Introduction

Painful sensory stimuli are detected and encoded by peripheral sensory neurons termed nociceptors, which can broadly be classified into unmyelinated C-fibre nociceptors and myelinated A-fibre nociceptors (Dubin and Patapoutian, 2010; Lewin and Moshourab, 2004). Both populations comprise a variety of subpopulations that are finely tuned to detect different types of noxious physical and chemical stimuli (Fig. 1). The vast majority of all nociceptors, except for a small proportion termed silent nociceptors, are activated by mechanical stimuli. Nociceptors that exclusively respond to mechanical stimuli are termed C-fibre and A-fibre mechanonociceptors (C-Ms and AMs), respectively. Those that additionally respond to noxious thermal stimuli are often collectively termed A-fibre or C-fibre polymodal nociceptors, though they can be further sub-classified according to their sensitivity to noxious heat and/or cold. In addition to nociceptors, noxious mechanical stimuli inevitably also activate low-threshold mechanoreceptors (LTMRs), which primarily detect various types of tactile stimuli (Fig. 1) (Abraira and Ginty, 2013). Hence, painful stimuli – in particular mechanical stimuli – activate a wide range of functionally diverse sensory afferent subtypes, thereby generating a plethora of sensory information that is simultaneously transmitted to the spinal cord. Accordingly, there is an intense debate about whether different types of pain result from the processing and integration of multiple sensory inputs (i.e. pattern theory) or from neural activity in specific subtypes of sensory afferents that are particularly sensitive to certain types of stimuli (labelled lines theory) (Ma, 2010; Prescott et al., 2014). In the following sections, I will provide an overview of the different types of sensory neurons and their proposed contribution to different forms of pain. Moreover, I will discuss how pain related sensory information is processed in the dorsal horn of the spinal cord with an emphasis on recently delineated neural circuits that mediate pain hypersensitivity in the setting of nerve injury and inflammation.

Peripheral sensory neurons involved in heat-, cold- and mechanical pain

The cellular and molecular basis of cold and heat pain is quite well understood. Mice lacking the menthol-sensitive ion channel TRPM8 are insensitive to innocuous cooling, only partially avoid painfully cold temperatures and show significantly attenuated cold pain associated with inflammation and nerve injury (Bautista et al., 2007; Dhaka et al., 2007). Interestingly selective diphtheria toxin (DTX) mediated ablation of the sensory neurons that express TRPM8 results in an even more pronounced phenotype – i.e. an almost complete loss of cold pain (Knowlton et al., 2013). Similar results were obtained for heat sensitivity. Thus, mice lacking the heat-sensitive ion channel TRPV1 only develop little heat hyperalgesia during inflammation and only show reduced sensitivity to noxious heat at very high temperatures (Caterina et al., 2000), whereas ablation or silencing of TRPV1-positive neurons using various different techniques resulted in an almost complete loss of heat pain (Brenneis et al., 2013; Cavanaugh et al., 2009; Mishra and Hoon, 2010). Likewise, ablation of nociceptors that express the neuropeptide CGRP and the Nerve Growth Factor (NGF) receptor TRKA, which includes most TRPV1-expressing neurons, also produced a profound loss of heat pain sensitivity (McCoy et al., 2013). Hence, it is now widely accepted that TRPM8-positive neurons and TRPV1-positive neurons are the cellular sensors for cold and heat pain, respectively. However, the above-mentioned studies also clearly demonstrated that TRPM8 and TRPV1 are not the only sensors for noxious cold and heat but that additional yet unknown ion channels must also contribute to thermal nociception.

