The Study of C-Low Threshold Mechanoreceptors: the Case Of

UNIVERSITE D’AIX-MARSEILLE ECOLE DOCTORALE Des Sciences de la Vie et de la Santé IBDM – UMR 7288

Thèse présentée pour obtenir le grade universitaire de docteur

Spécialité : Neurosciences

Présentée et soutenue publiquement par

Manon Bohic

The study of C-low threshold mechanoreceptors : the case of bhlha9 in somatosensation

Soutenue le 31/01/2019 devant le jury :

Sophie Chauvet IBDM - Marseille Présidente de jury Håkan Olausson Linköping University Sweden Rapporteur David Bennett Oxford University UK Rapporteur Irène Marics IBDM - Marseille Invitée Aziz Moqrich IBDM - Marseille Directeur de thèse

Numéro national de thèse/suffixe local : 2017AIXM0001/001ED62 1

Remerciements :

Aziz, je te remercie de m’avoir accueillie dans ton équipe pour mon tout premier stage dans le milieu de la recherche il y a de ça presque 8 ans, et ensuite en master et en thèse ! Toi et Irène avaient été mes mentors et mon inspiration dans mes études et ma carrière de jeune chercheuse et je vous en serai éternellement reconnaissante. Mais plus que ça, tu as partagé avec moi non seulement ta vision de la science et de la recherche, mais aussi ta vision de la vie, ton humour, ton sourire, ta sagesse et ton extrême générosité. Choukrane.

Irène, le sourire, la bonne humeur, une épaule pour les chagrins d’amour, du recul et de la sagesse quand la thèse ne se passe pas comme prévu, un soutien jusque dans le détail du micro pour mon séminaire interne et puis des discussions fringues et maquillage ☺ Je te remercie pour tout Irène, tu vas beaucoup me manquer, bisous

Pascale, merci de m’avoir aidé tous les jours de la thèse d’une façon ou d’une autre, comme Aziz et Irène à la fois pour l’aspect scientifique (l’enseignement de la biologie moléculaire, la rigueur dans le travail et l’organisation du labo) mais aussi humain, pleins de magnifiques photos, le gospel (j’ai vraiment adoré !), des anecdotes de ta vie personnelle qui m’ont aidé dans la mienne et puis une réelle bienveillance. Toi, Aziz et Irène, c’est vous les piliers du labo qui en faites un endroit aussi familial que j’ai toujours eu plaisir à retrouver le matin pendant toutes ces années.

Stéphane et Ana, non je ne vais pas vous mettre dans le même paragraphe ! Chacun le sien, juste pour vous dire que vous êtes tout aussi importants pour moi que les trois piliers du labo que je viens de remercier ^^

Stéphane, j’ai adoré te découvrir petit à petit pendant ces années de thèse, on a commencé doucement et puis on a découvert que tous les deux ‘on n’aime pas les gens’ mais on aime les Spartiates et on a fini employeur – employée. Je te remercie pour ton calme, ta gentillesse, ta générosité et ta curiosité scientifique permanente qui m’a beaucoup inspiré et puis ton humour, le petit sourire taquin et le clin d’œil qui va avec. Merci pour tout

Ana, Ah Ana la magnifique, grande prêtresse des huiles essentielles et des doubitchous roulés sous les aisselles ^^ on en a passé quelques-unes des soirées à discuter de tout pendant des heures, au labo ou avec une petite pression au Blackstone. Et puis heureusement que tu étais là pour m’offrir un repas maison, chaud et équilibré de temps en temps. Bien sûr, comme Stéphane, vous m’avez beaucoup apporté humainement mais aussi scientifiquement. C’est avec vous deux que j’ai appris à « discuter science », pas forcément que dans un cadre formel, dans un labmeeting, mais toujours avec passion. Merci. De même, tu vas beaucoup me manquer.

Au milieu de tous ces Moqrich, je me dois de remercier Rosanna Dono, qui m’a aussi pris sous son aile en master et m’a enseigné beaucoup pendant ces tout premiers stages dans la recherche, avec douceur, calme et aussi avec le sourire.

Les autres membres de la fine équipe : Lucie et Anissa, les deux derniers ‘bébés Moqrich’ :p et Catarina. Je suis bien contente d’avoir partagé le bureau avec vous ces dernières années, d’avoir partagé avec vous un peu de ce que j’ai appris en science (si si ça nous arrive de parler science dans le bureau des jeunes ^^ ), et puis aussi des discussions de fringues, de maquillage, de célébrités, horoscopes, et autres objets vendus ou non chez sephora et monoprix brefs des discussions plus ou moins scientifiques mais passionnantes aussi !

Francis, je suis bien contente que tu aies décidé de rejoindre l’équipe parce que j’ai beaucoup appris en discutant avec toi, comme avec Aziz, non seulement de science mais de la vie en général, la co**erie uniformément répartie, la politique, les sushis, je suis vraiment contente de te connaître ! Et puis il faut dire que tu nous as offert un délicieux méchoui d’agneau, dans un coin reculé et magnifique de Marseille que je ne connaissais pas (avec pleins de virages comme je les aime sur la route) ça ne s’oublie pas !

Nadine et Sunjae, même si on a peu échangé, je me trouve chanceuse d’avoir été dans la même équipe que vous, et d’avoir découvert grâce à vous un petit peu de cette chose mystérieuse qu’est pour moi l’électrophysiologie. Bonne continuation à tous les deux.

Et en-dehors de l’équipe, parce qu’une thèse, dans les bons et dans les mauvais moments, on la partage avec tout plein de gens et c’est ça qui rend la recherche si belle : Raphaël et Alain ça remonte à bien avant la thèse ! En licence déjà on partageait les sudokus, les mots croisés et les parties de cartes pendant la pause de midi, au soleil aux tables du CROUS. On ne pensait pas vraiment ni à la thèse, ni même au master ^^ et puis Anaïs nous a rejoint quelques années plus tard puis Jules, et j’ai adoré nos pauses cafés, nos discussions, nos sessions vidage de sac, et tout le reste à l’extérieur de Luminy (Amsterdam, bientôt la Russie…). En ce qui me concerne, ça m’a définitivement aidé à tenir le cap et vous allez beaucoup me manquer mais j’espère bien qu’on se reverra de temps en temps, soit dans mon futur chez moi dans le New Jersey, soit chez vous où que ce soit ☺

Et puis il y a les amitiés qui ont commencé à la fac ou au labo et qui continuent et continueront à l’extérieur : Isis, Agnès, Amel et Eunice love you

Enfin, les trois personnes les plus importantes dans ma vie, ma maman, mon papa et ma petite sœur. Merci pour tout, merci d’être toujours là pour moi, de croire en moi même quand je n’y crois pas trop, de m’aimer infiniment comme je vous aime infiniment en retour. Je suis impatiente de vous faire découvrir et de découvrir avec vous les Etats-Unis. Oui c’est pas la porte à côté mais je suis sûre que ça va être sympa. On a fait pas mal des pays de l’Europe tous les 4, il est temps de passer à un autre continent ! Tous pour un…

Résumé :

Notre système nerveux somatosensoriel joue un rôle clé dans notre vie quotidienne, dans nos interactions avec notre environnement, les gens et les choses qui nous entourent, ce qui le rend indispensable pour notre bien-être physique et mental. En effet, il nous permet de percevoir aussi bien l’effet relaxant d’un massage que la douleur d’une cheville foulée. Les neurones somatosensoriels qui innervent notre peau permettent la détection de ces stimuli et sa transformation en un signal qu’ils transmettent à la moelle épinière. Cette moelle épinière joue un rôle de filtre, ou de « portillon » qui ne laisse passer que les informations les plus pertinentes en direction des structures supérieures telles que le cerveau. Cependant, une défaillance à n’importe quel niveau de cette chaîne peut entraîner l’établissement de douleurs chroniques, un mal qui perturbe la vie d’un cinquième de la population mondiale.

Les neurones somatosensoriels présentent une vaste hétérogénéité à la hauteur de la variété des sensations qu’ils sont capables de transmettre. Ainsi, ils peuvent être subdivisés en populations distinctes en fonction de leur anatomie, leur physiologie, les modalités qu’ils détectent mais aussi leur profil moléculaire. Une bonne connaissance de ces différents profils moléculaires est capitale pour construire les outils génétiques nécessaires pour étudier chaque population et ainsi comprendre leurs spécificités. Même si les neurones somatosensoriels sont l’objet de recherches depuis plusieurs décennies maintenant, une population toute particulière du nom de mécanorécepteurs à bas seuil à fibre C (C-Low Threshold MechanoReceptors C-LTMRs) n’est revenue que récemment sur le devant de la scène. Ces neurones innervant strictement la peau poilue ont été très conservés au fil de l’évolution des espèces (Seal et al., 2009; Vallbo et al., 1999; Zotterman, 1939). Ils peuvent transmettre à la fois le toucher affectif et affiliatif (Loken et al., 2011), le refroidissement de la peau (Li et al., 2011), mais aussi moduler la douleur chez l’homme et chez la souris (Delfini et al., 2013; Francois et al., 2015; Liljencrantz and Olausson, 2014; Nagi et al., 2011; Urien et al., 2017). Cependant, à l’heure actuelle peu d’études permettent de construire des hypothèses quant aux mécanismes qui expliqueraient des fonctions aussi diverses.

Etant donné les défis de santé publique auxquels fait face notre société, à savoir l’augmentation des troubles mentaux et le manque de traitements efficaces des douleurs chroniques, une meilleure connaissance d’une population de neurones impliquée à la fois dans les interactions sociales et dans la modulation de la douleur paraît indispensable. Aussi, le but de ma thèse a été de développer des outils génétiques et de les utiliser pour mieux connaître les C-LTMRs. Pour ce faire, j’ai mis à profit les résultats d’une étude effectuée précédemment par l’équipe qui a grandement enrichi notre connaissance du profil moléculaire des C-LTMRs (Reynders et al., 2015). J’ai ainsi pu étudier un gène dont l’expression est particulièrement forte dans les C-LTMRs, au sein du système nerveux de la souris. Ce facteur de transcription nommé bhlha9 apparaît dans les C-LTMRs peu après la naissance ce qui en fait un marqueur intéressant de cette population. Des animaux génétiquement modifiés pour ne plus l’exprimer montre deux phénotypes intrigants. Le premier apparait en lien avec le rôle des C-LTMRs dans la détection du refroidissement de la peau puisque ces souris mutantes n’adaptent plus leur comportement à une variation de température du sol sur lequel elles marchent. Le deuxième

5 phénotype présenté par ces souris évoque un autre aspect de la fonction des C-LTMRs qui est la modulation des douleurs induites par une lésion. En effet, les mutants bhlha9 montre une douleur exacerbée dans un modèle de douleur inflammatoire. Plus intriguant encore, cette douleur exacerbée est insensible à l’effet normalement analgésique d’un modulateur positif de la signalisation GABAergique ionotropique. Ceci suggère un rôle des C-LTMRs dans la modulation de la douleur inflammatoire via le système GABAergique ionotropique. De plus, ces deux phénotypes sont associés à un dimorphisme sexuel puisque seuls les mâles mutants et non les femelles sont touchés. Ceci est un nouvel aspect de la recherche sur les C-LTMRs qui méritera d’être pris en compte dans des études futures. En conclusion, le profil d’expression de bhlha9 en fait un potentiel nouvel outil génétique pour l’étude des C-LTMRs par la communauté scientifique. De plus, j’ai découvert que ce facteur de transcription est nécessaire pour une perception de la température et une modulation de la douleur inflammatoire appropriées.

Abstract

Somatosensory neurons are highly heterogeneous, to par with the wide array of sensations they allow us to detect. As such they can be subdivided into distinct populations according to their anatomy, physiology, the modalities they detect but also their molecular profile. A good knowledge of these molecular identities is the key to develop appropriate genetic tools to study each population and understand their specificities. Even though somatosensory neurons have been the subject of studies for decades now, one particular subpopulation of neurons called C-Low Threshold MechanoReceptors (C-LTMRs) has only recently been at the forefront of sensory and pain research. These hairy-skin innervating neurons are highly conserved across species (Seal et al., 2009; Vallbo et al., 1999; Zotterman, 1939). They mediate affective and affiliative touch (Loken et al., 2011) and modulate pain in mouse and in human (Delfini et al., 2013; Francois et al., 2015; Liljencrantz and Olausson, 2014; Nagi et al., 2011; Urien et al., 2017). Interestingly, they are also activated by cooling (Li et al., 2011). Few studies hint at the spinal cord networks that could explain C-LTMRs having such different functions: transduction of innocuous sensation and pain modulation (Kambrun et al., 2018).

Considering the current health issues faced by our society, be it the chronic pain and related pain-killer abuse epidemic or the increase in mental health issues, a better understanding of a population of neurons necessary both for proper pain modulation and everyday social interactions seems hugely relevant. Thus, the aim of my thesis studies was to learn more about this particular population of neurons. I took advantage of the C-LTMR molecular repertoire recently generated in our laboratory (Reynders et al., 2015) to show that the transcription factor bhlha9 is highly enriched in this population compared to the rest of the nervous system. Its expression appearing only after birth makes it an interesting genetic tool to study C-LTMRs. Male mice lacking BHLHA9 show a wide-ranging thermotaxis defect compared to littermate controls.

Moreover, the same male mice present an exacerbated inflammatory pain in the formalin test paradigm. Intriguingly, this exacerbated pain is insensitive to the normally analgesic effect of GABAA receptor positive modulation. Recording of lamina II inner neurons, the dorsal horn region receiving C- LTMR inputs, show an increased excitability in naive condition. In conclusion, I uncovered bhlha9 as a new molecule required for correct perception of temperature and for inflammatory pain processing, in a sexually dimorphic manner.

Keywords : sensory nervous system, C-LTMRs, temperature, pain, genetic inactivation, bhlha9

Table des matières Introduction ...... 10 I - Peripheral sensory neurons...... 11 Anatomy of somatosensory neurons ...... 11 Developmental origin of somatosensory neurons in the mouse ...... 14 Somatosensory neuron populations ...... 16 C-LTMRs ...... 22 II - Spinal processing of peripheral input ...... 48 Gate Control Theory ...... 50 C-LTMR spinal circuitry ...... 52 Results ...... 62 Bhlha9 and gm7271 KI-KO mice ...... 62 Tafa4 transgenic mouse ...... 67 Bhlha9-KO mouse ...... 71 Publication ...... 71 Additional results of the study of bhlha9-KO mice ...... 94 Bhlha9 conditional KO mouse ...... 113 Discussion: ...... 118 Bhlha9 total KO versus bhlha9 cKO DRG specific ...... 118 Sexual dimorphism of the sensory phenotypes observed in bhlha9 KO male mice ...... 119 The study of C-LTMRs ...... 122 Annexe: ...... 136

Introduction

Our somatosensory nervous system is key to our daily interactions with the outside world, to our mental wellbeing and to our survival. It allows us to feel changes in or applied to our body, be it a sprained ankle or the relaxing effect of a massage. It is composed of somatosensory neurons responsible for the detection of stimuli of varied nature, innervating our skin as well as our inner organs such as viscera and muscles. Once activated, these neurons send the information to the dorsal horn of the spinal cord, where occurs an assembly, a filter and a relay of the signals coming from the periphery to supraspinal structures. This step allows an encoding of the information involving different parameters such as location, intensity, modality of the stimulus. The spinal cord is also the place where higher structures can modulate the integration of incoming signal in a process called descending control. Once the signal is encoded in the spinal cord, projection neurons will send it to the brain where the somatosensation will be associated with our emotional state and personal history to create a conscious feeling (Figure 1).

Figure 1: Scheme of the pathway followed by a peripheral stimulus detected by peripheral receptors in the skin, encoded by these peripheral sensory neurons (PSNs), transmitted and processed by spinal cord networks of the dorsal horn before being relayed to higher structures such as the thalamus by projection neurons, for example of the spino- thalamic tract.

In conclusion, understanding the somatosensory nervous system in physiology and in pathology requires that each of these steps be taken into account: detection at the periphery, integration at the spinal cord level and processing in supraspinal structures. During my PhD, I focused on a subclass of peripheral somatosensory neurons, the C-low threshold mechanoreceptors (C-LTMRs) and the first level of integration of peripheral signals coming from these neurons in the spinal cord. Hence in this introduction I will give an overview of our current knowledge of cutaneous primary sensory neurons with a focus on C-LTMRs, their functions and the very basic understanding we have of their spinal cord circuitry.

I - Peripheral sensory neurons

Anatomy of somatosensory neurons

Somatosensory neurons or peripheral/primary sensory neurons (PSNs) are pseudo- unipolar neurons: a single process emerging from their cell body splits into a peripheral and a central branch, and they lack dendrites. The peripheral branch innervates specific peripheral organs such as skin, muscles and viscera, where it can detect a wide range of stimuli. The area of the skin innervated by a cutaneous PSN is called its receptive field, the size of it varies depending on the type of neuron considered.

The detection of a stimulus generates an action potential (AP) which travels antidromically towards the cell body, located in an organ called dorsal root ganglion (DRG). DRGs are distributed by pairs on each side of the spinal cord along the rostro-caudal axis (Figure 2). They are vascularized organs constituted of the cell body of PSNs, glial cells and macrophages, creating a particular micro-environment itself the subject of study.

Figure 2: Representation of DRGs localized all along the spinal cord

The AP then travels in an orthodromic manner along the central branch of the axon toward the spinal cord (Figure 3) where the release of glutamate transmits the signal to first-order spinal cord interneurons. Of note, all mouse PSNs express the vesicular glutamate transporter 2 gene (Vglut2) whereas only myelinated LTMRs express Vglut1 and only C-LTMRs express Vglut3 (Usoskin et al., 2015).

Figure 3: Drawing of a DRG with an action potential (AP) traveling along the axon

The somatotopic columnar organization of the central projections of PSNs into the dorsal horn of the spinal cord (Abraira and Ginty, 2013) is reminiscent of the somatotopy found in the somatosensory cortex. Indeed, it has been shown that neighboring neurons in terms of their receptive field in the skin overlap extensively in the spinal cord (Olson et al., 2017) (Figure 4).

Figure 4: Adapted from (Abraira and Ginty, 2013), LTMRs innervating the same patch of hairy skin and the columnar organization of their central afferences targeting different laminae of the dorsal horn of the spinal cord in a somatotopical manner, C-LTMRs in red.

Developmental origin of somatosensory neurons in the mouse Neural crest cell progenitors delaminate from the neural tube from embryonic day 9 to 10.5 (E9-E10.5) and start forming the DRGs. At this step, the cells start exiting the cell cycle and losing their expression of the multipotent marker sox10 (Kim et al., 2003). This process ends around E13.5. The post-mitotic neurons subdivide into two groups as they start expressing one of two basic helix-loop-helix transcription factors : Neurogenin1 and 2 (Ngn1 and Ngn2) (Ma et al., 1999), as well as Pou4f1 and Isl1 in both cases (Figure 5). The next step is a subdifferentiation into 4 groups depending on the expression of four different neurotrophic factor receptors, that are still routinely used to categorize PSNs in adult: Ntrk1 (TrkA), Ntrk2 (TrkB), Ntrk3 (TrkC) and Ret (Figure 5). These receptors bind respectively Nerve Growth Factor (NGF), Brain-Derived Growth Factor (BDNF) and Neurotrophin-4 (NTF4), Neurotrophin-3 (NTF3), Glial cell line-derived Neurotrophic Factor (GDNF). The proper expression of each of these receptors and the corresponding neurotrophins is crucial for the survival of DRG neurons during development (Crowley et al., 1994; Moqrich et al., 2004) and for the proper innervation of their central (Patel et al., 2000) and peripheral targets (Gorokhova et al., 2014).

At E11.5, the first subgroup co-expresses TrkB, Ret and one of Ret’s co-receptors Gfrα2, they are called “early” Ret neurons (Luo et al., 2009). The second subgroup co-expresses TrkC and Runx3, one of two Runt-related transcription factors of high importance for the correct specification of a number of DRG neurons (Kramer et al., 2006). The third subgroup expresses TrkB and the last expresses TrkA and Runx1 (Kramer et al., 2006). All these subpopulations continue expressing Pou4f1 and Isl1, which is necessary for proper expression of neurotrophin receptors (Dykes et al., 2011). The populations expressing TrkB, TrkC and/or Ret mostly originate from the first wave of Ngn2 expression. They also need expression of Maf (Wende et al., 2012) and will give rise to myelinated LTMRs and proprioceptors. Conversely, the neurons expressing TrkA and Runx1 at this stage derive from the Ngn1 wave of neurogenesis and will give rise to small-diameter neurons such as nociceptors, pruriceptors and C-LTMRs (Marmigere et al., 2006; Moqrich et al., 2004). By E17.5 TrkA positive neurons again subdivide: about 50% downregulate Runx1 but keep expressing TrkA and Met (Gascon et al., 2010). These will become peptidergic nociceptors. In the other 50% of neurons, Runx1 represses the expression of Met (Gascon et al., 2010), TrkA expression decreases and they start expressing Ret. These neurons will become non-peptidergic “late” Ret positive neurons including mechano-nociceptors expressing MrgprD and C-LTMRs (Figure 5).

