Temporal and spatial dynamics of spinal sensorimotor processing in an intersegmental cutaneous nociceptive reflex

Abstract

The cutaneus trunci muscle (CTM) reflex produces a skin “shrug” in response to pinch on a rat’s back through a three-part neural circuit: 1) A-fiber and C-fiber afferents in segmental dorsal cutaneous nerves (DCNs) from lumbar to cervical levels, 2) ascending propriospinal interneurons, and 3) the CTM motoneuron pool located at the cervicothoracic junction. We recorded neurograms from a CTM nerve branch in response to electrical stimulation. The pulse trains were delivered at multiple DCNs (T₆–L₁), on both sides of the midline, at two stimulus strengths (0.5 or 5 mA, to activate Aδ fibers or Aδ and C fibers, respectively) and four stimulation frequencies (1, 2, 5, or 10 Hz) for 20 s. We quantified both the temporal dynamics (i.e., latency, sensitization, habituation, and frequency dependence) and the spatial dynamics (spinal level) of the reflex. The evoked responses were time-windowed into Early, Mid, Late, and Ongoing phases, of which the Mid phase, between the Early (Aδ fiber mediated) and Late (C fiber mediated) phases, has not been previously identified. All phases of the response varied with stimulus strength, frequency, history, and DCN level/side stimulated. In addition, we observed nociceptive characteristics like C fiber-mediated sensitization (wind-up) and habituation. Finally, the range of latencies in the ipsilateral responses were not very large rostrocaudally, suggesting a myelinated neural path within the ipsilateral spinal cord for at least the A fiber-mediated Early-phase response. Overall, these results demonstrate that the CTM reflex shares the temporal dynamics in other nociceptive reflexes and exhibits spatial (segmental and lateral) dynamics not seen in those reflexes.

NEW & NOTEWORTHY We have physiologically studied an intersegmental reflex exploring detailed temporal, stimulus strength-based, stimulation history-dependent, lateral and segmental quantification of the reflex responses to cutaneous nociceptive stimulations. We found several physiological features in this reflex pathway, e.g., wind-up, latency changes, and somatotopic differences. These physiological observations allow us to understand how the anatomy of this reflex may be organized. We have also identified a new phase of this reflex, termed the “mid” response.

INTRODUCTION

Nociceptive reflexes have been studied from a variety of perspectives, including pain physiology, pain pharmacology, spinal cord physiology, and sensorimotor physiology. Examples of nociceptive reflexes include the flexor reflex, tail flick reflex, and paw withdrawal reflex. They have the advantage of being easier to measure and more quantitative than many complex behavioral measures of nociception, e.g., facial grimace, low appetite, and reduced exploratory behavior.

The cutaneus trunci muscle (CTM) reflex (Fig. 1) is one such nociceptive reflex, consisting of a localized skin “shrug” in response to pinch on a rat’s back. It is also called the panniculus reflex in larger mammals (Radostits and Done 2007). It was first studied in the 1830s in the hedgehog (Hall 1833) and more generally in dogs and horses by the late 1800s (Wilson 1898), including a mention by C. S. Sherrington in 1906 (Sherrington 1906). Despite this early start, the neuroanatomy of the CTM reflex was not known until the 1980s, when the motor nucleus was found to be located in the C₇–T₁ spinal cord segments (Baulac and Meininger 1981; Holstege et al. 1987; Krogh and Towns 1984; Theriault and Diamond 1988a) and a branch of the lateral thoracic nerve was found to innervate the muscle (Krogh and Towns 1984), which we refer to as the CTM nerve.

Fig. 1.

Simplified cutaneus trunci muscle (CTM) reflex neural circuit. The CTM neurogram recording from a CTM nerve branch is illustrated with an example stimulation on the L₁ dorsal cutaneous nerve (DCN), which is oversized for display purposes. DRG, dorsal root ganglion; PSN, propriospinal neuron; MN, motoneuron. Question marks represent possible routes signals could take crossing the midline. Given the latency differences in the reflex based on contralateral vs. ipsilateral stimulation, this simple formulation should not preclude different populations of interneurons or additional polysynaptic pathways.

The CTM reflex has a large intersegmental organization and can be activated by stimuli applied to dermatomes as far caudally as L₄ (Petruska et al. 2014), while the motoneuronal pool is at C₇–T₁. This intersegmental organization has made the CTM reflex attractive as a model for studying interventions after spinal cord injury, including applied electrical fields (Borgens et al. 1990), polyethylene glycol (Borgens et al. 2002), neurotrophic substances (Bohnert et al. 2007), and other pharmacological agents (Yates et al. 2006). There is no direct parallel of the CTM reflex in humans, as humans lack the CTM. The most similar reflex in function or behavior is the abdominal or erector spinae reflexes (Hagbarth and Kugelberg 1958; Kugelberg and Hagbarth 1958).

The CTM reflex can be activated by cutaneous mechanical stimulations along the back surface or electrical stimulations of Aδ or C afferents of the dorsal cutaneous nerves (DCNs), the spinal segmental nerve branches with primary afferents that innervate the back cutaneously (Blight et al. 1990; Petruska et al. 2014; Theriault and Diamond 1988a). In rats and mice the CTM reflex is activated preferentially with such a noxious stimulation of the DCNs in an anesthetized preparation (Krogh and Denslow 1979; Krogh and Towns 1984; Petruska et al. 2014; Theriault and Diamond 1988b), but in the awake rat nonnoxious stroking can elicit the reflex in a high-stress condition (Duarte et al. 2005). The degree of sensitivity to nonnoxious stimulation appears to be species dependent, as it is easily activated in the guinea pig with light touch (Blight et al. 1990).

The interneuronal pathway for the CTM reflex has not been conclusively identified, but the short delay [10–15 ms (Petruska et al. 2014; Theriault and Diamond 1988b)] between ipsilateral stimulation and the Aδ-mediated early phase of the reflex response suggests time for only a few synapses. Previous research (Blight et al. 1990) has suggested a myelinated monosynaptic pathway from spinal interneurons to the area of the CTM motor nucleus at C₇–T₁ as a candidate ascending pathway previously thought to be related to interlimb coordination in locomotion (Matsumoto et al. 1976; Miller et al. 1973). In Fig. 1, we show these candidate connections in a simplified wiring diagram adapted from early studies (Blight et al. 1990; Petruska et al. 2014).

Several groups have described aspects of the CTM reflex anatomy and physiology previously, including the neuroanatomical subcomponents (Baulac and Meininger 1981; Holstege and Blok 1989; Krogh and Denslow 1979; Krogh and Towns 1984; Theriault and Diamond 1988a), somatotopy (Theriault and Diamond 1988a, 1988b), rostrocaudal extent (Holstege et al. 1987; Holstege and Blok 1989; Petruska et al. 2014), stimulus strength dependence of the late phase of the reflex (Petruska et al. 2014; Theriault and Diamond 1988b), and existence of the contralateral and ipsilateral aspects of the reflex (Blight et al. 1990; Holstege and Blok 1989; Petruska et al. 2014). We were recently able to quantify selectively labeled DCN A and C fibers, demonstrating that their central projection patterns rather than peripheral axon numbers contributed to the somatotopic organization of the CTM reflex (Lee et al. 2017). However, many other aspects of the physiology have not been well explored, including segmental differences in the size of the reflex, stimulation history dependence, e.g., wind-up or habituation, frequency dependence, and latency. We believe a detailed quantification of the normal physiological behavior could help us understand the reflex in a more graded sense, adding to our understanding of how small changes in the reflex might be affected by various perturbations, e.g., peripheral nerve or spinal cord injury, and pharmacology. We hypothesize that the CTM reflex will demonstrate quantitative features consistent with other previously described pain reflexes.

Therefore, we thoroughly characterized the temporal, segmental, lateral, and afferent type differences in this reflex in the uninjured, anesthetized animal. We quantified sensitization and habituation and compared them with the stimulus-to-stimulus changes in latency. We analyzed the DCN and frequency differences, compared the contralateral-to-ipsilateral differences, and explored stimulation strength differences. All of these quantifications will help further characterize the reflex to better understand spinal nociception and segmental organization, in addition to establishing a baseline for injury studies.

