Relating Spinal Injury-Induced Neuropathic Pain and Spontaneous Afferent Activity to Sleep and Respiratory Dysfunction

Abstract

Spinal cord injury (SCI) can induce dysfunction in a multitude of neural circuits including those that lead to impaired sleep, respiratory dysfunction, and neuropathic pain. We used a lower thoracic rodent contusion SCI model of neuropathic pain that has been shown to associate with increased spontaneous activity in primary afferents and hindlimb mechanosensory stimulus hypersensitivity. Here we paired capture of these variables with chronic capture of three state sleep and respiration to more broadly understand SCI-induced physiological dysfunction and to assess possible interrelations. Noncontact electric field sensors were embedded into home cages to non-invasively capture the temporal evolution of sleep and respiration changes for six weeks after SCI in naturally behaving mice. Hindlimb mechanosensitivity was assessed weekly, and terminal experiments measured primary afferent spontaneous activity in situ from intact lumbar dorsal root ganglia (DRG). We observed that SCI led to increased spontaneous primary afferent activity (both firing rate and the number of spontaneously active DRGs) that correlated with increased respiratory rate variability and measures of sleep fragmentation. This is the first study to measure and link sleep dysfunction and variability in respiratory rate in a SCI model of neuropathic pain, and thereby provide broader insight into the magnitude of overall stress burden initiated by neural circuit dysfunction after SCI.

Introduction

Neuropathic pain can develop after peripheral or central nervous system injury, resulting from trauma or disease.¹ Persistent and stimulus-independent neuropathic pain (i.e., spontaneous pain) develops in more than half of patients with spinal cord injury (SCI),² yet there remains limited understanding of its underlying mechanisms³ and few effective therapies for this debilitating condition.

Many studies explore neurobiological mechanisms and management for spontaneous pain using experimental assessment tools optimized for stimulus-evoked allodynia and hyperalgesia.^4–10 Clinically, pain is commonly scored through questionnaires that can be highly subjective and variable.^11,12 Pre-clinically, most modalities of pain are inferred through changes in evoked mechanical or thermal sensory sensitivity, regardless of whether the pain being explored has features independent of stimulus-evoked responses.^13,14

Stimulus-evoked assessment methods are considered the semiquantitative gold standard, but their assessment limitations may include temporal and behavioral state-dependent shifts in sensitivity that could prevent identification or misdiagnosis as spontaneous pain.³ Conversely, there is increasing evidence that spontaneous primary afferent activity represents one initiation site and chronic driver of maladaptive sensory processing, consistent with spontaneous pain perception.^6,7,15–17 Spontaneous activity in both A and C fibers has also been observed in many neuropathic pain conditions.^7,15,18,19 In SCI, this peripheral hypersensitivity can predict the development of central pain.^20,21

Key studies in rodent contusion SCI models reveal potential links between stimulus-evoked hypersensitivity (hyperalgesia/allodynia) and the incidence of spontaneous firing in cell bodies of primary afferents below the level of injury.^6,16,22,23 These studies assessed spontaneous activity from C fiber dorsal root ganglia (DRG) neurons after DRG dissociation that may artifactually enhance hyperexcitability, including increased spontaneous firing that mirrors the effects seen after chronic compression of DRGs.²⁴ These limitations highlight the need to study DRG activity in situ to further support increased SCI-induced spontaneous afferent activity increases with emergent dysfunction.⁶

To better understand and measure stimulus independent spontaneous pain, non-evoked assessment methods have been developed. For example, the conditioned place preference test^16,25–27 and grimace scale²⁸ have been shown to differentiate animals with neuropathic pain from those without. Potential confounds of these tests include efficacy of analgesics, animal training, and experimenter interactions.^14,29 There is a need for additional physiological and behavioral indices of ongoing spontaneous pain during natural behavior.³⁰

Impaired sleep is a significant, although often overlooked, component of suffering after SCI with meaningful impacts on life quality.³¹ Although patients with SCI commonly experience reduced sleep quality,^32–35 research on cause-effect relations between impaired sleep after SCI and additional important comorbid conditions is limited.