Figure 5: Adapted from (Olson et al., 2016), illustration of the molecular programs involved in the specification of mammalian cutaneous Low-Threshold MechanoReceptors, with C-LTMRs outlined in orange.

Somatosensory neuron populations

PSNs are highly heterogeneous and can be classified a number of ways: based on their molecular profile, anatomy, physiology, the modalities they detect… Cutaneous somatosensation ranges from innocuous to noxious sensations. That includes, but is not limited to, light touch, vibration, stretching, indentation of the skin on the innocuous side of the spectrum, and acute pain such as noxious cold, heat, chemical, mechanical stimulation but also chronic pain on the noxious side of the spectrum. Chronic pain itself can be subdivided into neuropathic pain, where neurons are affected, or inflammatory pain, where the tissue surrounding neurons is injured and leads to an intense excitation of peripheral neurons. The vast heterogeneity of PSNs is the reason we can sense such varied stimuli. And it is important to take into account that one stimulus rarely activates only one subtype of PSN but the interplay of the activation of various PSNs is what leads to a specific sensation.

I will next describe PSNs based on their anatomical features and explain later how the simultaneous activation of such different subclasses of neurons can be integrated at the level of the spinal cord.

In terms of their anatomy we can distinguish PSNs based on the diameter of their axon and cell body, on their peripheral terminals in hairy or glabrous skin and on the location of their central projection in the dorsal horn (Abraira and Ginty, 2013).

Aβ fibers Aβ neurons have the largest axon diameter and cell body, about 30 micrometers for the latter. Their fiber is heavily myelinated, which allows the fastest conduction velocity among PSNs between 16 and 100 m/s. They all transduce light mechanical stimulation, an innocuous type of information on the innocuous-noxious spectrum, as such they are all called Aβ low- threshold mechanoreceptors (Aβ LTMRs). But they can be subdivided based on the type of

16 end organ they associate with, in the hairy or the glabrous skin, and their electrophysiological properties: whether they adapt slowly or rapidly to stimulation and their firing rate. In the hairy skin they associate with hair follicles to sense movement and pressure whereas in the glabrous skin, they associate with non-neuronal end organs necessary for the detection of the stimulus they transduce, such as Merkel cells. These particular non-neuronal organs are however also present in the hairy skin but in a more concentrated fashion around the base of the hair follicle. The adaptation rate of an Aβ fiber correlates with the type of stimulus they can sense: slow adaptation can be associated with an increased sensibility to a static stimulus, as opposed to rapidly-adapting Aβ-LTMRs more adapted to the detection of moving stimuli.

Aβ SAI-LTMRs Some Aβ fibers innervate end organs called Merkel cells present in both hairy and glabrous skin, they respond to indentation, adapt slowly, spike irregularly and convey a high spatial resolution. They are called slowly-adapting type I Aβ fibers or SAI-LTMRs (Figure 6).

Aβ SAII-LTMRs Another subpopulation of Aβ fibers is associated to a different end organ: Ruffini corpuscles. They are less sensitive to indentation than SAI-LTMRs but more sensitive to stretching of the skin, they are also slow-adapting but spike more regularly. They are called slowly-adapting type II Aβ fibers or SAII-LTMRs (Figure 6).

Aβ RA-LTMRs Where slowly-adapting neurons spike during the entire stimulation, rapidly-adapting neurons spike only at the beginning and at the end of the stimulation. We can distinguish rapidly- adapting type I from type II Aβ LTMRs. Both respond to vibration and are associated with end organs: low-frequency vibration like a flutter for RAI-LTMRs associated with Meissner corpuscles and higher-frequency vibrations for RAII-LTMRs associated with Pacinian corpuscles, which are extremely sensitive (in the nanometer range) but offer a low spatial resolution (Figure 6).

Hairy skin Aβ LTMRs Finally, some Aβ-LTMRs are specific to hairy skin due to their peripheral terminals innervating specifically hair follicles. These neurons can again be subdivided into slowly and rapidly- adapting Aβ LTMRs. Hairy Aβ SA-LTMRs associated with Merkel cell complexes or touch domes are in close proximity to a subtype of hair follicles, guard hairs which represent 1% of total

17 hair types in the mouse (Li et al., 2011). Hairy Aβ RA-LTMRs form longitudinal lanceolate endings around guard hair which allow their activation by hair follicle deflection (Figure 6).

Figure 6: From (Abraira and Ginty, 2013), schematic representation of hairy and glabrous skin innervation by LTMRs, C-LTMRs in red.

Central projections of Aβ fibers Aβ LTMRs are known to project to laminae III and IV of the deep dorsal horn of the spinal cord, though a recent study by Abraira et al shows that they actually also innervate part of the innermost lamina II. Their central projections present certain specificities as compared to those of other PSNs.

1- Upon exiting DRGs, Aβ-LTMRs axons extend their central branches in both rostral and caudal directions and then enter the spinal cord at several different levels, as opposed to other PSNs central branches entering the spinal cord at one level close to their DRG of origin (Abraira and Ginty, 2013). 2- Among LTMRs, Aβ-LTMRs central branches are the only ones synapsing directly onto projection neurons on some occasions, forming “direct pathways”, whereas information coming from other PSNs typically has to go through a network of spinal interneurons before it can be transmitted to projection neurons(Abraira et al., 2017). Projection neurons receiving inputs from Aβ-LTMRs are either part of the Post- Synaptic Dorsal Column pathway (PSDC) or the SpinoCervical Tract (SCT). The PSDC pathway receives direct excitatory inputs from Aβ-LTMRs but also indirect information

first processed by local interneuron network. The PSDC and SCT projection neurons are located in laminae III to V. 3- Aβ-LTMRs not only play a role in innocuous touch but, through the integration of their input in the spinal cord, they are also involved in the modulation of the pain signal. The working model for the modulation of nociceptive information in the spinal cord by LTMR input is called the gate control theory (GCT) first described by Melzack and Wall in 1965, which I will discuss later on (Melzack and Wall, 1965).

Aδ fibers Aδ neurons have an axon diameter and cell body of intermediate size, about 20 micrometers. Their fiber is lightly myelinated, hence their conduction velocity is slower than Aβ LTMRs, between 5 and 30 m/s. Aδ neurons include a population involved in the transduction of noxious mechanical or thermal stimuli and a population that transduces an innocuous signal, light mechanical touch (Abraira and Ginty, 2013).

Aδ nociceptors The first population is called Aδ nociceptors due to them transducing noxious stimuli. More specifically they can be A-mechano-heat receptors (A-MH), responding both to noxious thermal and mechanical stimuli, or high threshold mechanoreceptors, HTMRs responding specifically to high intensity mechanical stimuli (Roberts and Elardo, 1985). Sensitivity to noxious cold has also been observed (Simone and Kajander, 1997). Nociceptors innervate both hairy and glabrous skin, a feature crucial for individuals to be able to sense potentially damaging stimuli anywhere on the skin and react accordingly.

Central projections of Aδ nociceptors Aδ nociceptors transduce sharp, acute pain leading to a reflex response, from their peripheral free nerve endings in the skin to spinal cord interneurons located in the most superficial lamina of the dorsal horn and to lamina V. The projection neurons relaying noxious heat or mechanical stimuli are part of the anterolateral tract system, their cell bodies are located in lamina I and III to V. Because this information needs to be transmitted quickly to higher structures, Aδ nociceptors sometimes synapse directly onto relay neurons. However, it can be submitted to local modulation in the dorsal horn, be it from descending pathways, inhibitory and excitatory local interneurons or input from other PSNs.

Aδ-LTMRs The second population of Aδ neurons are called Aδ-LTMRs. They were first found to innervate D-hairs in the cat and as such they are sometimes called D-hair afferents (Brown and Iggo, 1967). These are the most sensitive mechanoreceptors in the hairy skin of many species, although their existence in human is still unproven. In terms of their electrophysiological signature, they are rapidly-adapting and discharge at the beginning and at the end of a mechanical stimulation as light as 0.07mN. Interestingly, they exhibit direction sensitivity due to their terminals being concentrated on one side of the hair follicle (Rutlin et al., 2014) and also respond to rapid cooling of the skin they innervate (Li et al., 2011). In the periphery, their terminals form longitudinal lanceolate endings around the base of both zigzag and awl/auchen hair follicles, hair types that represent respectively 76% and 23% of total hairs in the mouse (Li et al., 2011) (Figure 6). Recent findings extended their territory of innervation by showing that hind paw ventral skin, the skin of the palm usually considered as glabrous skin, actually presented hair follicles in some rodent species along with “glabrous D-hair afferents” innervating them. These terminals showed a similar asymmetrical innervation of the follicle in a horseshoe shape lending them the same directional sensitivity to hair deflection than “hairy D-hair afferents” (Walcher et al., 2018).

Central projections of Aδ-LTMRs Aδ-LTMRs central projections innervate quite largely the innermost lamina II layer as well as lamina III. The latter lamina also receives information from Aβ-LTMRs, but their conduction being faster than that of Aδ-LTMRs, the integration in one sensory column of the dorsal horn of the different inputs coming from a similar region of the skin includes a temporal component. Inputs coming from Aδ-LTMRs, just as in the case of most Aβ-LTMRs, are first processed by a dense network of local spinal interneurons before the information is transmitted to projection neurons of the PSDC. And, again similarly to Aβ-LTMRs, Aδ-LTMR activation plays a role in the modulation of the nociceptive input at the level of the spinal cord.

C fibers C fibers have the smallest axon diameter and cell body among PSNs, their fibers are unmyelinated and as such exhibit the lowest conduction velocity, around 1m/sec. Like Aδ neurons, they are highly heterogeneous in terms of the modalities they transduce ranging

20 from innocuous light touch to burning pain. C fibers involved in gentle touch are called C-low threshold mechanoreceptors (C-LTMRs) and I will address in detail their specificities in part 4 of my introduction. The rest of C fibers are free nerve ending C nociceptors and can be subdivided into two categories based on their molecular profile. More specifically one subclass expresses and releases peptides when activated, those are peptidergic C fibers often referred to as Trpv1 or CGRP positive fibers. By opposition the other class of neurons is called non peptidergic C nociceptors, characterized by their expression of Mas-related G-protein coupled receptor member D (MrgprD). A recent study showed one notable difference between nociception induced by activation of peptidergic or non-peptidergic nociceptors (Beaudry et al., 2017). Optogenetic activation of Trpv1 positive fibers induces paw withdrawal and licking, hinting at a consciousness of pain, whereas activation of MrgprD positive fibers induces “only” aversive behavior such as paw lifting. The authors further show that conditioned place aversion is induced by Trpv1 positive fiber activation but not MrgprD positive fiber activation. This example of the subtle differences between fibers that seem rather identical in the modality they transduce goes to show how important it is to study the overall behavior induced by the activation of a subpopulation of PSNs. Indeed, differences in the activation of these two types of C-fiber free-nerve ending nociceptors are probably due not only to particularities of their afferences but also to the processing happening in the spinal cord and in higher structures.

Peptidergic C nociceptors Most C fiber nociceptors are polymodal according to electrophysiological studies, they can transduce chemical, mechanical and thermal noxious stimuli (McCoy et al., 2013). Interestingly, some strictly mechano- or thermo-nociceptors actually become sensitive to other noxious modalities following sensitization (Schmidt et al., 1995). PSNs being pseudo- unipolar makes them capable of receiving and sending information from both peripheral and central endings. And for example, phenomenon of central sensitization could lead to higher sensibility to stimulation in the periphery. Peptidergic C nociceptors express neuropeptides such as substance P and CGRP. These substances can be released after stimulation of nociceptors and lead to the creation of an “inflammatory soup” composed also of cytokines and chemokines released by keratinocytes and immune cells at the site of injury. Neurons exposed to the inflammatory soup and expressing receptors to these molecules will see their excitability enhanced. For that reason, when studying peripheral cutaneous activation of

PSNs, it is important to take into account the surrounding tissues and cell types involved, which always includes the immune system but also the skin and the vascular system (Fromy et al., 2018; Marics et al., 2014).

Non-peptidergic C nociceptors Non-peptidergic C nociceptors are highly heterogeneous in terms of molecular and functional identity, they are characterized by their expression of the Ret neurotrophin receptor and various members of the Mas-related genes (Mrgs) family which makes them sensitive to noxious mechanical stimulation and/or itch-inducing chemicals (Dong et al., 2001; Guan et al., 2010; Liu et al., 2009; Zylka et al., 2003). This includes MrgprD-positive mechano-nociceptors that bind isolectin IB4 and express the protein GINIP (Gaillard et al., 2014). Where peptidergic neurons innervate the basis of the epidermal layer of the skin, free nerve endings belonging to non-peptidergic neurons innervate the skin at a slightly more superficial level.

Central projections of C nociceptors Peptidergic C fibers target laminae I and II outer of the dorsal horn where spinal cord interneurons will relay the information to projection neurons of the anterolateral tract, though similarly to Aδ nociceptors, some peptidergic C nociceptors synapse directly onto projection neurons. Non-peptidergic C fibers send their projections slightly farther in the spinal cord in the outer lamina II, underneath lamina I peptidergic projections. Although most nociceptors are polymodal when studied in electrophysiological experiments, it has been shown that ablation of specific subpopulations of nociceptors in adulthood can lead to loss of a specific modality (Cavanaugh et al., 2009). This suggests that both PSNs and spinal networks are necessary for the discrimination of various noxious modalities transduced by the same group of neurons. Spinal networks are further involved not only in the discrimination but also in the selection of input coming from peripheral nociceptors.

C-LTMRs History of C-LTMRs

C-LTMRs were first discovered almost a century ago, in 1939 in Sweden by Zotterman in the cat (Zotterman, 1939). They were then discovered in other species including primates (Bessou et al., 1971; Douglas and Ritchie, 1957; Iggo and Kornhuber, 1977; Kumazawa and Perl, 1977a, b). But for a long time, they were thought to have disappeared along evolution and so did not exist in human. In 1988 Johansson showed their presence by microneurography experiments

22 on the human face around the eye (Johansson et al., 1988) and soon after in the arm and leg hairy skin (Vallbo et al., 1993; Vallbo et al., 1999). Microneurography is a technique that allows microstimulation and recording of skin single-unit fibers identified thanks to their electrophysiological profile, first developed in Sweden in 1965 by Karl-Erik Hagbarth and Åke Vallbo. It is extremely useful to study PSNs in human especially when combined with psychophysical assessment of the affective component of a stimulation (Vallbo, 2018).

Anatomy of C-LTMRs

C low-threshold mechanoreceptors or C-LTMRs are a unique class of C fibers. They have a small-diameter fiber and cell body and are unmyelinated. Their conduction velocity is in the same range as C nociceptors but they transduce pleasant mechanosensation such as light touch involved in social bonding. Their peripheral projections innervate specifically hairy skin and more precisely form longitudinal lanceolate endings around Zigzag and Awl/Auchen hair which represents 99% of mouse hair (Figure 6). In other words, in the mouse, these terminals target the same hair follicles as Aδ-LTMR longitudinal lanceolate endings and actually interdigitate with them. But one needs to remember that Aδ-LTMRs have not yet been proven to exist in human, possibly lending even more importance to C-LTMRs. On the central side, C- LTMRs send projections to the innermost part of lamina II (Figure 4). That is to say that when subdividing lamina II into a dorsal and a ventral part, C-LTMRs actually innervate the inner part of the ventral lamina II, one sublayer deeper than C non-peptidergic nociceptor projections but more superficially than Aδ-LTMRs. Hence in the dorsal horn, they are at the border between the integration of noxious and innocuous input. And they are involved in both modalities: directly when it comes to transducing light touch, and indirectly in the gate control theory model when it comes to processing noxious input (Cheng et al., 2017).

Molecular identity of C-LTMRs

As mentioned earlier, C-LTMRs originate from the second wave of Ret positive or “late” Ret neurons expressing Runx1 (Kramer et al., 2006; Lou et al., 2013; Lou et al., 2015) (Figure 5). Currently, their molecular markers in adult are Slc17a8 (Vglut3) (Seal et al., 2009), Tyrosine Hydroxylase (TH) (Li et al., 2011) and Tafa4 (fam19a4) (Delfini et al., 2013). Few studies have explored the mechanisms behind the acquisition of this molecular identity. But in 2013, Lou et al hypothesized that Runt domain transcription factor Runx1, known to coordinate the

23 development of a number of unmyelinated DRG neurons, played a role in the specification of C-LTMRs. Vglut3 expression in these neurons can be detected from neonatal stages onward (P4, (Lou et al., 2013)). In this publication, Lou et al show that conditional knock-out of Runx1 specifically in Vglut3 positif DRG neurons leads to a loss of expression of Vglut3 and TH in C- LTMRs, and a loss of the longitudinal lanceolate endings they usually form around the base of hair follicles supposedly instead taking the shape of circumferential endings. Hence, Runx1 seems necessary for C-LTMRs to express Vglut3 and TH, and for them to innervate properly their peripheral targets. This latter aspect of their specification seems to happen around P4- P7, stage where Runx1 expression disappeared in this conditional knock-out.

Moreover, the knock-out of Runx1 in C-LTMRs also leads to a strong reduction in mechanosensitivity of these neurons due to a down-regulation of the mechanically gated ion channel Piezo2, normally expressed in C-LTMRs at high levels. More specifically, when mechanically stimulating DRG neurons in culture and performing cell-patch recording, Lou et al find that small Vglut3-positive neurons (transient and persistent) respond with a mix of rapid, mixed and slowly adapting currents. And loss of expression of Piezo2 seems to lead to a switch from a majority of rapid and mixed adapting currents to slowly adapting currents and mostly non-responder neurons. In conclusion, Runx1 expression in prospective C-LTMRs is mainly responsible for the acquisition of their mechanoreceptor phenotype, through correct Piezo2 expression. Despite a significantly decreased mechanosensitivity in this model, the authors could not see any marked change in acute or chronic mechanical pain nor in the development of mechanical allodynia, in the behavioral tests performed which unfortunately were mostly on glabrous skin not innervated by C-LTMRs. It would be very interesting to perform more appropriate behavioral tests to assess the importance of piezo2-dependent C- LTMR mechanosensitivity in innocuous hairy skin mechanosensation and pain modulation.

In a follow-up publication on the molecular control of C-LTMR specification published in 2015, Lou et al show that the transcription factor Zfp521 is necessary for the proper segregation of prospective non-peptidergic cutaneous C-fibers into MrgprD positive neurons and C-LTMRs (Lou et al., 2015) (Figure 5). Both of these subpopulations maintain their expression of Runx1 in adult and as such are called Runx1-persistent but they show different molecular profiles and innervation patterns in the periphery and the spinal cord. Zfp521 expression by prospective C-LTMRs seems to be one of the reasons explaining this divergence during

24 differentiation. Zfp521 expression is quite restricted to C-LTMRs among DRG neurons. Lou et al interestingly describe that a conditional knock-out of Zfp521 in C-LTMRs leads to their expression of what they consider to be MrgprD positive molecular markers: MrgprD itself, mGluR5, TRPC3 and P2X3. Hence in the absence of Zfp521 expression in C-LTMRs, molecular markers of MrgprD neurons seem to be extended to C-LTMRs. Conversely, expression of C- LTMR molecular markers: Vglut3, TH and Tafa4, is significantly reduced in absence of Zfp521 expression, although, it is mentioned, not as strongly as in the conditional knockout of Runx1 in C-LTMRs. Interestingly, C-LTMR specific knockout of Zfp521 does not affect piezo2 expression nor does it affect skin innervation. In conclusion, Zfp521 controls partially molecular specification of C-LTMRs by repressing the expression of MrgprD positive neuronal markers and favoring the expression of Vglut3, TH and Tafa4. But Runx1 controls expression of Piezo2 by C-LTMRs and their peripheral innervation pattern through Zfp521-independent pathways.