METHODS

All procedures were conducted with the approval of the Emory University Institutional Animal Care and Use Committee.

Surgery.

Female Long-Evans rats (225–250 g, n = 24; Charles River Laboratories, Wilmington, MA) were anesthetized with intraperitoneal pentobarbital sodium injection (50 mg/kg, Nembutal; Ovation Pharmaceuticals, Deerfield, IL) and kept on a warming pad until they showed no paw withdrawal reflex to pinch or blink reflex before surgery. The back skin was shaved and then incised along the midline from the skull base to the iliac crest. DCNs at different spinal segmental levels (T₆–L₁) were isolated from underlying fascia and cut distally. Emerging from the lateral thoracic nerve, the CTM nerve divides into four or five intramuscular branches. One large branch of the CTM nerve (generally the third) was dissected free of connective tissue from within the CTM belly. This approach left other CTM branches intact, such that DCN stimulation could still evoke visible CTM contractions. The rats were placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA) to maintain the position of the rat and electrodes. All exposed tissues were bathed in mineral oil, and core body temperature was maintained with a warming pad regulated by a feedback controller (TCAT-2; Physitemp Instruments, Clifton, NJ) monitoring a rectal temperature probe. Supplemental pentobarbital (10% of the initial dose) was given intraperitoneally to maintain anesthetic level as needed (usually once an hour), as judged by the reemergence of a flexor reflex to paw pinch or a blink reflex.

Stimulation and neurogram recording.

The rats (n = 24) were split into three groups with different stimulation paradigms: 1) DCNs at consecutive spinal segmental levels (T₆–L₁) ipsilateral to the recording site were stimulated at a strength of 0.5 mA (n = 8); 2) ipsilateral DCNs (T₆–L₁) were stimulated at a strength of 5 mA (n = 8); and 3) both ipsilateral and contralateral DCNs (only T₆, T₈, T₁₀, T₁₂, and L₁) were stimulated at a strength of 5 mA (n = 8). A single DCN at each spinal level was placed on a bipolar silver wire electrode at a time and stimulated with pulse widths of 250 µs for either 0.5 mA or 5 mA. The pulse trains were delivered at frequencies of 1, 2, 5, or 10 Hz for 20 s each. The fixed stimulation duration was chosen to take advantage of both analyzing effects of different frequencies over the same time period and analyzing effects of the number of stimulations by selecting a certain number of stimulations from the beginning at each frequency. Responses over the different stimulation numbers at each frequency were weighted during normalization (see below). In each stimulation protocol, we began stimulation at the caudal DCNs, continuing on to the rostral DCNs in all experiments. For the studies in which both sides were stimulated, the left DCN (ipsilateral to the CTM nerve recording electrode) was stimulated first and then the right (contralateral) DCN at the same spinal level. At each DCN, the pulse trains were given from low frequency to high frequency to reduce habituation effects, waiting 2 min between pulse train onsets. Because of the duration of each stimulation protocol and the consistency between protocols, all the protocols were tested in different groups of animals. The stimulation strength and pulse duration were chosen based on empirical testing in which only the early (A fiber dependent) phase of the CTM reflex was evident at 0.5 mA, whereas 5 mA was consistently capable of eliciting both the early and the late (C fiber dependent) phases of the reflex. Previous work explored the relationship between response and afferent population recruited by the stimulation (Doucette et al. 1987; Nixon et al. 1984; Petruska et al. 2014; Theriault and Diamond 1988b). Although 0.5 mA rarely evoked a late phase, switching from 0.5 mA to 5 mA could result in some increases in the early phase of the evoked response. In many preparations 3 mA was sufficient to elicit the maximal late response, but in others 5-mA stimulation was required. In order to be consistent, 5 mA was used in all of the preparations.

The CTM nerve branches were placed over a bipolar recording electrode. The CTM responses (see Fig. 2) were amplified at 10,000× (16-Channel Differential Amplifier, model 3500; A-M Systems, Sequim, WA) and filtered with a Humbug (Digitimer, Letchworth Garden City, UK). The CTM neurograms were recorded continuously during the entire stimulation protocol for each animal on a USB data acquisition board (USB 6259 BNC; National Instruments, Austin, TX) at a 10,000 Hz sampling rate. The stimulus signals were recorded to the same file to locate the stimulus responses.

Fig. 2.

Cutaneus trunci muscle (CTM) neurogram data processing. Top: a representative unfiltered, unprocessed CTM neurogram response to a single stimulus with a 250-µs pulse width (5th Stim) from repeated stimulus at 2 Hz at 1 dorsal cutaneous nerve (DCN; T₁₃) in a single animal (animal 2) at a 5-mA stimulus strength. Middle: a filtered, wavelet-denoised, and rectified signal. The time-windowed phases of the response are shown as Early (3.5–25.5 ms), Mid (25.5–45.5 ms), Late (45.5–95.5 ms), and Ongoing (110.5–195.5 ms). As shown at bottom, waveforms were averaged with the other animals (n = 8) to produce a mean waveform of the response.

To confirm that this ordered approach did not introduce any significant bias to our recordings, in several animals we retested various DCNs at various stimulation frequencies at the end of the recording session, and these proved to generate responses equivalent to the initial recordings (data not shown). In addition, for control purposes, in a subset of animals we tested lengthening and shortening the pulse width to confirm that there was not an additional larger C-fiber phase recruited with longer pulse widths (there was not) when stimulating at 5 mA.

Neurogram signal processing.

The continuous neurogram recording data were imported into MATLAB (MathWorks, Natick, MA) for artifact elimination, filtering (smoothing), and analysis. In the stimulus response data, the stimulus artifacts were zeroed first (from stimulus to 2 ms after the stimulus). Then, the response data were comb filtered at 60 Hz to eliminate high-frequency harmonics from 60-Hz noise and high-pass filtered at 200 Hz. All filters were noncausal filters, i.e., running both forward and backward in time to eliminate phase artifacts. Wavelet denoising (Sym8, soft thresholding, MATLAB) was applied to the data. Finally, the continuously recorded data were separated into individual evoked responses for 200 ms from each stimulus onset using the stimulus signal, rectified, and time-windowed (Fig. 2) into defined phases (Early, Mid, Late, Ongoing), as discussed below. For the false-color plots (see Fig. 5), there was an additional filtering step (low pass, Butterworth, 1,000 Hz), applied after normalizing (discussed below) and averaging the animals.

Fig. 5.

Sequential neurograms for dorsal cutaneous nerve (DCN) groupings stimulated at 5 mA at different frequencies. Left: false-color plots show the relative size of the sequential responses (with bright defined as larger values and dark as smaller values). Right: a comparison between the Early and Late phases of the DCN-evoked cutaneus trunci muscle (CTM) neurogram response over time for the caudal DCN grouping. Signal sizes are averaged [n = 24, 8 animals, 3 segments (T₆–T₈, T₉–T₁₁, and T₁₂–L₁) per animal], and CTM neurogram data are shown for each stimulation repeated for 20 s at each frequency at a stimulus strength of 5 mA. Before averaging across animals, all neurograms from each animal were normalized. The false-color images are comprised of individual horizontal lines, where each horizontal line represents a single averaged (across animals) stimulus response for that stimulus number in the stimulus train. The first stimulus response in the train is in the top row; the last stimulus response is in the bottom row. The 10-Hz response is truncated to 100 ms to remove the next stimulus from the image. In the comparison between the Early and Late phases of the responses on right only the caudal segments are shown, but they exemplify the trends in the other DCNs: the sizes of the Early and Late phases of the response largely track together, with increases and decreases (facilitation and habituation) occurring in a frequency-dependent manner (although the 1st-to-2nd stimulus response changes are unique; see results and discussion about early habituation). Exaggerated trend lines with arrows are drawn to guide the reader as to the direction of change over time in the comparisons. The sizes of the circles correspond to 30% of the standard deviation at that time in the stimulation train for the designated DCN grouping and frequency.

Stimulation response time-windowing.

