The challenge of treating chronic pain remains considerable. Some estimates place the incidence of chronic pain at 40% or more (Dahlhamer et al., 2018; Fayaz et al., 2016; Cohen et al., 2021) and the most recent Global Burden of Disease data indicate that musculoskeletal pathologies, including lower back pain and osteoarthritis, are particularly prevalent worldwide (GBD 2019 Diseases and Injuries Collaborators, 2020). Accelerated by the poor outcomes, side-effect profiles, and addiction problems (e.g., Scholl et al., 2018) associated with pharmacological therapy, coupled with the lack of progress in delivering novel analgesic drugs, the use of neuromodulation has expanded as an alternative strategy for pain relief (Knotkova et al., 2021).

Since its introduction more than 50 years ago (Shealy et al., 1967), conventional spinal cord stimulation (SCS) has been used to treat chronic pain. Stimulating current pulses delivered via epidural electrodes at a moderate tonic frequency of 40-70 Hz are intended to activate large-diameter, myelinated Aβ fibers that typically carry information for non-painful stimuli such as touch and pressure. This activation eventually leads to inhibition of incoming signals from small-diameter, nociceptive, weakly myelinated Aδ or unmyelinated C fibers, as predicted by the Gate Control theory of pain (Melzack and Wall, 1965). Thus, by activating Aβ fibers, conventional SCS generates concomitant paraesthesia over the appropriate locus of pain, with the aim of reducing nociceptive signaling in the dorsal horn of the spinal cord via activation of low-threshold, non-nociceptive elements (Taylor et al., 2014).

Novel forms of SCS, including burst and high-frequency paradigms, have been developed in an attempt to generate paraesthesia-free reduction of pain signaling and improve outcomes (Kapural et al., 2015; Deer et al., 2018; Lempka and Patil, 2018). In contrast to these stimulation strategies, which predominantly utilize high frequency and/or short pulse width trains, we have recently described a novel form of neuromodulation that employs an ultra-long pulse, ultra low frequency (ULF) biphasic current waveform (Jones et al., 2021). Each phase of the ULF waveform uses a slowly ramped rise, avoiding neural excitation, to a plateau phase, several seconds in duration and directly inhibits the passage of action potentials along axons. In our laboratory experiments in anesthetized rats, we showed that this waveform produces a rapidly-developing, fully reversible conduction block in sensory nerve fibers excited by peripheral stimulation. Our modeling work described how this inhibition is achieved immediately by inactivation of sodium channels during the plateau phases of the waveform. The slow ramped rise to the plateau phase avoids activation of axons. We also identified a delayed conduction block, developing over several minutes, that arises from profound changes in intracellular and extracellular potassium ion concentrations, leading to a depolarising shift and prolonged action potential block. In clinical studies, the ULF waveform is highly effective in reducing pain scores in patients suffering chronic back or leg pain (Jones et al., 2021).

With conventional metal electrodes, such long current biphasic pulses, approximating to short periods of direct current, would be damaging to neural tissue. However, the use of electrodes composed of material with a very high charge capacity circumvents any such effects. Such material allows the ultra long biphasic pulses to follow the primary rules of traditional biphasic neuromodulation pulses to avoid tissue or electrode damage (i.e., avoid the voltage range that causes water electrolysis and electrode potentials that would cause metal dissolution). The waveform has undergone extensive bench and animal testing to examine these risks and establish stability (Harris et al., 2022; Verma et al., 2025). ULF neuromodulation, therefore, differs fundamentally from other approaches that employ very short pulses designed to stimulate fibers within the spinal dorsal horn. To further characterize the effects of ULF neuromodulation, it is critical to investigate its impact at higher levels of the neuraxis.