C-LTMRs and MrgprB4-positive neurons

In 2013, Vrontou et al describe a newly-discovered population of somatosensory neurons expressing the Mrg family member MrgprB4 (Vrontou et al., 2013; Zylka et al., 2003). These neurons show a peripheral pattern of innervation restricted to hairy skin and functional properties similar to those of CT-afferents described in human, such as “massage-like” stroking-induced firing in vivo. Because contrary to Vglut3-positive C-LTMRs these neurons are not activated by mechanical stimulation ex vivo, they do not qualify as C-Low Threshold MechanoReceptors. Furthermore, similarly to MrgprD-positive non-peptidergic mechano- nociceptors, these Ret-expressing C-fibers bind IB4, do not associate with any specialized cutaneous sensory structures and their central terminals target outer lamina II, a more superficial region than that innervated by classically described C-LTMRs (Liu et al., 2007). Finally, whereas C-LTMRs respond to mechanical stimulations of a force as light as 1 mN and do not distinguish between smooth probe indentation and pinprick but respond equally well to both (Liljencrantz and Olausson, 2014), MrgprB4 neurons do not respond to pinprick and are activated by brush strokes within the 20-90 mN range (Vrontou et al., 2013).

Functions of C-LTMRs

Caress C-LTMRs are different from other LTMRs in many ways. Where Aβ-LTMRs see their firing rate increase with the speed of the stimulus, independent of the temperature, C-LTMRs firing rate peaks at 100 times per second for a warm stimulus at a very specific speed of 3 cm/sec in human, speed described as most pleasant, and their firing rate slows down whether this speed increases or decreases. Where myelinated LTMRs allow fine tactile discrimination in terms of form, texture and location of the stimulus in a rapid manner thanks to their high conduction velocity, C-LTMRs do not allow such discrimination. Indeed, in humans lacking large myelinated LTMRs, stimulation of C-LTMRs does not allow fine discrimination (localization, direction) of the stimulus and is not associated with activation of the somatosensory cortex (Bjornsdotter et al., 2009; Olausson et al., 2002).

It has been shown that the optimal speed, pressure and temperature to activate C-LTMRs correlate with pleasantness in human studies (Ackerley et al., 2014). More specifically, these fibers, known as C-tactile or CT afferents in human, are best activated when stimulated at a speed included between 1-10 cm/sec, with a pressure of 0.3-2.5 mN and at a temperature of 32°C, human skin temperature, rather than a cooler 18°C or a warmer 42°C. These are parameters best describing gentle human touch and caress.

The function of these neurons is incredibly interesting in many ways. It has been established that among LTMRs, C-LTMR activation, on the arm for example, is associated with a pleasant feeling, whereas the same stimulation of the palm, where C-LTMRs are absent, doesn't induce such an emotionally pleasant feeling but more of a sensory discriminative one (McGlone et al., 2012). But, one study showed that stimulation of the palm after stimulation of C-LTMRs of the arm actually increases the pleasantness of the second stimulation. This would imply that C-LTMRs, and the central network involved when they are activated, can turn an innocuous touch into a pleasant one (Loken et al., 2011). The notion of an impact of C-LTMR activation on a later activation of other LTMRs actually fits with the electrophysiological properties of C- LTMRs. More specifically, CT afferents adapt in an intermediate fashion to a stationary stimulus and can show after-discharges, that is they continue discharging after the end of the stimulation, for up to several seconds, unlike other LTMRs (Nordin, 1990). Another possible argument in favor of this lasting effect of C-LTMR activation is their specific expression of the

26 chemokine-like protein TAFA4, which has been studied, for now, under pain conditions showing its analgesic effect (Delfini et al., 2013) but still needs to be studied to assess a potential « feel-good » effect of its putative release by central terminals in the spinal cord after stimulation of C-LTMRs in the periphery.

Social bonding More than just a pleasant way to interact, stimulation of C-LTMRs seems like an effective way to gain someone’s attention and that from an early age according to studies performed in 9- months-old infants (Fairhurst et al., 2014). In this study, the authors showed that not only did light brushing at CT optimal rate, 3 cm/sec, and not 0.3 or 30, correlated with a decrease in the infant’s heart rate but it also correlated with an increased attentional engagement, that is duration of looks and gaze shift toward the caress. Furthermore, the infant’s response in terms of heart rate correlated with the caregiver’s own sensitivity to social touch. In another study investigating the role of CT activation in affiliative behavior, the authors showed that humans instinctively apply a stimulation optimal for CT activation when asked to stroke their baby or their partner, as opposed to applying a higher speed non-optimal for CT activation when asked to stroke an artificial arm (Croy et al., 2016). In conclusion, pleasant touch seems to play an important role in human social interactions at all age.

Temperature sensation C-low threshold mechanoreceptors are not only involved in detection of mechanical stimulation but also temperature decrease (Douglas et al., 1960; Iggo, 1960). This is observed in the robust response to rapid cooling, applied either in electrophysiological experiments on DRG cell culture, from 31°C to 15°C (Francois et al., 2015) or skin-nerve preparations (Zimmermann et al., 2009), both from 50°C to 30°C or 30°C to 2°C (Li et al., 2011). C-LTMRs are known to exhibit fatigue, that is a decreased firing rate after repeated mechanical stimulations, not to be confused with adaptation of the firing rate to a single sustained stimulus. Cooling of the skin seems to also produce fatigue in C-LTMRs and reduce sensitivity to mechanical touch, hence we can talk about cross-modal fatigue of C-LTMRs (Hahn, 1971). The only example of an in vivo role for C-LTMRs in temperature sensation was published by François et al, 2015 in Cell reports. In this publication, the conditional knockout of the voltage- gated calcium channel Cav3.2 in C-LTMRs lead to a preference for cooler temperatures in the

27 gradient test, among other phenotypes underlining the role of C-LTMRs in mechanical and thermal pain hypersensitivity.

Tickling and mechanical itch When they were first described by Zotterman in 1939, C-LTMRs were thought to transduce “very light touch” of hairy skin and the tickling - faint itching - associated with it. Later psychophysical tests by Liljencrantz and Olausson in 2014 on neuronopathy patients lacking Aβ-LTMRs showed that when soft brush stroking was applied to the hairy skin of the arm, patients could feel pleasant touch, which they could not when the same stimulation was applied to the palm. This feeling however was not associated with any painful, tickling or itching sensation (Liljencrantz and Olausson, 2014). In parallel, two studies by (Bourane et al., 2015; Fukuoka et al., 2013) hypothesize that C-LTMRs mediate mechanically-evoked itch, as opposed to chemical itch which is associated with a burning - stinging sensation. Fukuoka’s hypothesis is based on loose correlations between the characteristics of mechanical itch and activation of C-LTMRs such as intermediate adaptation, fatigue and after-discharge. The study by Bourane and colleagues however shows more tangible proof of an anatomical pathway linking C-LTMRs and NPY-expressing inhibitory spinal cord interneurons, involved in the gating of mechanical itch behavior in the mouse. In conclusion, the role of C-LTMRs in itch still needs to be investigated.

C-LTMRs and pain modulation

Pain definition Pain is the conscious and unpleasant feeling associated with the activation of nociceptors in the periphery (nociception), it is both a sensory and emotional experience. The role of transient acute pain is to prevent or limit tissue injury, as such it is critical for our survival. But pain can become chronic: even though the injury has healed the person will continue to suffer, in which case pain has no beneficial protective value anymore. Chronic pain of various origins now affects over 1 in 5 people in the world and currently available pain treatment does not always provide efficient pain relief and/or with serious side-effects. Dysfunction at each level of the circuitry (periphery, spinal cord, brain) can give rise to chronic pain in the form of mechanical and/or thermal hypersensitivity. Thus, it is important to investigate each aspect of pain and for that to use different pain models. Depending if the nerve or the tissue surrounding the nerve is injured we will talk about neuropathic or inflammatory pain

28 respectively. The two words most used when describing pain in research are allodynia and hyperalgesia. Both refer to a heightened sensitivity of the sensory nervous system to thermal or mechanical stimulation. But allodynia is a painful response to a stimulation that would normally be perceived as innocuous in a healthy individual, whereas hyperalgesia is an increased/extreme response to a stimulation that would also be painful for a healthy individual but less so.

Pain models C-LTMRs have been shown to play a role in the modulation of pain in the context of both neuropathic and inflammatory pain (Delfini et al., 2013; Francois et al., 2015; Urien et al., 2017) and I will now describe the mouse pain models used in these studies and mine.

1-Neuropathic pain models CCI Chronic Constriction Injury (CCI) is a model of unilateral peripheral mononeuropathy combining nerve compression, Wallerian degeneration and epineurial inflammation. It consists in three loose ligatures of the common sciatic nerve (Figure 7) which leads to the development of mechanical allodynia (Bennett and Xie, 1988) due to inflammation and swelling. In this model, severe reduction of large myelinated fibers had been observed as well as loss of a number of unmyelinated neurons, though a number of these survive and are responsible for the transmission of the pain signal. This model is relevant to assess the role and the pattern of expression of a number of molecules following CCI, or the role of a population of neurons in the establishment and recovery from pain, with return to baseline levels of mechanical sensitivity after 30 days.

Figure 7: From (Ueda, 2006), Chronic Constriction Injury (CCI) and Spared Nerve Injury (SNI) mouse models of neuropathic pain.

SNI The Spared Nerve Injury (SNI) model was developed to study the genes and neuronal populations involved in persistent neuropathic pain characterized by spontaneous pain, allodynia and hyperalgesia in human (Seltzer et al., 1990), and to test the potential therapeutic effects of drugs on alleviation of persistent pain (Malmberg et al., 1997). Indeed, in this model, the ligation and transection of the tibial and common peroneal nerve (Figure 7) induces thermal and irreversible mechanical hypersensitivity. This model of neuropathic pain is more easily reproducible and presents less variability within cohorts than CCI, but is also more drastic in the irreversibility of mechanical pain. Single nerve fiber recordings from skin-nerve preparations of SNI-injured animals show that both Aδ and C mechanoreceptors, including but not limited to nociceptors, fire more following surgery (Smith et al., 2013). Cold hypersensitivity is a hallmark of neuropathic pain, which is nicely recapitulated by a heightened response to cooling of the skin of the injured hindpaw during two weeks following SNI, with response later coming back to baseline levels. We talk about allodynia in the case of acetone application, which in itself is not noxious, and hyperalgesia in the case of noxious stimulation by dry ice adapted from (Brenner et al., 2012). SNI-induced cold hypersensitivity

30 can be measured with acetone application to the ventral side of the injured hindpaw, this has a cooling effect, or with the cold plantar test. When the latter is performed, mice are allowed to settle down in plastic enclosures on a glass plate and dry ice is applied to the glass underneath the center of the injured hindpaw.

In both models of neuropathic pain, mechanical sensitivity is measured by stimulation of the plantar hindpaw ipsilateral to the surgery with Von Frey filaments, made up of nylon with varying degrees of thickness and so varying degrees of rigidity which correspond to a certain force of application usually represented in grams. Their application and the measure of the response of the mouse, ranging from removal to biting of the paw, allows to determine first a baseline mechanical sensitivity before surgery and then a follow-up of the establishment and maintenance of pain and sometimes recovery from it with a return to baseline levels.

2-Inflammatory pain models Carrageenan The injection of the inflammatory irritant carrageenan in the plantar side of the hindpaw of the mouse leads to an initial acute inflammation followed by a chronic one inducing both thermal and mechanical hypersensitivity during two weeks. Similarly to the neuropathic pain models cited above, mechanical hypersensitivity is measured by application of Von Frey filaments to the site of injection. Thermal hypersensitivity is measured with the Hargreaves test where an infrared heat stimulus is applied to the glabrous skin of the injected hindpaw and baseline sensitivity is compared to sensitivity after Carrageenan injection with the contralateral paw used as control.

Formalin The formalin injection is the faster of these four pain models and also the only one where spontaneous pain is exhibited by the animal which reduces experimenter-dependent variability. Injection of formalin in the ventral side of the hindpaw induces immediate nocifensive behavior in the animal such as lifting, guarding, shaking, biting and licking of the paw (Figure 8). Where lifting and guarding behavior may not require the nociceptive information to be processed by the brain and could depend on a local spinal cord reflex, biting and licking implies that the animal holds its injured paw with the forepaws and that suggests that the nociceptive information has been processed and integrated all the way to the brain. There are several subtleties to the formalin pain test. It was originally designed as a subcutaneous injection in the dorsal part of the mouse hindpaw of 10 μl 1% diluted formalin 31

(Hunskaar and Hole, 1987). Since then, researchers have mostly been injecting the formalin underneath the ventral skin of the hindpaw and at various dilutions. Though the most frequent dilution is the one we use of 2% which induces the two typical phases of nociception, some publications mention rather extreme modalities of injection in mouse of volumes up to 40 μl and of dilution up to 5%.

Figure 8: Pictures of mice following hindpaw intraplantar formalin injection.

Prior to the injection, the animal is habituated to the settings of the test for 30 minutes, usually a Plexiglas cylinder on a table so the animal can be observed from all angles. The animal starts shaking and licking its injured paw within seconds of the injection and both the latency of the first response and the duration of the response per periods of 5 minutes are taken into account. The early phase is one of acute inflammatory pain and is supposed to last 5 minutes. In the literature it is described as the direct result of the activation of primary sensory fibers of the paw by the injection of the formalin and by the inflammation produced by it (the paw very rapidly reddens and swells up). Fibers including myelinated LTMRs and Aδ and C nociceptors have been shown as activated. Then during 5 to 10 minutes the pain seems to subside and the animal explore its environment again. This is called the interphase and would be due to a desensitization of the primary afferent fibers directly affected by the formalin and to a descending modulation from higher structures onto the spinal network (Franklin and Abbott, 1993; Matthies and Franklin, 1992). One could hypothesize that it is potentially due to the release of inhibitory neurotransmitters such as GABA in the spinal cord after such an intense activation of various populations of PSNs including nociceptors.

The onset of the second phase corresponds to the reappearance of signs of nocifensive behavior which may start off milder than actual biting and shaking with signs of guarding and lifting of the paw off the surface of the table. Globally speaking the mouse starts walking more gingerly on its injured paw than on the control one again. The second phase has been observed to last from 20-30 minutes to 45 depend on the experimenter and the dosage. Commonly, the mouse is observed for an hour with the first phase including the first 10 minutes and the second phase the next 50 minutes. The time course of the nocifensive behavior is usually represented per periods of 5 minutes and as cumulated time for each phase. The cellular and molecular mechanisms underlying the second phase have been for long the subject of debate. Initially thought to be mostly the result of the sensitization of spinal cord networks due to the initial intense nociceptive stimulus, several subsequent studies showed that : 1- ablation of certain PSN subpopulations affects more or less strongly the first and the second phase, hence peripheral activation is still at play during the second phase, I will detail these studies in the next paragraph; 2- there is evidence of an important role played by the release of pro- inflammatory molecules in the periphery during the second phase such as bradykinin (Correa and Calixto, 1993), Calcitonin Gene-Related Peptide and substance P (Rogoz et al., 2014)... and many others like prostaglandin, histamine are involved in formalin pain though it is not clear if they act at a peripheral and/or central level due to the recurrent use of total knockouts to study genes with a broad pattern of expression and drugs that cross the blood-brain barrier. Moreover, formalin pain is composed of a sensory-discriminative aspect but also of an affective component. Identifying the cellular and molecular pathways involved in each of these two components requires the design of specific behavioral tests such as the conditioned place aversion test (Johansen et al., 2001).

One of the interesting aspects of the formalin pain paradigm is the ability to test the antinociceptive effect of number of drugs: opioids, channel blockers, GABA and glutamate signaling modulators, on acute inflammation mechanisms in the periphery and central sensitization both at the spinal cord and brain level. Again, the downside is the difficulty to identify the targets of the drug of interest. For example, the animal is typically observed for an hour after injection which gives enough time for the formalin to diffuse within the paw and affect different fibers in the paw from the ventral to the lateral and dorsal side at different concentrations. Indeed, although it has never been quantified, there are clues suggesting that

33 the whole limb could actually be affected with the animal not only licking the paw but also most of the hindlimb at peak nocifensive behavior. As a consequence, a drug which modulates nociception only during the second phase could either act on the spinal sensitization/descending control aspect of the pain which comes into play after the initial high-intensity nociceptive input. And/or the drug could act on primary afferents which were only reached by the formalin “wave” after the end of the first phase, or on primary afferents close to the injection site, desensitized by the formalin at first, but which start firing again later on. When it comes to the central aspect of formalin pain modulation, several pathways come into play which involve both excitatory and inhibitory cells most of them expressing receptors to glutamate and/or GABA. Although several regions of the brain and laminae of the spinal cord have been shown as activated by formalin (Tan et al., 2017; Wei et al., 2006), our knowledge of the exact populations of neurons involved and the different subunits of glutamate and GABA receptor they express is still lacking. Hence we must be cautious not to over-interpret results involving the antinociceptive effects, or lack thereof, of a drug on global formalin-induced pain behavior.

Role of C-LTMRs in pain modulation: building blocks from various studies Apart from a direct activation by mechanical or thermal stimulation, C-LTMRs can also play a role in pain modulation in different neuropathic and chemical pain models both in mouse (Delfini et al., 2013; Francois et al., 2015; Urien et al., 2017) and human (Liljencrantz and Olausson, 2014; Nagi et al., 2011).

First of all, their central projections in innermost lamina II colocalize with spinal cord interneurons expressing the gamma isoform of protein kinase C (PKCγ), thought to be involved in neuropathic pain (Malmberg et al., 1997). PKCγ knock-out mice show normal acute pain responses but reduced thermal hypersensitivity and no mechanical hypersensitivity following sciatic nerve ligation, a model of neuropathic pain. Moreover, these mice showed a decreased nociceptive response to formalin injection in the hindpaw, a model of inflammatory pain. A possible conclusion of this study is that PKCγ spinal cord interneurons are necessary for nerve injury-derived thermal and mechanical hypersensitivity, in the case of sciatic nerve ligation, and for the prolonged phase of inflammatory pain due to formalin injection, but not for acute pain. One caveat of this study is the interpretation of the results considering that this is a total knock-out of PKCγ which is not expressed only in this subpopulation of spinal cord excitatory

34 interneurons but also in cortical excitatory neurons (Liu et al., 2018; Saito and Shirai, 2002). Moreover, its expression during development may be more widespread than in adult in which case a total knockout could affect the specification of a number of neurons, and indeed seems to affect motor coordination slightly (Malmberg et al., 1997; Saito and Shirai, 2002). These limitations in the interpretation of the results of a total knockout mouse model are a recurrent problem in many studies. They can be overcome by the generation of conditional knockout mice where not only spatial but also temporal control of the deletion of the gene can be achieved.

The next studies address more directly the role of C-LTMRs in acute mechanical sensitivity and mechanical and thermal pain modulation. To study the function of a neuronal population one can activate, silence or ablate it. If such tools are not available, one can also study the role of a gene that is selectively expressed in these neurons. If the gene’s pattern of expression is specific to these neurons, the potential phenotypes of a total knockout can be attributed to functional defects of this particular population. But if the gene’s pattern of expression is broader be it temporally or spatially, which is often the case, then a conditional knockout is required and needs to be carefully designed. At the beginning of my thesis, the molecular markers defining C-LTMRs were limited to 3 genes: TH, Vglut3 and Tafa4.

TH Tyrosine Hydroxylase is specific to a subset of C-LTMRs (Brumovsky et al., 2006). About 60% of Vglut3-positive neurons express TH in adult DRG neurons according to (Lou et al., 2013) but it is more broadly expressed in the DRG during development (Lagerstrom et al., 2010; Lindeberg et al., 2004) and in sympathetic neurons at much higher levels. TH is a rate-limiting enzyme normally responsible for the synthesis of L-DOPA using tyrosine as a substrate in the catecholaminergic synthesis pathway (Brumovsky, 2016). According to (Lou et al., 2013; Lou et al., 2015), its expression in C-LTMRs is Runx1 and Zfp521-dependent. The absence of expression of the dopamine transporter DAT1 (slc6a3) by C-LTMRs and the lack of TH expression in rat and human DRGs are probably the reasons why no knockout of the gene in C-LTMRs has ever been reported.

The case of Vglut3 knockouts Vglut3 is one of three glutamate transporters (Vglut 1, 2 and 3 also known as slc17a7, slc17a6 and slc17a8 respectively) expressed in the nervous system. As mentioned earlier, their pattern

35 of expression in the DRG can be used to identify different subtypes of fibers with Vglut2 displaying the broadest pattern of expression at various levels in basically all DRG neurons, Vglut1 essentially expressed by myelinated LTMRs and Vglut3 solely found in C-LTMRs in adult. Now, as I alluded to earlier, the pattern of expression of a gene should be thoroughly assessed before interpreting the consequences of a knockout. The study of the Vglut3 gene is an example of over-interpretation of a global knockout.