While it is well established that sleep disruption and chronic pain are bidirectionally and negatively reinforcing,^34–36 their association in SCI has not been detailed clinically or in pre-clinical animal models. This is at least partly because of the difficulty in measuring rodent sleep using methods that rely on invasive electroencephalographic recordings. Moreover, invasive surgery undoubtedly introduces added stress burden that may complicate behavioral changes and pain after SCI.

Similarly, while there are several studies describing respiratory-related sleep dysfunction caused by cervical SCI,^34,37,38 including sleep apnea,³⁹ pre-clinical work aimed at understanding whether alterations in respiration are associated with SCI-induced neuropathic pain are limited. Using lower thoracic SCI models that do not overtly impair respiratory neural circuits, studies in a restraint cylinder showed respiratory rate increases after thoracic contusion or hemisection injury that associate with the emergence of allodynia.^40,41 This important observation on respiration needs to be reproduced during natural behavior that has traditionally relied on respiratory measures obtained with whole-body plethysmography.^42,43

In the thoracic contusion SCI model of neuropathic pain, this study sought to understand interrelations between indices of spontaneous and evoked pain with changes in sleep architecture and respiratory function. These variables were captured non-invasively during natural behavior in the animal's home cage using electric field sensors^42,44 and were compared with changes in hindlimb mechanosensitivity and spontaneous afferent activity recorded from DRG neurons in situ.

Obtained results identified a correlated interplay between excitability increases in sensory systems, and both respiratory and sleep dysfunction, providing a broader perspective on possible bidirectional negative impact between sensory and homeostatic circuit dysfunction after SCI.

Methods

Animals

All procedures were approved by the Emory University Institutional Animal Care and Use Committee. Adult C57/Bl6 mice (n = 20, female) were pair housed in home cages under standard 12:12 hour light-dark cycles with ad libitum food and water.

Animal surgeries

The SCI was induced via moderate contusion of the T10 spinal cord segment.⁴⁵ Fourteen adult mice (day 205) were prepared for aseptic surgery and deeply anesthetized with 2–3% isoflurane. A midline skin incision was made to dissect the muscle and fascia before exposing the spinal cord at T8–T10 via a dorsal laminectomy. In six animals (Sham Group, n = 6), wounds were closed at this point. In the eight remaining animals (SCI Group n = 8), a moderate contusion SCI was made dorsally at T9/10 with the Infinite Horizon impactor (50 kdynes, IH-400, Precision Systems and Instrumentations) before wound closure.

All mice received post-operative pain relief (2 mg/kg meloxicam), antibiotic agents (2.5 mg/kg Baytril), and supplemental hydration (0.5 mL sterile saline daily). Manual bladder expression was undertaken twice daily until each animal voided independently. One animal was excluded from these data because of meeting end-point criteria. Six additional mice served as naïve controls (Naïve, n = 6).

Physiobehavioral monitoring

Home cages (32  × 18 × 14 cm, Super Mouse^TM microisolator), described previously,^42,44 were instrumented with electric field (EF) sensors (Plessey Semiconductors, PS25251, 1 cm², +/-5V, 1 kHz sampling) able to passively translate disruptions of the local electric field caused by movement into a voltage trace. The EF sensors have been validated against respective gold standard techniques to reliably quantify resting respiration, sleep-wake staging (including differentiating rapid eye movement [REM] sleep from Non-REM sleep), and animal motor activity non-invasively from outside the animals' home cage.^42,44

Animals were acclimated to their instrumented home cages for several days before recording (Fig. 1A). The dividing insert was only present during physiobehavioral recordings; cage-mates shared the full home cage at all other times. Three baseline recordings were collected before surgery (baseline). For six weeks post-surgery, recordings were collected twice weekly at night (6 pm–6 am) for SCI animals and once weekly for sham and naïve animals.

FIG. 1.

Methodology for non-invasive monitoring of animal sleep and respiration. (A) Home cage instrumentation for electric field (EF) sensing. A total of four individual EF sensors were used per cage, with two dedicated to each animal. (B) Representative spectrogram (top panel) and overlaid hypnogram (black trace) obtained from EF sensor recording (bottom panel) for quantification of sleep events. (C) Representative recording of breathing with electric field sensor.