There is considerable evidence of abnormal spontaneous activity in sensory fibers in patients with chronic pain. This activity appears to be important not only in the initiation of central sensitisation, but also in the maintenance of the sensitized state and in the generation of spontaneous pain—one of the most troublesome features associated with many types of chronic pain (Finnerup et al., 2021; Roza and Bernal, 2022). Accordingly, there is also evidence that chronic pain can be transiently relieved by suppression of activity in peripheral sensory fibers (Kuner and Flor, 2016).

Under normal conditions, somatic nociceptive signals, processed at the spinal level, are relayed to the brain predominantly via neurons projecting through lateral and anterolateral tracts (ventral and ventrolateral in animals). Many axons ascending in these tracts terminate in the thalamus, where further signal processing occurs and determines if the information is relayed on to the somatosensory cortex. Thus, the thalamus represents a key relay station for pain signals ascending from the spinal level to higher brain centers (Figure 1). Under conditions of chronic pain arising from disease or injury, the total amount of sensory input arriving at the thalamus is likely to be elevated, amplifying thalamic neurone activity (Guilbaud et al., 1990; Patel and Dickenson, 2016) and thereby increasing thalamic signaling to other brain areas, including the sensory cortex.

In this study, therefore, our aims were twofold. Given the recognized importance of thalamic processing in the ascending transmission of sensory signals, we first wished to determine whether epidural ULF neuromodulation could acutely impact thalamic neurone responses to signals generated by stimulation of peripheral sensory receptive fields under normal conditions. Second, we wanted to examine the effect of epidural ULF current on previously reported pathological changes in thalamic activity (Guilbaud et al., 1990; Patel and Dickenson, 2016) in a model of neuropathic pain.

Recordings were made from a total of 131 thalamic neurons in 33 rats. Recording sites of some of these neurons are shown in Supplementary Figure S1. Most cells recorded in the VPL fell into the LT classification, responding preferentially to low intensity electrical or mechanical stimulation. Fewer HT cells, responding only to high intensity stimulation, were encountered; and an even smaller cohort of WDR-type cells was found.

The experiments described here demonstrate that ULF current, applied epidurally over the spinal cord, reduces evoked responses of thalamic neurons to stimulation of peripheral receptive fields in the non-pathologic state and preferentially reduces ongoing, aberrant firing in the pathologic state. Whilst it remains uncertain upon which spinal neural elements ULF current is exerting its primary effect, these results cement our previous findings that ULF neuromodulation produces a dose-dependent inhibition of the nervous system and continues to support the utility of ULF current for the reduction of chronic pain signals.

Our previous study described for the first time the inhibitory effect of the novel ULF waveform on conduction of action potentials along primary sensory afferent fibers, such that both evoked action potentials in normal fibers and ectopic spontaneous impulses in axotomised fibers could be prevented from reaching the spinal cord (Jones et al., 2021). While others have suggested that long pulse widths may have differential effects on different ion channels (Yang et al., 2018) leading to a potential mechanism to affect normal vs. pathological fibers, we have not examined this aspect (normal vs. pathological ion channel expression and ULF’s effects thereon) in this work. Our previous computational studies demonstrated that the ULF waveform results in sodium channel inactivation and extracellular potassium accumulation in fibers close to the ULF electric field, effectively preventing the passage of action potentials. We showed that larger diameter, myelinated, faster-conducting fibers are more susceptible to inhibition by ULF current. To further investigate this finding, we wanted to understand the supraspinal effects that may follow from ULF inhibition at the spinal level.

As the thalamus is an important waystation on the route of sensory signals to the cortex, we investigated how ULF current applied at spinal level may alter supraspinal signaling. In our findings, we achieve proof of principle in showing that ULF current, applied epidurally at lumbar cord level in laboratory animals, impacts transmission of signals from a variety of peripheral sensory receptor types and fields to brain centers. We show this to be the case in both normal healthy animals and in a model of nerve injury, where sensory input to the thalamus is pathologically enhanced. Why is this important?