We used time windows to isolate specific phases of the stimulus response to analyze them independently. The time window cutoffs used were 3.5–25.5 ms after stimulus for the Early phase, 25.5–45.5 ms for the Mid phase, 45.5–95.5 ms for the Late phase, and 110.5–195.5 ms for the Ongoing phase. The Ongoing phase was quantified for 1, 2, and 5 Hz stimulation data, but not 10 Hz, because of the timing of the next stimulus. These time windows were chosen based on the location of the peaks of the respective responses across all DCNs, without requiring special time windows for each spinal level and stimulation frequency. In all cases, boundaries of the time windows have been extended and smoothed with a spline interpolation (Spline, MATLAB), the time windows gradually reduced down over 2.5 ms at both ends of each time window. The value of the time-windowed signal was calculated by averaging the rectified neurogram signal over the relevant phase. In simulations, this method correlated better to the total number of artificial spikes in the neurogram than other methods tested, such as the sum of the squared signal or the square root of the sum of the squared signal (data not shown).

Normalization.

For combining multiple animals’ data with the same stimulus strength, frequency, DCN, and side (ipsilateral vs. contralateral), individual animal data were normalized with each animal’s mean of all evoked responses before any cross-animal calculations (see Fig. 2). We chose mean normalization in each animal because it has three advantages over max-amplitude normalization: 1) it guarantees that individual animals’ data are equally weighted in the final output, 2) it considers variations between each response component in each animal, and 3) it is less sensitive to noise processes because of not relying solely on just a single maximal value from the entire data set. In cases in which different stimulation frequency data were averaged, the higher-frequency responses with a larger sample size (e.g., 200 stimulations at 10 Hz compared with 20 stimulations at 1 Hz for the 20-s stimulation period) were weighted with stimulus numbers so the different frequencies would have the same final weighting in the outcome variable. In cases in which different stimulus strength data sets were compared, the data were normalized with respect to the first stimulation in each train to avoid the differential behaviors later on, i.e., sensitization or habituation. This approach to stimulus strength normalization should, if anything, have underestimated the differences between high (5 mA) and low (0.5 mA) stimulus strength reflex responses.

Sensitization/habituation.

To quantify sensitization and habituation (Groves et al. 1969), we used ratios of stimulus responses. Sensitization was calculated as the first-to-maximum response ratio, and habituation was quantified as the last-to-maximum ratio of the mean response. The use of the mean response across animals, instead of individual responses, avoided many of the issues associated with noise. In all cases, we calculated sensitization and habituation on a phase-by-phase basis (Early, Mid, Late, and Ongoing).

Latency.

All latency calculations involved first averaging the rectified stimulus response across multiple animals and DCNs. Then, the averaged, rectified responses were low-pass filtered at either 250 Hz for Early-phase latency calculations or 100 Hz for Late-phase latency calculations, with Butterworth filters applied noncausally. The low-pass frequency values were found empirically. Latency for the Early phase of the CTM neurogram was measured from the stimulus onset either to the maximal amplitude (peak latency) or to the point before the peak (onset latency) that corresponded to 50% of the peak amplitude. The Late-phase latency was measured for the peak latency. The maximal amplitude values and their time points were found with the “findpeaks” function, a standard peak-finding function in MATLAB that finds everywhere the derivative changes from >0 to <0 and returns the x, y coordinate of the peak values.

Statistical comparisons.

In cases where two groupings were compared (Fig. 3, Fig. 6, Fig. 7), Student’s t-tests were used to establish significance. In groupings where more than two were compared (Fig. 4), one-way analysis of variance (ANOVA) was used to establish significance. In addition, we used Cohen’s d to measure all effect sizes where results were significantly different at a confidence level of 99% or more. In all figures, error bars represent standard deviation. Significance was generally shown at >95%, >99%, or >99.9%. Following standard conventions for Cohen’s d (Cohen 1992), a small effect is noted at >0.2, a medium effect at >0.5, and a large effect at >0.8. A Cohen’s d of >1.1 is a very large effect size.

Fig. 3.

Effects of stimulus strength on the dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram response (DENR). Left: DENRs at both low (0.5 mA to activate only A fibers) and high (5 mA to activate both A and C fibers) stimulus strengths, averaged across all animals and all DCNs at 2-Hz stimulation. Time windows of Early, Mid, Late, and Ongoing phases are shown by horizontal bars above the waveforms. The high stimulus strength is needed to generate the maximal Early response as well as the Mid and Late responses. Right: averaged, normalized amplitudes in each phase from 8 animals, 9 DCNs, with 40 stimuli per animal (n = 2,880). Statistical differences (***P < 0.001) were found in all comparisons between high- and low-stimulus strength stimulation (horizontal lines above bars). Cohen’s d for effect size is indicated below the horizontal bars. Error bars represent SD.

Fig. 6.

Stimulation frequency and history-dependent (stimulus number) sensitization. Top to bottom: the Early, Mid, Late, and Ongoing response phases, as described in the previous figures. Left: the average size of the respective phase of the dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram response (DENR) vs. stimulation frequency (all DCNs and all stimulation numbers were averaged) for both low and high stimulus strengths. Center and right: the quantification of each phase of the DENR over the first 10 stimuli at 1 Hz and 5 Hz, respectively, also for both low and high stimulus strengths. The last values at the 10th stimulation are indicated over the bars for both stimulus strengths. †Mid phase does not exist at the more rostral DCNs, so only the 4 most caudal DCNs (T₁₁–L₁) were analyzed. Ongoing phase is not analyzed for the 10-Hz stimulation. Error bars represent SD. Statistical significance was calculated by t-tests between the low- and high-stimulus strength data: *P < 0.05, **P < 0.01, ***P < 0.001. n = 72 (8 animals × 9 segments). Numbers below asterisks indicate the effect size, calculated with Cohen’s d. Dark gray bars, the higher stimulus strength (5 mA), capable of stimulating both A and C fibers; light gray bars, low stimulus strength (0.5 mA), capable of stimulating only A fibers.

Fig. 7.

Differences in the cutaneus trunci muscle reflex when evoked from ipsilateral and contralateral dorsal cutaneous nerve (DCN) stimulation at 5 mA. Top: false-color plots where the response sizes are defined as brightness being brighter with larger values, each line of pixels representing a single stimulus response, with the stimulus number on left. Stimulations at 5 mA stimulus strength at 5 Hz of representative DCNs (T₆, T₁₀, and L₁, in rostral to caudal order) are shown; each false-color image represents the average of 8 animals. Bottom: statistical differences between the contralateral and ipsilateral mean values across all 100 stimulations at 5 Hz from the data at top (t-test, n = 800, ***P < 0001). Numbers below asterisks indicate the effect size, calculated with Cohen’s d. Error bars represent SD.

Fig. 4.

Averaged responses from rostral (T₆–T₈), middle (T₉–T₁₁), and caudal (T₁₂–L₁) dorsal cutaneous nerve (DCN) groupings (columns), stimulated at 5 mA at different frequencies (rows). For each plot or bar shown, evoked responses are averaged across 20, 40, 100, and 200 stimulations at the defined frequencies of 1, 2, 5, 10 Hz, respectively, across 3 DCNs at the defined DCN grouping (rostral, middle, caudal), and across 8 animals (n = 480–4,800 depending on stimulation numbers). The 12 trace plots show the averaged neurogram responses with the phase boundaries (E, Early; M, Mid; L, Late; O, Ongoing). The quantification of those phases is shown in bar graphs at right and bottom. Right: the response size for different DCN groupings (R, rostral; M, middle; C, caudal) are shown for each frequency. Bottom: the response size for different frequencies (1, 1 Hz; 2, 2 Hz; 5, 5 Hz; X, 10 Hz) are compared for each rostrocaudal DCN grouping. Within each of the bar graph panels (at right or bottom), 3 DCN groupings or 4 frequencies were compared for statistical significance by ANOVA, shown with a bar above the set compared. All group tests showed that each set of frequencies or DCN groupings compared has significant differences (***P < 0.001, n = 480–4,800). The number below the significance (represented by asterisks) is the effect size of moving from the smallest to the largest value in each set (e.g., the shortest bar compared with the tallest bar in each set of 3 or 4), calculated with Cohen’s d. With 10 Hz, no data after 100 ms are included (the Ongoing phase), as the next stimulus began after 100 ms.