(Seal et al., 2009) The story starts in 2009 with a publication in Nature about the role of Vglut3 in pain modulation, with the use of a total knockout. Seal et al announce in a much talked about publication that “Injury-induced mechanical hypersensitivity requires C-low threshold mechanoreceptors”. In this publication, mice lacking Vglut3 in their entire nervous system show a defect in acute noxious mechanosensation and fail to develop mechanical hypersensitivity in the context of : the inflammatory model of carrageenan injection, capsaicin injection, and the neuropathic pain models of spared nerve injury and Brennan postoperative pain (a model where an incision in the plantar side of the paw leads to mechanical hypersensitivity, allowing to assess mechanical allodynia development and recovery).

At that time, Vglut3 was considered to be specific to C-LTMRs by Seal et al. Its expression in adult is specific to DRG as shown by the loss of Vglut3 staining in the adult spinal cord following rhizotomy; and an absence of detectable colocalization with Vglut1 or Vglut2 in C-LTMR central terminals led Seal et al to believe that in a Vglut3 knockout context, C-LTMRs can no longer transmit any glutamatergic signal to the spinal cord. This publication was published 5 years before the Tafa4 study and it would be interesting to see if TAFA4 is still released by C- LTMRs in this context, assuming that C-LTMRs release TAFA4 at the central level which has not been proven yet. Moreover, knowing that a light expression of Vglut2 does exist in C-LTMRs, it would be relevant to check the consequence of their stimulation on inner lamina II spinal networks, in a skin-nerve setting for example, to assess whether in a Vglut3 knockout context C-LTMRs truly can no longer release glutamate.

Because of the complete lack of mechanical allodynia in both inflammatory and neuropathic pain, Seal et al conclude that Vglut3-positive C-LTMRs are the neurons responsible for the transmission of a painful signal in response to innocuous stimuli in these contexts and show a change in the sensation conveyed, from pleasant to painful, after injury.

(Lou et al., 2013) In 2013, Qiufu Ma’s lab in Harvard published a publication, which I already mentioned earlier, where they make use of the expression of Vglut3 in C-LTMRs to conditionally knockout the master transcription factor Runx1. They describe this transcription factor as necessary for the correct acquisition of their peripheral innervation pattern and molecular identity by C-LTMRs at neonatal/early postnatal stages. In this publication the authors use a reporter mouse, that is a mouse expressing cre recombinase under the control of the Vglut3 promoter crossed with a mouse carrying a lox-stop-lox cassette followed by a cassette encoding the red fluorescent molecule Tomato, where any neuron that has ever expressed Vglut3 will be red. In consequence, they had to perform a thorough analysis of the expression of Vglut3 both in spinal cord and DRG neurons during development and in adult to determine the identity of tomato-positive neurons in adult reporter mice.

They found that 15% of tomato neurons in the adult DRG had a large or medium diameter cell body and myelinated fiber, whereas C-LTMRs are characterized by their small cell body size and unmyelinated fiber. These neurons did not show adult expression of Vglut3, only at neonatal stages around P0.5 when Vglut3 expression is initiated, hence they were called Vglut3-transient DRG neurons. Even though it was not assessed in this study, transient expression of Vglut3 in these myelinated neurons may play a role in their differentiation/acquisition of correct features and a total knockout of Vglut3 may affect them. Moreover, they found that only 60% of small Vglut3-persistent express TH. These neurons form longitudinal lanceolate endings around hair follicles in the hairy skin similarly to what was shown by (Li et al., 2011). But Lou and colleagues also mention other Tomato-positive endings in both hairy and glabrous skin, some unmyelinated free-nerve endings in close juxtaposition with hair follicles, some unmyelinated fibers in the thick glabrous skin associated with Meissner corpuscles and some myelinated fibers associated with Merkel cells, themselves expressing Vglut3. Now stainings published in the publication do not show which of these endings would be persistent or transient Vglut3-expressors, but Lou and colleagues hypothesize that free-nerve endings are small Vglut3-persistent neurons and that myelinated fibers are large and medium Vglut3-transient neurons. In conclusion, this study is the first one to establish a broad expression pattern for Vglut3 in DRG, be it transient or persistent, one much larger than assumed in the Seal publication.

The second finding of the study is transient expression of Vglut3 by spinal cord neurons around P4-P7 which has disappeared by P56 in adult. Although the authors do not further analyze the future identity of these neurons, it is well known that maturation is an ongoing process at postnatal stages in spinal cord circuits and one can assume that, just as for Vglut3-transient DRG neurons, Vglut3-transient spinal neurons may present defects in their fate acquisition in a total Vglut3 knockout.

(Draxler et al., 2014) Mentioning the Lou et al study, Draxler and colleagues publish in 2014 a publication where they assess both the role of Vglut3 using the total knockout and the role of Vglut3-positive afferents through their stimulation via channelrhodopsin activation, in thermal and mechanical hypersensitivity in various pain paradigms. The pain modalities assessed are slightly different from those of the Seal publication: both mechanical and thermal noxious stimulation in the context of Carrageenan for inflammatory pain, Chronic Constriction Injury and oxaliplatin treatment for neuropathic pain. Carrageenan injection leads to a strong increase in mechanical hypersensitivity in control animals, slightly less so in Vglut3 knockouts. Interestingly, knockout animals in the Seal publication did not develop any mechanical hypersensitivity and had a much lower mechanical baseline, even though both teams used the same mouse line and supposedly the same method to measure mechanical sensitivity (up/down stimulation). In the case of CCI-induced neuropathic pain, knockout of Vglut3 does not affect the development of thermal or mechanical pain. Finally, in the oxaliplatin model of neuropathy, which recapitulates the nerve damage usually found in chemotherapy patients treated with the drug, Vglut3 KO mice developed reduced thermal and mechanical hypersensitivity, as compared to control mice. Just as in the case of the Seal and colleagues’ study, it is impossible to determine the actual contribution of C-LTMRs among other Vglut3- positive neurons, transient or persistent, in the development of mechanical or cold hypersensitivity.

The other more innovative aspect of the study involves the stimulation of Vglut3-positive fibers in the periphery: the authors use a mouse expressing Channelrhodopsin in all cells expressing or having ever expressed Vglut3, that is to say C-LTMRs, free nerve ending TH- negative C-fibers and A-fibers associated to Merkel cells in the DRG, and transiently expressing spinal cord interneurons. By applying the blue light activating channelrhodopsin directly to the

38 plantar hindpaw of the mouse, the authors avoid activating the spinal Vglut3-positive subpopulation of neurons. But they still activate at least three different fiber types and considering that the light is applied to the plantar glabrous skin of the hindpaw, it is difficult to estimate the proportion of hairy-skin innervating C-LTMRs they activate. In this paradigm, Vglut3-positive peripheral afferent stimulation in naive mice only gives rise to minor reactions such as glancing toward the paw or weak withdrawal. In a CCI-induced neuropathic pain context, blue light did not affect paw withdrawal latency compared to baseline levels. However, in the case of oxaliplatin treated mice, stimulation by blue light of Vglut3-positive fibers led to reduced paw withdrawal latency, and even shaking and licking of the paw several seconds after the end of the light stimulation. In conclusion, the Vglut3-positive peripheral fibers activated by blue light in this paradigm, are sensitized by oxaliplatin treatment and could be involved in the cold and mechanical hypersensitivity associated with this neuropathy.

(Peirs et al., 2015) To follow up on the results of the Seal publication of 2009 and taking into account the broader than expected expression of Vglut3, Cedric Pears, in Rebecca Seal’s team, used various Cre lines to assess the contribution of the different Vglut3-positive neuronal subpopulations in the phenotype described in (Seal et al., 2009): significantly reduced mechanical hypersensitivity following Carrageenan or capsaicin injection and following SNI or Brennan. The results allowed to clarify the role of C-LTMRs in this phenotype or lack thereof. They also increased our understanding of spinal cord microcircuits and their specificity in the transmission of different types of pain. The first cre mouse Piers and colleagues used was one specific to DRG: Advillin- cre. Advillin is expressed in all DRG neurons from around P16.5 onward (Hasegawa et al., 2007). Analysis of this conditional knockout did not show any altered development of mechanical hypersensitivity in the case of carrageenan inflammation nor in SNI-induced neuropathic pain. Subsequent analysis of Nav1.8-specific knockout of Vglut3 in unmyelinated and some myelinated fibers showed hypersensitivity levels similar to control animals too. In conclusion, Vglut3 expression in C-LTMRs is not required for the development of mechanical allodynia in carrageenan or SNI pain models.

The authors go on to use several other cre lines to define further which subpopulations on Vglut3-transient spinal neurons are required for injury-induced mechanical hypersensitivity in

39 an etiology-dependent manner. I will describe this elegant publication and its important findings for spinal cord circuitry in the part of my introduction dedicated to the spinal cord.

In conclusion, the example of the study throughout the years of the involvement of Vglut3- expressing neurons in injury-induced mechanical hypersensitivity goes to show that: 1- it is important to determine the pattern of expression of a gene of interest not only in the organ studied but also in surrounding tissues, especially those that interact with and are involved in the process we are interested in. In the case of DRG neurons, it is paramount to check for expression of a gene of interest not only in the rest of the nervous system (spinal cord and higher structures) which are just as involved in sensory processing, but also in skin cells, sensory end organs such as Merkel cells, in adult and during development. 2- very few genes are actually specific of one population of cells during both adulthood and development, and the use of conditional knockouts, and especially the development of intersectional genetic tools, will allow in the future to tightly control where and when a gene is silenced.

Tafa4 fam19a4 is a gene encoding the chemokine-like protein TAFA4 in a pattern rather specific to C-LTMRs. In 2013, (Delfini et al., 2013) in our lab showed that knockout of this gene in the mouse leads to a strong increase in neuropathic and inflammatory pain using several pain models, that injection of the molecule rescues the phenotype in all of these pain models, and that this pain modulation defect due to the loss of TAFA4 expression is linked to an increase in excitability and a decrease in inhibitory tone of spinal cord lamina II inner networks responsible for the processing of C-LTMR incoming input.

To go into more detail, the pain models used were: chronic constriction injury for the neuropathic pain model and carrageenan and formalin injection for inflammatory pain models. Concerning the neuropathic pain model, the KO mice develop a more important and longer-lasting mechanical allodynia and hyperalgesia, from day 14 post-surgery onward, which is rescued by intrathecal (IT) injection of TAFA4 at day 30 post-surgery. This long-term increase in hypersensitivity points to a role of TAFA4, and therefore C-LTMRs, in the “healing process” during neuropathy, probably by toning down the injury-induced excitation of spinal networks through activation of the inhibitory system. TAFA4 seems to allow for a quicker return to normal balance of excitation/inhibition in the spinal cord and progressive disappearance of pain in the mouse.

As for the inflammatory models of pain, the authors assessed the role of this molecule expressed by C-LTMRs in the context of acute chemical inflammation by injection of 2% formalin and long-term inflammatory pain with injection of 1% carrageenan. In both cases, KO animals show increased nociception. In the formalin paradigm, KO mice have a behavior similar to control animals during the first phase but a highly significant increase in nocifensive behavior during the second phase (Figure 9); hinting at a role of TAFA4 in the processing of nociceptive input after the first phase, possibly by limiting sensitization of the spinal cord networks involved. In the case of carrageenan inflammation, KO mice showed signs of mechanical allodynia and hyperalgesia faster than control mice, and this hypersensitivity lasted 20 days longer on average. Moreover, not only did IT injection of TAFA4 at day 6 post- injection of carrageenan reverse the long-lasting and abnormal hypersensitivity exhibited by KO mice, it also reversed the “normal” carrageenan-induced hypersensitivity of control mice at day 1 post-injection. This result opened the door to the use of TAFA4 as a potential analgesic drug in a number of pain types, not only those involving a dysfunction of C-LTMRs.

The fact that the IT injection of TAFA4 reverses pain or at least brings it back to control levels in all three models (Figure 9), and at different time points (15 minutes before formalin injection, 1 day post-injection of carrageenan and at day 30 post-surgery) underlines the major contribution of Tafa4 and therefore C-LTMRs in different etiologies of pain involving quite probably various anatomically and molecularly distinguishable central pathways, independently of whether the pain is recent or if it has set in and perturbed pathways in a long-term manner.

Figure 9: From (Delfini et al, 2013), the left panel shows Tafa4 KO mice exhibiting increased formalin-induced pain during the second phase compared to wild type male

41 littermates. The right panel shows that TAFA4 intrathecal injection reverses this increased inflammatory pain, which is thus due to perturbed C-LTMR function.

Ginip Ginip is a Gαi-interacting protein which identifies two subpopulations of Ret-positive DRG neurons: MrgprD-positive mechano-nociceptors and C-LTMRs. The molecule was discovered following the work of Aziz Moqrich on a mouse expressing the TrkC gene at the TrkA locus, and lacking all nociceptors. The result of a microarray comparing the transcriptome of these trkAtrkC/trkC mice and control mice resulted in the discovery of a number of genes with an intriguing pattern of expression including Tafa4 and Ginip. The role of Ginip itself in the neurons that express it has been described and published by (Gaillard et al., 2014). The publication shows that Ginip expression is important for timely recovery from mechanical hypersensitivity in models of inflammation (carrageenan) and neuropathy (CCI). Moreover, the prolonged mechanical hypersensitivity exhibited by Ginip knockout mice is baclofen- insensitive, leading to the conclusion that Ginip expression is necessary for correct GABAB receptor signaling in non-peptidergic C-fibers.

The role of Ginip-positive neurons was assessed in a second study thanks to the ablation of these neurons in adult mice. In this publication (Urien et al., 2017), the authors use a panel of behavioral tests to assess the sensory defects due to ablation of both MrgprD-positive mechano-nociceptors and Tafa4-positive C-LTMRs. Ablated mice show a significant decrease in innocuous mechanical sensitivity (a result that I obtained and that I will describe further in the Results part of my thesis) and an almost complete disappearance of the second phase nocifensive behavior in the formalin test. Even though two subpopulations were ablated in these mice, C-LTMRs are quite probably the neurons most involved in these sensory defects and I will now describe why. Mechano-nociceptors are high-threshold mechanoreceptors activated by high intensity mechanical stimuli in normal conditions, hence their absence probably does not account for the defect in innocuous mechanical sensitivity. The explanation behind the supposed contribution of C-LTMRs to formalin pain is more complex and necessitates a review of the literature. Two studies reporting neuronal ablations and formalin pain need to be cited: (Abrahamsen et al., 2008) involving a rather global Nav1.8-dependent developmental ablation of DRG neurons; (Shields et al., 2010) involving a more restricted and adult ablation of MrgprD and Trpv1-positive neurons.

Abrahamsen et al show that developmental ablation of all Nav1.8 (scn10a) -positive neurons, through their conditional expression of diphtheria toxin early on, eliminates the second phase nocifensive behavior of formalin pain, without affecting the first phase. They conclude from these results that nociceptors are responsible for second-phase nociception. Another conclusion could be that they are dispensable for the first phase. The original aim of this study by John Wood’s team was to determine the contribution of nociceptors to a large array of pain processes but a few caveats need to be mentioned. First of all, Nav1.8 expression is broader during development, starting from around E18 onward than in adult where it is restricted to unmyelinated fibers. Hence, contrary to what the publication says, the neurons ablated in this model include not only prospective nociceptors but also prospective C-LTMRs, unmyelinated fibers that are not nociceptors, and a portion of prospective myelinated fibers (Shields et al., 2012). Furthermore, the formalin test was performed by injecting a 5% dilution, and there is no mention of the volume used. These are rather extreme conditions considering 0.5 to 2% dilutions of formalin are more often used, with the first responsible for the activation of TrpA1 receptors (McNamara et al., 2007) and the latter sufficient to induce a second phase lasting up to 60 min. Nonetheless, an interesting conclusion of this study is that, contrary to what was previously thought at the time, peripheral neurons seem to be at least partially responsible for the second phase of nocifensive formalin-induced behavior.

In 2010 Shields and colleagues use a different strategy to perform the ablation of subpopulations of neurons with the expression of the receptor of the human diphtheria toxin (DTR) and subsequent injection in adult mice of the toxin (DTX). In these conditions, much less functional plasticity is at play and the ablation is more specific to the targeted subset of neurons. The authors from Allan Basbaum’s lab use the MrgprD-cre mouse to express DTR specifically in these neurons to ablate them, and use high dose capsaicin (TrpV1 agonist) injections to ablate TrpV1-positive peptidergic nociceptors. They also assess formalin pain in ablated animals with two different paradigms: injection of 0.5% or 2% formalin. Interestingly, the contribution of each subset of neuron is different depending on the dilution used: TrpV1 nociceptors are required for low-dose (0.5%) formalin pain, both for the first and the second phase with a drastic 80% reduction in nocifensive behavior, but not at all for high-dose (2%) formalin pain. MrgprD-positive neurons however are not required for low-dose nor high-dose formalin pain.

In light of these studies, the fact that Ginip-ablated mice display an almost complete lack of second phase pain following injection of 2% formalin suggests that C-LTMR ablation and not MrgprD-positive neuron ablation is the main contributor to this phenotype.

Cav3.2 C-LTMR-specific knockout In 2015, François et al published the latest of the few studies where a sensory defect can clearly be attributed to C-LTMRs in mouse. In their publication published in Cell Reports, the authors describe the knockout of the voltage-gated calcium channel Cav3.2 (cacna1h) and its role in cold and mechanical sensitivity in acute conditions and following tissue-injury. This gene is expressed in Aδ-fiber and C-fiber LTMRs in adult and, using a conditional approach, the authors show that Cav3.2 expression in C-LTMRs is necessary to fine-tune their light touch sensitivity in naive mice, but is also required for C-LTMR contribution to cold and mechanical allodynia (Francois et al., 2015). Electrophysiological recordings on DRG cell cultures and more physiologically relevant skin-nerve preparations (Zimmermann et al., 2009) show that this activity-dependent calcium channel is required to fine-tune the low-threshold mechanosensitivity and excitability of Aδ and C-LTMRs. More specifically, blockade of Cav3.2 by application of TTA-A2, a specific Cav3.2 channel blocker, leads to a higher mechanical threshold of excitability, lower firing frequency and lower conduction velocity, in short a shift of C-LTMR firing properties toward those of high threshold mechanoreceptors (nociceptors). In terms of behavior, this translates into sensory defects both in naive animals and after injury. C-LTMR-specific knockout of Cav3.2 reduces mechanical and temperature sensitivity under acute conditions: notably knockout mice display a decreased cold sensitivity both when acetone is applied to their hindpaw, and when performing the thermotaxis assay called gradient, where mice presented with a variation of temperature ranging from 10 to 53°C normally show a preference for a warm temperature around 31°C but C-LTMR specific knockouts prefer cooler temperatures. Under pathological pain conditions, mice with altered C-LTMR signaling present decreased nocifensive behavior during both phases of the formalin test (10μl 2%) and reduced mechanical and an absence of cold allodynia in the spared nerve injury paradigm. This latest finding proves that C-LTMRs also, and not only myelinated LTMRs as is commonly thought, play a role in injury-induced hypersensitivity to light touch in mice.

C-LTMR contribution to pain modulation: clues from human studies Proofs of the involvement of C-LTMRs in both pain modulation and allodynia come not only from animal models but also from human studies. Pain studies in human are tightly regulated

44 and the participants are financially compensated. They are of prime importance because they allow us to gain hindsight into the translational value of rodent studies across species. For example, in the case of Low-Threshold MechanoReceptors, Aδ-LTMRs have yet to be discovered in human hairy skin, thus lending even more importance to Aβ and C-LTMRs. Hairy skin stimulation tends to activate all types of LTMRs (Cole et al., 2006). In mouse, genetic tools such as optogenetics and intersectional genetics that allow specific activation of a subclass of neurons start to appear, but in human there are only two ways of selectively activating Aβ or C-LTMRs. The first is to study them in patients suffering from neuronopathies of various origins, which lead to a loss of a subclass of neurons. Studies have reported cases of people lacking either most myelinated fibers (Cole and Sedgwick, 1992; Fernandez et al., 1994; Low et al., 1978) or nociceptors (Morrison et al., 2011). The second way to test selectively one class of human hairy skin neuron in healthy subjects is to anaesthetize one of them either by performing compression nerve block to create a short-term ischemia of myelinated fibers, or by low-dose intradermal anesthesia to silence unmyelinated cutaneous afferents (Nagi et al., 2011).