Many reports have described changes in thalamic neurone sensitivity and activity in models of chronic pain, including nerve injury (Guilbaud et al., 1990; Vos et al., 2000; Syré et al., 2014; Patel and Dickenson, 2016), diabetic neuropathy (Fischer et al., 2009), polyarthritis (Gautron and Guilbaud, 1982), inflammation (Guilbaud et al., 1986) and spinal cord injury (Miki et al., 2000; Hains et al., 2005, 2006). In such cases, including the SNL model used in this study, where damaged sensory afferent fibers have developed spontaneous discharge, at least some of this peripherally generated afferent signal is likely to reach the brain via ascending spinal pathways. Activity originating in pathological nociceptive fibers may be relayed to the thalamus via second-order nociceptive-specific or WDR-type neurons projecting in the spinothalamic tracts. Alternatively, the dorsal column-lemniscal pathway may provide a route for ectopic action potentials arising in normally non-nociceptive sensory fibers that become active in disease or following injury (Miki et al., 1998b; Miki et al., 2000).

Let us first consider the prevalence of different response types observed in our experiments and the effect of ULF current on evoked thalamic activity under normal, non-pathological conditions. The relative proportions of LT neurons to HT neurons reported in different studies varies considerably. Some studies have reported somewhat higher incidence of HT neurons than we encountered (Guilbaud et al., 1980; Casey and Morrow, 1983; Chung et al., 1986). There are multiple factors that may account for this difference, including presence of anesthetic (some groups have recorded only in awake animals), type of anesthetic, species investigated, and the search method used to locate neurons. Our view is that the search method is likely to be of primary importance with respect to our findings. We took care to apply high-intensity stimuli very sparingly to minimize the likelihood of sensitizing the skin or deeper tissues by repeatedly pinching the toes, for example. Thus, whilst most cells with LT responses were also tested to see if they responded to more intense stimuli (i.e., WDR-type), we seldom employed high-intensity stimuli as a primary search stimulus. This may have biased our sample to include more LT neurons. With this approach, in common with many others (e.g., Angel and Clarke, 1975; Kenshalo et al., 1980; Apkarian and Shi, 1994; Montagne-Clavel and Olivéras, 1995), we encountered many more thalamic cells that were LT responsive than HT responsive. We found that LT responses of thalamic neurons in naïve animals appeared more susceptible to ULF inhibition than HT responses—91% of LT responses compared to only 44% of HT responses were significantly attenuated by ULF current. As highlighted above, interpretation of this result mandates some caution due to the small number of HT cells compared to the LT group.

Our earlier findings with ULF current applied to dorsal roots indicated a preference for inhibition of evoked signals carried by LT fibers (Jones et al., 2021), a result that is supported by these observations on thalamic responses with epidural ULF current. We must then consider the significance of this in terms of the types of fibers likely to contribute to the generation and maintenance of chronic pain. To further investigate this, we examined the effect of epidural ULF neuromodulation on thalamic activity in the SNL neuropathic pain model.

Sensory afferent fibers that convey signals from LT receptors are predominantly (though not exclusively—see Liljencrantz and Olausson, 2014) large, myelinated, fast-conducting Aβ-type fibers. Conversely, signals generated by high-intensity stimuli are usually carried by smaller diameter, weakly myelinated Aδ and non-myelinated C type fibers. Our previous work showed that these fibers required higher current amplitudes to inhibit, and our present findings further support this. Traditional dogma holds that many aspects of pathological chronic pain should be mediated by signals originating in the fibers that are responsible for conveying pain-related signals under normal, healthy conditions (i.e., those Aδ- or C-type afferents with typically higher thresholds both for activation and for inhibition by ULF current). The animal findings present a possibly counter-intuitive picture, as human pain patients receiving ULF treatment demonstrated no loss of normal sensation but did experience a significant improvement in chronic pain (Jones et al., 2021).