Unravelling the cellular and molecular basis of mechanical pain has proven much more difficult. Thus, the ion channel that transduces noxious mechanical stimuli is still elusive (Ranade et al., 2014) and only little is known about the contribution of different sensory afferent subtypes to various types of mechanical pain. Some of the above-mentioned studies found no difference in mechanically-evoked pain in mice lacking TRPV1-or TRPM8-expressing fibres (Cavanaugh et al., 2009; Knowlton et al., 2013; McCoy et al., 2013; Mishra and Hoon, 2010). Cavanaugh et al. however showed that ablation of another subset of polymodal C-fibre nociceptors, which are characterized by the expression of the Mas-related G-protein coupled receptor D (MRGPRD) (Rau et al., 2009), only increases mechanical pain thresholds but not heat pain thresholds. Thus, together these studies quite unexpectedly suggested that only MRGPRD-positive neurons but not TRPV1- or TRPM8-positive neurons are required for mechanical pain, despite all three being sensitive to mechanical stimuli. It should be noted however that in all these studies pain sensitivity was assessed several days after the respective cell populations had been killed, which raises the possibility that the outcomes of these studies were affected by compensatory changes in the pain pathways. Indeed, using a more rapid pharmacological approach to selectively silence TRPV1-positivefibres, Brenneis and colleagues (2013) found that TRPV1-positive neurons in addition to heat pain also contribute to mechanical pain. One possible explanation for this discrepancy is that Brenneis et al. used the pinch test and pinprick stimuli to assess mechanical pain, whereas Cavanaugh et al. (2009), Mishra et al. (2010) and Knowlton et al. (2013) used punctate mechanical stimuli applied with von Frey filaments and McCoy et al. (2013) used a tail clip assays. Interestingly, Brenneis et al. only found differences in pinch-evoked pain but not in pinprick-evoked pain, indicating that different submodalities of mechanical pain may be mediated by different subsets of nociceptors. Hence another possible explanation for the discrepancy of the above-mentioned studies is that MRGPRD-positive nociceptors are more efficiently activated by von Frey filaments while TRPV1-expressing neurons are particularly sensitive to pinch stimuli. The hypothesis that different types of mechanical stimuli are detected by different subsets of nociceptors is further supported be recent work from my lab. We showed that pinprick stimuli are detected by a subset of nociceptors that are not required for detecting other types of noxious mechanical stimuli. Thus, we found that ablation of A-fibre nociceptors that express the neuropeptide Y receptor type 2 (NPY2R) results in significantly prolonged paw withdrawal latencies in response to pinprick stimuli but does not affect von Frey-evoked behavioural responses and heat pain (Arcourt et al., 2017). However, using optogenetics we further showed that selective activation of NPY2R-positive neurons evokes abnormal pain behaviour and that simultaneous activation of NPY2R-positive nociceptors and LTMRs is required to mimic pinprick-evoked pain behaviour. Thus, our studydirectly demonstrated that the integration of multiple sensory inputs from functionally distinct afferent subtypes is required to generate a certain submodality of mechanical pain. Accordingly, it is possible that the loss of a single afferent subpopulation may solely change the way a certain stimulus is perceived but may not necessarily change the behavioural response evoked by that stimulus, and hence one should be cautious when interpreting results from cell ablation studies.

Fig. 1

Primary somatosensory neurons are classified into nociceptors (red and orange), which terminate as free nerve endings in the skin and detect various types of noxious stimuli (indicated at the bottom), and low-threshold mechanoreceptors (LTMRs; green and blue), which innervate specialised structures in the skin, sometimes referred to as end-organs, and detect different tactile stimuli (indicated at the bottom). Abbreviations: non-peptidergic C-fibre mechanoheat nociceptor (non-pep. C-MH), peptidergic C-fibre mechanocold nociceptor (pep. C-MC), peptidergic C-fibre mechanoheat nociceptor (pep. C-MH), A-fibre mechanonociceptors (AM), C-fibre low threshold mechanoreceptor (c-LTMR), Aδ-fibre LTMR (D-hair), Aβ-fibre slowly-adapting type I and type II LTMR (SA-I LTMR and SA-II LTMR), Aβ-fibre rapidly-adapting type I and type II LTMR (RA-I LTMR and RA-II LTMR).