Sleep analysis

Sleep was analyzed from the EF sensor home cage recordings. As described previously,^44,46 wake, REM sleep, and non-REM sleep were quantified through changes in animal movement and respiration (Fig. 1B). Weekly averages were calculated for total sleep time, percent time spent in each arousal state, average sleep bout duration, average REM sleep event duration, REM sleep latency, percentage of sleep bouts containing REM sleep, sleep fragmentation index (total awakenings per hour of sleep), and microarousal index (total microarousals <60 sec per hour of sleep).

Respiration analysis

Respiration was detected during non-REM sleep and quiet wake by the EF sensors (Fig. 1C).⁴⁴ Each breath was identified by characteristic fluctuation in sensor voltage waveform, with measure peak amplitude used to calculate the instantaneous frequency in Hz for each breath (1/interpeak time). Average respiratory rate (RR) was calculated from captured instantaneous frequency values, while the calculated standard deviation was used as a measure of breathing consistency and described as RR variability (RRV). Weekly values for RR and RRV were calculated using a custom MATLAB script that filtered the raw EF sensor voltage trace using a 0.1 Hz HP and 10 Hz LP 4th order Butterworth filters.

Mechanical sensitivity

Mechanical sensitivity was measured using the Chaplan up-down protocol to calculate 50% paw withdrawal threshold (PWT).⁴⁷ The 50% PWT was assessed three times before surgery (baseline) and once weekly afterward for six weeks on the same day as overnight physiobehavioral recordings were taken. Single-animal values were obtained by averaging the 50% PWT for both hindpaws.

Isolation of multi-segmental DRGs

An isolated spinal cord mouse preparation⁴⁸ was extended to include DRG from segments C8 through S1. Dorsal and ventral roots were cut to isolate the DRGs and prepared as described previously.⁴⁸ All recordings were undertaken at room temperature.

Extracellular DRG recordings

Dura and pia were removed around the DRG and glass suction electrodes (200–250 μm tip diameter) were positioned on lumbar (L1–L6) DRGs to record spontaneous afferent population activity for a minimum of 25 sec (50 kHz, Digidata 1322A 16 Bit DAQ, Molecular Devices, Fig. 2). Before the experimental recordings, electrode suction was optimized to minimize the noise floor, enabling post hoc threshold detection of spikes at three times the noise amplitude. Signals were amplified (5000x) and filtered (LP 3 kHz at recording, LP 5Hz post hoc).

FIG. 2.

Methodology for electrophysiology for isolated dorsal root ganglia (DRG) preparation. (A_i) A schematic of the isolated DRG preparation. After exposure of the spinal cord and peripheral aspects of primary afferents (DRG, dorsal root, spinal nerve) from C8-S1 segments, the dorsal roots (DRs) and ventral roots (VRs) were cut at each segment, leaving only a section of the DRs with the attached DRGs and the surrounding ribs and musculature (not shown). (A_ii) Example recordings of spontaneous activity in the lumbar (L) L3 DRG from the spinal cord injury (SCI, red) and L1 DRG from the Sham (black) populations. Displayed is a single sweep from maximally active DRGs from two different animals. Single units marked with * are expanded in the inset (scale bars are 2 msec and 10 μV).

Spontaneously active (SA) and silent DRG spike counts were measured with Clampfit software (v 10.7 Molecular Devices). The DRGs were considered SA (SA DRGs) with >1 spike within 25 sec sampling period. The DRG firing frequency was determined by averaging spike frequency within five- 5 second bins. Incidences of SA DRGs were determined by calculating the percentage of DRGs with observable spontaneous activity, relative to the total number of lumbar DRGs recorded within an animal. Of a total possible 12 DRG recordings (bilateral DRGs for L1-L6), the average number of sampled whole DRGs from the SCI group and Sham group were similar at 7.6 ± 2.5 and 6.8 ± 1.6, respectively.

Statistical analysis

Statistical differences in behavioral metrics of sleep, respiration, and mechanical sensitivity were determined using two-way analysis of variance (ANOVA), with a post hoc Sidak multiple comparison test (α = 0.05). A two-tailed Mann-Whitney test was used to analyze DRG firing frequencies and to assess differences in incidence of SA DRGs. The RRV and 50% PWT were correlated with sleep metrics using Pearson correlation analysis. Pearson correlations were also conducted for week 6 between post-mortem electrophysiology measures and each measured behavior (RRV, 50% PWT, and sleep measures). All statistical analyses were performed using Statistica (Tibco Software Inc. Palo Alto, California) and SigmaPlot (Systat Software). All data are presented as mean ± standard deviation.