To better understand this result, it is instructive to consider alternative roles for Aβ fibers in signaling pain. While Aβ-type afferents are less readily associated with nociception, there is plentiful evidence of a population of receptors with fast-conducting Aβ-type fibers that encode HT, nociceptive stimuli (Burgess and Perl, 1967; Lynn and Carpenter, 1982; Djouhri and Lawson, 2004; Fang et al., 2005; Devor, 2009). It seems likely that a proportion of such Aβ nociceptors could contribute to HT responses of some thalamic cells. If this is indeed the case, it may explain why some HT thalamic responses (4 of 9 cells tested in this study) could be diminished by ULF current, possibly affecting conduction in large, myelinated fibers. Additionally, Aβ nociceptors undergo specific changes in action potential dynamics in some pain models (e.g., Wu and Henry, 2009). Such changes appear to manifest only at later time points (up to 2 months after induction of the model) and therefore seem likely to reflect changes that are important in the chronification of disease, as may be seen in human patients.

Another important role for Aβ fibers in the development of chronic pain is central sensitisation, whereby facilitation of responses to primary afferent inputs occurs within the central nervous system (Woolf, 1983). Such facilitation can be produced by an intense input from nociceptors, acutely activated by tissue injury (i.e., in conditions of nociceptive pain; see Woolf, 2011). However, there seems little doubt that central sensitisation may also arise due to a prolonged afferent barrage into the CNS provided by chronic ectopic spontaneous activity of sensory fibers in a neuropathic state (see further discussion below). The relative contributions of abnormal spontaneous activity in A- and C-type fibers to the different aspects of chronic pain remains the subject of debate (see North et al., 2018; Raja et al., 2020). Ectopic firing in A fibers seems likely to underly the phenomenon of touch-evoked allodynia, as selective peripheral block of A-fibers in human patients (Campbell et al., 1988) or interruption of ipsilateral Aβ fiber-specific dorsal column pathways in animals (Sun et al., 2001) alleviates allodynia. Pathological conditions may cause some A fibers to remain ectopically active on a chronic timescale, whereby they then contribute significantly to maintenance of a sensitized state. Additionally, there is a possibility of phenotypic switch of LT Aβ fibers or the neural circuits that they feed into, such that they begin to serve a nociceptive function. Injured A fibers have been shown to begin synthesizing peptides normally associated only with nociceptive sensory afferent fibers, such as substance P (Noguchi et al., 1994; Noguchi et al., 1995; Neumann et al., 1996; Weissner et al., 2006) and calcitonin gene-related peptide (Miki et al., 1998a; Ma et al., 1999). The significance of this lies in a novel ability acquired by Aβ fibers to drive a different type of activity in neurons of the spinal dorsal horn and dorsal column nuclei by releasing pain-related transmitters. This change may explain the nociceptive responses to normally innocuous stimuli that characterize many forms of chronic pain (Pitcher and Henry, 2004), though the phenotypic switch theory has been questioned (Hughes et al., 2007). Another mechanism by which Aβ fiber input can functionally switch from normal mechanosensation to nociceptive-like drive is through disinhibition, which disproportionately affects excitatory neurons (Lee et al., 2019).

Direct investigation of a specific Aβ mechanism was outside of the scope of this study, but we utilized the SNL model to examine the pathological state of thalamic activity levels and the effect of ULF current thereon. Notably, the spontaneous afferent fiber activity seen at early time points after nerve injury occurs almost exclusively in A-fibers (Liu X. et al., 2000; Liu C. N. et al., 2000). SNL is a well-studied model of pain, and the results we present here lend support for an effect of ULF current upon large-diameter fibers. Following experimental axotomy-type nerve lesioning, spontaneous activity in sensory fibers commences soon after the injury. The incidence of firing, and the fiber types which become active, appear to depend to some extent on the model. However, most studies agree that firing in A fibers predominates, with very few spontaneous C fibers, and very low rates of discharge in those C fibers that do show activity (Govrin-Lippmann and Devor, 1978; Wall and Devor, 1983; Kajander and Bennett, 1992; Michaelis et al., 1995; Tal et al., 1999).

