Influencers in the Somatosensory System: Extrinsic Control of Sensory Neuron Phenotypes

Abstract

Somatosensory neurons in dorsal root ganglia (DRG) comprise several main subclasses: high threshold nociceptors/thermoceptors, high- and low-threshold mechanoreceptors, and proprioceptors. Recent years have seen an explosion in the identification of molecules that underlie the functional diversity of these sensory modalities. They also have begun to reveal the developmental mechanisms that channel the emergence of this subtype diversity, solidifying the importance of peripheral instructive signals. Somatic sensory neurons collectively serve numerous essential physiological and protective roles, and as such, an increased understanding of the processes that underlie the specialization of these sensory subtypes is not only biologically interesting but also clinically relevant.

Introduction

The sensory nervous system serves to sample the internal and external environment to alert us of dangerous or opportunistic situations, and—if need be—to elicit appropriate action. Smell, vision, and audition for instance can warn of dangerous circumstances and activate escape mechanisms; taste, smell, or touch aid in distinguishing between nutritious or toxic foods, and touch and nociception generate rapid withdrawal reflexes on skin contact with harmful stimuli. Many of these sensory modalities also serve important roles in regulating normal physiological functions. For instance, a diverse array of somatic and autonomic sensory neurons maintain gastrointestinal motility (Camilleri 2021; Williams and others 2016); a large number of mechanoreceptive and chemosensory airway and baroreceptors monitor and regulate respiration and blood pressure (Min and others 2019; Prescott and others 2020); and distinct sets of mechanoreceptors that innervate skin and muscle provide sophisticated cutaneous and proprioceptive feedback to refine object interaction and dexterous movement (Handler and Ginty 2021; Johnson 2001; Zampieri and de Nooij 2021). The discriminative abilities of each of these sensory modalities to distinguish between different sounds, smells, tastes, temperatures, or types of discriminative touch, rest in large part on the vast number of specialized sensory neuron subtypes. Given the importance of these senses, a better understanding of the developmental mechanisms that underlie this diversity in mammalian sensory neuron subtypes remains a topic of high relevance.

Similar as in other parts of the nervous system, the acquisition of the three main somatosensory neuron classes (high-threshold nociceptors/thermoceptors), low-threshold mechanoreceptors [LTMRs], and proprioceptors) appears to depend on cell-autonomous gene regulatory programs (Florio and others 2012; Marmigère and Ernfors 2007). However, the morphological and physiological diversification of each of these main sensory classes into a vast repertoire of distinct sensory phenotypes does not just involve the stepwise execution of an intrinsic transcriptional program but is guided by extrinsic cues throughout development (Albers and Davis 2007; Conover and Yancopoulos 1997; Ernsberger 2009; Lewin 1996; Sieber-Blum and Zhang 1997). Indeed, some recent studies suggest that environmental signals may impose specific sensory fates (Faure and others 2020; Moqrich and others 2004; Oliver and others 2021; Sharma and others 2020; Wu and others 2019; Wu and others 2021). Drawing mainly from the somatosensory system in the mouse that serves the trunk and limbs, this review highlights some of the work that supports this idea. Specifically, after a general description of sensory neurogenesis, I will attempt to catalog the role of environmental signaling factors in the acquisition and subtype specification of the various sensory neuron classes. I will then describe some of the recent studies that have begun to offer a mechanistic basis for the refractory nature of cell type diversification within each of these sensory neuron cohorts. The review will end with a discussion on the potential relevance of these findings in a clinical context and suggestions for areas of future investigation. Studies of somatosensory neuron subtypes originate from the late 19th century and have since generated an enormous body of literature—too expansive to permit an inclusion of all the foundational studies in this review. Apologies to all those whose work I was unable to include.

Early Specification in the Mammalian Somatosensory System

Most somatosensory neurons derive from the neural crest, a transient cell population that delaminates from the dorsal aspect of the neural tube (Fig. 1Ai) (Hovland and others 2020). The majority of migrating neural crest cells coalesce in dorsal root ganglia (DRG) or in sympathetic chain ganglia (SG) (serving trunk and limb spinal levels) but they also contribute to cranial ganglia, most notably the trigeminal ganglion (TG), which serves the face (Fig. 1Aii). DRG are located at discrete bilateral segmental positions along the entire rostral-caudal extent of the spinal cord, while SG are located more ventrally and restricted to axial segmental levels. Each of these neural crest–derived populations is subject to the influence of several signaling factors, including members of the Wnt, bone morphogenetic protein (BMP), retinoic acid (RA), and fibroblast growth factor (FGF) families (Hovland and others 2020). Variant exposure to these signals culminates in the expression of distinct sets of proneural transcription factors (TFs), the Neurogenins (Ngns) 1 and 2, or Mash1 (Figure 1Aii) (Marmigère and Ernfors 2007). Expression of the Ngns or Mash1 biases multipotent neural crest cells toward a sensory or an autonomic sensory fate, respectively (Ma and others 1999; Zirlinger and others 2002).

Figure 1.

Development of sensory neuron subtypes in the dorsal root ganglia (DRG). (A) Early environmental signaling factors (e.g., BMP, Wnt, RA, FGF) orchestrate the induction of the neural crest marked by Ap2α and Sox10 (i), the choice between a DRG sensory (Ngn⁺) or sympathetic chain ganglion (SG; Mash1⁺) proneural identity (ii), and the expression of nascent sensory neuron subtype markers, the neurotrophin receptors (TrkC, TrkB, Ret, and TrkA), labeling proprioceptive (blue), low-threshold mechanoreceptor (LTMR; green), and small diameter nociceptors, mechanoreceptors, thermoreceptors, and puriceptors (sdTrkA⁺ lineage; red), respectively (iii). Signaling and trophic factors are labeled in purple; transcription factors and receptors molecules in black. Neurotrophic signals (NT3, BDNF, NRTN, NGF) are present in the surrounding mesenchyme and in peripheral and central afferent target areas and required for sensory neuron survival and differentiation. (B) Main classes of DRG sensory neurons and their distinct peripheral (skin, muscle) and central projection patterns. Spinal targets of sdTrkA⁺ lineages (in red), LTMRs (in green), and proprioceptors (in blue) are primarily situated in dorsal horn, intermediate spinal lamina, and ventral horn, respectively. Sensory subtypes shown serve as representatives for all the subtypes of an individual class.