In 2011, Nagi and colleagues took advantage of the second strategy to show that C-Tactile afferents, C-LTMRs in human hairy skin, mediate mechanical allodynia. The model of pain they use consists of injecting a hypertonic saline solution into the tibialis anterior muscle, the front calf muscle on the lateral side of the tibia, which gives rise to muscle pain but also secondary allodynia at the level of the skin overlying the painful muscle when a vibration is applied. Previous results from this team showed that vibration-induced LTMR activation in this context led to an increase in pain intensity (Nagi et al, 2009 Proc Austral Neurosci Soc). In 2011, using compression block of the sciatic nerve and low-dose anesthesia of unmyelinated afferents, they show that myelinated cutaneous afferent nerve block abolished the sensation of vibration but not vibration-evoked allodynia, whereas 0.25% Xylocaine anesthesia of CT- afferents abolished vibration-evoked allodynia but did not affect the perception of vibration itself (Nagi et al., 2011). Of note, gentle brushing, a stimulus activating more specifically CT than Aβ fibers compared to vibration, also produced tactile allodynia in healthy patients with myelinated nerve block in this publication. Similar results underlining the role of CT-afferents in mechanical allodynia were obtained by the same authors in a follow-up publication in 2013, in a model of exercise-induced muscle soreness and in one clinical subject suffering from

45 pathological pain (Nagi and Mahns, 2013). In conclusion, CT-afferents are involved in injury- induced tactile allodynia even when the stimulus is optimal for activation of Aβ fibers. Moreover, this study once again shows that CT-mediated input has a “pluripotent central effect”: it elicits a pleasant sensation in healthy subjects but, in a context of concurrent activation of muscle nociceptors, their stimulation generates allodynia, hinting at a switch between activation of innocuous-touch and noxious-touch pathways.

In 2014, Liljencrantz and Olausson publish a study where they investigate the role of CT- afferent stimulation in a model of capsaicin injection-induced tactile allodynia in patients lacking Aβ fibers. Comparing the effect of CT stimulation in the allodynic zone and a non- allodynic zone in terms of pleasantness and intensity of the evoked sensation, the patients report a reduced hedonic value of touch in the allodynic zone, suggesting that, in this particular model of tactile allodynia, the processing of CT-mediated signaling is affected (Liljencrantz and Olausson, 2014). Following up on that observation, the authors propose a working model where CT involvement in tactile allodynia is actually through a loss of their pain-inhibiting properties, as opposed to the theory proposed previously by Nagi and colleagues where CT activation itself would be painful. However, the authors make room in their discussion for a third possibility: a duality in the consequences of CT activation where their release of glutamate in the spinal cord would be pronociceptive but their release of the chemokine-like protein TAFA4 (Delfini et al., 2013) is anti-nociceptive. This latest hypothesis was based at the time on the false conclusion by Seal et al that the reduced hypersensitivity of global Vglut3 KO animals was solely due to the lack of glutamate signaling by C-LTMRs. A dual function of C-LTMRs is discussed here: signaling pleasantness or modulating pain networks. It raises the question of the first-order spinal targets of C-LTMRs and the following pathways which must be multiple to allow such duality. It is worth noting that the two patients lacking selectively Aβ fibers do not feel any pain following tactile stimulation of the allodynic zone, as opposed to healthy subjects, which supports the hypothesis that Aβ fiber-signaling in pathologic conditions becomes painful and is at least partly responsible for tactile allodynia (Liljencrantz and Olausson, 2014).

After publishing in 2011 on the C-fiber-mediated dual nature of vibratory stimulation (neutral in healthy individuals and painful in allodynic conditions), David Mahns’ team states in 2015 that a similar vibratory stimulation is actually unpleasant in healthy individuals and hence

46 gives rise to pain when reproduced in allodynic conditions. In their study published in PLoS One in 2015, in a context of tactile allodynia, Shaikh and colleagues assess the effect of a stimulus on pain sensation, depending on whether it was originally pleasant or unpleasant in healthy conditions (Shaikh et al., 2015). They use a human pain model identical to the one they used in their previous 2011 publication: injection of hypertonic saline solution into the tibialis anterior muscle leading to pain in this compartment followed by assessment of tactile sensitivity of the overlying skin. They apply a variety of stimulations, ranging from pleasant to unpleasant in naive conditions. Velvet-stroking and low-frequency vibration (20 Hz) are described as pleasant by subjects, whereas sandpaper-stroking and high-frequency vibration (200 Hz) are unpleasant though not painful. They report that previously pleasant stimuli tend to alleviate muscle pain somehow, whereas unpleasant stimulations increase pain intensity. More importantly, this tendency is significant even in the case of myelinated nerve block which supports previous results on the implication of CT-afferents in affective processing, tactile allodynia and pain relief.

Also in 2015, David Mahns’ team publishes a publication focused on thermal allodynia in the same model of hypertonic saline injection in the lateral calf muscle (Samour et al., 2015). This is a nice correlate to the publication published by François and colleagues on the role of the Cav3.2 channel in cold allodynia in the mouse. François and colleagues show that conditional knockout of Cav3.2 in C-LTMRs significantly reduces mechanical allodynia and completely abolishes the development of cold allodynia in the SNI mouse model of neuropathy. Here Samour and colleagues explore muscle pain-induced thermal allodynia. Following their usual protocol of hypertonic saline infusion into the tibialis anterior muscle, the authors show that subjects develop allodynic symptoms to dynamic cooling but not warming of the overlying skin. Furthermore, the cooling stimulation, which is innocuous when applied on non-allodynic skin, is still pronociceptive following myelinated nerve block, meaning low-threshold C fibers must be involved in this form of allodynia. C-Low Threshold Mechanoreceptors have been shown in the literature to express the activity-gated calcium channel Cav3.2 (Cacna1h) (Francois et al., 2015). To ascertain that these are the fibers involved in this C fiber-mediated thermal allodynia, the authors use sequentially inhibitors of TrpV1 and TrpM8 channels respectively involved in noxious heat and cold sensation and Cav3.2 channels. TrpV1 channels are activated or desensitized, depending on the dose used, by capsaicin; TrpM8 by menthol

47 and Cav3.2 channels are selectively inhibited by TTA-A2. TrpV1-positive and TrpM8-positive neurons are excluded from the list of neurons involved here because their inhibition does not affect the pain intensity during the application of the cooling stimulus. Inhibition of Cav3.2 however abrogates the increase in pain intensity due to cooling of the skin overlying the painful muscle. This goes to prove the implication of Cav3.2-expressing C-LTMRs in cold allodynia in the context of muscle pain. Interestingly, the authors performed the blockade of Cav3.2 channels in absence of myelinated nerve block and while the subjects report an absence of cold allodynia they still detect the cooling sensation. This suggests that, although at least one other population of neurons is activated by dynamic cooling of the skin, the one involved in cooling-induced allodynia in this model is Cav3.2-expressing C-LTMRs. The subjects also specify that following Cav3.2 blockade, there is no cooling-induced allodynia but the cooling stimulus itself is “now clearly discernible”, implying that the other population activated by cooling lends a more discriminative aspect to the sensation than C-LTMRs. This specificity is reminiscent of the fact that, when it comes to light mechanical stimulation too, C-LTMRs lend a more affective than discriminative aspect to the sensation.

II - Spinal processing of peripheral input

To fully grasp the subtlety of the somatosensory nervous system, one needs to take into account the major role the spinal cord plays in the modulation and the integration of signals coming from the periphery, and the complexity of the networks located there which allow a confrontation with descending information from higher structures. One does not fully understand how a population of PSN works without knowing which higher-order circuitry is at play when these PSNs are activated and under which circumstances. Every piece of information coming from PSNs has to go through the dorsal horn of the spinal cord where it is either transmitted directly to projection neurons, as is sometimes the case for C and Aδ nociceptive input and Aβ-LTMR input, or it has to go through a network of inhibitory and excitatory interneurons. Recent publications have shown that these local interneurons often receive information from different subtypes of primary afferent fibers, other local interneurons and neurons belonging to various descending control pathways (Abraira et al., 2017). Moreover, one primary afferent often contacts several different types of interneurons

48 and in consequence can activate different networks of inhibitory and/or excitatory nature at the same time.

Just like PSNs, spinal cord neurons can be distinguished based on a multitude of features: their anatomy, the shape, size and orientation among laminae of their arborisation, the neurotransmitter(s) they secrete, their molecular markers during development and in adulthood, which can be different, as well as of course their localization in the different laminae of the dorsal horn. Anatomically, the spinal cord has been classified into 10 dorso- ventral laminae by Swedish neuroscientist Dr Rexed in 1952 based on cytological features such as cell size (Rexed, 1952). This classification is still in use and currently the dorsal horn of the spinal cord is described as spanning laminae I to VI with lamina I being the most superficial one and with the mean size of neurons globally increasing from lamina I to VI. Generally, it is considered that laminae I and II are involved in the integration and transmission of noxious thermal and mechanical information with lamina I projection neurons of the anterolateral tract responsible for the message reaching higher structures (Figure 10), whereas laminae III to VI are globally involved in innocuous touch signaling with projection neurons of the PostSynaptic Dorsal Column (PSDC) and the SpinoCervical Tract (SCT), located in laminae III to V, being the two main output structures (Figure 10). However, this is a very simplified view of the organization of the dorsal horn. For example, C-LTMRs are involved in gentle innocuous touch signaling and project to inner lamina II, conversely Aδ nociceptors project not only to superficial laminae I and II but also to the deeper lamina V.

Figure 10: On the left, from (Rexed, 1954), laminar organization of the spinal cord as described by Rexed in 1954; on the right, from (Abraira and Ginty, 2013), the three major output pathways of the spinal cord dorsal horn: the anterolateral tract in yellow, the Post- Synaptic Dorsal Column (PSDC) in blue and the SpinoCervical Tract (SCT) in pink.

In conclusion, one cannot consider the processing happening in the dorsal horn simply as superficial versus deep or noxious versus innocuous. Indeed, local interneuron networks play a major role in connecting these different regions: activation of superficial laminae by nociceptive signal probably always involves a concurrent LTMR-mediated activation of deeper laminae. And the resulting sensation we get is due to the integration of both.

Gate Control Theory A possible model of these networks was first described in the Gate Control Theory proposed by Melzack and Wall in 1965 in Science, with a revised version published in 1978 (Wall, 1978). This model offers an alternative view to the opposing prevailing theories at the time: specificity and pattern, described by Von Frey and Goldscheider respectively in 1894. Von Frey discovered what he called pain points on the human skin and developed a theory of receptor sensitivity where the receptor activated by a specific type of stimulus in the periphery is associated with a pathway, transmitting the information all the way to the brain, that is dedicated to the relay of that type of information only (Max Von Frey, 1894 Die Gefühle und Verhältnis zu den empfindungen). The opposing view at the time was the intensive theory, or

50 pattern theory extended by Goldscheider in 1894 (Alfred Goldscheider, 1894 Ueber den Schmerz in physiologischer und klinischer Hinsicht), where the feeling of pain is not actually specifically transmitted by one type of fiber but is due to an excessive activation of non- specific cutaneous fibers because “there are no specific fibers and no specific endings” (Sinclair, 1955). In this theory, the modality is specified by the spatio-temporal pattern of firing of the fibers. Where the first postulate discounts any involvement of the spinal cord in the processing of incoming stimuli and any cross-talk between the different modality-specific pathways, the second postulate disregards the large body of work already available at that time supporting a functional specialization of peripheral fibers. The model proposed by Melzack and Wall takes into account elements of reasoning from both sides but also recent scientific discoveries on a potential role of the dorsal horn, and more specifically the region called substantia gelatinosa corresponding anatomically to lamina II, in the processing of incoming peripheral information. This region of the spinal cord contains mostly small- diameter neurons and seems to lack any projection neurons compared to lamina I and laminae III to V. Hence it is mostly composed of peripheral afferents, which we now know transduce both noxious and innocuous modalities (for example Mrg-expressing nociceptors and C-LTMRs respectively), and locally projecting interneurons (notably PKCγ-positive in inner lamina II but also many others). In a context of transmission of pain, Melzack and Wall propose that the integration of the message involves: 1- small (S) and large (L) afferent fibers, 2- substantia gelatinosa (SG) interneurons, 3- spinal nociceptive transmission T cells, 4- dorsal-column fibers projecting to the brain (“the action system”). In this model (Figure 11), both small and large afferents connect with T cells, but large fibers are submitted to a feedforward inhibition whereby inhibitory SG interneurons normally prevent them from activating T cells. In conclusion, SG interneurons represent a gate that prevents innocuous stimulation of large fibers from activating the pain network (T cells). However, it makes room for the possibility that, when the system is perturbed, large fibers can actually reach transmission neurons. This would account for the hallmark of neuropathic pain that is allodynia where innocuous stimulation induces pain: following injury and perturbation of the inhibitory-excitatory balance in the spinal cord, light-touch activated neurons such as LTMRs manage to activate a “pain pathway”. This attractive model also proposes a descending control of pain processing through modulation of SG interneurons, following the activation of higher structures by dorsal-column fibers (Melzack and Wall, 1965; Wall, 1978). Subsequent studies focusing on

51 the characterization of such interneurons and their place in the dorsal horn circuitry involved in pain and touch processing demonstrated the actual complexity of the system with an incredible variety of interneuron populations and the constant cross-talk happening at that level between neurons encoding different modalities, directly or through activation of common inhibitory or excitatory neurons.

Figure 11: From (Melzack and Wall, 1965), schematic diagram of the gate control theory of pain mechanisms.

C-LTMR spinal circuitry

Limitations In the last two decades, a number of studies have investigated the anatomy, molecular profile and/or function of spinal cord neurons. In particular, interneurons of the superficial laminae I and II have been the subject of most publications in this field because they receive direct inputs from C and Aδ nociceptors and as such are most directly involved in pain processing and modulation (Todd, 2010, 2017). However, there are two ways to consider pain modulation networks in the spinal cord: interneurons directly receiving nociceptive input from the periphery and neurons receiving input from other PSNs involved in the modulation of pain, as proposed by the gate control theory, namely LTMRs. The spinal cord circuitry activated by LTMRs has been much less investigated and as a consequence very few publications are

52 available, especially on C-LTMR circuitry. In most electrophysiological studies, one can identify the type of afferent fiber activated (Aβ, Aδ or C) based on conduction velocity and the average electrical intensity of stimulation to which they respond (≤ 25 μA for Aβ, 30-100 μA for Aδ and 150-500 μA for C fibers). However, the exact identity of the fiber such as C-nociceptor versus C-LTMR cannot be determined. For example, in the case of spinal cord preparation where the dorsal root can be electrically stimulated, it is difficult to ascertain which fiber is being activated. One way to circumvent these issues is to use optogenetic tools targeted specifically to one population of neurons. Indeed, if Aβ fibers are optimally activated by an electrical stimulation of an intensity as low as 20 μA, they can also be activated at higher intensities. As a consequence, 300 μA stimulations can activate C fibers, but the resulting changes in the network will be masked by concurrent Aβ inputs, especially under certain recording conditions such as disinhibition. Hence, in most publications, hypotheses about the nature of the afferent fiber involved are mostly based on a correlation between the lamina of the spinal cord being recorded, the latency of the evoked response and the identity of the afferent fibers terminating there. In that scenario, it is important to remember that innermost lamina II actually receives afferences not only from C-LTMRs but from all three types of LTMRs, according to a recent publication by (Abraira et al., 2017), and that C-fiber MrgprD-positive and MrgprB4-positive afferents are close neighbors. I will now describe the rare publications hinting at spinal targets of C-LTMRs, most of them published after the beginning of my thesis.

(Lu and Perl, 2003): feed-forward inhibition of C-fiber input in SG

A study by Lu and Perl in 2003 is often referenced to as the first hint that C-LTMRs are involved in spinal gating processes. It is actually rather on the order of circumstantial evidence. The study was performed on rat spinal cord slices, and more specifically on substantia gelatinosa (SG). In the rat, lamina II receives afferences not only from Aδ peptidergic nociceptors, C non- peptidergic nociceptors and C-LTMRs, as in the mouse, but also from C-fiber TRPV1-positive nociceptors. The authors nicely performed simultaneous recordings of two SG neurons located at the border between outer and inner lamina II. They looked for paired neurons at a maximum distance of 250 μm of each other and showing monosynaptic characteristics (the delay between an action potential in the presynaptic cell and any evoked postsynaptic activity has to be at most of 0.2 msec). The conclusions of the study are fourfold. 1-The monosynaptic connection between some SG neurons is not random, it was actually observed in 10% of all

53 randomly recorded SG neurons. 2-The synapse is of inhibitory nature: it is sensitive to the

GABAA antagonist bicuculline. 3-It usually involves a presynaptic islet cell and a post-synaptic central cell as described by Ramon y Cajal (1909, Histologie du système nerveux de l'homme et des vertébrés) the central cell was also described as a small islet cell by Andrew Todd (Todd and McKenzie, 1989; Todd and Spike, 1993). 4-Both SG neurons receive glutamatergic input from C-fibers, sensitive to the AMPA receptor antagonist CNQX. However, the exact identity of the primary afferents is probably different because the “large” C-fiber connecting the presynaptic cell always displays a faster conduction velocity than the “small” C-fiber connecting the post-synaptic SG cell following dorsal root stimulation (Figure 12). This, combined with a shorter latency of dorsal root stimulation-evoked response and a lesser “threshold stimulus intensity”, would indicate a lower threshold of activation for the C-fiber innervating the presynaptic cell. All these elements lead the authors, and many after them, to presume that the C-fiber innervating islet cells would be a low-threshold C-fiber conveying innocuous stimulations whereas the C-fiber connecting the post-synaptic cell would be a nociceptor. Indeed, it has been observed several times (Douglas and Ritchie, 1957; Douglas et al., 1960) that in the wide spectrum of C-fibers, nociceptors tend to have a relatively smaller diameter and a lower conduction velocity than low-threshold mechanoreceptors. In conclusion, this study supports a model where innocuous C-fibers inhibit noxious C-fiber input by activating a GABAA-dependent inhibitory islet cell-central cell SG neural module.

Figure 12: From (Lu and Perl, 2003), scheme of the two SG cells recorded simultaneously, which are monosynaptically connected. The presynaptic islet cell receives input from a “large” C-fiber, possibly a C-LTMR, and the post-synaptic transient central cell receives input from both the islet cell and the “small” C-fiber, possibly a nociceptor.

(Bourane et al., 2015): LTMR-gating of mechanical itch

In 2015, a publication published by Martyn Goulding’s team identified NPY-cre spinal cord inhibitory interneurons (a heterogeneous population of NPY-transient and persistent neurons located from lamina I to IV) as being involved in the gating of mechanical itch (Bourane et al., 2015). Mechanical itch is defined as itch induced by light mechanical stimulation such as an insect or a parasite walking lightly on your skin, as opposed to chemical itch induced by the injection/application of a pruritic compound on the skin or the bite of a mosquito. Bourane et al show that ablating or silencing NPY-cre neurons in mice leads to spontaneous itch restricted to hairy skin, without any interference with chemical itch or pain thresholds. Hence, NPY-cre spinal cord neurons represent a mechanical itch-specific gating mechanism. To investigate which PSNs are upstream of NPY-cre neurons, the authors used different tracing techniques including retrograde labeling of cutaneous afferents by injection of cholera toxin subunit B

(CTB) in the hairy skin followed by single-synapse trans-synaptic rabies tracing to confirm the synaptic nature of the contact. They find that all three types of LTMRs contact NPY-cre neurons. Hence mechanical itch is gated by LTMR-activated NPY-cre inhibitory spinal interneurons and C-LTMR spinal targets include NPY-cre neurons.

