chronaxie of unmyelinated axons (Idlett et al., 2019) and constant-current delivery with stable stimulus-artifact amplitude throughout trains excludes a decline in the physical stimulus delivered to the tissue. A separate possibility is that activity-dependent hyperpolarization elevates the threshold for excitation at the stimulation site during a train, such that a nominally supramaximal stimulus becomes inadequate for a subset of axons (Ritchie and Straub, 1957; Bostock and Grafe, 1985). Because CAP recordings are made downstream of the stimulation site, they cannot localize whether a given failure reflects elevated threshold for initiation or reduced safety factor for propagation; both plausibly reflect the same underlying activity-dependent hyperpolarization. Hence, we cannot fully exclude a contribution from stimulation-site threshold elevation with the present data. This ambiguity does not apply to responses exceeding the original control amplitude, including facilitation of fast IGN and slow splanchnic components at 36°C (Figs. 4, 5; Tables 3, 4), since threshold-dependent recruitment failure cannot produce responses larger than baseline. For the depression seen, two observations argue against a stimulation-site mechanism as the primary explanation: depression differed systematically between IGN and splanchnic pathways despite common stimulation conditions (Fig. 3B), consistent with a downstream, pathway-dependent process; and recruitment was comparable across a twofold range of stimulus intensity and duration at rest (Fig. 1D), indicating a safety margin that would need to be substantially eroded to account for the reported reductions. (c) Frequency dependent spike amplitude modulation. History-dependent AP amplitude modulation is a known phenomenon in unmyelinated fibers (Weidner et al., 2000; De Col et al., 2008; Obreja et al., 2010; Hoffmann et al., 2016). Such changes in AP waveform are associated with a reduced safety factor for propagation and could contribute mechanistically to the observed frequency-dependent changes in CAP magnitude (and assumed conduction failure) at the population level. However, it is possible that global reductions in voltage changes in individual spikes due to membrane potential changes contribute to observations of a reduced CAP reflect without the need for conduction failure. This is unlikely as DC shifts during stimulus trains were absent or of small value. More definitively, examination of recordings where individual putative spikes were differentiable (Fig. 3A) clearly shows a history-dependent loss of individual axons without apparent change in the amplitude of individual units. Moreover, a complementary paper (Halder et al., 2026) finds individual unit recruitment variability with conduction failures even in the absence of frequency-dependent conditioning. Thus, the major finding of the paper that quantifies summed population responses (presented as changes in the rectified integral) is consistent with probabilistic frequency-dependent failure of spike generation and/or propagation in individual units. Additional evidence against simple AP-amplitude depression as the sole explanation includes the differential effects across slow vs. fast axons, and the post-train potentiation at 36°C in slow axons. (3) Classification of axonal populations was based on conduction latency rather than direct morphological identification; the relationship among conduction velocity, axon diameter, and branching pattern therefore remains inferential. (4) Experiments were performed ex vivo and may not capture ongoing in vivo neuromodulatory influences. For example, mice are nocturnal and experiments were undertaken during daytime, and ion channel expression/activity undergoes diurnal variation to reduce excitability during inactive periods (Itri et al., 2010; Colwell, 2011; Steponenaite et al., 2024). (5) Firing frequencies and durations tested are high. Reported analyses focused on total population effects over the first 9 events, and events separated into fast and slow conduction velocities at 1s into the train, at the end of the 20s train and 2s after the train. Our quantified effects over the first second are likely to have physiological relevance based on the following. In vivo recordings of preganglionic activity identify firing frequencies typically at 0.5–3 Hz under resting/baseline conditions, with phasic bursting that rarely exceeds ~10 Hz even during reflex activation (Janig, 2022). Still, higher than average firing rates (5-25 Hz for ~0.5-1sec) are seen in SPNs during inspiration phases of respiration in anesthetized and decerebrate unanesthetized cats (Preiss et al., 1975; Polosa et al., 1988). Additionally, activation of somatic or visceral afferents can significantly elevate SPN firing rates above average (Wyszogrodski and Polosa, 1973), while hypoxic conditions prompt SPN somas to fire at rates surpassing 5-10 Hz (Rohlicek and Polosa, 1986). Thus, brief firing at 5–20 Hz can occur under selected physiological or reflex conditions. Importantly, we observed measurable depression within the first several stimuli, before completion of the prolonged trains. The 20 Hz data extend the characterization of SPN conduction at near-maximal frequencies. Prolonged repetitive stimulation undertaken at higher frequencies (e.g. 20 Hz for 15s) (Morris et al., 2005) has been shown to generate prolonged spiking in sympathetic ganglia via non- cholinergic transmission (Bronk and Pumphrey, 1935; Janig, 2005). Still, longer duration trains have limited physiological analogy and are better interpreted as a stress test of frequencyfollowing capacity, that may have relevance to future electrotherapeutic approaches.

Summary.

The present findings indicate that SPN axons operate with limited conduction reserve that is strongly modulated by temperature and activation history. Preferential vulnerability of slower-conducting, likely small-diameter and branching axons indicates that conduction failure represents a significant determinant of effective sympathetic output as a decrease in the magnitude of population responses would lead to diminished synaptic activity on postganglionic neurons. These results identify axonal conduction reliability as a physiologically regulated variable that can shape the strength and distribution of autonomic signaling.