In the mouse, sensory neurogenesis takes place roughly between embryonic day (e) 9 and 13. Within sensory progenitors, Ngn2 expression coincides with the generation of early born (e9-10.5) sensory neurons that give rise to proprioceptors and LTMRs, including Aδ LTMR nociceptors (Bachy and others 2011; Ma and others 1999; Ventéo and others 2019). Expression of Ngn1 is associated with later born (e9.5-e13) sensory neurons that primarily yield small diameter nociceptive, mechanoreceptor, thermoceptive, and pruriceptive sensory neurons (Liu and Ma 2011; Ma and others 1999). Despite their divergent expression patterns, Ngn2 and Ngn1 were shown to be functionally interchangeable (Blanchard and others 2015; Ma and others 1999). Moreover, fate mapping of either Ngn2 or Ngn1 progenitors labels all sensory neuron subtypes, indicating that the mere expression of these proneural genes does not define early born (proprioceptor, LTMR) or late born (e.g., nociceptor, thermoceptor) sensory classes (Sharma and others 2020; Zirlinger and others 2002). Indeed, analyses of Ngn mutant mice suggests that Ngn activity may primarily regulate the timing of sensory progenitor differentiation competence (Ventéo and others 2019). The transcriptional activities of the Ngns, and other downstream TFs, culminates in the activation of expression of the TFs Pou4f1 (encoding Brn3a) and Islet1 (Isl1) (Fig. 1Aii). Pou4f1 and Isl1 act together to force an exit from the neurogenic program, repress spinal neuronal fates, and activate a general sensory phenotypic differentiation program (Dykes and others 2011; Sun and others 2008). As part of this program, post-mitotic neurons begin to express specific sets of neurotrophin receptors (e.g., the tropomyosin kinases [Trk] A, B, C, and the receptor tyrosine kinase Ret) and start extending axons to peripheral target domains (Figure 1Aiii) (Dykes and others 2011; Sun and others 2008). The patterns of neurotrophin receptors broadly predict distinct sensory fates, with proprioceptors classically defined by TrkC expression (the receptor for neurotrophin [NT]3 and encoded by Ntrk3), LTMRs by TrkB (the receptor for brain-derived neurotrophic factor [BDNF] or NT4, encoded by Ntrk2), or Ret (the receptor for glial cell–derived neurotrophic factor [GDNF] family ligands), and small diameter nociceptor, mechanoreceptor, thermoceptor, and pruriceptor neurons expressing TrkA (the receptor for nerve growth factor [NGF], and encoded by Ntrk1), and/or Ret (Conover and Yancopoulos 1997; Ernsberger 2009; Lewin 1996). While these traditional neurotrophin receptor–based definitions of sensory subtypes remain widely in use, it is important to note that these classifications are oversimplifications and that neurotrophin receptor expression can be dynamic during sensory neuron development.

Sensorigenesis is relatively well understood, yet how generic postmitotic sensory neurons differentiate into many specialized sensory identities remains less clear. The two postmitotic TFs, Pou4f1 (encoding Brn3a) and Isl1, appear to exert influence over the initial selection and consolidation of distinct sensory lineages. In Isl1 mutant mice, for example, molecules associated with a nociceptive/thermoceptive sensory identity (e.g., TrkA, Runx1) rapidly decline following their initial induction, while those marking a proprioceptive identity (e.g., TrkC, Runx3) are mostly unaffected (Sun and others 2008). Conversely, in Pou4f1 mutant mice, the expression of proprioceptive molecules is markedly reduced, while the initial expression of molecules indicative of LTMR or nociceptor fates is largely unaffected (Dykes and others 2010; Lei and others 2006). Thus, despite their pan-sensory expression, Pou4f1 and Isl1 transcriptional programs appear to bias post-mitotic sensory neurons (direct or indirectly) toward different sensory lineages (Fig. 1B). However, the mechanisms that predispose generic post-mitotic Ngn⁺Pou4f1⁺Isl1⁺ neurons to distinct sensory fates remain largely unknown.