RESULTS

To understand the contribution of input parameters to the multisegmental CTM reflex, the DCN-evoked CTM neurogram response (DENR) was analyzed by stimulating DCNs at various spinal segmental levels (T₆–L₁) with different stimulation parameters including frequencies (1, 2, 5, or 10 Hz) and stimulus strengths (0.5 or 5 mA).

Typical CTM responses and Mid and Ongoing phases.

Previous reports on the CTM reflex (Doucette et al. 1987; Nixon et al. 1984; Petruska et al. 2014; Theriault and Diamond 1988b) have identified the large Early phase of the DENR as the A fiber-dependent phase and the prolonged Late phase as the C fiber-dependent phase of the reflex. The DENR recorded in this study has those responses in consistent time ranges, the highest-amplitude phase—the Early phase—from the stimulus onset to 25.5 ms and the long-lasting Late phase from 45.4 ms to 95.5 ms (Fig. 2 and Fig. 3). A previously unreported response was found sometimes and termed the Mid phase based on the time window (25.5–45.5 ms) between the Early and Late phases (Fig. 3). The Mid phase was evoked by more caudal DCN stimulation (Fig. 4). After the Late phase of the DENR, an unstructured increase in activity, termed the Ongoing phase, is observed lasting for up to a few seconds with no identifiable peaks. Existence of the Mid phase and the Ongoing phase, the increased background activity, suggests complex temporal and spatial features of the CTM reflex that have not been reported previously.

Stimulus strength dependence.

The convention of afferent population/stimulus strength dependence in the DENR was adopted (Petruska et al. 2014; Theriault and Diamond 1988b) and modified: an Early phase-activating stimulus strength of 0.5 mA to activate only A fibers and a stronger Late phase-activating stimulus strength of 5 mA to activate both A and C fibers. Whereas the low-stimulus strength (0.5 mA) DCN stimulation generally only evoked the Early phase of the DENR, high-stimulus strength (5 mA) DCN stimulation was sufficient to activate the Early, Mid, Late, and Ongoing phases of the DENR (Fig. 3, left). The high stimulus strength (5 mA) generated a significantly larger Early phase than the low stimulus strength (t-test, P < 0.001; Fig. 3, right), with a very large effect size (Cohen’s d = 1.2). The differences between the Mid, Late, and Ongoing phases evoked by the low and high stimulation strengths were also statistically significant (t-test, P < 0.001), with large to very large effect sizes (Cohen’s d: 0.9–1.4). This supports that the chosen stimulation strengths of 0.5 and 5 mA were proper to activate afferent populations selectively in this reflex and to investigate the stimulation strength dependence in further analysis.

DCN (spinal segmental level) dependence.

The size of the DENR was measured with stimulations on DCNs from the different spinal segmental (rostral, middle, caudal) levels at different stimulation frequencies (1, 2, 5, 10 Hz) (Fig. 4). We found significant differences in the size of all phases of the DENR based on which DCN (spinal segment) was stimulated.

Generally, the middle DCNs produced the largest Early, Late, and Ongoing phases of the DENR, whereas the caudal DCNs tended to produce the largest Mid phase of the DENR (Fig. 4, right). Although the middle and caudal DENRs were often of comparable size, the rostral DCNs consistently produced the smallest DENRs, regardless of which phase of the DENR was examined. DCN differences of each phase were statistically significant (1-way ANOVA, P < 0.001) at all stimulation frequencies, and the DCN-based effect sizes ranged from small to very large (Cohen’s d: 0.2–1.4), higher stimulation frequencies (2–10 Hz) tending to show larger effect sizes of the differences (Fig. 4, right).

Frequency dependence.

In our experiments, stimulation frequency was a major determinant of the size of the DENR. The Early phase of the DENR was largest at either 1-Hz or 2-Hz stimulation, depending on which DCN was stimulated (Fig. 4, bottom), and the largest overall Early phase from any DCN-frequency combination was at 2 Hz on the middle DCNs. Ten-hertz stimulation produced the smallest Early-phase response of the DENR of any of the stimulation frequencies tested, showing a very large effect size, with Cohen’s d values of 1.2, 1.5, and 1.6 in rostral, middle, and caudal segmental levels, respectively, compared with the largest Early-phase response from 1- or 2-Hz stimulation frequency at the same DCN level (see Fig. 4, bottom). Finally, the effect sizes (Cohen’s d values 1.2–1.6; Fig. 4, bottom) quantifying the significant differences of the Early phase across different stimulation frequencies tended to be greater than the effect sizes (Cohen’s d values, 0.8–1.4; Fig. 4, right) quantifying the significance of the Early-phase differences with which DCN was stimulated.

The latter phases of the DENR (Mid, Late, and Ongoing) showed larger frequency dependence than the Early phase (Fig. 4). Two- or five-hertz stimulation tended to produce the largest values for the latter phases of the DENR, depending on the exact combination of DCN-grouping and DENR-phase under examination. The differences attributed to stimulation frequency ranged from small to medium effects for the Mid phase (0.2–0.7), compared with medium to very large effects (0.6–1.9) for the other phases of the DENR (Fig. 4, bottom). The frequency dependence tended to be largest in the middle DCNs. Overall, the middle DCNs stimulated at 5 Hz produced the largest Late phase of the DENR.

Stimulation history dependence.

Strong stimulation history dependence was found in all phases of the DENR (Early, Mid, Late, and Ongoing) in terms of both increases and decreases in the responses. Figure 5 illustrates the effects of different DCN (spinal segmental) groups (caudal, middle, and rostral) and stimulus frequencies (1, 2, 5, and 10 Hz) on the DENRs. Several features are apparent in Fig. 5: 1) The Late phase of the DENR does not appear until the third to fifth stimuli—most apparent in the 1-Hz stimulation frequency data—and does not reach its full size until later stimulation numbers in the stimulus train. 2) The Mid phase of the DENR is most readily apparent at 2 Hz, in the middle and caudal DCNs (see also Fig. 4). 3) In the 10 Hz data (Fig. 5, bottom left), both the Early and Late phases of the DENR decrease in size with increasing stimulation number in all DCN groups. This phenomenon is apparent to a lesser degree in the 5 Hz data. 4) An increase in the Ongoing phase of the DENR is apparent at the 2- and 5-Hz stimulation frequency data. It was not possible to quantify whether this increase existed for the 10 Hz data because of the timing of the next stimuli (at 100 ms after the last stimuli). 5) The peak latency of the Mid and Late phases of the DENR is increasing in both 5- and 10-Hz stimulation with increasing stimulation number.

The Early and Late phases of the DENR both sensitize and habituate together (Fig. 5, right), overlaying one another when they increase and decrease together over the course of sequential stimuli. An important exception to this phenomenon is the first-to-second-stimulus reduction in the Early DENR as seen in the 10 Hz data, discussed below in Habituation.

Habituation.

Habituation is a phenomenon where the nervous system responds progressively less to an identical subsequent stimulus (Groves et al. 1969). Late in the stimulus train, the DENR habituates (reduces in size). This phenomenon is most apparent at 5 and 10 Hz stimulation frequencies (Fig. 5 and Table 1). Habituation generally occurred at both low and high stimulus strengths, and more strongly at higher frequencies with longer pulse trains.

Table 1.

Quantification of habituation

Stimulation Type	Early Phase	Mid Phase	Late Phase	Ongoing Phase
Ipsi, 0.5 mA
1 Hz	0.85	—*	0.95‡	0.97†
2 Hz	0.92	—*	0.88†	0.92†
5 Hz	0.86	—*	0.84†	0.89†
10 Hz	0.63	—*	0.85†	—‡
Ipsi, 5 mA
1 Hz	0.99	0.94	1	1
2 Hz	0.92	0.82	0.91	0.92
5 Hz	0.81	0.66	0.77	0.8
10 Hz	0.79	0.7	0.67	—‡
Contra, 5 mA
1 Hz	0.88	—*	1.00†	0.96†
2 Hz	0.86	—*	1.00†	1.00†
5 Hz	0.84	—*	0.85†	0.82†
10 Hz	0.6	—*	0.81†	—‡

We quantified habituation in the dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram response (DENR) with last-to-maximum ratio of the mean value (averaged across animals and DCNs) for each phase of the DENR. Smaller values indicate more habituation, and a value of 1.00 would mean no habituation.