Results

SCI resulted in hindpaw mechanosensitivity

Mechanical sensitivity was assessed to confirm the injury caused mechanical hypersensitivity in agreement with previous findings.⁵⁰ Persistent heightened mechanical sensitivity developed in the animals with SCI beginning at week 2 relative to Naïve and Sham groups and SCI baseline (Fig. 3A; Supplementary Fig. S1).

FIG. 3.

Spinal cord injury (SCI) leads to increases in mechanical sensitivity and afferent spontaneous activity. Graphs in A, Bi and Bii are probability density violin plot of animal averages (circles) with group averages (horizontal lines). (A) Mechanical Sensitivity. Mechanical sensitivity (50% paw withdral threshold [PWT]) is heightened after SCI. The p values are: SCI versus Sham at week 6- *p < 0.05, **p < 0.005, ***p < 0.0005; baseline versus week 6 in the SCI group- ^^^p < 0.0001. (B) Spontaneous Afferent Firing. Graphs are probability density violin plots of animal averages (circles) with group averages (horizontal lines). (B_i) The maximal observed firing frequency and (B_ii) incidence of SA lumbar dorsal root ganglia (DRGs) in the SCI group are significantly greater than that of the Sham group. SCI (n = 7), Sham (n = 5). The p values are: SCI versus Sham at week 6- *p < 0.05, **p < 0.005, ***p < 0.0005; baseline versus week 6 in the SCI group- ^^^p < 0.0001. (B_iii) Line plot of % SA DRG at each DRG segmental level. Each point represents the sum total of SA DRGs of all animals in that respective group averaged by the total number of recorded DRGs in that respective group. Note that the SCI population was characterized by having increased incidence of bursting in more caudal lumbar segments below the T10 contusion SCI (black lines and circles) relative to sham animals (gray line and circles) (p < 0.001 chi square).

Greater spontaneous DRG afferent activity emerged after SCI

To assess below-level spontaneous primary afferent activity, spontaneous nerve population activity in L1–L6 DRGs was recorded bilaterally from SCI and Sham animals six weeks after injury. Compared with Sham (1.2 ± 1.1 Hz), animals with SCI exhibited greater spontaneous activity (2.6 ± 1.2 Hz), as characterized by the maximal DRG firing frequency (p < 0.05, Fig. 3B_i). The incidence of DRGs showing SA was significantly greater in the SCI (53.5 ± 31.2%) than Sham population (20.9 ± 6.5%, p < 0.05, Fig. 3B_ii).

We also compared incidence of SA in DRG from L1 to L6 (Fig. 3B_iii). The SA in the Sham group was preferential to rostral lumbar segments nearer the T8–10 laminectomy, while the SCI population also expressed spontaneous activity in more caudal lumbar segments (p < 0.001; chi-square). This is consistent with reports seen previously.⁶

Respiration became more erratic after SCI

The average resting RR did not show significant differences between treatment groups or across time (Fig. 4A). At baseline, all treatment groups exhibited a consistent RRV averaging 0.21 ± 0.02 Hz. The RRV, however, increased in the SCI group relative to SCI baseline and Naïve and Sham groups at all time periods after two weeks (Fig. 4B; Supplementary Fig. 1B).

FIG. 4.

Respiration changes after spinal cord injury (SCI). Graphs are probability density violin plots of animal averages (circles) with group averages (black and white horizontal lines). (A) Average respiratory rate appears unaffected by SCI, while (B) respiratory rate variability differs significantly when comparing SCI versus Sham at post-operative week 6; and baseline versus week 6 in the SCI group. The p values are: SCI versus Sham at week 6- *p < 0.05, **p < 0.005, ***p < 0.0005; baseline versus week 6 in the SCI group- ^^^p < 0.0001.