Extrinsic Signals Involved in the Generation of TrkA Lineage Sensory Subtypes

The largest and functionally most diverse population of somatosensory neurons are unmyelinated or thinly myelinated small to medium diameter TrkA-lineage sensory neurons which provide feedback regarding temperature, tissue acidity, mechanical pressure, or itch, from skin and internal organs (e.g., bone, heart, muscle) (Fig. 1B) (Li and others 2016; Sharma and others 2020; Usoskin and others 2015). This sensory class derives from both waves of sensory neurogenesis and constitutes roughly 75% to 80% of all DRG neurons. Ngn2-derived medium diameter (md)TrkA⁺ neurons include Aδ nociceptors, while Ngn1 or Krox20⁺ boundary cap-derived small diameter (sd)TrkA neurons include Mrgprd⁺ wide-dynamic range mechanoreceptors, C-LTMRs, thermoceptors, and pruriceptors (Abdel Samad and others 2010; Bachy and others 2011; Luo and others 2007; Ma and others 1999; Maro and others 2004). For lack of a better name these later born neurons are often referred to as “nociceptors” yet they comprise many neurons that do not elicit a nocifensive response when activated. Therefore, we here will refer to these neurons as sdTrkA-lineage neurons. Evident from TrkA and NGF mutant mice, TrkA-lineage neurons depend on NGF for their survival starting from e12.5 (Conover and Yancopoulos 1997; Patel and others 2000). Mice that lack NT3 also show a ~50% reduction in TrkA sensory neurons in trigeminal ganglion (TG) (Huang and others 1999). This suggests that a proportion of TrkA-lineage neurons initially may (also) rely on NT3 for their survival. The number of TrkA TG neurons is normal in TrkC mutant mice, however, suggesting that NT3 may signal through TrkA at these early stages (Huang and others 1999). There are three key developmental transitions that progressively mediate the diversity of the TrkA⁺ neuronal cohort: the acquisition of the md- and sdTrkA lineages, the generation of the peptidergic/non-peptidergic division within the sdTrkA lineage, and the refinement of the individual nociceptive, thermoceptive, mechanoreceptive, and pruriceptive sensory phenotypes. Extrinsic signals are important drivers in each of these steps.

Acquisition of the TrkA-Lineages

Aδ mdTrkA⁺ and sdTrkA-lineage neurons are specified in large part through the actions of the transcriptional modulator Prdm12 (Fig. 2A). In Prdm12 mutant mice, expression of TrkA is not detected, while ectopic expression of Prdm12 is associated with a repression of TrkC and TrkB expression (Fig. 2A) (Bartesaghi and others 2019; Desiderio and others 2019). Prdm family proteins exhibit intrinsic histone lysine methyltransferase activity and can recruit lysine methyltransferases to specific enhancer domains. Prdm12 onset coincides with Ngn expression and appears to act in concert with these TFs to initiate neurogenesis of the TrkA lineage (Bartesaghi and others 2019; Desiderio and others 2019) (Fig. 2A). In Xenopus placodes and in P19 cells, Prdm12 expression appears induced by low levels of retinoid signaling (Desiderio and others 2019; Yang and Shinkai 2013). Although the influence of retinoids on Prdm12 expression remains to be demonstrated for neural crest–derived sensory neurons, this suggests that Pdrm12 onset may be coupled to the timing of intra- or extra ganglionic retinoid signaling. Continued expression of Prdm12 is mainly observed in neurons that co-express the nociceptive marker Nav1.8 and includes both peptidergic (TrkA⁺) and non-peptidergic (Runx1⁺) nociceptive neurons (Desiderio and others 2019; see also following text). The maintenance of Prdm12 in maturing TrkA-lineage neurons was shown to be required for TrkA maintenance but may also indicate additional roles for Prdm12 in the diversification of nociceptor fates (Bartesaghi and others 2019; Desiderio and others 2019). Thus, while the direct mechanism by which TrkA⁺ fates emerge is not fully understood, Prdm12/Ngn activity is an important event that introduces an asymmetry among immature NGN/Pou4f1/Isl1 sensory neurons and suggests that epigenetic mechanisms contribute to the acquisition and maintenance of TrkA⁺ sensory lineages.

Figure 2.

Sequential specification of sdTrkA⁺ lineage subtype identities. (A) Generation of the TrkA⁺ sensory lineages by virtue of the onset of TrkA expression. The Prdm12 transcription factor acts in concert with Ngns, and possibly KLF7, to promote expression of TrkA (expressed by the Ntrk1 gene) while repressing expression of TrkC and TrkB (encoded by Ntrk3 and Ntrk2, respectively). (B) proposed signaling pathways that may underlie the molecular segregation of peptidergic and non-peptidergic sdTrkA⁺ lineage neurons which innervate the stratum spinosum (SS) and stratum granulosum (SG) of the epidermis, respectively. NGF/TrkA signaling in sdTrkA⁺ neurons (pink nuclei) initiates the expression of the Runx1 transcriptional co-factor Cbfβ; whereas expression of Runx1 requires the activity of Islet1 (possibly in combination with TFs that act downstream of FGF signaling). Activity of the Runx1/Cbfβ transcriptional complex is thought to drive the maturation of non-peptidergic sdTrkA⁺ lineage neurons (orange nuclei) by inducing the expression of various molecules biased to non-peptidergic neurons. Peptidergic neurons (dark red nuclei) have been proposed to originate from both TrkA⁺Runx1⁻ and TrkA⁺Runx1⁺ lineages. Arrows denote activating (green) or repressing (red) signaling cascades or direct transcriptional activities. Target derived signals (e.g., NGF, HGF, GDNF, and NRTN) may reinforce specific sdTrkA⁺ lineage subtype identities. See text for details.

Another transcriptional regulator important for TrkA⁺ lineage neurons is the Kruppel-like transcription factor KLF7 (Fig. 2A). KLF7 was found to bind a minimal promotor of TrkA, and KLF7 mutant mice exhibit a ~50% loss in TrkA⁺ neurons (Lei and others 2001; Lei and others 2005). The TrkA minimal promotor contains at least 9 conserved TF binding sites, which besides KLF7 also include two Pou4f1 sites (Ma and others 2000). In vitro co-transfection studies in PC12 cells suggest that Pou4f1 acts synergistically with KLF7 to directly activate TrkA expression (Ma and others 2003). Indeed, while the initiation of TrkA expression is normal in the absence of both TFs, expression of TrkA fails to be maintained (Lei and others 2006). The notion that CGRP⁺ neurons are absent while the number of Ret⁺ neurons is unchanged in p0 KLF7^−/− mutants suggest that KLF7 is especially important for the peptidergic sdTrkA⁺ lineage in which TrkA expression is maintained (see following text) (Lei and others 2006). Together, these data suggest that within sensory progenitors, Pdrm12 acts together with Ngns (1 and 2) to promote TrkA-lineage fates, while the subsequent activities of Prdm12, KLF7, and Pou4f1 are required to maintain TrkA expression.