Pain processing neural circuits in the dorsal horn of the spinal cord

The spinal cord is more than just a relay station for sensory information. Rather, it is a major site of sensory information processing, where afferent inputs are modulated by local and supraspinally-derived excitatory and inhibitory signals (Braz et al., 2014; Todd, 2010). The importance of the dorsal horn in signal processing is reflected by that fact that the great majority of dorsal horn neurons are local interneurons and only a very small proportion of cells – so-called projection neurons – relay sensory input to supraspinal targets. Projection neurons, 80% of which express the neurokinin 1 receptor (NK1R), are most abundant in lamina I, where they account for about 5% of all neurons. They are absent form lamina II and are scattered throughout lamina III-VI. Retrograde and anterograde tracing studies revealed that NK1R⁺ neurons in lamina I project to multiple supraspinal targets including the periaqueductal grey, the lateral parabrachial area, the caudal ventrolateral medulla and the nucleus of the solitary tract, suggesting that NK1R⁺ neurons are involved in multiple aspects of pain perception (Todd, 2010). Indeed, selective ablation of NK1R⁺ projection neurons reduces thermal hyperalgesia and mechanical allodynia induced by capsaicin, nerve injury or inflammation (Mantyh et al., 1997). NK1R⁺ projection neurons receive monosynaptic input from putative peptidergic C-fibre nociceptors, which also project to vertical and transient central cells in lamina II. Aδ-fibre nociceptors, by contrast, mainly provide input to vertical cells in lamina II (Lu and Perl, 2003, 2005) and non-peptidergic MRGPRD-positive C-fibre nociceptors project to various types of morphologically distinct interneurons in lamina II including vertical cells and central cells but not islet cells (Wang and Zylka, 2009; Yasaka et al., 2014). Patch-clamp recordings from pairs of neurons in spinal cord slices further showed that in addition to projecting to excitatory interneurons, different types of C-fibre nociceptors also activate different subsets of inhibitory interneurons in lamina II (Zheng et al., 2010). Thus, cooling-sensitive C-fibres specifically activate GABAergic central cells in lamina II, which in turn inhibit GABAergic islet cells and most importantly vertical cells in lamina II. Considering that vertical cells receive input from all types of nociceptors, this circuit provides a possible explanation for why cooling has an analgesic effect. While classical neurophysiological studies mainly characterized interneurons by means of morphology and electrophysiological properties, more recent studies have started to unravel the molecular identity of these cell types. Duan and colleagues(2014), for example, have characterized the function of a subset of excitatory interneurons that express Somatostatin (SOM). They show that SOM⁺ neurons include PKCγ-expressing neurons located at the border between lamina II and III as well as vertical cells located in outer lamina II. Consistent with the proposed role of MRGPRD- and Aδ-fibre nociceptors, which are connected to projection neurons via vertical cells in lamina II, in mechanical pain (see above), selective ablation of SOM⁺ neurons completely abolished pain behaviour evoked by various types of noxious mechanical stimuli such as von Frey filaments, pinprick stimuli and pinch stimuli. Heat pain, which is mediated by peptidergic C-fibre nociceptors that are directly connected to NK1R neurons, was not altered following ablation of SOM neurons.

In the setting of injury or inflammation pain sensitivity usually increases, such that normally innocuous stimuli evoke pain (allodynia) and painful stimuli evoke exacerbated pain (hyperalgesia). Structural and functional changes in the spinal circuits of the dorsal horn have been proposed to underlie pain hypersensitivity, in particular mechanical allodynia. The Gate Control Theory (GCT) of Pain (Melzack and Wall, 1965) proposed that Aβ-LTMRs, which signal touch and project to lamina III-IV, are also connected to pain processing neurons in the superficial dorsal horn, but that this connection is normally ”gated” by inhibitory interneurons. The GCT further proposed that in the setting of nerve injury or inflammation the gates are opened, hence allowing Aβ-LTMR input to excite pain-processing neurons, thereby turning gentle touch into pain. Indeed several studies have shown that compression block of Aβ-fibres reduces mechanical allodynia in patients suffering from chronic pain syndromes and that activation of Aβ-fibres triggers action potential firing in neurons in lamina I and II in animal models of neuropathic pain (Sandkühler, 2009).