Sleep became fragmented after SCI

Total sleep time was not significantly different between treatments or across time (Fig. 5A). Progressive fragmented sleep developed in animals with SCI, beginning at week 2, as measured by decreased sleep bout duration, increased sleep fragmentation index, and increased microarousal index relative to Naïve and Sham groups and SCI baseline (Fig. 5B-D; Supplementary Fig. 1C–E).

FIG. 5.

Sleep changes after spinal cord injury (SCI). Graphs are probability density violin plots of animal averages (circles) with group averages (black and white horizontal lines). (A) While time spent sleeping was unaffected by SCI, (B) average sleep bout duration, (C) sleep fragmentation index, and (D) microarousal index during a 12-hour dark cycle were significantly different when comparing SCI versus Sham at postoperative week 6; and baseline versus week 6 in the SCI group. (E) REM sleep duration, (F) REM sleep latency, and (G) the percentage of sleep bouts containing REM sleep were all reduced in SCI animals at post-operative week 6 relative to their baseline and sham animals at post-operative week 6 (except for REM sleep duration) suggesting premature arousal to wake, an increase in sleep pressure, and an overall reduction in sleep quality. The p values are: SCI versus Sham at week 6- *p < 0.05, **p < 0.005, ***p < 0.0005; baseline versus week 6 in the SCI group- ^^^p < 0.0001.

In animals with SCI, REM sleep duration, REM latency, and the percentage of sleep bouts containing REM sleep decreased at later time points relative to Naïve and Sham groups and SCI baseline (Fig. 5E–G; Supplementary Fig. 1F,G). Specifically, changes in REM sleep developed later than both mechanical sensitivity and sleep fragmentation.

SCI-induced spontaneous afferent excitability increases associate with respiratory and sleep changes

Correlative relationships between behavioral measures and spontaneous activity were assessed in individual animals with identified relations shown in Figure 6. Increased RRV, denoting more erratic breathing, trended toward a positive correlation with increased incidence of SA in DRGs (R² = 0.29, p = 0.072) driven largely by changes in animals with SCI at week 6 (Fig. 6A).

FIG. 6.

Spontaneous hyperexcitability metrics associate with changes in respiration and sleep after spinal cord injury (SCI). Data are individual animal averages for each measure at week 6 for sham laminectomy (circle) and SCI (square) animals. Pearson correlation regression lines (gray dotted line), R², and p values are given in each plot. (A) Incidence of SA dorsal root ganglia (DRGs) trend with respiratory rate variability. (B) Increased firing rate of SA DRGs correlate with reduced REM duration. (C) Increased firing rate of SA DRGs correlate with reduced sleep bout duration.

Decreasing REM sleep duration (R² = 0.43, p = 0.020) and sleep bout duration (R² = 0.40, p = 0.028), indicating fragmented sleep, both correlated with increasing DRG firing frequency (Fig. 6B, 6C) in both the sham laminectomy and SCI animals at week 6. Together, results indicate significant correlation between increased afferent excitability and poor sleep quality.

SCI-induced hindpaw mechanosensitivity, increased respiratory rate variability, and impaired sleep correlated across time

While Figure 3 compared changes between naïve, sham, and SCI populations at baseline through six weeks post-injury, we further assessed emergent temporal differences after SCI (Fig. 7A) to explore relative timing onset and correlations between emergence of changes to RRV and mechanosensitivity (50% PWT) to changes in measured sleep variables.

FIG. 7.

Comparing longitudinal changes in sleep, respiration, and mechanosensitivity in mice with spinal cord injury (SCI). Weekly group means (black circles) and standard deviations (gray bars) for SCI animals only with respiration, mechanical sensitivity, and sleep measures plotted against each other. The dotted line follows weekly sequential time points from baseline to week 6 (labeled). R² and p values reflect linear regression results. Although some regressions are significant, low R² values indicate additional variance could arise from other factors. Further, the largest changes for most measures was seen between weeks 1 and 2, with slower changes at the later weeks. (A) Overall, respiratory rate variability (RRV) changes are seen at week 1 and precede paw withdrawal threshold (PWT) changes that begin at week 2 (Ai), but note that in 2/6 mice, PWT changes were seen at week 2 and 3 that preceded subsequent increase in RRV (Aii) demonstrating that these changes are not necessarily coupled. (B) Sleeping measures plotted against 50% PWT show that both change in common at week 2. (C) Sleep measures plotted against RRV show that RRV changes precede sleep.