Generation of Peptidergic and Non-Peptidergic sdTrkA⁺ Lineage Fates

Shortly after the emergence of the sdTrkA⁺ lineage, the majority of TrkA neurons initiate expression of the Runt-domain TF Runx1 (Chen and others 2006; Gascon and others 2010). The induction of Runx1 expression appears independent of NGF signaling, however, Runx1 transcriptional activity is controlled by NGF/TrkA signaling, which regulates the expression of Cbfβ, a transcriptional co-factor of Runx1 (Fig. 2B) (Huang and others 2015). Runx1 expression itself appears to require the pan-sensory TF Islet1 but it is unclear how this activity of Islet is restricted to the sdTrkA⁺ lineage (Huang and others 2015; Luo and others 2007). Possibly there is a role for TFs that act downstream of FGF2, which was independently shown to promote the expression of Runx1 in early sdTrkA⁺ lineage neurons (Fig. 2B) (Hadjab and others 2013). NGF signaling is required for the maintenance of Runx1 expression (Luo and others 2007). Nascent sdTrkA neurons subsequently segregate into two main subtypes: peptidergic neurons that have been proposed to derive from TrkA⁺Runx1⁻ and/or TrkA⁺Runx1⁺ neurons, and non-peptidergic nociceptors that originate from TrkA⁺Runx1⁺ neurons (Fig. 2B) (Chen and others 2006; Gascon and others 2010). In some TrkA⁺Runx1⁺ neurons, NGF signaling (thought to act through Runx1/CBfβ) promotes the expression of the GDNF receptors Ret, Gfra1, and Gfra2, as well as other transcripts typically associated with a nonpeptidergic nociceptor identity (e.g., TrpM8, Mrgprd) (Chen and others 2006; Huang and others 2015). As these neurons begin to innervate their peripheral target domains, the original TrkA⁺Runx1⁺ population progressively segregates into NGF-dependent and GDNF-dependent sensory subclasses (Fig. 2B) (Chen and others 2006). TrkA⁺Runx1⁺Ret⁺ neurons innervate the superficial stratum granulosum of the epidermis and, in response to target derived Neurturin (NRTN) or GDNF, maintain Runx1 and Ret, but extinguish TrkA (Fig. 2B) (Luo and others 2007; Molliver and Snider 1997). NGF-dependent TrkA⁺Runx1⁺ neurons, instead, innervate the stratum spinosum (as well as internal organs) and maintain TrkA while extinguishing Runx1 (Fig. 2B) (Chen and others 2006). The repression of Runx1 in TrkA⁺ neurons is believed to involve another extrinsic signal, hepatocyte growth factor (HGF; expressed in dermal papilla cells), which signals through Met receptors (induced by NGF/TrkA signaling) (Fig. 2B) (Gascon and others 2010). Runx1, in turn, is thought to safeguard the Ret⁺Runx1⁺ nonpeptidergic lineage by a direct transcriptional repression of Met (Gascon and others 2010). Thus, through a balancing act between NGF, GDNF, NRTN, and HGF signals, TrkA⁺Runx1⁺ neurons segregate into TrkA⁺ peptidergic and Runx1⁺Ret⁺Gfra⁺ non-peptidergic sdTrkA⁺ lineages.

Diversification of sdTrkA⁺ Lineage Sensory Phenotypes

The largest increase in sdTrkA⁺ lineage diversity emerges during the second phase of their development, when peptidergic and non-peptidergic neurons mature into a myriad of specialized sensory phenotypes (Usoskin and others 2015). This process appears to be guided by many of the same signaling factors as described above. The notion that TrkA signaling is required beyond mediating neuronal survival was first demonstrated through the use of Bax1/NGF and Bax1/TrkA double mutant mice in which the loss of the pro-apoptotic protein Bax1 preserves TrkA neurons in the absence of NGF (Patel and others 2000). These and other studies since have demonstrated that NGF/TrkA signaling is not essential for initial central or peripheral axonal outgrowth but is required for normal innervation and branching within the epidermis (Patel and others 2000; Wickramasinghe and others 2008). Moreover, it was shown that NGF/TrkA signaling (and the transient expression of Runx1) is necessary for the induction or maintained expression of many molecules associated with a peptidergic nociceptive, noxious thermoceptive, or pruriceptive phenotype (e.g., Tachykinin 1, CGRP, TrpV1, Mrgpra3, Mrgprb4) (Abdel Samad and others 2010; Patel and others 2000). In contrast, signaling through NRTN/Ret and the persistent expression of Runx1 is needed to establish the biochemical phenotype of non-peptidergic neurons (e.g., expression of TrpA1, TrpC3, TrpM8, Mrgprd) and for developing C-LTMR phenotypes (Abdel Samad and others 2010; Luo and others 2007). Thus, different extracellular signals, NGF (expressed in the stratum spinosum of the epidermis) and NTRN/GDNF (expressed in the stratum granulosum), result in sdTrkA⁺ lineage neurons with distinct molecular phenotypes and response properties. sdTrkA⁺ lineage neurons that innervate different regions of skin or internal tissues (e.g., bladder or muscle) can exhibit distinct functional and morphological features, indicating that they may further specialize dependent on their precise target location (Olson and others 2017). To what extent this additional refinement in phenotypes depends on quantitative differences in signaling strength of known ligands (e.g., NGF, GDNF, NRTN, Artemin) (Albers and others 2006; Elitt and others 2006; Jankowski and others 2017), or on yet to be identified extrinsic signals, remains unclear.