Spontaneous activity begins as early as 3 h following axotomy (Kirk, 1974), increases rapidly, reaching a peak within 3 days (Liu X. et al., 2000; Ma et al., 2003) and declines thereafter, but is maintained at a more modest level for weeks. This intense abnormal barrage, continued over several days, is likely to activate at least some elements of the central sensitisation process at multiple sites in the pathway of the signals, including the spinal dorsal horn (Woolf, 1983), gracile nucleus (Miki et al., 1998b; Miki et al., 2000; Wang et al., 2004; Fukuoka et al., 2015) and thalamus itself (Wang et al., 2020). In the sensitized state, interruption of the driving input (i.e., the afferent impulse barrage) may not result in immediate complete cessation or normalization of neuronal firing that has been pathologically elevated for many hours or days, but may nevertheless attenuate it. Our results align with this idea. This provides a possible explanation for the lack of complete inhibition of thalamic cell firing by short-term ULF current in SNL animals.

An alternative interpretation of the observation that ULF current did not silence thalamic neurone firing completely is that the magnitude of the ULF current was insufficient to inhibit conduction of all pathological spontaneous impulses arriving at the spinal cord. To reiterate, the number of spontaneously active fibers observed in the first week after L5 spinal root axotomy is typically very high compared to sham-operated or naïve animals. Liu X. et al. (2000) reported spontaneous fibers in 56% of the filaments they examined; Ma et al. (2003) found an incidence of 89%; our own unpublished findings indicate a similar high incidence of ectopic activity in axotomized fibers. It seems unlikely that ULF current at intensities we used would inhibit conduction in every active fiber throughout the entire L5 spinal root or surrounding neural tissue. The current was applied epidurally in these experiments and there is some separation in the form of dura and cerebrospinal fluid between the electrodes and the underlying neural tissue. To achieve an electric field sufficient to inhibit conduction across the entire root may necessitate increasing the current to a level where even a slowly ramped rise could cause undesired activation of some fibers.

The first week after an injury is a relatively short period of time, and it is established that activity in injured A fibers falls to a lower level as the model matures. However, some degree of spontaneous activity has been shown to persist for many weeks after the initial injury, at which point the model may be regarded as more representative of a chronic neuropathic condition. In our experiments, we showed greatly increased levels of ongoing activity in thalamic cells in the first week after SNL injury, much of which was susceptible to reduction by ULF current applied at lumbar spinal level. Whilst it was not possible to determine fully the physiological role of the thalamic cells we recorded from here (because of deafferentation in the SNL animals), nevertheless ULF current was able to restore thalamic activity to a level similar to that observed in normal uninjured animals (see Tables 2, 3). This finding suggests that such ectopic activity, arising in damaged afferent fibers, can drive ongoing activity in many normally quiescent thalamic neurons, and that ULF current applied at the spinal level is capable of reducing the level of that activity. We cannot be certain about the exact site of action of ULF neuromodulation. We targeted application toward the side of the spinal cord ipsilateral to the nerve injury, at the level where the injured root joins. The reduction of neuronal firing in VPL nucleus may well represent a composite result of inhibitory effects on an afferent barrage in both primary sensory afferents approaching the root entry zone and on fibers ascending in the dorsal columns.

The work here is consistent with our previous experiments, whereby ULF inhibits a wide range of neural signals (Jones et al., 2021). In this work, we continue to expand on the abilities of ULF to inhibit neural signals associated with pain. The development, mechanisms, and maintenance of chronic pain, involving many peripheral and central mechanisms, are still widely debated. Regardless of those particular mechanisms and pathways, in this work, we showed that ULF applied to the epidural space of the spinal cord can have a marked effect on supraspinal signals both in non-pathologic (evoked footshock) and pathologic (SNL) models. Thus, epidural ULF application in a pathological pain model demonstrated a normalizing effect on thalamic activity to resemble more closely the latent activity pattern of the thalamus.