^*There was no Mid phase of the response in the contralateral or low-stimulus strength data.

^†Ratios closer to 1.00 here indicate very small responses, which could not undergo much habituation.

^‡The next stimulus coincided with the Ongoing phase at 10 Hz. Contra, contralateral; Ipsi, ipsilateral.

Habituation is stronger in the low-stimulus strength (0.5 mA) DENRs than the high-stimulus strength (5 mA) DENRs (Table 1) and is most prevalent in the 10 Hz data. In the ipsilateral DENRs, where the all phases of the DENR consistently appear, the Late phase undergoes more habituation than the Early phase of the DENR.

There was an additional type of short-term habituation observed very early in the DENRs: a decrease in the Early phase of the DENR between the first and second stimuli, instead of the previously discussed type of habituation appearing much later in the stimulus train. This first-to-second-stimulus habituation was evident at all frequencies but was most prevalent at 10-Hz stimulation (a 40% decrease in the Early phase between the 1st and 2nd DENRs). In addition, this first-to-second-stimulus habituation only appeared in the Early phase of the DENR, not the Mid, Late, or Ongoing phases. This type of habituation either was very short term or was overwhelmed by sensitization (see below), as the Early phase of the DENR was always larger by the third to fifth stimuli.

Sensitization.

Sensitization is a phenomenon where the nervous system evokes an initial increase in the response to an identical subsequent stimulus (Groves et al. 1969). This phenomenon was found in the CTM reflex earlier in the stimulus trains, with higher frequencies leading to more sensitization (Fig. 6). Qualitatively, the high-stimulus strength DENRs increase with more stimuli, whereas the low-stimulus strength DENRs decrease or remain the same size (Fig. 6). Quantitatively, the high-stimulus strength (5 mA) Early phase grew by >50% by the 10th stimuli of the train in both 1- and 5-Hz stimulation DENRs (Fig. 6, top, center, and right). The low-stimulus strength DENRs showed no such growth, staying at approximately the same value at the 10th stimulus as they were at the 1st stimulus. The Mid and Late phases of the DENRs showed larger differences than the Early phase: by the 10th stimuli, the 5-mA-stimulated Mid and Late phases were often twice as large as the 0.5-mA-stimulated Mid and Late phases. The Ongoing phase of the DENR was the most frequency dependent, with 1-Hz stimulation showing relatively small values whereas 5 Hz showed large values. To quantify the sensitization in the phases of the DENR, we compared the first response (the response from the 1st stimulus in the stimulus train) to the maximal response for each phase during each stimulus train, shown in Table 2.

Table 2.

Quantification of sensitization

Stimulation Type	Early Phase	Mid Phase	Late Phase	Ongoing Phase
Ipsi, 0.5 mA
1 Hz	1	—*	1.48	1.25
2 Hz	1	—*	1.75	1.64
5 Hz	1	—*	1.98	1.87
10 Hz	1	—*	1.87	—†
Ipsi, 5 mA
1 Hz	1.44	2.08	3.5	2.02
2 Hz	1.69	2.76	4.03	3.09
5 Hz	1.61	2.53	3.98	3.65
10 Hz	1.14	2.12	2.95	—†
Contra, 5 mA
1 Hz	1.21	—*	1.5	1.19
2 Hz	1.34	—*	2.33	1.81
5 Hz	1.13	—*	2.33	2.34
10 Hz	1	—*	2.06	—†

We quantified sensitization in the dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram response (DENR) with max-to-first ratio of the mean value (averaged across animals and DCNs) for each phase of the DENR. Larger values mean more sensitization, and a value of 1.00 means that the first response was the max response (i.e., no sensitization).

^*There was no Mid phase of the response in the contralateral or low-stimulus strength data.

^†The next stimulus coincided with the Ongoing phase at 10 Hz. Contra, contralateral; Ipsi, ipsilateral.

The DENR from the first low-stimulus strength stimuli in the stimulus train was consistently the largest, demonstrating a lack of sensitization at the low stimulus strength (Table 2). In the high stimulus strength, the first DENR was never the largest. In the high-stimulus strength trains, the Late phase of the DENR grew much more than the Early phase (2.95–4.03 vs. 1.14–1.69), and 2 Hz produced the strongest sensitization (followed closely by 5 Hz) among the frequencies tested (Table 2).

Lateral differences.

The contralateral DENR is smaller, generally consisting of only the Early phase of the DENR, compared with the ipsilateral side with Early, Mid, Late, and Ongoing phases (Fig. 7). The contralateral DENRs were significantly smaller than the ipsilateral DENRs (Student’s t-test, P < 0.001) in 97% of the 49,600 compared pairs in direct pairwise comparisons of DENRs of the same phase (3 or 4 phases), in the same animal (8 rats), from stimulation at the same DCN (5 DCNs), with the same stimulation frequency (4 frequencies) and stimulus strength (5 mA), at the same stimulation number (20–200). There was no case in which the contralateral DENR was significantly greater than the ipsilateral DENR.

In the Early phase of the DENR, the average contralateral DENR was 29 ± 27% (SD) of the Early phase of the ipsilateral DENR across all DCNs (spinal segments) tested. However, there were DCN-to-DCN differences in this ratio (Fig. 7). The T₆ DCN had the smallest effect size when the contralateral DENR was compared to the ipsilateral DENR, and the relative maximum contralateral-to-ipsilateral ratio was at T₆ at 2 Hz (mean: 54%) for the Early phase of the DENR. It was not possible to compare the Mid, Late, or Ongoing phases of the contralateral side, because the contralateral DENR only rarely had those phases (Fig. 7).

There were lateral differences in both sensitization and habituation. Contralateral stimulation produced less sensitization than ipsilateral stimulation (1.00–1.34 vs. 1.14–1.69) in the Early phase, as shown in Table 2. As in the case of ipsilateral DENRs, contralateral DENRs from 2-Hz stimulation showed the most sensitization of any stimulation frequency (Table 2). Habituation of the Early phase of the DENR was similar or even stronger on the contralateral side than the ipsilateral side (0.60–0.88 vs. 0.79–0.99), and both ipsilateral and contralateral DENRs had the most habituation from 10-Hz stimulation.

Latency.

We quantified the latency of the DENR in order to better understand the neural connections that allow this reflex to function in the spinal cord. Longer latencies imply more synapses, or unmyelinated pathways, whereas shorter latencies put an upper limit on the number of synapses possible and may imply myelination of the axons involved.

The onset latency defined as the time point at 50% of the maximal amplitude for the Early phase of the DENR was 9.8 ± 2.0 ms (or 14.6 ± 1.6 ms for the peak latency) on the ipsilateral side. There were small but statistically significant differences in onset latency of the Early phase across ipsilateral DCNs, but they all fell between 9.5 ms and 10.5 ms. In general, the more caudal ipsilateral DCNs’ Early phase onsets were slightly more delayed (T₆ at 9.6 ms, T₈ at 9.8 ms, T₁₀ at 10.3 ms, T₁₂ at 10.4 ms), but L₁ was slightly earlier than its neighbors, at 9.8 ms. Contralateral DENRs, on the other hand, were delayed, with an additional 4.3 ms compared with the ipsilateral side (onset: 14.1 ± 3.9 ms, peak: 18.9 ± 3.7 ms) on average.

Changes of the peak latencies of the Early and Late phases of the DENR with repetitive stimulation at 5 Hz were quantified on both ipsilateral and contralateral sides (Fig. 8). With ipsilateral DCN stimulation, the peak latency of the Early phase of the DENR decreases with more stimulation (~10% earlier by the end of the stimulus train), following a parallel trend with sensitization of the DENR size for the first half of the stimulus train (0–50 of stimulation number; Fig. 8, middle). However, the ipsilateral peak latency stays decreased even during the late habituation period at the second half of the stimulus train. In addition, the short-term habituation between the first and second stimuli discussed above is reflected in the Early latency data with the short-term increase, as the second DENR is shortly slower than the first DENR (1.0–1.6 ms, depending on frequency), despite the long-term decrease.

Fig. 8.