Extrinsic Cues and Low-Threshold Mechanoreceptive Sensory Neurons

LTMRs that innervate the skin, tendons, or ligaments are comprised of several subtypes, each dedicated to a different aspect of discriminative touch (Handler and Ginty 2021). Adult LTMR classes are variably referred by their main physiological features or their anatomically distinct sensory endings (Fig. 3A). For instance, rapidly adapting (RA) Aβ LTMRs are fast conducting myelinated mechanoreceptors that can be associated with Pacinian or Paciniform corpuscles, Meissner corpuscles, or circumferential and longitudinal lanceolate endings surrounding hair follicles (Fig. 3A) (Wende and others 2012). Slowly adapting (SA) Aβ LTMRs are also myelinated and fast conducting but typically are associated with Merkel cells and possibly Ruffini endings (Fig. 3A). Instead, Aδ LTMRs and C-LTMRs are thinly myelinated RA, and non-myelinated intermediately adapting (IA) afferents, respectively, that associate with awl and zigzag hair lanceolate endings (Handler and Ginty 2021; Zheng and others 2019). Recent investigations demonstrate that the diversity among LTMRs may still increase with further scrutiny, as several of the receptor organs (e.g., Merkel cells, Meisner corpuscles) can be innervated by more than one LTMR with distinct molecular identities and physiological properties (Bachy and others 2011; Bai and others 2015; Jenkins and others 2019; Neubarth and others 2020; Niu and others 2014).

Figure 3.

Development of low-threshold mechanoreceptors (LTMRs). (A) LTMR subtypes can be distinguished based on neuronal size and conduction velocity (Aβ, Aδ, and C), sensory receptor endings, and physiological properties. LTMR afferent response properties range from rapidly adapting (RA), to intermediate adapting (IA), to slowly adapting (SA). (B) Transcriptional mechanisms that direct nascent proprioceptor or AβLTMRs toward different developmental trajectories. (C) Sensory end-organs harbor various cell types that express signaling factors and can influence the specialization and/or maturation of developing AβLTMR sensory neurons (see text for details).

Similar as for TrkA lineage neurons, signaling through neurotrophin receptors is important for LTMR survival and phenotypic specialization throughout their development. In contrast to Aδ and C-LTMRs (which derive from md- and sdTrkA⁺ lineage neurons, respectively), AβLTMRs originate from transient hybrid TrkC⁺/TrkB⁺ or Ret⁺/TrkB⁺ neurons but undergo dynamic changes in neurotrophic receptor expression between e10 and e14 (Fig. 3B) (Bachy and others 2011; Bai and others 2015; Bourane and others 2009; Kramer and others 2006; Luo et al 2009; Wende and others 2012). The rapid changes in neurotrophin receptor expression within this short developmental window has made it a challenging task to deduce the exact developmental origin of mature AβLTMRs. Given the influence of neurotrophins on LTMR differentiation, a key question is how neurotrophin receptor expression is regulated. A few studies have begun to shed some light on this. For instance, the transcriptional activities of Runx3 and Shox2 guide hybrid (e11) TrkC⁺/TrkB⁺ neurons into proprioceptor and AβLTMR subtypes. In prospective proprioceptors and AβSA LTMRs, Runx3 promotes expression of TrkC and represses Shox2. Conversely, Shox2 is required to promote expression of TrkB in presumptive Aβ RA and AδLTMRs (Fig. 3B) (Abdo and others 2011; Appel and others 2016; Kramer and others 2006). The Maf TF may serve a similar function as Runx3 in hybrid Ret⁺/TrkB⁺ neurons by repressing Shox2 (and thereby TrkB) in prospective Ret^only and Ret⁺/TrkC⁺ neurons that derive from these neurons (Fig. 3B) (Bourane and others 2009; Lecoin and others 2010; Wende and others 2012). Given the importance of Runx3, Shox2 and Maf in regulating neurotrophin receptor profiles in nascent LTMRs, a better understanding of the developmental signals that drive these TFs should provide additional insight into LTMR lineage selections.

Unlike nociceptors that terminate in free endings, LTMRs associate with elaborate receptor structures that often co-develop with the innervating afferent. These sensory end-organs frequently consist of multiple cell types including dermal cells, specialized keratinocytes, fibroblasts, or terminal Schwann cells, and can serve as signaling centers that drive LTMR functional maturation and maintenance (Fig. 3C) (Cobo and others 2021; Jenkins and others 2019). Indeed, touch dome development and Merkel cell innervation appears to be guided by BMP expressed by specialized dermal cells that abut the guard hair follicle, and by (yet to be identified) signals from Keratin-17⁺ keratinocytes (Fig. 3C) (Doucet and others 2013; Jenkins and others 2019). Similarly, epidermal cells associated with hair follicles secrete the extracellular matrix component EGFL6, which appears necessary for the normal morphological development of lanceolate endings (Fig. 3C) (Cheng and others 2018). BDNF, secreted from hair follicle–associated keratinocytes, is required for orientation selectivity of Aδ LTMR lanceolate endings (Fig. 3C) (Rutlin and others 2016). The extent to which these receptor target-derived factors drive subtype specification or merely promote phenotypic maturation is difficult to discern, but in light of the notion that sensory identities remain plastic until fairly late in their development (see following text), end-organ-derived cues may deserve more attention.