While SCI-induced RRV changes correlated significantly with 50% PWT and several measures of sleep fragmentation (Fig. 7A, 7B), their coefficient of determination (R²) was relatively low, meaning little variance in the data was explained by their relationships. Conversely, 50% PWT correlated significantly with sleep measures with relatively high R² values for raw behavioral data (Fig. 7C).

Temporal emergence of different behavioral modalities give some insight into the order of pathogenesis when transitioning from acute to chronic SCI. The SCI-induced RRV increases appear early (in 4/7 mice) in the first week after SCI, whereas heightened 50% PWT does not emerge until the second week after injury (in 6/7 mice) at which time both RRV and PWT changes are significant (Supplementary Fig. 1). The RRV similarly precedes all indices of increased sleep disruption (Fig. 7B) while sleep disruption indices occur at the same time as 50% PWT (Fig. 7C).

Discussion

The present findings provide the first detailed study on sleep impairment after SCI in an animal model, demonstrating the emergence of sleep fragmentation and subsequent impaired REM sleep. Importantly, changes were captured non-invasively during natural behavior in the animal's home cage.⁴⁴

Our estimate neuropathic pain magnitude (50% PWT) correlated with magnitude of sleep disruption. This was expected because poor sleep and chronic pain have a well-established association.^50,51 This index of neuropathic pain magnitude also correlated with RRV. Interestingly, respiration changes preceded both mechanosensitivity and sleep changes after SCI.

Whether correlated changes reflect independent responses impacted by SCI or are causally associated is unknown. Further studies will particularly benefit from a greater understanding of the cause-effects relations between impaired sleep and neuropathic pain. Below we discuss results of SCI -induced changes in sleep, respiration, and sensory hyperexcitability in separate sections before providing an overall summary perspective.

Sleep changes after SCI

Currently there are very limited pre-clinical animal studies devoted to assessing sleep changes after SCI. Only one study looked at and observed sleep dysfunction independent of respiration.⁵² This study was undertaken in the rat and followed changes for 15 days after complete lower thoracic spinal transection; they authors observed a transient reduction in sleep efficiency during the light cycle.

The technological advance of the EF sensors has enabled non-invasive quantification of sleep changes in rodents after SCI.⁴⁴ Although the time spent asleep during the 12h dark cycle (6 pm–6 am) was unchanged, animals with SCI had more fragmented and low quality sleep relative to their sham and naïve counterparts. There were meaningful reductions in average sleep bout duration, REM sleep duration, REM sleep latency, and the percentage of sleep bouts containing REM sleep.

Sleep fragmentation index—representing total number of awakenings/hour of sleep—and microarousal index—representing total number of <60 sec microarousals/hr of sleep—were increased after SCI. The increased brief arousals and reduced duration of sleep bouts were associated with a compensatory increased number of sleep events after SCI, presumably reflecting experienced increased sleep pressure.

Because REM sleep occurs at the end of a sleep cycle, it is intuitive to link these observed reductions in REM sleep to reduced sleep bout duration. If sleep events are shortened and without observed REM sleep, it is possible the animal did not achieve normal transitions for restorative sleep. Factors responsible for sleep compromise may include increased pain pathway recruitment of sympathetic drive, given known correspondence with mechanical hypersensitivity and in other studies showing evidence of ongoing spontaneous pain.^6,53 Regardless of mechanism, sleep dysfunction would contribute to overall stress burden after SCI and, through actions on common stress axis pathways, contribute to overall disease burden in SCI.

Respiration changes after SCI

While cervical SCI clearly impairs respiration because of direct injury to respiratory neural circuits that impact sleep,^54–56 there is little consideration on whether SCI-induced neural dysfunction in other areas interacts with and impacts respiration. Nonetheless, recent work has shown that, when recorded in restraint cylinders, resting respiratory rate (RR) in awake animals increases after thoracic contusion SCI that may precede and correspond with the emergence of allodynia.^40,41

Noble and associates,⁴⁰ recorded resting RR in awake, restrained rats with a lower thoracic contusion during daylight hours while we assessed RR and RRV in naturally behaving sleeping mice. For three weeks post-SCI, Noble and coworkers⁴⁰ report consistent elevated resting RR in restrained rats immediately before evoked nociceptive stimulus but only increased RR during an evoked nociceptive event at one time point (day 21 post-SCI). Further, resting RR did not correlate with hindpaw mechanical sensitivity, but evoked RR inversely correlated such that mechanically hypersensitive animals exhibited lower evoked RR.