Early and Late phase latencies compared with Early and Late phase response sizes over the stimulus train. All dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram responses shown are from 5-Hz stimulation data. Top: an example of the original, rectified average signal of the 40th stimulus response from 5-Hz ipsilateral stimulation; within middle and bottom panels, the top plots show that signal filtered by a low-pass Butterworth filter (at either 250 Hz for the Early phase or 100 Hz for the Late phase). Middle: the Early-phase peak latencies were measured on both the contralateral and ipsilateral sides over the entire stimulus train. As can be seen in the plot, the Early phase’s latency decreases during the train, whereas the Early phase’s response size increases and then decreases later in the stimulus train (farther right). The only exception to these trends is in the 1st-to-2nd stimulus response change, where both the latency increases and the size of the response is reduced. This early, 1st-to-2nd, stimulus change is referred to as early habituation in results and discussion. Bottom: the Late-phase peak latency was measured only for the ipsilateral side, as the contralateral side has only a minimal Late phase of the response. The Late-phase peak latency consistently increased with more stimuli, whereas the Late-phase response size increased in the beginning and then decreased later in the stimulus train. In all cases, the data sets were averaged across DCNs and animals.

The peak latency of the Late phase of the DENR changed over the entire course of the train, but in the opposite direction to the Early phase (Fig. 8, bottom): the Late phase consistently became more delayed instead of getting earlier. In addition, the trend did not follow the sensitization or the habituation of the Late phase, nor did it follow the trend of the Early phase getting earlier. This Late-phase-getting-later phenomenon was most apparent in the 5 Hz and 10 Hz data. In Fig. 8, bottom, the peak latency for the 5 Hz data went from roughly 63 ms to 75 ms by the end of the stimulus train (a 20% increase).

Refractory period/inhibition.

Activity-induced quiet periods on neurons are common for two reasons: 1) Neurons will not fire again after a period of firing because of depletion of some key chemical or returning to homeostasis (refractory period). 2) Alternatively, neurons may have a reciprocating connection with an inhibitory neuron, diminishing activity on the original neuron if it activates the inhibitory neuron (e.g., Renshaw cells). Both of these phenomena can appear as a postactivity quiet period. This type of quiet period was found in the DENR in two forms, both of which are illustrated qualitatively in Fig. 9. In Fig. 9, left, the prestimulus activity for 50 ms before each stimulation at 5 Hz, which includes a late part of Ongoing phases (150–195.5 ms) increased gradually at the early stimulation train of repeated stimulations and remained throughout the rest of the stimulation train. However, there is a quiet period between the Early and Late phases of the DENR where the activity level is lower than the prestimulus activity (inhibition/refractory period). In the more caudal DCNs, this quiet period appears between the Early and Mid or between the Mid and Late phases. To relate the quiet period to the latency of Late phases in rostral DCNs, the averaged responses are shown across DCNs in Fig. 9, right. The latency of the Late phase of the DENR is delayed by more caudal DCN stimulations when a strong Mid phase presents. The delays for the different DCNs for the Late phase are all in the range of the 60–70 ms average until L₁, which has a peak latency of 80 ms (Table 3).

Fig. 9.

Inhibition/refractory periods in the cutaneus trunci muscle (CTM) reflex. Left: a false-color image of the average response across all dorsal cutaneous nerves (DCNs) for each stimulus over the repeated stimuli at 5 Hz at 5 mA, beginning 50 ms before the stimuli to show the prestimulus activity; “prestimulus activity,” “stimulus artifact,” and “inhibition/refractory” time points are indicated. Right: the evoked responses averaged for each DCN across all repeated stimuli at 5 Hz at 5 mA to investigate effects of inhibition/refractory periods on the DCN-evoked CTM neurogram response Mid phase. The Late phase is delayed in the more caudal DCNs when the Mid phase exists, with a dashed line highlighting the delay.

Table 3.

Late DENR latency from ipsilateral stimulation demonstrating refractory/inhibition

DCN (Spinal Segment)	5 Hz Late-Phase Latency, ms
T6	60.0
T8	60.5
T10	68.0
T12	63.8
L1	80.2

Latency was calculated as the peak of the Late phase of the dorsal cutaneous nerve (DCN)-evoked cutaneus trunci muscle neurogram response (DENR) after it was rectified and low-pass filtered.

DISCUSSION

The novelty of this study is the detailed temporal, stimulus strength-based, lateral and segmental quantification of the CTM reflex and our confirmation that this reflex shares signal processing features (wind-up and habituation) seen in other pain reflexes. 1) The Mid phase of the CTM reflex, between the Early and Late phases, was identified and characterized. Analogs of this Mid phase can be seen in other nociceptive reflexes. 2) We identified segmental and lateral differences in the DENR, compared them to latency differences, and found results suggestive of how the reflex is anatomically organized. 3) Detailed characterizations of stimulation frequency and stimulation history dependence were performed to quantify sensitization (wind-up) and two kinds of habituation. Although many of the phenomena (e.g., wind-up) observed in this reflex are consistent with observations from other nociceptive reflexes, some (e.g., segmental comparisons) are unique to this reflex. In addition to these physiological findings, our detailed characterization of the “normal” reflex will help lay further groundwork for this reflex to be used in the study of neural development, plasticity, pharmacology, and peripheral or central nervous system injury/disease.

CTM reflex exhibits wind-up.

Herrero et al. (2000) define wind-up as “a progressive, frequency-dependent facilitation of the responses of a neuron observed on the application of repetitive (usually electrical) stimuli of constant intensity.” It has historically been considered to be a C fiber-dependent phenomenon (Herrero et al. 2000; Mendell and Wall 1965) and could be considered to be one of the most well-studied and important features of spinal nociception. The CTM reflex response demonstrates this phenomenon, and this facilitation can be observed in the CTM reflex at similar stimulation frequencies as wind-up in the flexor reflex (Arendt-Nielsen et al. 1994).

Wind-up was observed in both Early (A fiber) and Late (C fiber) phases of the DENR with high stimulus strength. Wind-up was most easily observed at high-frequency stimulation on the ipsilateral side. Low-stimulus strength (A fiber only) stimulation, however, did not produce wind-up in otherwise identical conditions (same frequency, DCN, stimulus number). Contralateral DENRs had very weak wind-up compared with ipsilateral DENRs (despite identical stimulus strength). When wind-up does occur, both Early and Late phases of the DENR increase together, supporting the concept that the same process at the same neurons may be responsible for at least inducing the wind-up of all phases (Early, Mid, Late, and Ongoing) of the response. This concept that wind-up is occurring at the same neuron for both Early and Late phases of the DENR suggests that wind-up as a process resides at the second (most likely) or subsequent neuron in the neural circuit underlying the CTM reflex (Fig. 10).

Fig. 10.

Proposed neural organization of the cutaneus trunci muscle (CTM) reflex based on physiological features. Peripheral inputs start at bottom, going up into the ipsilateral and contralateral spinal cord. The thicker line shows the myelinated fibers, and the thinner line shows the unmyelinated fibers. There exists a connection from the contralateral side to the ipsilateral side, but it is unknown whether it is polysynaptic or just unmyelinated (or both). The CTM motor nucleus is shown at top with its ipsilateral connection. Contra- and ipsilateral propriospinal interneurons may in fact be multiple/different interneurons, but for simplicity in the diagram we drew them as an individual group. Different physiological features of the reflex are localized in the diagram as discussed in results and discussion.

Although wind-up was both stimulus strength and frequency dependent in all phases (Early, Mid, Late, and Ongoing) of the DENR, the Ongoing phase showed the largest difference from stimulation frequency (Fig. 6). The differences between the phases of the DENR demonstrate the complexity of wind-up: It is a graded response affecting different phases of the same reflex differently, not just an on-or-off phenomenon.

The C fiber dependence of wind-up is also notable, as a parallel to the feeling of pain. In a human arm flexion reflex study, it was found that the C fiber-dependent late phase corresponded much better with perceived pain than the early phase of that reflex (Willer 1977), suggesting that the DENR’s Late phase is more salient as a “pain” signal than the DENR’s Early phase.

CTM reflex exhibits habituation.