(Cheng et al., 2017): inner ventral lamina II gated C pathway sensitized by nerve injury

Two years later, Qiufu Ma’s team publishes a publication on spinal gate control, combining electrophysiology and behavior with the use of intersectional genetics and pharmacogenetics (Cheng et al., 2017). This work was achieved through a close collaboration with Martyn Goulding who opened the way to the use of dual recombinase conditional alleles in the study of spinal cord dorsal and ventral networks (Britz et al., 2015; Duan et al., 2014; Li et al., 2013; Stam et al., 2012). In this context, it is possible to triply restrict the expression of a gene (DREADD, DTR…) to a population of cells which must express both the cre recombinase and the flippase. In their article, Cheng and colleagues explore the role of Vglut3-cre and Somatostatin-cre spinal cord neurons in dynamic and punctate mechanical hypersensitivity in different contexts of inflammatory and neuropathic pain, and in a more virtual context of total disinhibition with the use of both bicuculline and strychnine, respectively antagonists of

GABAA and glycine receptors. In order to unmask C-fiber input when stimulating the dorsal root at high intensities in an electrophysiological set-up, the authors took advantage of the fact that Aβ fibers enter the dorsal horn centrally, through the dorsal funiculus, and cut the spinal cord more laterally than is usual. In this preparation, high intensity stimulation (500 μA) leads to C fiber-evoked currents in inner ventral lamina II Vglut3-cre negative neurons under “normal” conditions, without giving rise to action potentials (APs). However, under disinhibition conditions, the same stimulation gives rise to APs in 72% of these cells. Moreover, 68% of these C fiber-evoked APs are morphine-sensitive, with only few spinal neurons hyperpolarized following the bath application of morphine, suggesting that morphine- inhibition acts mainly on primary afferents. Even more interestingly, following SNI, C-fiber stimulation generated APs in 50% of inner ventral lamina II neurons compared to none under naive conditions. In conclusion, C-fiber input to inner ventral lamina II Vglut3-cre negative neurons is the subject of feed-forward inhibition. This C-fiber input is partly morphine- sensitive (68%) and partly morphine-insensitive. And this gated C-fiber pathway is highly sensitized in the case of nerve injury. Finally, 90% of C-fiber evoked APs in inner ventral lamina

II neurons under disinhibition conditions are lost following ablation of SomatostatinLbx1 interneurons, suggesting that these neurons are necessary for the occurrence of most C-fiber evoked APs in inner ventral lamina II neurons under disinhibition or nerve injury condition. In conclusion, the results of this article are of high interest for the scientific community: it highlights the importance of inhibitory networks in tightly controlling which peripheral input can or cannot give rise to APs in the dorsal horn, it also shows the sensitivity and loss of efficiency of these gating mechanisms following nerve injury. However, the use of developmental markers of spinal interneurons (vglut3, somatostatin) which correspond to broad, diverse and poorly-defined populations in adult does not allow the confrontation of these results with those of other publications, in terms of the exact spinal neurons involved. This is a recurrent problem which was nicely addressed in the next publication I will describe, where Abraira and colleagues generated an array of genetic tools based on the identification of genes uniquely expressed by adult spinal subpopulations of interneurons (Abraira et al., 2017). In the same way that the adult spinal neurons mentioned in this publication are difficult to identify, so are the C-fibers evoking APs in inner ventral lamina II neurons under disinhibition conditions. Indeed, even though anatomically speaking, the C-fibers best known to target this sublamina are C-LTMRs, the fact that 68% of these would be morphine-sensitive is surprising. Neither C-LTMRs, nor MrgprD-positive outer lamina II innervating C-fibers, are known to express the μ opioid receptor MOR (oprm1) (Gaillard et al., 2014). Peptidergic C- fiber nociceptors however do express it. Hence, even though it seems logical that disinhibition of a gated C pathway normally involved in innocuous sensation would lead to allodynia in a context of nerve injury, there is no direct proof in this publication that the C-fibers whose input is relayed by SomatostatinLbx1 interneurons are C-LTMRs.

(Abraira et al., 2017): circuitry of C-LTMRs in the touch-recipient zone

The studies cited previously mainly focus on the superficial laminae directly involved in pain and itch peripheral input reception and processing. However, there is a severe lack of interest in deeper dorsal horn interneurons, which, for a long time, were thought to be only involved in innocuous touch processing. As in the case of DRG neurons, the lack of molecular markers specific to particular subpopulations of spinal interneurons in adult severely hinders their study. A very recent single cell RNA sequencing-based article now describes the existence of 15 inhibitory and 15 excitatory populations of dorsal horn neurons based on their molecular

57 profile (Haring et al., 2018). And in 2017, David Ginty’s lab published a study aimed to fill the gap in knowledge as to the identity of innocuous touch-deep dorsal horn interneurons. Laminae II inner to V are the recipient of innocuous touch-transducing LTMR input. The study by Abraira and colleagues focused on identifying molecular markers specific of the 11 subpopulations of spinal interneurons receiving direct input from LTMRs. The authors generated an array of corresponding mouse molecular-genetic tools, now available for the scientific community to explore the function of these neurons. This study also described which of the 11 newly identified spinal subpopulations received input from which LTMR subtype. This thorough investigation of the touch spinal circuitry led to the discovery that the newly identified spinal interneurons receive input from 1 to up to all 3 LTMR subtypes. All of these spinal interneurons also receive cortical input. Conversely, LTMRs synapse onto 4 to all 11 deep spinal interneuron subtypes, 7 of which are excitatory and 4 inhibitory. Interestingly, the authors estimate that all three LTMR subtypes form an equivalent number of synapses onto deep dorsal horn neurons, but only some Aβ fibers synapse directly onto the post-synaptic dorsal column pathway (PSDC) constituted of projection neurons. Moreover, the common location of primary afferents and interneurons in the same lamina does not necessarily correlate with a connection between the two. Concerning C-LTMR circuitry, the results show that they form synapses with 3 subtypes of deep dorsal horn interneurons: PKCγ-positive excitatory neurons, the excitatory subpopulation of parvalbumin-positive neurons and RORβ- positive inhibitory neurons (Abraira et al., 2017). These results are of high importance in terms of connectivity and they open the way to future functional studies of the C-LTMR circuitry.

(Kambrun et al., 2018): Tafa4-modulated interaction between microglia and spinal inhibitory neurons

In 2018, Kambrun and colleagues published a follow-up to the 2013 Tafa4 article published by our team (Delfini et al., 2013). The original article established the role of Tafa4 in modulating pain of various origins, inflammatory and neuropathic, by showing that knockout of Tafa4 in mice leads to increased mechanical hypersensitivity in various paradigms, which is reversed by acute intrathecal injection of tafa4 both in the case of CCI-induced neuropathic pain and formalin-induced inflammatory pain (Figure 9). Even more interestingly, intrathecal injection of TAFA4 is also able to acutely and completely reverse inflammatory pain six hours after injection of carrageenan in wild type mice. Later, Kambrun and colleagues investigate by electrophysiological analysis how, at the level of the spinal cord, a hypothetical secretion of

TAFA4 could affect local networks’ excitability. In this study, the authors look at effects of TAFA4 in wild type mice, either in naive or inflammatory pain conditions, to assess a role of TAFA4 when spinal networks are at “baseline” levels of activity and after they have been stimulated by an important influx of nociceptive information. The authors show that bath application of TAFA4 increases inhibition which in turn decreases excitation in inner lamina II neuronal networks both in naive condition and in complete Freund’s adjuvant (CFA)-induced inflammatory condition. They further show that this effect is mediated by GABAergic transmission and it is glia-dependent: blocking GABA transmission with the use of bicuculline and phaclofen, and microglial activation with the use of minocycline, both prevented TAFA4- induced modification of IPSC and EPSC frequency in spinal cord slices of naive and inflamed animals. Based on immunostainings, the authors propose that this microglial component of the response to TAFA4 application involves tafa4-induced microglial retraction. This would allow synaptic transmission of GABAergic inhibition. In conclusion, this study reinforces the notion that C-LTMR modulation of pain involves a GABAergic system-mediated decrease of spinal excitation (Figure 13). It also puts forward the novel idea that this is at least in part a microglia-dependent process.

Figure 13: From (Kambrun et al., 2018), scheme of the interplay between the central afferences of Tafa4-expressing C-LTMRs, microglial cells and spinal GABAergic neurons, first relay neurons for C-LTMR input. This interplay is responsible for mechanical pain alleviation by Tafa4 in CFA-induced inflammation.

In my introduction, I presented the basics of the somatosensory nervous system in general and a more in-depth review of what we know so far about the subpopulation of PSNs called C-Low Threshold MechanoReceptors, in terms of the modalities they detect and the clues that can be found in the literature as to what their spinal circuitry might look like. Briefly, these neurons transduce two types of innocuous sensations which are light affective touch and cooling of the skin. They have also been shown extensively, both in mouse models and in human, to mediate and/or modulate injury-induced pain. The latter function underlines the two ways we can consider these neurons: 1- as direct detectors and mediators of innocuous and maybe even noxious modalities in the periphery or 2- as modulators of peripheral input at the spinal level, by way of activation of inhibitory pathways that could dampen pain signals, as is suggested to be the role of LTMRs in the gate control theory and of C-LTMRs in the Tafa4 KO mouse.

This raises many questions. Which molecules do they express that could allow them to sense light touch, cooling and maybe noxious stimuli? Considering that they are activated both by caress and cooling, modalities that have a very different emotional valence, what is the difference in the transduction of those two modalities? Is it at the peripheral level, at the level of the neurons they connect with in the spinal cord, or at the level of the brain areas activated following their stimulation? If we manage to answer these questions, could we uncouple these different functions so as to use the pain-modulating aspect of C-LTMRs without affecting their role in social behavior? Or the reverse, if they are indeed pain mediators as suggested by Nagi and colleagues (Nagi et al., 2011), could we activate them at will pharmacologically to use them as “feel-good” neurons without also developing allodynia?

Considering the current health issues faced by our society, be it the chronic pain and related pain-killer abuse epidemic or the increase in mental health issues, a better understanding of C-LTMRS, a population of neurons necessary both for proper pain modulation and everyday social interactions, seems hugely relevant. Thus, the aim of my thesis studies was to learn more about this particular population of neurons. To do this, I took advantage of the C-LTMR 60 molecular repertoire recently generated in our laboratory (Reynders et al., 2015) to study the role of a transcription factor called bhlha9 highly expressed in these neurons among the nervous system in general. I will now describe the results I obtained through behavioral analysis, RNA sequencing experiments, conditional knock-out of the gene and electrophysiological studies, and in a third part I will discuss these results and how they open up new perspectives for the study of C-LTMRs.

Results

Bhlha9 and gm7271 KI-KO mice

In 2015, Ana Reynders in the team published a study which expanded greatly the molecular repertoire of two important populations of non-peptidergic neurons: GINIP-expressing and IB4-binding mechano-nociceptors and GINIP-expressing C-LTMRs (Reynders et al., 2015). To cite only a few of the interesting results of this study, LTMRs in general seem to have a number of genes in common, but very few are actually specific of C-LTMRs. Among the list of less than 10 genes which showed a pattern of expression interesting for the study of C-LTMRs were gm7271 and bhlha9. gm7271 is a long-non coding RNA, a type of RNA which has been shown to regulate the function of other RNAs by binding to them and either “buffering” them or leading to their degradation. And bhlha9 is a transcription factor of the basic helix-loop-helix family, a family which also counts NeuroD1, Ngn1, Olig2 and other genes of prime importance for the development of the nervous system (Roybon et al., 2010; Zhou and Anderson, 2002).

I decided to characterize more thoroughly their expression during development and in the DRG and found that neither was expressed in the spinal cord during development or in adult. I also found that gm7271 showed a dynamic pattern of expression in DRG in the embryo, and was highly expressed in C-LTMRs and few large neurons in adult (Figure 14 and 15). Whereas bhlha9 seemed even more restricted in its expression than gm7271 because it was not expressed during development, appearing only at around P7 and was found in a subset of C- LTMRs in adult (Figure 16).

Figure 14: gm7271 in situ hybridization in red combined with TrkA, TrkB, TrkC or GINIP immunostainings on embryonic E13 and E15 DRG slices. gm7271 is expressed in the DRG during development at E13 in GINIP-positive neurons and at E15 in TrkA and TrkC-positive neurons.

Figure 15: gm7271 in situ hybridization in red combined with TrkB or Tafa4 in situ hybridization or TrkA, TrkC, GINIP and/or IB4 staining on adult thoracic DRG slices (four upper panels), gm7271 in situ hybridization in red combined with GINIP and IB4 staining on an adult thoracic spinal cord slice. gm7271 is highly enriched in C-LTMRs as observed in the first picture of C-LTMRs triple-positive for gm7271, Tafa4 and GINIP, a few large neurons are simple- positive for gm7271. gm7271 is not expressed in the spinal cord.

Figure 16: bhlha9 in situ hybridization in red combined with GINIP and IB4 staining on an adult thoracic DRG slices (left) and on adult thoracic spinal cord slice (right). bhlha9 is highly enriched in a subset of C-LTMRs, and is not expressed in the spinal cord.

We decided to take advantage of these interesting patterns of expression to develop genetic tools to study C-LTMRs through the generation of knock-in mice presenting in their construct the following cassettes: WGA-Venus and hDTR (Figure 17). WGA stands for Wheat Germ Agglutinin, a protein capable of crossing the synapse by being secreted by the presynaptic neurons that express it then uptaken by next-order post-synaptic neurons (Yoshihara, 2002). As such it is a useful tool for neuronal tracing. The WGA-Venus cassette being flanked by two loxP sites, the hDTR cassette can only be expressed after cre recombination. hDTR is the human receptor for diphteria toxin, which, when expressed by a population of cells and combined with the injection of diphteria toxin (DT), allows the ablation of this population. Hence, these mice would allow us to trace the spinal connections of C-LTMRs and to ablate them, either partially or totally at a chosen time point, and study the potential defects presented by these mice through behavioral and electrophysiological studies. Finally, the homozygous animals would be knocked-out for the gene of interest which would allow us to study the role of the gene itself. As such, I will now call these mice KI-KO for knock-in - knockout mice.

Figure 17: Scheme of the genetic construct designed in the team to generate bhlha9 and gm7271 KI-KO mice. The top panel shows the construct surrounded by 5’ and 3’ homology arms to target the locus of the bhlha9 and gm7271 genes for homologous recombination. The first cassette codes for Wheat Germ Agglutinin fused to the fluorescent protein Venus. It is flanked by two loxP sites that allows the excision of the cassette when combined with the expression of a cre recombinase. The first cassette ends with a polyA sequence (3xpA), which means that the following cassette will only be expressed if the first one is excised. The original construct was to be used as described in the first case scenario described in the bottom left panel to assess WGA-Venus-expressing C-LTMR connectivity in the spinal cord. Mice carrying the construct on both alleles (homozygous recombinant mice) do not express the targeted gene anymore. This would allow us to study the role of the gene itself in C-LTMR physiology, as described in the bottom middle panel. Finally, the bottom left panel shows the use of the expression of the last cassette of the construct. Crossing animals expressing cre recombinase specifically in DRG neurons leads to excision of the first cassette specifically in those neurons. This allows expression of the human Receptor of the Diphteria Toxin encoded by the hDTR- Sv40pA cassette in C-LTMRs. This expression coupled to injection of Diphteria Toxin in adult mice leads to specific ablation of C-LTMRs, either all of them, following the pattern of expression of gm7271 or a subset of them in the case of the bhlha9 KI-KO mouse.

Pascale Malapert in the team generated the construct for the gm7271 KI-KO mouse, and I participated in the generation of the bhlha9 KI-KO mouse, including generation of the construct, selection of positive ES cell clones by Southern Blot, genotyping and selection of potential chimeras. Unfortunately, the bhlha9 chimera was not fertile and never produced any offspring.

I characterized the gm7271 KI-KO mouse and showed a nice expression of WGA-Venus specific to C-LTMRs (Figure 18). However, my sequencing experiments and quantitative PCRs showed that the construct did not insert correctly: the hDTR cassette was not inserted in full and did not allow its correct expression. Moreover, the mouse was not actually a KI but a transgenic mouse with an intact gm7271 coding sequence, which did not allow the possibility to study the KO of the gene.

Figure 18: GFP, GINIP and IB4 staining on thoracic DRG slice from adult chimeric gm7271WGA-Venus/+ mice. WGA-Venus expression is correctly restricted to C-LTMRs (GINIP- positive in blue, IB4-negative neurons) in gm7271WGA-Venus/+ mice.

Tafa4 transgenic mouse

In parallel, we decided to make good use of the WGA-Venus – hDTR construct and combine it with the specificity of expression of tafa4 in C-LTMRs by creating a genetic tool where these cassettes are inserted downstream of the tafa4 promoter in a BAC (Bacterial Artificial Chromosome), the tafa4 promoter being quite long. A new cassette was added: the

67 channelrhodopsin-mCherry cassette by Pascale Malapert to allow for optogenetic activation of C-LTMRs (Figure 19). Again I participated in the screening of recombinant animals and characterized a very strong expression of WGA-Venus specifically in C-LTMRs in the DRG, as well as in the habenula, a brain region known by our team to express tafa4. The expression of WGA-Venus seemed actually much stronger in tafa4 transgenic mice than I had observed in gm7271 KI mice, with no need for any immunostaining to visualize Venus and the appearance of what seems like inclusion vesicles containing WGA-Venus (Figure 20 right panel). This allowed a nice visualization of C-LTMR central afferences in the dorsal horn of the spinal cord (Figure 20 left panel). But I very rarely observed occurrences of actual spinal cord neurons positive for Venus that may have uptaken the WGA-Venus secreted by C-LTMRs. In conclusion, the expression of the first cassette nicely followed the pattern of expression of tafa4.

Figure 19: Scheme of the construct used to generate Tafa4 transgenic mice. As opposed to KI-KO mice mentioned previously, this construct does not contain homology arms so that the tafa4 locus stays intact, and the construct inserts randomly somewhere in the mouse genome. Moreover, this construct contains an additional cassette in second position that codes for Channelrhodopsin fused to mCherry followed by an Internal Ribosome Entry Site (IRES) and the hDTR cassette, so that both Channelrhodopsin and hDTR proteins are

68 synthesized simultaneously from a common messenger RNA. As described for the original construct in Figure 17, the first WGA-Venus cassette needs to be excised by cre recombination for the following ones to be expressed. Crossing the Tafa4 transgenic mouse with a mouse expressing the Cre recombinase specifically in the DRG restricts the expression of Channelrhodopsin and hDTR specifically to the DRG, avoiding expression in the habenula. The expression of Channelrhodopsin would allow specific activation of C-LTMRs by blue stimulation. This optogenetics technique coupled with electrophysiology would allow us to characterize C-LTMRs better, and the impact of their stimulation on spinal cord networks.

Figure 20: GINIP, Venus and IB4 staining on thoracic spinal cord and DRG slices from adult Tafa4 transgenic mice. Venus staining matches closely expression of Tafa4 in C-LTMRs, both in the DRG and at the level of central afferences in inner lamina II of the dorsal horn.

Similarly to the original construct used to generate KI-KO mice, to obtain expression of the following cassettes (channelrhodopsin-mCherry and hDTR), we need to cross these transgenic mice with mice expressing the cre recombinase. As such, I crossed tafa4 transgenic mice with Nav1.8-cre mice (as described in Figure 19 bottom right panel) and expected 1- a loss of expression of the WGA-Venus, 2- the appearance of the expression of Channelrhodopsin- mCherry and hDTR in C-LTMRs. However, crossing mice possessing the tafa4 transgene and

69 expressing the cre recombinase did not show any decrease in expression of the WGA-Venus nor expression of hDTR by immunostaining (Figure 21). In line with this, injections of DT even at increasingly high doses at P20 were inefficient in ablating C-LTMRs in these mice.

Figure 21: GINIP, Venus and IB4 staining on thoracic DRG slices from adult Tafa4Tg/+ (left panel) and Tafa4Tg/+::Nav1.8cre/+ mice (right panel). Cre expression from the nav1.8 locus does not seem to excise properly the WGA-Venus cassette.

It is known that Advillin (avil) is another gene strongly expressed in DRG and earlier on than Nav1.8. So I decided to cross tafa4 transgenic mice with avil-cre mice and again check for expression of WGA-Venus, hDTR and sensitivity to DT injections in pups expressing both the transgene and the cre recombinase. In these mice, there was an important decrease in expression of WGA-Venus already at P0, and an almost complete disappearance by P20 (Figure 22). However, I could not see any expression of hDTR by immunostaining, nor by in situ hybridization. As a consequence, I could not succeed in ablating C-LTMRs.

Figure 22: GINIP and Venus staining on thoracic DRG slices from adult Tafa4Tg/+ and Tafa4Tg/+::Avilcre/+ mice. Conversely to crossing Tafa4Tg/+ mice with Nav1.8cre/+ mice, crossing them with Avilcre/+ mice resulted in complete excision of the WGA-Venus cassette and absence of expression of WGA-Venus.

To look for an explanation at the genomic level, I performed a series of PCR amplifications of portions of the transgene followed by sequencing and confirmed that the construct was correct and completely inserted in the genome of transgenic mice. However, considering the strong expression of WGA-Venus before cre recombination, it seems highly probable that the transgene got inserted several times in tandems, as it tends to happen with BAC constructs. And when recombination occurs, an important portion of DNA is excised and only the very first promoter and the last cassettes would remain. In that case, the hypothesis is that if the first promoter was not inserted in full, the remaining cassettes will not be expressed.

Bhlha9-KO mouse Publication Fortunately, during Fall 2014, Dr Ben Arie’s team published a study where they show that bhlha9 is expressed in the limb between E10 and E14, and that KO of the gene induces hindlimb malformation in the mouse with a syndrome recapitulating the features of split- hand/foot malformation and long-bone deficiency (SHFLD) in human (Schatz et al., 2014) 71

(Figure 23). Schatz and colleagues were willing to collaborate with us, and after rederivation of the mouse which took a year, I was able to start the study of the total KO of bhlha9 and its impact on somatosensation.