Extrinsic Influences on Proprioceptive Sensory Neurons

Proprioceptors share many similarities with AβLTMRs but selectively innervate skeletal muscle and provide information about limb position and movement (Fig. 1B). They diversify at two different levels: the peripheral muscle targets that they innervate (referred to as “muscle-type identity”), and the association with functionally distinct intramuscle receptors (here referred to as “functional identity”) (Fig. 4A) (Zampieri and de Nooij 2021). With regard to their functional identity, two main subtypes exist: Muscle spindle (MS) and Golgi tendon organ (GTO) afferents. MSs are embedded in the bulk of the muscle, in between the extrafusal muscle fibers that are innervated by motor neurons and mediate muscle contraction. Each spindle consists of several intrafusal muscle fibers that serve as stretch receptors; they contain contractile elements, but their activation contributes little to muscle force output (Fig. 4A) (Banks and others 2021). MS afferents are further divided into group Ia and group II afferents, which are biased to the dynamic and static component of muscle stretch respectively (Fig. 4A) (Banks and others 2021). GTOs, in contrast to MSs, are located within myotendinous junctions and aponeuroses, the connective tissues that connect the extrafusal muscle fibers to muscle tendons (Jami 1992).

Figure 4.

Development of proprioceptor subtype identities. (A) Proprioceptors can be distinguished based on their functional or muscle-type identities. The three main functional subtypes include muscle spindle (MS) groups Ia and II afferents, and Golgi tendon organ (GTO) group Ib afferents. Afferent muscle-type identity varies based on peripheral muscle target and is reflected in spinal target selectivity, that is, MS afferents preferentially make synaptic contacts onto homonymous motor neurons that project to the same muscle target. (B) Proprioceptor muscle-type identity is influenced by mesenchymal and muscle-specific signaling factors. (C) Proprioceptor subtype identities may be influenced by (yet to be defined) signaling factors emanating from the intrafusal fibers of the muscle spindle or from the myotendinous junction.

All proprioceptors depend on NT3 for their initial survival and remain influenced by NT3 signaling into adulthood (Fariñas and others 1996; Patel and others 2003; Wang and others 2007). Emerging proprioceptors initially co-express TrkC and TrkB (Fig. 3B) (Kramer and others 2006). The onset of expression of TrkC or TrkB within these neurons has not been fully understood but the restriction into a solitary TrkC⁺ sensory phenotype depends on Runx3 (Fig. 3B) (Kramer and others 2006). Considering that the onset and activity of Runx3 is a driving force in the selection and consolidation of the TrkC lineage, what are the signals that induce Runx3? Expression of Runx3 in DRG is first detected at e11 and critically depends on Pou4f1; Runx3 is not detected in Pou4f1 mutants (Appel and others 2016; Eng and others 2007; Oliver and others 2021). Yet, since Pou4f1 is widely expressed in DRG, it seems likely that Pou4f1 primarily serves a permissive role for Runx3 expression. Also associated with Runx3 onset is Sox11, a TF that can be induced by retinoid signaling (Appel and others 2016; Jankowski and others 2006; Sock and others 2004; Wang and others 2019b). These observations may suggest that retinoid signaling activates Sox11, which together with Pou4f1 may help steer immature sensory neurons toward a Runx3⁺ proprioceptive identity.

As indicated above, proprioceptors possess superimposed “muscle-type” and “functional” identities (Fig. 4A). Our knowledge of how proprioceptor muscle-type identity emerges remains limited, yet several studies have begun to uncover some of the factors involved in regulating this aspect of their identity. First, it appears that Runx3 is necessary for the normal development of proprioceptors destined to innervate forelimb extensor muscles, but not or less so for forelimb flexor, axial, or intercostal muscles (Wang and others 2019a). This differential dependency on Runx3 correlates with its expression level: proprioceptors that innervate forelimb extensors express higher levels of Runx3 than those that innervate forelimb flexors. Runx3 expression levels at this stage appear to depend on the level of embryonic muscle NT3 (Wang and others 2019a). Thus, quantitative differences in (muscle) target-derived NT3 may diversify proprioceptors with respect to their muscle type identity (Fig. 4B) (de Nooij and others 2013; Fariñas and others 1996; Wang and others 2019a). However, previous studies indicated that the surrounding mesenchyme, not muscle, may provide the primary source for signals that instruct muscle-type identity (Poliak and others 2016). Whether these mesenchymal signals may also involve NT3, or may involve other signaling molecules (e.g., Wnt, retinoids), remains to be determined (Patapoutian and others 1999; Wang and others 2019b). Homeobox (Hox) TFs have also been implicated in the regulation of proprioceptor muscle-type identity given their regionally restricted patterns of expression in forelimb innervating proprioceptors. For instance, proprioceptors that innervate more proximal/dorsal limb muscles primarily express Hoxa5 while those innervating distal/ventral muscles express HoxC8 (Shin and others 2020). The relevance of Hox expression in regulating muscle type identity was demonstrated through conditional loss of function studies, which showed that proprioceptors that lack HoxC8 fail to make the appropriate synaptic connections to their normal motor neuron targets in spinal cord (Shin and others 2020). Although the initial induction of a given segmental combination of Hox genes may result from general anterior-posterior patterning events that are intrinsic to the sensory progenitors, the selective maintenance of a given Hox-profile in subsets of proprioceptors may be established when the afferents innervate their peripheral targets (Fig. 4B). Thus, additional studies are required to resolve how differences in the peripheral environment (e.g., levels of NT3 or other signaling molecules) result in qualitative (HOX factors) or quantitative (Runx3) differences in transcription factor expression, and how these factors act together to influence proprioceptor muscle-type identity.