Habituation is a phenomenon where the nervous system responds less to an identical subsequent stimulus of the same intensity. It is one of the ubiquitous types of short-term plasticity in the nervous system. We observed habituation in all responses, regardless of stimulus strength, side (contralateral/ipsilateral), DCN (spinal segment), or frequency. This observation is in contrast to some previous work by Blight and colleagues (Blight et al. 1990), likely explained by the relatively short stimulus trains they used (they only reported up to the 10th stimuli). Although habituation tends to be more apparent late in the stimulus trains (Fig. 5), it can start early in the responses that exhibit little to no wind-up (low stimulus strength and contralateral DENRs), supporting habituation as a separate process that competes with sensitization in the CTM reflex. This separation and competition between wind-up and habituation has been observed in the cat flexor reflex (Groves and Thompson 1970), where it was noted that (late) habituation eventually overwhelms sensitization. This competition and the ability of habituation to ultimately overcome wind-up suggest that it occurs at a later phase of the CTM reflex, perhaps on the output of the second neuron, or at the third neuron, as shown in Fig. 10. However, our data in this report cannot disprove habituation in any other parts of the neural pathway (e.g., C fibers).

Two types of habituation: early and late.

In addition to the slow, late habituation described previously, our data showed a second type of habituation occurring early in the stimulus trains (even before wind-up): the largest single stimulus-to-stimulus decrease in the response was often observed between the first and second stimuli in the train. This reduction was specific to the Early phase of the DENR, as the Late phase tended to grow between the first and second stimulation response in the train and was more pronounced at higher stimulus frequencies (5 and 10 Hz). The Early and Late phases tended to move together (wind-up or habituation) outside of this first-to-second stimulus response change (Fig. 5, 10 Hz bubble plot). This exception to the normal trend suggests that some change is occurring specific to the A fibers between the first and second stimulations. As discussed below in Stimulation changes reflex latency in divergent ways, this second type of habituation appears to be incorporated into the process affecting changes in Early-. phase latency (both onset latency and peak latency). In the proposed organization in Fig. 10, this phenomenon is shown in “Ipsilateral Interneurons.”

A Mid phase exists between Early and Late phases at caudal DCNs.

The latencies of the Early (4–25.5 ms) and the Late (45.4–95.5 ms) responses in this report were at least consistent with the measurement of CTM neurograms evoked by electrical stimulations of individual DCNs in rats as shown in a previous report (Petruska et al. 2014). CTM response latencies in EMG recordings appeared more variable depending on electrode locations (stimulating and recording) compared with the CTM neurogram recordings. EMG recordings with DCN stimulations in an early report (Theriault and Diamond 1988b) showed a slightly different onset of the early response (15–27 ms) but a quite late onset of the late responses (82–107 ms) in rats. However, the other EMG study showed completely different, variable latencies depending on the location of stimulating electrodes on the skin in guinea pigs (Blight et al. 1990).

We found a Mid phase in the reflex response from more caudal DCN stimulation at higher stimulation frequencies and strengths. While the Early and Late phases in the CTM reflex response have been identified as A fiber and C fiber dependent, respectively (Nixon et al. 1984; Theriault and Diamond 1988b), the Mid phase in the CTM reflex has not been described previously. The Mid phase may be a portion of the Early response that gets progressively delayed. Area under the curve analysis was considered, but the Early phase was actually increasing in size (Fig. 6) during the period that the Mid phase develops. Also, the fact that the Mid phase is localized to only the more caudal DCNs does not support it being a general delayed response across the whole system.

An equivalent to the Mid phase has been observed in the flexor reflex, considered to be Aδ in origin (Woolf and Swett 1984). The CTM reflex Mid phase is unusual, as it is wind-up dependent (Fig. 5 and Fig. 6). It also tends to habituate more quickly than the Late phase of the DENR (Fig. 5). The wind-up dependence and faster habituation support the Mid phase as being more fragile than the other phases of the reflex response. The fact that it sensitizes and habituates with both early and late makes it difficult to conclude whether it is a later A-fiber (perhaps Aδ) response or an earlier C-fiber response.

The rostral DCNs (entering rostral spinal segments) did not have this phase of the reflex response at all; despite those nerves having density of Aδ and C fibers similar to the caudal DCNs (Tansey et al. 2010), fiber densities cannot explain the existence of the Mid phase. The Mid phase only exists in the caudal DCNs, close to the location of the flexor reflex, and the flexor reflex has an analog of the Mid phase (discussed above). Therefore, we speculate that it may be a characteristic of the spinal neurons or neural circuits near the lumbar enlargement.

Ipsilateral responses partially reflect peripheral anatomical differences.

Segmentally, the largest DENRs were evoked from the middle or caudal DCNs, and there are two potential hypotheses for why that could be the case: 1) There could be differences in spinal processing in terms of the segment the afferents arrive at, such as interneuronal differences or the CTM motor nucleus giving more weight to specific afferents. 2) Every afferent input onto the motor nucleus was weighted the same, but the middle and caudal DCNs have more of the relevant types of afferents. Our laboratory has previously performed an anatomical analysis of the DCNs, finding a high variability in nerve diameters and density of the afferent numbers in the DCNs (unpublished data). There is not sufficient evidence to conclude definitively which hypothesis is true, but the evidence we have (e.g., contralateral DENRs stronger rostrally, Mid phase only exists caudally, etc.) suggests differences in spinal processing rather than differences in relevant numbers of afferents.

Stimulation changes reflex latency in divergent ways.

The peak latency of the DENR was quantified in order to better understand the neural connections that underlie this reflex in the spinal cord. We found the unexpected result that the latencies of the Early and Late phases of the DENR undergo opposite changes with more stimulation and these changes only partially follow the amplitude fluctuations (wind-up/habituation) of either the Early or the Late phases of the reflex, respectively. In Fig. 8, the peak latency of the Late phase consistently increases with more stimulation, regardless of whether habituation or sensitization is dominating, suggesting that the process causing the increased latency is not part of the core reflex circuit. C fiber conduction velocity is the most likely explanation for the Late phase changes, as unmyelinated fibers have been found to conduct more slowly with repeated action potentials (Thalhammer et al. 1994), likely because of changes in sodium channel inactivation (De Col et al. 2008).

Unlike the Late phase, the Early phase latency (both onset and peak latency) actually decreases with stimulation. Although Blight et al. (1990) did not discuss this phenomenon, their figure showed the Early phase latency decreasing in the first 10 stimuli in the guinea pig. Conduction velocity in myelinated fibers is also known to get slower with increased stimulation (Thalhammer et al. 1994), so conduction velocity cannot explain this phenomenon. In fact, the documented increases in myelinated fiber latency further emphasize this unusual trend, as the decreases in latency measured would have to overcome the slower conduction velocities.

The size of the Early phase only partially follows the latency changes. During the early (1st to 2nd stimuli) habituation and wind-up, the latency seems to match the changes in the size of the reflex, but the latency stayed short through the late habituation period (Fig. 8). These observations support wind-up, early habituation, and latency of the Early phase of the reflex all sharing the same part of the CTM reflex, perhaps even the same interneurons (Fig. 10). The observation that the late habituation seems to be independent of latency suggests that late habituation occurs at a common part further along in the pathway, such as in the motoneurons themselves (Fig. 10). It will take more work than the present set of experiments to make these determinations, though.

Ipsilateral propriospinal connections are well myelinated.

Latency can establish upper bounds on the possible number of connections or level of myelination in the neural circuit inside the spinal cord. The differences in onset latency between the rostral and caudal segments for the Early phase were very small, <1 ms. This small difference is notable because signals from caudal DCNs (T₁₃ or L₁) have to travel an additional distance of ~1 cm inside the spinal cord (Miller et al. 1973) compared with signals from rostral DCNs (T₆). Interestingly, the DENR from the L₁ DCN has about the same onset latency as T₆, despite L₁ being the most caudal of the segments tested. For onset latency differences to be <1 ms, and the distance to be ~1 cm, it would imply that at least the Early phase of the reflex is conveyed on fibers with a roughly 10 m/s conduction speed, suggesting large, well-myelinated fibers. This need for a high conduction speed could be slightly lessened by a motor nucleus extending up or down a few segments, but even if the requirements were reduced to 8 m/s, it would still require myelination in that pathway.