Cumulatively, these findings may have been confounded by anticipatory anxiety for the upcoming nociceptive test as much as for experiential pain. Alternatively, it is well established that nociceptive stimulation can increase sympathetic drives elevating heart rate and RR^57–59 such that the overall impaired nervous system after SCI, even lower thoracic contusion, may compromise motor and sensory processing sufficiently to affect respiration via altered abdominal and intercostal muscle function. In comparison, we observed no change in resting RR but increases in RRV. These changes also emerged coincident with measures of mechanical hypersensitivity.

Despite the differences between the studies, respiration changes were observed in both studies before development of mechanical sensitivity indicating that respiration measures may be a useful measure of stress and pain in chronic injury models. While the present studies characterize sleep and respiration under natural conditions, future studies would benefit from coupling these measures to natural measures of ongoing spontaneous pain to more directly link dysregulation in respiratory and sleep dysfunction to emergent pain.

While the EF sensors have shown great utility in capturing frequency-based respiratory features like RR and RRV,⁴² the EF sensors have not been validated to capture tidal volume^42,44 and therefore cannot report on whether minute ventilation is changed.^60–63 We do not expect such changes, however, given that recordings are made of the animal stationary in its nest so that factors associated with stress-behavior like hyperventilation will not be seen and contribute to differences in quantification or relevance of measured respiration during rest and sleep.

Neural basis of neuropathic pain after SCI

Spontaneous primary afferent activity has been identified as a driver of chronic pain in clinical^64–66 and pre-clinical^6,7,18,19,22 studies. Key studies conducted in rodent contusion models of SCI revealed potential links between stimulus-evoked hypersensitivity and spontaneous firing of primary afferents below the level of injury.^6,16,22,23 Here, we recorded activity from intact DRGs across multiple vertebral segments to confirm development of increased spontaneous afferent activity after SCI including in caudal lumbar DRG.

Recorded SA firing rates represented the combined activity of an unknown number of single neurons. Although our recording technique did not permit resolution to identify the class of primary afferent neurons within the signal population, there is evidence suggesting spontaneous nerve activity after SCI includes signals from A and C fibers that have been associated with neuropathic pain including in the T10 contusion SCI animal model.^{6,7,15,16,18,19,22,66,67} Because our whole DRG recording approach may not resolve spiking in unmyelinated C-fibers and a previous publication in rat thoracic contusion SCI showed greater SA increase in C fibers,⁶ our lower levels of SA likely underestimate the magnitude of recorded SA.

Another factor that may influence observed SA is choice of analgesics. Previous work on SA used the opioid buprenorphine for analgesia.^6,16 Recent assessment of buprenorphine on outcome in a comparable lower thoracic contusion SCI model observed a significant increase in self-biting behavior.⁶⁸ The opioid morphine worsens pain for many weeks in various pain models but with mixed results on whether it exacerbates post-SCI pain.^69,70

As previously shown in rat, SCI increased spontaneous afferent activity including increased incidence in more caudal lumbar DRG.⁶ More broadly, while increased primary afferent spontaneous activity has been previously observed in SCI models of neuropathic pain, here we found correlations between spontaneous lumbar DRG activity (incidence and firing frequency) and metrics of sleep and respiratory dysfunction—complications frequently observed in SCI patients.

In sham laminectomy and animals with SCI at week 6, the only behavioral time point at which terminal electrophysiology measures were captured, we showed a correlation between fragmented sleep measures (REM sleep duration, sleep bout duration) and elevated and DRG afferent firing frequency.

Observed correlations between measures are clearly prominently driven by the SCI animal changes. It is worth noting, however, that one sham laminectomy animal showed atypical elevated firing frequency that paired with sleep behavior measures typical of SCI changes, reinforcing the relationship between abnormal afferent activity and poor sleep quality.