Figure 23: From (Schatz et al., 2014), pictures of the forelimb syndactily affecting a subset of bhlha9 KO animals. No skeletal abnormalities have been observed, only reduced apoptosis in forelimb soft tissue during embryonic development.

The main results I obtained from the study of the total knock-out of bhlha9 are the subject of a publication submitted to Cell Reports. I will detail later the rest of the results of the total knock-out as well as those of the study of the conditional knock-out of bhlha9 in somatosensory neurons.

Loss of bhlha9 impairs thermotaxis and formalin-evoked pain in a sexually dimorphic manner

Manon Bohic1, Catarina De Pais Paiva Santos1, Ana Reynders1, Andrew J. Saurin1, Yves Le Feuvre2,3,

Nissim Ben-Arie4, Irène Marics1 and Aziz Moqrich1,*

1Aix-Marseille-université, CNRS, Institut de Biologie du Développement de Marseille, UMR 7288, case 907, 13288 Marseille Cedex 09, France.

2 Univ. Bordeaux, Interdisciplinary Institute for Neuroscience, UMR 5297, F-33000 Bordeaux, France

3 CNRS, Interdisciplinary Institute For Neuroscience, UMR 5297, F-33000 Bordeaux, France

4 Dept. of Cell & Developmental Biology, Institute of Life Sciences, Edmond J. Safra Campus at Givat- Ram, The Hebrew University of Jerusalem, Jerusalem 9190401, ISRAEL

*Correspondence should be addressed to Aziz Moqrich at [email protected]

SUMMARY

C-LTMRs have been described to convey affective aspects of touch and to modulate injury-induced pain in human and mice. However, a role of this population of neurons in temperature sensation has been suggested but not fully demonstrated. Here, we show that the transcription factor bhlha9 is expressed in a subset of adult C-LTMRs. Bhlha9 KO mice had normal responses to noxious thermal and mechanical stimuli under acute and pathological conditions. However, bhlha9 KO male but not female mice displayed impaired thermotaxis behavior in the temperature gradient paradigm and exhibited exacerbated pain hypersensitivity in the formalin test. Importantly, bhlha9 KO mice were insensitive to the analgesic effect of the positive allosteric modulation of the ionotropic GABAergic signaling. Our data consolidate the role of C-LTMRs in modulation of formalin-evoked pain and provide strong in vivo evidence of the role of this neuronal subpopulation in the discriminative aspects of temperature sensation.

INTRODUCTION

20% of the population in the world.

MechanoReceptors (C-LTMRs) has only recently been at the forefront of sensory and pain research.

These hairy-skin innervating neurons are highly conserved across species (Seal et al., 2009; Vallbo et al., 1999; Zotterman, 1939). They mediate affective touch (Loken et al., 2011) and modulate pain in mouse and in human (Delfini et al., 2013; Francois et al., 2015; Liljencrantz and Olausson, 2014; Nagi et al., 2011; Urien et al., 2017). Interestingly, they are also activated by cooling (Li et al., 2011). Few studies hint at the spinal circuitry that could explain such a duality: transduction of innocuous sensation and pain modulation (Kambrun et al., 2018).

Here, we take advantage of the C-LTMR molecular repertoire recently generated in our laboratory (Reynders et al., 2015) to show that the transcription factor bhlha9 is highly enriched in this population compared to the rest of the nervous system. Its expression appearing only after birth makes 76 it an interesting genetic tool to study C-LTMRs. Male mice lacking BHLHA9 show a wide-ranging thermotaxis defect compared to littermate controls. Moreover, the same male mice present an exacerbated inflammatory pain in the formalin test paradigm. Intriguingly, this exacerbated pain is

insensitive to the normally analgesic effect of GABAA receptor positive modulation. Recording of lamina

II inner neurons, the dorsal horn region receiving C-LTMR inputs, shows an increased excitability in naive condition. In conclusion, bhlha9 expression is required for correct perception of temperature and for inflammatory pain processing.

RESULTS:

Bhlha9 is dispensable for DRG neurons survival, maturation and targets innervation

Recent studies using RNA deep sequencing on single and sorted primary sensory neurons revealed the transcription factor bhlha9 among the few markers highly enriched in adult C-LTMRs (Reynders et al.,

2015; Usoskin et al., 2015). Triple staining experiments using bhlha9 RNA probe in combination with

GINIP antibodies and IB4 binding demonstrated that bhlha9 is expressed exclusively in GINIP+ neurons, more specifically the great majority of them are GINIP+/IB4- neurons which represent C-LTMRs (Gaillard et al., 2014) (Figures 1A-D). Quantification analysis showed that bhlha9 is expressed in 11% of total

DRG neurons and that high expression occurs in 60% of GINIP+/IB4- C-LTMRs and low expression in 6% of GINIP+/IB4+ neurons, which represent cutaneous free-nerve ending MRGPRD+ neurons (Gaillard et al., 2014) (Figures 1E). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) experiments showed that the highest expression of bhlha9 in the nervous system occurs in DRG neurons (Figure S1A) and that this expression starts just after birth and reaches adult levels by postnatal day 20 (Figure 1F).

To gain insight into the role of bhlha9 in sensory physiology, we took advantage of bhlha9 knock-out

(KO) mice previously generated and characterized by Schatz and colleagues (Schatz et al., 2014). The main phenotype observed in these mice consisted of asymmetrical syndactyly due to incomplete separation of soft tissues between forelimb digits 2 and 3 (Schatz et al., 2014). In our hands, this phenotype was observed in about 25% of homozygous male and female mice.

We first validated bhlha9 loss of expression in the DRG of bhlha9 KO mice using in situ hybridization for bhlha9 in combination with GINIP immunostaining and IB4 binding (Figures S1C-F). To examine the role of bhlha9 in the survival and maturation of DRG neurons, we carried out a thorough quantitative analysis of different thoracic DRG neuronal subpopulations in adult bhlha9 KO mice. Quantification of

TrkA-, TrkB-, Ret-, TrkC-, MRGPRD (GINIP+/IB4+)- and C-LTMRs (GINIP+/IB4-)-expressing neurons showed no difference between bhlha9 KO and their wildtype littermates (WT) (Figure S1B). We then asked whether loss of bhlha9 affects the central and peripheral projections of DRG neurons. Given the preferential expression of bhlha9 in C-LTMRs, we mainly focused on the central and peripheral projections of this particular population of neurons. To visualize the peripheral projections of C-LTMRs, we used GINIP immunostaining on whole mount hairy skin using the clearing method (Li et al., 2011).

In line with previous findings (Delfini et al., 2013; Li et al., 2011), GINIP+ axonal branches form the same types of longitudinal lanceolate endings around hair follicles in both genotypes (Figures S1G and S1H).

Similarly, a series of triple staining experiments on spinal cord (SC) sections showed that peptidergic

CGRP+, the subset of nonpeptidergic GINIP+/IB4+ (the cutaneous MRGPRD+ fibers) and GINIP+/IB4- (C-

LTMRs) afferents display normal central projections to their respective laminae in the dorsal horn of the SC (Figures S1I-L). Indeed, CGRP fibers mainly innervate lamina I, GINIP+/IB4+ afferents invade outer lamina II and GINIP+/IB4- terminals project to the most inner part of lamina II where they massively overlap with the PKCγ+ interneurons. Taken together, our results demonstrate that BHLHA9 is dispensable for the molecular maturation and anatomical organization of C-LTMRs.

Thermotaxis is impaired in bhlha9 KO male but not female mice

Since their discovery by Swedish physiologist (Zotterman, 1939), C-LTMRs’ role in conveying hedonic aspects of touch is well established. However, the role of these neurons in other somatosensory sensations is under intense investigation. Recent studies demonstrated that C-LTMRs display robust responses to rapid cooling but not warming (Francois et al., 2015; Li et al., 2011; Nordin, 1990), suggesting a role in temperature sensation. To gain insight into the role of bhlha9 in temperature sensation, we subjected bhlha9 KO male and female mice to a large battery of temperature tests, including the hot and cold plate, the dynamic hot and cold plate, the two temperature choice and the temperature gradient paradigms. We found no difference between WT and bhlha9 KO mice in the hot and cold plate (Figures S2A-D) as well as in the two temperature choice tests (Figures S2E-L). However, in the temperature gradient paradigm, male mice exhibited a net shift towards cooler temperatures in comparison to their WT littermates, whereas bhlha9 KO female mice displayed the same thermotaxis behavior as the WT mice (Figures 2A and 2B). Indeed, when we break down the behavior of the male mice into 3 consecutive 30 minutes, one can see that during the first 30 minutes, bhlha9 KO males exhibited the same exploratory behavior as their WT littermates (Figure 2C). During the second 30 minutes, WT mice started to show a clear bias towards the warmer area of the arena, whereas bhlha9

KO males occupy a much wider area with a bias towards cooler temperatures (Figure 2D). This behavior is strongly reinforced during the last 30 minutes with WT mice showing a strong preference around

33°C and bhlha9 KO male mice seemingly unable to make a difference between temperatures ranging from 24°C to 37°C (Figure 2E), suggesting that temperature sensation is impaired in bhlha9 KO male mice. We then asked whether this phenotype is due to impaired temperature sensation or to perturbed locomotor activity in bhlha9 KO male mice. To discriminate between these two possibilities, we measured the total distance travelled by WT and bhlha9 KO male mice under two conditions: (i) when the floor temperature of the arena was set at room temperature and (ii) when the floor

79 temperature was subject to a gradient varying from 15°C to 55°C (Figures 2F and 2G). Very interestingly, we found that WT and bhlha9 KO male mice travelled the same distance during the 90 minutes period when the floor temperature was set at room temperature (Figure 2F), demonstrating normal locomotor activity of bhlha9 KO male mice. However, when the temperature gradient was set between 15°C and 55°C, bhlha9 KO male mice travelled a much longer distance than their WT littermates. This demonstrates that the increased distance is due to a defect in temperature sensitivity and that loss of BHLHA9 strongly impairs the ability of bhlha9 KO mice to adapt their behavior to a change in temperature. This data is in line with the previously described response of C-LTMRs to cooling temperatures and supports a potential role of this particular population of neurons in discriminative aspects of temperature sensation.

Formalin-evoked pain is exacerbated in bhlha9 KO male but not female mice

Recent studies in humans and rodents provided evidence that C-LTMRs also play a role in modulation of injury-induced mechanical and chemical pain (Delfini et al., 2013; Francois et al., 2015; Kambrun et al., 2018; Liljencrantz and Olausson, 2014; Nagi et al., 2011; Urien et al., 2017). Given the enriched expression of bhlha9 in C-LTMRs, we first tested gross innocuous mechanical sensitivity of bhlha9 KO mice in the tape test (Figure S3A-D). We found no major difference between the two genotypes, suggesting that loss of BHLHA9 is dispensable for C-LTMRs to sense light touch. We then tested the mechanical and thermal sensitivity of bhlha9 KO mice under acute and two different pathological conditions: the chronic constriction injury of the sciatic nerve (CCI) (Bennett and Xie, 1988) and

Carrageenan-induced inflammation. Mechanical sensitivity was tested using the up-down method at basal conditions and at several time points after injury. In both neuropathic and inflammation-induced pain paradigms, mechanical thresholds at baseline and at all the time points tested after injury were similar in WT and bhlha9 KO mice (Figures S3E and S3G). Furthermore, bhlha9 KO mice had the same

80 responses to noxious cold stimuli before and after CCI and displayed the same Carrageenan-induced thermal hypersensitivity in the Hargreaves test (Figures S3F and S3H). Taken together, our results show that BHLHA9 is dispensable for acute and injury-induced mechanical and thermal hypersensitivity.

Previous studies from our laboratory demonstrated that C-LTMRs play a critical role in modulation of formalin-evoked pain (Delfini et al., 2013; Urien et al., 2017). To test whether loss of BHLHA9 affects the ability of bhlha9 KO mice to sense chemical pain, we subjected them to the formalin test. In this test, formalin injection triggers a biphasic pain-like response characterized by intense flinching, licking and biting behaviors. As we wanted to test males and females of both genotypes, we normalized the amount of formalin injected in function of the weight of the mice so that a mouse weighing an average of 25g would be injected with 10μl of 2% formalin with a variation of 1μl per 2.5g. As shown in Figure

3, intraplantar injection of 2% formalin triggers robust first and second pain responses in males and females of both genotypes. In line with the impaired thermotaxis results, bhlha9 KO male mice exhibited an exacerbated pain response to formalin injection during the second phase, suggesting a sexually dimorphic enhanced central sensitization in these mice (Figures 3A and 3B). NS11394 is a positive allosteric modulator of the ionotropic GABAergic signaling known for its pain relief effect during formalin inflammation in the rat (Munro et al., 2011). Very interestingly, oral administration of

NS11394 to mice 30 minutes prior to formalin injection was associated with a significant pain relief during both phases of formalin-evoked pain in WT but none in bhlha9 KO male mice (Figures 3E-G).

Taken together, our results demonstrate that loss of BHLHA9 exacerbates formalin-evoked pain through a dysfunction of the ionotropic GABAergic system.

Average amplitude of excitatory synaptic events is higher in bhlha9 KO mice

In order to assess whether the deficits in nocifensive behavior were mirrored by alterations in spinal neurons properties, we performed whole-cell recordings of unidentified inner lamina II interneurons to monitor afferent synaptic transmission in dorsal root-attached spinal cord slices from control (n=14) and bhlha9 KO mice (n=14). Recorded neurons were characterized for their passive properties (resting potential, membrane resistance and capacitance) (Figures S4A-E), as well as spiking behavior (Figures

S4A and S4F) and low threshold currents such as potassium inward rectifier (Figures S4B and S4G) and

Ih currents (Figures S4A and S4H). No difference was observed in the measured parameters between

WT and bhlha9 KO mice. By contrast, although the average inter EPSC interval did not differ significantly between WT and bhlha9 KO mice (Figures 4A and 4B), the average amplitude of excitatory synaptic events was significantly higher in bhlha9 KO mice than in WT littermates (Figures 4A and 4C,

WT 15.12+-1.9pA; KO 23.30+-2.127pA; p<0.01). Surprisingly, this increase in average EPSC amplitude was not accompanied by an alteration of the average EPSC decay (Figures 4D and 4E).

DISCUSSION

In this study, we report the identification of BHLHA9 as a novel marker of C-LTMRs and demonstrate that male mice lacking this transcription factor exhibit two major phenotypes: (i) a temperature sensation defect mainly revealed by the incapacity of the mutant mice to finely tune their thermotaxis behavior and (ii) an exacerbated response to formalin-evoked pain that was resistant to the pain relief effect of a positive modulator of the ionotropic GABAergic system.

C-LTMRs have been proposed to play a key role in temperature sensation. In vitro recordings show C-

LTMRs fire in response to cooling, both during skin-nerve experiments (Zimmermann et al., 2009) with a response to decreases in temperature ranging from 50°C to 30°C and from 30°C to 2°C (Li et al.,

2011), and in cultured primary sensory neurons (Francois et al., 2015). The wide-ranging thermotaxis

82 defect exhibited by bhlha9 KO mice is strongly reminiscent of this wide-ranging activation of C-LTMRs by cooling. The bhlha9 KO mice defect encompasses almost the whole range of innocuous temperatures (Figure 2: 24°C to 37°C) and underlines a possible deficiency in sensing a variation in temperature, but not detection of specific temperatures with noxious temperatures still correctly avoided by bhlha9 KO animals, as highlighted by their normal response to noxious heat and noxious cold (Figure S2). This C-LTMR thermosensitivity is quite intriguing because it is vastly different from other thermosensitive fibers characterized by a defined thermal threshold of activation, set for example by the TRP channels they express.

The actual molecular mechanism responsible for C-LTMR sensitivity to cooling is still unknown but there may be an evolutionary explanation. This characteristic response to cooling is conserved across species, with human C-LTMRs, called C-Tactile or CT afferents, displaying similar properties (Nordin,

1990). Indeed, human studies show that CT temperature and mechanical activation are tightly linked.

People perceive gentle touch, such as caress, as more pleasant when it is delivered at skin temperature, 32°C, rather than at a cooler or a warmer temperature (18 or 42°C) as shown by microneurography CT recordings combined with hedonic ratings (Ackerley et al., 2014). Remarkably, this heightened pleasantness of touch correlates with C-LTMR maximal firing frequency. Hence, one hypothesis could be that C-LTMRs are optimally tuned to detect and transduce affective social touch, both in terms of thermal and mechanical parameters. And their thermal sensitivity would have more of a social and hedonic valence than other sensory-discriminative thermosensitive fibers.

In the context of social behavior, this hedonic aspect of C-LTMR mechano-thermal activation could be particularly relevant to reinforce conspecific and parent-pup interaction and promote proper maternal/paternal care behavior such as pup retrieval, nest-building and huddling. It is interesting to note that only male mice lacking BHLHA9 show this profound defect in thermotaxis behavior.

Intriguingly, a study published in 1983 by Batchelder and colleagues shows that male but not female huddling behavior is affected by temperature: when housed at cold or warm temperatures, female mice maintain high levels of huddling whereas male huddling behavior significantly decreases at warm

83 temperatures compared to cold temperatures (Batchelder et al., 1983). This would suggest that mechano-thermal stimulation, for example of C-LTMRs, plays a different role in male and female mice social behavior and it could explain the sexual dimorphism exhibited by bhlha9 KO mice in our thermotaxis assay.

C-LTMRs present an intriguing particularity among C-fibers: they are involved in processes as different as mediating affective touch and modulating injury-induced pain. Delfini and colleagues show in 2013 that knock-out of the C-LTMR specific protein TAFA4 leads to increased pain responses in various neuropathic and inflammatory-induced pain paradigms (Delfini et al., 2013). In particular, tafa4 KO mice present an enhanced second phase nocifensive response to formalin injection. The formalin test of spontaneous inflammatory pain consists of injecting diluted formalin in the plantar side of the mouse hindpaw. This induces a nocifensive response where the animal guards, shakes, licks, bites the injured paw during a first phase of acute pain due to intense activation of peripheral neurons, followed by an interphase where the pain seems to subside, and a longer-lasting second phase of nocifensive behavior. The literature suggests that this second phase is due to ongoing peripheral activation of somatosensory neurons but also to central sensitization of spinal networks after an intense influx of peripheral input. The fact that tafa4 KO mice show increased pain behavior during the second phase but not the first phase suggests that this protein is involved in formalin pain modulation. This was confirmed by two facts. 1- Intrathecal injection of TAFA4 rescues the phenotype (Delfini et al., 2013).

2- Kambrun and colleagues show in 2017 that TAFA4 bath application on spinal cord slices increases inhibitory tonus which in turn decreases excitability in inner lamina II neurons, the dorsal horn region receiving C-LTMR inputs (Kambrun et al., 2018). As such, TAFA4-expressing C-LTMRs play an important role in formalin pain modulation. Although BHLHA9 is actually expressed in a subpopulation of C-

LTMRs, bhlha9 KO mice present a phenotype very similar to tafa4 KO mice with an exacerbated

84 formalin pain during the second phase. This suggests a similar deficiency in inhibitory modulation at the level of the spinal cord.

To test this hypothesis, we transposed to the mouse the experiment performed in the rat by Munro and colleagues in 2011 where oral administration of the positive allosteric modulator of ionotropic

GABA (GABAA) receptors NS11394 reduced formalin-induced pain during the second phase. Two different doses of the compound (1 and 3 mg/kg), efficient in their analgesic effect in the rat without affecting exploratory motility behavior, were used in the mouse. They both provided a highly significant pain relief for WT mice during the first and the second phase (Figure 3). However, there was no significant anti-nociceptive effect in bhlha9 KO mice during either phase. In conclusion, positive modulation of GABAA signaling in bhlha9 KO mice is inefficient in relieving formalin pain during both the acute and the central sensitization phase. To assess potential alterations in the physiological properties of bhlha9 KO mice spinal networks, we performed electrophysiological recordings of inner lamina II neurons and found significantly increased excitatory current amplitude in bhlha9 KO male mice compared to WT littermates. This lends support to the hypothesis that bhlha9 KO mice present an imbalance between inhibitory and excitatory signaling at the level of the spinal cord. Whether the increase in excitation is the consequence of the decrease in inhibitory tone, just as in the case of tafa4

KO mice, still needs to be assessed (Delfini et al., 2013; Kambrun et al., 2018).