Proprioceptors can also be distinguished based on their association with muscle spindle (MS) or Golgi tendon organ (GTO) receptors, which detect changes in muscle length or muscle force, respectively (Fig. 4A) (Zampieri and de Nooij 2021). Recent transcriptome analyses of adult proprioceptors have begun to provide insight into the molecular basis of these proprioceptor subtypes (Oliver and others 2021; Wu and others 2021). A remarkable finding of these studies came from the longitudinal molecular profiling of proprioceptors across their development. It appears that transcriptional distinctions between MS and GTO afferents emerge only after these neurons connect to their nascent receptor targets (Oliver and others 2021; Wu and others 2021). Similarly, many molecular differences between group Ia and group II MS afferents are not established until postnatal stages (Oliver and others 2021; Wu and others 2021). Given that prospective MS and GTO afferents that innervate the same peripheral muscle likely are exposed to the same (mesenchymal/muscle) environment, these studies may indicate that generic proprioceptors are instructed by MS- or GTO-derived signals to achieve their final identity (Fig. 4C). Consistent with this idea, blocking muscle spindle development results in a loss of molecular markers that are typical for MS afferents, thus indicating that the spindle environment can influence MS afferent identity (Fig. 4C) (Tourtellotte and Milbrandt 1998; Wu and others 2019). Together, these observations suggest that proprioceptors may respond to extrinsic cues at multiple stages of their journey, each promoting or repressing a specific differentiation choice. The nature of these signals is not yet clear, but clues to their identity (i.e., specific receptors) may be present in proprioceptor transcriptomic data (Oliver and others 2021; Wu and others 2021).

Intrinsic versus Extrinsic Regulation of Sensory Neuron Diversity

As described above, target-derived neurotrophic signals are critical for sensory neuron survival and have been implicated in sensory neuron subtype differentiation by activating preexisting cell-intrinsic transcriptional programs. However, several studies have questioned the extent to which nascent sensory neuron subtypes are in fact preordained to a specific fate when they first emerge (Moqrich and others 2004; Sharma and others 2020). Perhaps most significant among these was the demonstration that expression of a TrkC transcript from the TrkA locus could transform postmitotic TrkA⁺ nociceptive neurons into TrkC⁺ proprioceptive neurons (Moqrich and others 2004). This result implied that sensory neuron identity may not be intrinsically encoded in progenitor neurons but that emerging postmitotic sensory neurons—at the outset—remain competent to develop into many subtypes. Still, only a subset of prospective TrkA neurons were able to transform into proprioceptors, indicating some—possibly intrinsic—restrictions in the capacity of subsets of TrkA neurons to switch fate.

A second study supporting the critical role of extrinsic factors in specifying post-mitotic sensory identities relies on genome-wide DRG transcriptomics. This study demonstrates that shortly after cell cycle exit, emerging sensory neurons are transcriptionally unspecified (Sharma and others 2020). Instead, it was found that immature (~e11.5) sensory neurons co-express many of the transcription factors that are associated with distinct sensory phenotypes in adults (Sharma and others 2020). Some of these transcriptional regulators remain expressed in relatively broad neuronal populations, suggesting that they may control features shared between multiple sensory neuron subtypes, while others show a more limited expression pattern and become restricted to one or two specific adult sensory subtypes. For instance, Cux2 is expressed in most if not all early born/Ngn2 descendent neurons and may confer mechanosensory features shared by most low threshold mechanoreceptors (Bachy and others 2011). In contrast, Pou4f3, is selectively maintained in two subtypes of CGRP neurons (α and η), and proprioceptive GTO afferent neurons (Oliver and others 2021; Sharma and others 2020). Taking advantage of the same Bax1 mutant genetic strategy described earlier—permitting cell fate assessment when neurotrophic support is absent—it was demonstrated that extrinsic signals direct this gradual selection of subtype-specific gene expression patterns. In the absence of NGF and Bax1, the normal combinatorial pattern of transcription factors that defines subsets of adult TrkA⁺ nociceptors was completely altered (Sharma and others 2020). The signaling molecules that shape these transcriptional patterns and associated sensory identities likely derive from various locations along sensory axon trajectories, including from peripheral receptor targets. Despite the surprisingly overlap in transcriptional identities of emerging sensory neuron subtypes, it is important to note that, as seen for Runx1, TF transcript levels may not always correlate with TF activity levels (Huang and others 2015). Nevertheless, these studies begin to suggest a developmental mechanism in which each DRG neuron is born with roughly the same differentiation potential as its neighbors and is set up with overlapping sets of individual transcriptional factors. Depending on the environmental cues encountered along its peripheral path, each neuron may retain a progressively smaller subset of these transcription factors and, consequently, narrow its phenotypic potential (Fig. 5). The downstream transcriptional targets (e.g., cell surface molecules, ion channels) of the remaining TFs together may shape the sensory subtype identity of a given neuron. Qualitative or quantitative differences in TF expression levels in turn could influence the induction of the receptors for extrinsic signals, thus potentially reinforcing a given developmental trajectory. The extent to which the initial selection of TFs is equal across all post-mitotic sensory neurons is not yet clear. Since sensory neurons are born at different times, it is possible that rapid developmental changes in the nascent DRG environment may influence the repertoire of transcription factors these early postmitotic neurons can express. In addition, the observation that Prdm12 expression can predict a prospective TrkA⁺ nociceptive fate at proneural stages could indicate that an initial level of TF specification may already take place prior to when neurons become post-mitotic. The molecular mechanisms by which certain peripheral signals may promote the activity of one transcription factor over another remain largely unclear but could involve epigenetic, transcriptional, or post-translational mechanisms. At present, a lot of attention is directed at understanding the epigenetic changes that are imparted on growth factor exposure—as this has been seen to be important in other developing neural tissues (Frank and others 2015). Thus, the rapidly evolving technical and computational advances in genome-wide analyses of the epigenetic landscape are certain to also be relevant for the study of sensory neuron differentiation.

Figure 5.