Surprisingly, DRG firing frequency, incidence, and mechanosensitivity were not correlated, suggesting they may capture different elements of sensory dysfunction. For example, the incidence of spontaneously active DRGs may reflect broader aspects of pain sensitization because it correlates with sleep deficits that may arise from ongoing spontaneous pain. In comparison, observed increases in firing frequency with near-significant development of elevated RRV (i.e., erratic breathing) may encode other measures of physiological distress (e.g., metabolic status, baroreceptors, hypoxia).

While most correlations between behaviors and afferent hyperexcitability were not significant or had unclear relationships (data not shown), panels B and C in Figure 6 begin to quantify the relationship between poor sleep and neuropathic pain. The relationship between respiration changes and afferent activity did not quite reach significance; the trend is strong enough to merit further investigation into whether these measures can be used to indicate SCI pathology more fully.

Further, the use of only female mice used here may give an incomplete understanding of the pathophysiology of all patients with SCI because sex-specific differences in pain pathophysiology have been established.^71,72

Temporal changes in behavior give insight to SCI pathogenesis

Here we found correlations between heightened mechanosensitivity, fragmented sleep, and respiratory dysfunction—complications frequently observed in patients with SCI. Although this T10 contusion SCI model should not interfere with respiratory control, respiratory changes preceded development of mechanosensitivity and sleep fragmentation in some animals. These changes may reflect autonomic changes that are independent of those associated with mechanosensitivity and sleep changes at later time points.

In comparison, mechanosensitivity and sleep fragmentation shifts occurred along similar timelines (1–2 weeks post-injury) with plateau into a chronic phenotype from week 2 onward consistent with an expected interdependence.

Conclusion

Using an established SCI model of neuropathic pain, we compared the temporal emergence of hindpaw sensitivity to that of coemergent, sleep dysfunction and increased RRV over six weeks in the mouse. Results demonstrate considerable inter-animal variability in the temporal emergence of behavioral dysfunction, with strongest correlation between reduced mechanical sensitivity and sleep bout duration.

Emergent pathophysiology then stabilizes to a low-variability status over the last two weeks. At this point, an observed increase in spontaneous afferent activity is associated with increased RRV and sleep disruption, but not mechanosensitivity. Overall, these studies provide a broader understanding on the evolution and stabilization of important physiological variables that associate with stress burden after SCI. Derived insights may inform implementation of therapeutic strategies after SCI.

Authors' Contributions

Animal surgeries were performed by Shaquia Idlett-Ali and Mallika Halder. Tissue isolation, electrophysiology experimentation, and analysis were performed by Shaquia Idlett-Ali. Behavioral testing and analyses were conducted by Heidi Kloefkorn. Shaquia Idlett-Ali, Heidi Kloefkorn, and Shawn Hochman were involved in preparation of the manuscript.

Transparency, Rigor, and Reproducibility Summary

The analysis plan was not formally pre-registered, but the team members with primary responsibility for the analysis certify the analysis plan was pre-specified. Actual sample size was at least six subjects per group, the spinal cord injury group had a sample size of seven to ensure appropriate statistical power to account for any animal losses or exclusions premature to the completion of the study. To obtain a statistical power of at least 0.8 or greater, expected behavioral and electrophysiological effect sizes for multiple measures based on previous pilot data were used to calculate sample size; Six subjects per group met or exceeded the calculated sample sizes. One spinal cord injured animal met humane endpoint criteria and was euthanized prior to the completion of the study. These data were excluded from analysis. Animals were randomly assigned groups and experimenters were blinded to animal condition. Data were acquired weekly using the same equipment. Specific equipment and software used to acquire behavioral data in this study, including electric field sensors and LabVIEW .vi's, was custom built and may be purchased from the Emory University Cell Biology Electronics and Mechanical Department Core. Specific equipment and software used to acquire electrophysiological data can be purchased from Axon Instruments. There is no current or planned replication studies ongoing to our knowledge.

Fundng Information

Research supported by the Craig H Neilsen Foundation and the Paralyzed Veterans of America Research Foundation (S.H.). S.I. supported by National Science Foundation Graduate Research Fellowship Grant DGE-1650044 and the Alfred P. Sloan Foundation. H.K supported by NIH fellowship (5k12-Gm000680).

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Figure S1