Blight et al. (1990) found that the ipsilateral reflex was carried by interneuronal axons in the ventral half of the lateral funiculus of the spinal cord. Based on the short latency of the ipsilateral reflex, they also pointed out that a candidate spinal interneuron population with axons in the ventrolateral tract has monosynaptic connections to the region where the CTM motor nucleus is found in the cervicothoracic spinal cord. They suggested this pathway through a reinterpretation of a study that presumed that this pathway was related to interlimb coordination in locomotion (Miller et al. 1973). The latency of the pathways described is on the order of what would be needed for the latency results we have shown: they show the latency from L₁ to T₉ as being on the order of 1–2 ms. Overall, our latency results support the hypothesis that the interneuronal pathway carrying the CTM reflex is part of or similar to the pathway described by Miller et al.

Contralateral connection is polysynaptic.

As in Ipsilateral propriospinal connections are well myelinated, latency measurements can establish bounds on the possible number of synapses or level of myelination in a pathway. We found the contralateral response to be smaller and later than the ipsilateral response, in agreement with previous results in guinea pigs (Blight et al. 1990). The latency of the contralateral Early phase is ~4 ms slower than the ipsilateral equivalent, implying at least one neuron with synaptic delays and/or an unmyelinated connection across the spinal cord.

Previous research on this reflex in guinea pigs (Blight et al. 1990) hypothesized that the contralateral response was the result of decussation of the afferents upon entry into the spinal cord. Since these authors have not seen the C fiber-mediated late responses in both ipsi- and contralateral reflexes in the guinea pig, the presumed decussation should be limited to myelinated afferents. These authors were able to evoke the CTM reflex by a low-threshold mechanical stimulation at a single hair as well as by an electrical stimulation on the skin, implying the possible contribution of nonnociceptive afferents, e.g., Aβ fibers to the contralateral responses in the guinea pig. Although we also have activated all afferent types in DCNs, the CTM reflex in the rat is known to be nociceptive specific (Doucette et al. 1987; Nixon et al. 1984), rejecting the involvement of nonnociceptive afferents in our data except for the very small chance that nonnociceptive afferents are involved exclusively to the contralateral responses but not to ipsilateral responses.

Our results tended to support an alternate hypothesis. The contralateral DENR generally does not include a Mid, Late, or Ongoing phase even at high stimulus strengths and frequencies in the rat (Fig. 7). In this regard, the absence of the Mid, Late, and Ongoing phases supports the hypothesis that only myelinated afferents decussate upon entry into the spinal cord. Although this hypothesis is still possible, and some cutaneous afferents do cross the midline (Smith 1986), our latency data do not support this hypothesis as the mechanism of the primary contralateral pathway in this reflex. Signals on bilaterally projecting, myelinated A fibers would not take 3–5 ms to cross a rat spinal cord. In addition, our wind-up analysis shows that wind-up is C fiber dependent, and the contralateral response shows at least minimal wind-up (more than in our no-wind-up, low-stimulus strength data), so it must undergo wind-up at a neuron with some access to contralateral C fibers. Therefore, we hypothesize that an unmyelinated midline-crossing interneuron conveys the contralateral response, at least in the rat (shown in Fig. 10).

Habituation, on the other hand, appears to occur equally or even more strongly in the contralateral response, further supporting late habituation as being in a different part of the reflex than wind-up or latency changes (discussed above).

There also appear to be some segmental differences in the contralateral response. Rostral segments tended to have a stronger contralateral response, sometimes even including a weak Late phase. Contralateral caudal DCNs generated almost no response.

CTM reflex exhibits a refractory period or self-inhibition.

An activity-induced quiet period was found in our responses (Fig. 9). There are two common reasons for such a quiet period, the neuron’s refractory period or an inhibitory reciprocating connection. Although our data cannot conclude definitively which of these two reasons is responsible for the quiet period in the CTM reflex, the duration of the quiet period (~10 ms) supports the second reason—a reciprocal inhibitory connection. We have also seen that the reflex can inhibit itself across distant DCNs, from inputs as far distant as a tail pinch stimulus inhibiting a long-lasting DENR at T₁₀ (unpublished observation). This type of inhibitory connection is common in the nervous system (e.g., Renshaw inhibition) and more so in nociception (e.g., diffuse noxious inhibitory control, the gate theory of pain, etc.). Therefore, we hypothesize the existence of a reciprocal inhibitory connection in the CTM reflex, likely near the CTM motor pool.

Similarities to and differences from other cutaneous reflexes.

There are many nociceptive reflexes in the mammalian nervous system, and the similarities/differences between these reflexes can suggest commonalities of nociception. The flexor reflex is a well-studied example of a nociceptive reflex with several similarities to the CTM reflex. The relative timing of the phases from the CTM reflex and the flexor reflex are similar. Woolf and Swett (1984) demonstrated the flexor reflex by stimulating the sural nerve and recording from the hamstring efferents. The responses in that study appear very similar to those in our study, with the “C strength” flexor-reflex stimulation having Early, Mid, Late, and Ongoing equivalents. Herrero et al. (2000) recorded from a “multireceptive” deep dorsal horn neuron and found similar phases in the response.

The similarities to the flexor reflex mentioned above (e.g., Early/Late phases, sensitization, habituation, and stimulus strength similarities, Late phases getting later, etc.) suggest that the underlying spinal neural organization may be similar between the two reflexes. However, several differences exist that could be advantageous from an experimental perspective: 1) The CTM reflex can be evoked from different segments and offers a relatively similar contralaterally stimulated version as well (more similar than the crossed-extension-reflex analog for the withdrawal reflex). 2) The reflex’s motoneurons are separated from the sensory inputs by 6–13 segments (C₇–T₁ motoneuronal pool), which facilitates independent manipulation of the input and output parts of the reflex, for example, applying lesions, pharmacological agents, or other interventions to only the sensory side of the reflex. 3) The CTM reflex can be evoked at a higher level of anesthesia. The existence of the flexor reflex was used to determine when to give pentobarbital boosters, and the CTM reflex still produced a strong response at levels of anesthesia that abolished the flexor reflex. Therefore, the CTM reflex is less sensitive to pentobarbital anesthesia.

Limitations of approach.

Our experiments had to use anesthesia for ethical reasons, but anesthetics affect pain processing. Pentobarbital is one of the few anesthetics that do not abolish this reflex, but pentobarbital may have affected the reflex in minor ways nonetheless. Pentobarbital can change the responses of spinal neurons to be more like wide-dynamic range responses at a dosage lower than what we used (Collins and Ren 1987). In addition, pentobarbital can actually block AMPA receptors (Taverna et al. 1994), but the dose required is much higher than what we used (10 mM, compared with the ~0.2 mM we used). Nevertheless, the choice to use a reduced preparation allowed us to study the effects of both stimulation strength (afferent populations activated) and stimulation frequency in a precise manner.

Conclusions.

This study presents a thorough characterization of a nociceptive cutaneous reflex temporally and spatially, with different afferent populations. We identified and characterized the Mid phase of the CTM reflex. We identified segmental and lateral differences in the DENR, compared them to latency differences, and found results suggestive of how the reflex is anatomically organized. We identified and quantified wind-up as well as two types of habituation. This characterization builds on the previous work in this reflex, with the hope of laying down a foundation for better understanding of spinal cord physiology, spinal cord injury, and pain physiology.

GRANTS

This work was funded by start-up funds from the Department of Physiology, Emory University School of Medicine and the Spinal Cord Injury Research Program at Shepherd Center in Atlanta, GA as well as National Institute of Biomedical Imaging and Bioengineering Grant EB-006179 and a Khalifa University Research Grant.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.E.T. conceived and designed research; J.M.W., H.J.L., P.M., and K.E.T. performed experiments; J.M.W., H.J.L., P.M., S.P.D., and K.E.T. analyzed data; J.M.W., H.J.L., and K.E.T. interpreted results of experiments; J.M.W. prepared figures; J.M.W., H.J.L., and S.P.D. drafted manuscript; H.J.L., P.M., S.P.D., and K.E.T. edited and revised manuscript; K.E.T. approved final version of manuscript.