The polysynaptic excitatory connection that links Aβ-LTMRs with the superficial dorsal horn was originally thought to begin with PKCγ-expressing neurons located at the border between lamina II and III, which receive monosynaptic input from Aβ-fibres (see circuit A in Fig. 2) (Miraucourt et al., 2007). PKCγ-positive neurons were shown to be linked to lamina I projection neurons via transient central cells in inner lamina II and vertical cells in outer lamina II (Lu et al., 2013). The vertical cells in outer lamina II were subsequently shown to receive direct monosynaptic input from almost all types of sensory afferents including Aβ-LTMRs, suggest a prominent role of these neurons in signal integration (Yasaka et al., 2014). Consistent with these findings, Duan and colleagues (2014) showed that SOM⁺-neurons at the border between lamina II and III, most of which co-express PKCγ, and SOM⁺ neurons in outer lamina II, which include many vertical cells, both receive direct input from Aβ-fibres (see circuit A and D in Fig. 2). More recently additional excitatory connections were described. Thus Aβ-fibres were shown to be connected to lamina III neurons that transiently express VGLUT3 and that relay sensory input to PKCγ neurons via a yet not further characterized population of glutamatergic interneurons (circuit B in Fig. 2). Interestingly, these glutamatergic interneurons are also connected to calretinin-expressing cells in inner lamina II, which further relay Aβ-fibre-derived input to the same transient central cells as PKCγ-positive neurons (circuit C in Fig. 2) (Peirs et al., 2015). Hence, a total of four different excitatory connections that link Aβ-fibres to lamina I pain-processing neurons have been described. These pathways are usually, as predicted by the GCT, silenced by inhibitory interneurons. Dynorphin-expressing (DYN⁺) interneurons in inner lamina II, for example, inhibit PKCγ neurons (circuits A and B) and DYN⁺ neurons in outer lamina II inhibit SOM⁺ vertical cells (circuit D). Accordingly, ablation of DYN⁺ interneurons results in spontaneous development of mechanical allodynia (Duan et al., 2014). In addition PKCγ neurons (circuits A and B) were found to be inhibited by parvalbumin-expressing (PV) neurons (Petitjean et al., 2015) and by DYN-negative glycinergic neurons (Duan et al., 2014; Lu et al., 2013). Finally, calretinin-expressing neurons (circuit C) also appear to be under inhibitory control, but the cellular identity of this inhibitory gate is still unknown (Peirs et al., 2015).

Fig. 2

Schematic illustration of the neural circuits in the dorsal horn of the spinal cord that are involved in pain-processing. The borders between lamina I, II and III are indicated by dashed lines. Cells that are mainly involved in acute pain signalling are depicted in red and cells that link LTMR-derived sensory input to pain-processing neurons in lamina I and II are shown in blue. (A), (B), (C) and (D) mark the four major pathways that connect LTMRs with the superficial dorsal horn. Abbreviations: neurokinin 1 receptor (NK1R), somatostatin (SOM), dynorphin (DYN), calretinin (CR), protein kinase C-γ (PKCγ), glycine (GLY), parvalbumin (PV), vesicular glutamate transporter (V3).

Taken together, a lot of progress towards delineating the spinal circuits that mediate pain has been made in recent years. However, several important open questions remain.

The presence of multiple circuits that link Aβ-LTMR input to lamina I projection neurons, for example, raises the question as to whether each of these circuits is activated by a different subtype of Aβ-LTMRs or whether different circuits are recruited in different pathological pain states. Support for the latter hypothesis was provided by Peirs et al. (2015), who showed that the circuit containing CR-neurons becomes active in the Carrageenan model of inflammatory pain, whereas the PKCγ-neuron containing circuit is activated in the spared nerve injury model of neuropathic pain. Moreover, most of the above-mentioned interneuron subpopulations are neurochemically and morphologically heterogeneous and partially overlap with one another (Duan et al., 2014; Gutierrez-Mecinas et al., 2016; Peirs et al., 2015), suggesting that pain-processing circuits in the spinal dorsal horn may be even more complex than depicted in Fig. 2. The exact identity of the input and output neurons is also still elusive. Thus, it is unclear which particular subtype of Aβ-LTMRsmediates touch-evoked painand it is not known whether different types of projection neurons relay pain-signals to higher order brain regions in different pathological states.Last but not least the signalling mechanisms that lead to the unsilencing of the normally inactive dorsal horn circuits in the first place and hence to the chronification of pain are also still unknown. Addressing these open questions and translating these recent discoveries into effective therapies for humans will be one of the central challenges in pain research in the future.