Mechanisms of sensory neuron subtype specification. Sensory neuron subtype specification is thought to rely on molecular mechanisms in which intrinsic genetic determinants interact with extrinsic signaling molecules to specify mature somatosensory phenotypes (e.g., peripheral and central targets and biochemical properties). The relevance of the intrinsic versus extrinsic signals may range from being entirely dependent on intrinsic factors (i), to increasing influences from extrinsic signaling molecules (ii), to a complete reliance on extrinsic signals (iii). (Blue, green, and red colored cell bodies demarcate proprioceptor, LTMR, and sdTrkA⁺ lineage subtypes, respectively; colored nuclei indicate a role for intrinsic subtype-specific developmental mechanisms). In (ii) afferent subtype identities are broadly specified with respect to their class-identity by intrinsic factors, while intraclass subtype identities depend on environmental signaling sources. In (iii), afferent subtype is plastic at the outset and emerges entirely through sequential influences of extrinsic signals present along the trajectory of the afferents. Future studies are required to distinguish which of these models is most appropriate in the development of individual somatosensory neuron subtypes.

Relevance to the Clinic

The observation that some sensory afferents assume their mature phenotype well after reaching their peripheral target supports the idea that sensory neuron identity may remain plastic after they become post-mitotic (Oliver and others 2021; Sharma and others 2020; Wu and others 2019; Wu and others 2021). Several studies indicate that human sensorigenesis relies on many of the same molecules as in rodents and may similarly be influenced by environmental signaling molecules (Quinn and others 2021; Rayon and others 2021). The notion that human sensory phenotypes may also be—and perhaps remain—plastic in their development toward a specific sensory subtype, could have important translational implications. First of all, following injury, regeneration of specific sensory phenotypes could be aided by the application of subtype relevant signaling molecules at defined locations along their axonal trajectory. Along similar lines, sensory neurons and/or sensory endorgans exhibit morphological alterations or degenerate during aging (Feng and others 2018; Fundin and others 1997; García-Piqueras and others 2019). Such age-related changes in sensory endorgans have also been associated with changes in end-organ-derived signals (Bergman and others 2000). Thus, if target-derived signals can influence neuronal identity, this could mean that a change in the (balance of) environmental signals during senescence could lead to altered sensory phenotypes. Indeed, aging Merkel cells in Merkel cell neurite complexes of the whisker mystacial pad are known to lose expression of NT3 (Bergman and others 2000). The Merkel cell afferents that innervate these structures consequently also lose expression of TrkC, but maintain expression of TrkA (Bergman and others 2000). These TrkC^offTrkA⁺ Merkel afferents may still respond to nearby sources of NGF or NT3, but since TrkA uses a different intracellular pathway than TrkC (and elicits distinct cellular responses) they potentially can adopt a more “TrkA⁺ lineage-like” phenotype (Fig. 6) (Gorokhova and others 2014). Although such observations have thus far been limited to studies in rodents, it is not unreasonable to speculate on similar age-related changes in human sensory neuron target tissues. Thus, understanding the developmental signals that drive neurons to their unique phenotype may be utilized to help prevent or alleviate injury- or age-related sensory dysfunction. Finally, a better understanding of the developmental signals that unlock the specific transcriptional code for a certain sensory subtype may offer important insights for in vitro derivation studies aimed at generating these sensory subtypes for disease modeling.

Figure 6.

Sensory neuron subtype plasticity and aging. Aging-associated changes in peripheral target environments could potentially result in pathological changes in sensory neurons. As schematically depicted, in rodents some Merkel cell afferents express the neurotrophin receptors TrkC and TrkA and primarily rely on NT3 expressed in Merkel cells. In senescence, Merkel cells can be depleted or show diminished expression of NT3, and Merkel afferents fail to maintain TrkC. Hypothetically, a relative increase in signaling through TrkA receptors (in response to nearby sources of nerve growth factor [NGF]) could alter Merkel cell afferent character and central target selectivity. Note that such aging-related changes in Merkel cell afferents remain to be demonstrated in human.

Conclusions

As outlined in this review, somatosensory neurons acquire their sensory identity through progressive broad-to-restricted patterning of transcription factor expression. Peripheral signals, along the peripheral trajectory or within the target areas of the sensory axons, have been suggested to play a major role in this selection process, although the molecular mechanisms are not yet fully understood. The continued influence of extrinsic target derived signals may sustain some level of sensory neuron plasticity throughout life, with implications for the injured or aging somatosensory system.

Despite these advances in our understanding of sensory neuron development, it remains unclear to what extent this “modus operandi” of sensory differentiation is universal across other sensory systems, or why it may present a developmental advantage. Possibly, differentiation through extrinsic signals permits greater adaptability in response to changes in peripheral tissues. Other questions relate to the type of extrinsic cues that sculpt the diverse sensory subtype phenotypes. Many of these developmental signals—in particular, members of the neurotrophin family of signaling factors—function in multiple aspects of sensory neuron differentiation and frequently are dynamically (co-)expressed in or near sensory neuron targets. This presents an enormous challenge in teasing out the precise role of a given extrinsic factor at a specific peripheral location. Moreover, signaling molecules other than neurotrophins (e.g., BMPs, FGFs, HGFs) are also likely involved, either acting alone or in conjunction with neurotrophin signaling (Ji and Jaffrey 2012). Insight into the nature and role of each of the relevant extrinsic signals may be derived from existing or future longitudinal sensory neuron transcriptomic analyses, by virtue of the receptor molecules that are expressed in individual sensory cohorts or subtypes at specific differentiation divergence points. Such studies may not only reveal the identity of the extrinsic cues but will also permit an assessment of the corresponding transcriptional changes following specific choice points (i.e., which transcription factors prevail, and which do not). Last, it will be important to discern the developmental cues that delineate functional sensory subtype identities from those that impart a regional identity. Thus far, such studies are largely focused on proprioceptors given the direct implication for their integration with the appropriate spinal motor circuits, but similar studies will be equally relevant for other sensory modalities. Future studies that enable the manipulation and isolation of specific sensory neuron subsets based on more regionally restricted genetic labeling strategies may help begin to address these questions.