A defect in GABAA signaling in bhlha9 KO mice could be due to different causes. Electron microscopy experiments performed by Kambrun and colleagues highlight direct synaptic contacts between C-

LTMR central terminals and inner lamina II GABAergic spinal interneurons. This would make GABAergic interneurons first relay neurons for C-LTMR input and key players in the pain modulation acted by C-

LTMRs. A defect in GABAergic transmission could be due to a fine change in arborization of C-LTMR central terminals in bhlha9 KO mice compared to WT animals, leading to a decrease in inhibitory synapses. It could also be the result of a difference in expression of GABAA receptors at the membrane of C-LTMRs. Indeed, the literature provides compelling examples of the importance of presynaptic

85 inhibition in pain transmission (Witschi et al., 2011; Zeilhofer et al., 2012). In their study, Witschi and colleagues elegantly show that the anti-nociceptive effect of diazepam, another positive allosteric modulator of GABAA receptors, happens via increased presynaptic inhibition. More specifically, this mechanism requires proper expression of the GABAA α2 subunit in peripheral afferents. A defect at this level of the GABAergic system in bhlha9 KO mice could explain their impaired analgesic response to GABAA positive modulation.

In conclusion, we show here that bhlha9 is a new molecular marker of a subset of C-LTMRs. Its expression is required for proper temperature perception and normal inflammatory pain modulation.

Moreover, we show that the ionotropic GABAergic system plays an important role in this inflammatory pain modulation process, possibly through presynaptic inhibition at the central terminal of peripheral afferents.

EXPERIMENTAL PROCEDURES

A full description of methods is found in the Supplemental information.

Supplemental information

Supplemental information includes experimental procedures, 4 figures and 4 figure legends

ACKNOWLEDGMENTS:

We are grateful to the Moqrich lab members for fruitful scientific discussions. This work has been funded by the ERC-Starting grant paineurons 260435 to A.M. M.B PhD thesis was funded by the

"Ministère de l’éducation nationale et de la recherche" and by "la Fondation pour la recherche médicale (FRM)".

AUTHORS CONTRIBUTIONS:

A.M, I.M and M.B conceived the project and designed the experiments. M.B performed most of the molecular, histological and behavioral experiments, C.S performed the injury-induced mechanical pain experiments, A.R contributed to the in situ hybridization experiments. A.S contributed to unpublished

RNA seq data. Y.L performed electrophysiological experiments on SC neurons. N.B provided the Bhlha9

KO mice. I.M performed unpublished data, A.M and M.B co-wrote the manuscript.

COMPETING FINANCIAL INTERESTS:

The authors declare no competing financial interests.

FIGURE LEGENDS:

Figure 1. bhlha9 is highly expressed in C-LTMRs In situ hybridization for bhlha9 on thoracic DRG section showing colocalization with GINIP+ IB4- neurons, known as C-LTMRs, bhlha9+ neurons in red (A),

GINIP+ neurons in blue (B) and IB4 stained neurons in green (C). Arrows indicate bhlha9+ C-LTMRs (A,

B, D), arrowhead indicates a low bhlha9 expressor GINIP+ IB4+ neuron known as MrgprD+ neuron (A,

B, C, D). (E) Percentage of bhlha9+ neurons among C-LTMRs and MrgprD+ neurons. (F) Expression of bhlha9 in the DRG at postnatal stages P1, P8, P14, P20 and in adult.

Figure 2. bhlha9 KO male, but not female, mice show a wide-ranging temperature perception defect with a preference for cooler temperatures in the thermotaxis assay. Overall behavior of WT and bhlha9 KO male mice (A) (WT n=16, KO n=13). Overall behavior of WT and bhlha9 KO female mice (B)

(WT n=16, KO n=17). Behavior of WT and bhlha9 KO male mice during the first (C), second (D) and last

(E) 30 min of the gradient test. Total distance traveled by male mice during the thermotaxis test (F)

87 and total distance traveled when the floor of the apparatus was set at room temperature (G) (WT n=9,

KO n=9). Statistical tests: two-way ANOVA for (A), (B), (C), (D), (E), one-way ANOVA for (F) and (G); * p<0.05 *** p<0.001

Figure 3: bhlha9 KO male, but not female, mice exhibit PAM-insensitive exacerbated chemical pain in the formalin test during both the first and the second phase. Nocifensive behavior of WT and bhlha9 KO male mice during the first and the second phase of the formalin test (A) (WT n=13, KO n=12).

Nocifensive behavior of WT and bhlha9 KO female mice during the first and the second phase of the formalin test (B) (WT n=12, KO n=17). Time course of the nocifensive response of WT and bhlha9 KO male mice during the formalin test (C). Time course of the nocifensive response of WT and bhlha9 KO female mice during the formalin test (D). Nocifensive behavior of WT and bhlha9 KO male mice during the formalin test following per os administration of NS11394, positive allosteric modulator (PAM) of

GABAA receptors at 1 or 3 mg/kg or its vehicle (E). Time course of the nocifensive response of WT (F) and bhlha9 KO (G) male mice during the formalin test following per os administration of NS11394.

Statistical tests: two-way ANOVA for (C), (D), (F), (G) one-way ANOVA for (A), (B) and (E); * p<0.05 ** p<0.01 *** p<0.001

Figure 4: Excitatory synaptic events’ amplitude is higher in lamina IIi neurons of bhlha9 KO animals compared to WT. Voltage clamp recordings of spontaneous excitatory synaptic currents in WT (top) and bhlha9 KO animals (A). Distribution of spontaneous EPSC average interval (B) and amplitude (C).

Average unscaled EPSC recorded in WT and bhlha9 KO animals (D) (see methods). Distribution of spontaneous EPSC average decay (E); ** p<0.01

REFERENCES

Ackerley, R., Backlund Wasling, H., Liljencrantz, J., Olausson, H., Johnson, R.D., and Wessberg, J. (2014). Human C-tactile afferents are tuned to the temperature of a skin-stroking caress. J Neurosci 34, 2879- 2883.

Batchelder, P., Kinney, R.O., Demlow, L., and Lynch, C.B. (1983). Effects of temperature and social interactions on huddling behavior in Mus musculus. Physiol Behav 31, 97-102.

Bennett, G.J., and Xie, Y.K. (1988). A peripheral mononeuropathy in rat that produces disorders of pain sensation like those seen in man. Pain 33, 87-107.

Delfini, M.C., Mantilleri, A., Gaillard, S., Hao, J., Reynders, A., Malapert, P., Alonso, S., Francois, A., Barrere, C., Seal, R., et al. (2013). TAFA4, a chemokine-like protein, modulates injury-induced mechanical and chemical pain hypersensitivity in mice. Cell Rep 5, 378-388.

Francois, A., Schuetter, N., Laffray, S., Sanguesa, J., Pizzoccaro, A., Dubel, S., Mantilleri, A., Nargeot, J., Noel, J., Wood, J.N., et al. (2015). The Low-Threshold Calcium Channel Cav3.2 Determines Low- Threshold Mechanoreceptor Function. Cell Rep.

Gaillard, S., Lo Re, L., Mantilleri, A., Hepp, R., Urien, L., Malapert, P., Alonso, S., Deage, M., Kambrun, C., Landry, M., et al. (2014). GINIP, a Galphai-Interacting Protein, Functions as a Key Modulator of Peripheral GABAB Receptor-Mediated Analgesia. Neuron 84, 123-136.

Kambrun, C., Roca-Lapirot, O., Salio, C., Landry, M., Moqrich, A., and Le Feuvre, Y. (2018). TAFA4 Reverses Mechanical Allodynia through Activation of GABAergic Transmission and Microglial Process Retraction. Cell Rep 22, 2886-2897.

Li, L., Rutlin, M., Abraira, V.E., Cassidy, C., Kus, L., Gong, S., Jankowski, M.P., Luo, W., Heintz, N., Koerber, H.R., et al. (2011). The functional organization of cutaneous low-threshold mechanosensory neurons. Cell 147, 1615-1627.

Liljencrantz, J., and Olausson, H. (2014). Tactile C fibers and their contributions to pleasant sensations and to tactile allodynia. Front Behav Neurosci 8, 37.

Loken, L.S., Evert, M., and Wessberg, J. (2011). Pleasantness of touch in human glabrous and hairy skin: order effects on affective ratings. Brain Res 1417, 9-15.

Munro, G., Erichsen, H.K., Rae, M.G., and Mirza, N.R. (2011). A question of balance--positive versus negative allosteric modulation of GABA(A) receptor subtypes as a driver of analgesic efficacy in rat models of inflammatory and neuropathic pain. Neuropharmacology 61, 121-132.

Nagi, S.S., Rubin, T.K., Chelvanayagam, D.K., Macefield, V.G., and Mahns, D.A. (2011). Allodynia mediated by C-tactile afferents in human hairy skin. J Physiol 589, 4065-4075.

Nordin, M. (1990). Low-threshold mechanoreceptive and nociceptive units with unmyelinated (C) fibres in the human supraorbital nerve. J Physiol 426, 229-240.

Reynders, A., Mantilleri, A., Malapert, P., Rialle, S., Nidelet, S., Laffray, S., Beurrier, C., Bourinet, E., and Moqrich, A. (2015). Transcriptional Profiling of Cutaneous MRGPRD Free Nerve Endings and C-LTMRs. Cell Rep.

Schatz, O., Langer, E., and Ben-Arie, N. (2014). Gene dosage of the transcription factor Fingerin (bHLHA9) affects digit development and links syndactyly to ectrodactyly. Hum Mol Genet 23, 5394- 5401.

Seal, R.P., Wang, X., Guan, Y., Raja, S.N., Woodbury, C.J., Basbaum, A.I., and Edwards, R.H. (2009). Injury-induced mechanical hypersensitivity requires C-low threshold mechanoreceptors. Nature 462, 651-655.

Urien, L., Gaillard, S., Lo Re, L., Malapert, P., Bohic, M., Reynders, A., and Moqrich, A. (2017). Genetic ablation of GINIP-expressing primary sensory neurons strongly impairs Formalin-evoked pain. Scientific reports 7, 43493.

Usoskin, D., Furlan, A., Islam, S., Abdo, H., Lonnerberg, P., Lou, D., Hjerling-Leffler, J., Haeggstrom, J., Kharchenko, O., Kharchenko, P.V., et al. (2015). Unbiased classification of sensory neuron types by large-scale single-cell RNA sequencing. Nat Neurosci 18, 145-153.

Vallbo, A.B., Olausson, H., and Wessberg, J. (1999). Unmyelinated afferents constitute a second system coding tactile stimuli of the human hairy skin. J Neurophysiol 81, 2753-2763.

Witschi, R., Punnakkal, P., Paul, J., Walczak, J.S., Cervero, F., Fritschy, J.M., Kuner, R., Keist, R., Rudolph, U., and Zeilhofer, H.U. (2011). Presynaptic alpha2-GABAA receptors in primary afferent depolarization and spinal pain control. J Neurosci 31, 8134-8142.

Zeilhofer, H.U., Wildner, H., and Yevenes, G.E. (2012). Fast synaptic inhibition in spinal sensory processing and pain control. Physiol Rev 92, 193-235.

Zimmermann, K., Hein, A., Hager, U., Kaczmarek, J.S., Turnquist, B.P., Clapham, D.E., and Reeh, P.W. (2009). Phenotyping sensory nerve endings in vitro in the mouse. Nat Protoc 4, 174-196.

Zotterman, Y. (1939). Touch, pain and tickling: an electro-physiological investigation on cutaneous sensory nerves. J Physiol 95, 1-28.

Supplemental data

Experimental procedures

Mice

Mice were maintained under standard housing conditions (22°C, 40% humidity, 12 h light cycles, and free access to food and water). Special effort was made to minimize the number as well as the stress and suffering of mice used in this study. All protocols are in agreement with European Union recommendations for animal experimentation.

In situ hybridization and immunofluorescence

In situ hybridization and immunofluorescence were performed following standard protocols (Li et al.,

2011; Reynders et al., 2015).

To obtain adult DRG and spinal cord, mice were deeply anesthetized with a mix of ketamine/xylazine and then transcardially perfused with an ice-cold solution of paraformaldehyde 4% in PBS (PAF). After dissection, DRG were postfixed for at least 2 to 24 h in the same fixative at 4°C.

Tissues were then transferred into a 30% (w/v) sucrose solution for cryoprotection before being frozen and stored at -80°C. Samples were sectioned (12 to 40 μm) using a standard cryostat (Leica). RNA probes (bhlha9, TrkA, TrkB, Ret) were synthesized using gene-specific PCR primers and cDNA templates from adult mouse DRG.

In situ hybridization was performed using digoxigenin labeled probes. Probes were hybridized overnight at 55°C, and the slides incubated with the horseradish peroxidase anti-digoxigenin antibody

(Roche). Final detection was achieved using cy3 TSA plus kit (Perkin Elmer).

The following oligonucleotides were used for nested PCRs for probe synthesis: bhlha9-F1: AACCGGGCTGAGGATTTTGT, bhlha9-R1: CCACGAAACCGGTCGAACA,

95 bhlha9-F2: TTTTGTGGAGGACTTGGGGC, bhlha9-R2+T7: TAATACGACTCACTATAGGGGACATTGACGCCAGTTTCCC,

TrkA-F1: TTTGCTCTCCTCTCATCTCTCC,

TrkA-R1: AAATTGATTCCAGAGACCCAGA,

TrkA-F2: GGCGATCGAGTGTATCACG,

TrkA-R2+T7: TAATACGACTCACTATAGGGTAATTGCTATTATGATGGATGCTG,

TrkB-F1: CTGAGAGGGCCAGTCACTTC,

TrkB-R1: CATGGCAGGTCAACAAGCTA,

TrkB-F2: CAGTGGGTCTCAGCACAGAA,

TrkB-R2+T7: TAATACGACTCACTATAGGGCTAGGACCAGGATGGCTCTG,

Ret-F1: TTAGATCCCCTTTCCCTTTAGC,

Ret-R1: GAGTGTCTGTGGCTACAACTGC,

Ret-F2: CTGCTCATCACTAGCCACCA,

Ret-R2+T7: TAATACGACTCACTATAGGGTGTGCCTTTCACACAAGCTC

For immunofluorescence, primary antibodies were diluted in PBS - 10% donkey or goat serum (Sigma)

- 3% bovine albumin (Sigma) - 0.4% Triton X-100 and incubated overnight at 4°C. Primary antibodies used in this study are as follows: goat anti-TrkC 1:1000 (R&D Systems), rabbit anti-CGRP 1:1000

(Chemicon), rabbit anti-PKCγ 1:1000 (Santa Cruz Biotechnology), anti-S100 1:400 (Dako), rat anti-GINIP

1:1000 (generated in the lab,(Gaillard et al., 2014)). Corresponding donkey or goat anti-rabbit, anti-rat, and anti-goat Alexa 488, 555, or 647 (Invitrogen or Molecular probe antibodies) were used for secondary detection. Isolectin IB4 Conjugated to Alexa Fluor 488 dye was used at 1:200 (Invitrogen).

For whole mount skin immunostaining as described in (Li et al., 2011), back hairy skin was shaved using a commercial hair remover, wiped clean with PBS and tissue, tape stripped and dissected. Excess fat and skin was removed, the remaining tissue cut in small pieces and fixed with 4% PFA at 4°C for 2h, rinsed with PBS then washed with PBS+0.3% triton every 30 min for 6h at room temperature (RT).

Tissue incubation with primary antibodies was done over 3 days at room temperature in 0.3% PBS

Triton + 5% Donkey serum + 20% DMSO. Tissues were washed with PBS+0.3% triton every 30 min for

6h at RT, incubated with secondary antibodies for 4 days at room temperature in 0.3% PBS Triton + 5%

Donkey serum + 20% DMSO, washed with PBS+0.3% triton every 30 min for 6h at RT, dehydrated in

50% methanol for 5 min at RT, then in 100% methanol for 20 min at RT and finally cleared in BABB

(benzyl alcohol + benzyl benzoate 1:2) for 20 min at RT.

Cell counts and statistical analysis.

12 μm serial sections of thoracic (T12) DRGs were distributed on six slides and subjected to different markers including the pan-neuronal marker SCG10. This approach allowed us to represent all counts as percentage of the total number of neurons (SCG10+). For each genotype, two DRGs were counted in at least three independent mice. Statistical significance was set to p <0.05 and assessed using one- way ANOVA followed by unpaired t test. qRT-PCR

Thoracic pairs of DRGs were dissected from P1, P8, P14, P20 and adult WT male mice. Cerebellum, hippocampus and cortex were dissected from adult WT male mice. RNA was extracted by using the

RNeasy Mini Kit (Quiagen), according to manufacturer's instructions. RNA samples were reverse- transcribed into cDNA (ImpromII Reverse Transcriptase, Promega) and used as template for qRT-PCR quantification of bhlha9 expression using the following oligos:

Fw: GCATCCTGGATTACAACGAGGC, Rv: CAAGGATAGAGCAGTGATGCGG.

Behavioral assays

All behavioral assays were conducted on WT and bhlha9 KO littermates of 8-12 weeks in age. Animals were acclimated for 30 minutes to their testing environment prior to all experiments that are done at room temperature (~22°C). Experimenters were blind to the genotype of the mice during testing.

Students’ T test was used for all statistical calculations. All error bars represent standard error of the mean (SEM). General behavioral (Locomotor and learning activity) was measured using Rotarod apparatus (LSI Letica Scientific Instruments). Gradient, Thermal plates, openfield, Hargreaves and Von

Frey apparatus were from Bioseb instruments.

General behavioral assays

Rotarod test

A rotarod apparatus (LSI Letica Scientific Instruments) was used to explore coordinated locomotor, balance and learning function in mice. Mice were placed on a rod that slowly accelerated from 4 rpm to 44 rpm constant speeds of rotation over 5min, and the latency to fall off during this period was recorded. The test was done 4 consecutive days. Each day, the animals were tested three times separated by at least 5 min resting period.

Thermal sensitivity

Response to temperature choice test and Response to temperature Gradient assay were performed as described in (Moqrich et al., 2005) but using Bioseb apparatus.

Hot plate

To assess heat sensitivity, mice were confined individually to a metal surface maintained at 48°C, 50°C,

52°C and 55°C by a Plexiglas cylinder 20 cm high, and the latency to nociceptive responses (licking, shaking or jumping of hind paws) measured. To prevent tissue damage, mice were removed from the

98 plate immediately after a nociceptive response or after a maximum of 90s, 60s, 45s and 20s respectively. Each mouse has been tested three times at 48°C, 50°C and 52°C and twice at 55°C with a latency of 5 min between each test; the withdrawal time corresponds to the mean of all measures. A latency of at least 1h between each tested temperature was respected.

Cold plate

This test was performed as described by Brenner and colleagues (Brenner et al., 2012).

To test cold sensitivity, mice were allowed to acclimate on a glass plate, placed individually in plastic chambers for an hour. Then crushed dry ice, piled in a syringe with the tip cut off, was applied right under the middle of the hindpaw of the mouse. We measured the latency until lifting of the paw with a cut-off of 20 s to avoid tissue damage. Each mouse is exposed six times with a minimum of five minutes resting period between trials. In the case of the cold test following CCI, the results in Figure

S2F show the result of stimulation of the injured hindpaw.

Thermal Gradient test

This test has been described previously (Moqrich et al., 2005). Briefly, mice were individually tracked for 90 min in four separate arenas of the thermal gradient apparatus (Bioseb). A controlled and stable temperature gradient of 14°C to 53.5°C was maintained using two Peltier heating/cooling devices positioned at each end of the aluminium floor. Each arena was virtually divided into 15 zones of equal size (8 cm) with a distinct and stable temperature. The tracking was performed using a video camera controlled by the software provided by the manufacturer.

Two-temperature choice tests

Two mice were placed simultaneously in each lane of the two temperature choice apparatus (Bioseb).

Mice were tracked for 10 min using the Bioseb software. During the first day, both plates were kept at

20°C during 10 min. Days after this acclimatizing period, 2 plates were individually warmed or cooled

99 to different temperature (42°C to 16°C) and kept at the appropriate temperature for 10 min test. A 1h time lapse was respected between 2 different tests.

Thermal nociceptive threshold (Hargreaves’s test)

To assess hind paw heat sensitivity, Hargreaves’ test was conducted using a plantar test device

(Bioseb). Mice were placed individually into Plexiglas chambers on an elevated glass platform and allowed to acclimate for at least 30 minutes before testing. A mobile radiant heat source of constant intensity was then applied to the glabrous surface of the paw through the glass plate and the latency to paw withdrawal measured. Paw withdrawal latency is reported as the mean of three measurements for both hindpaws with at least a 5 min pause between measurements. IR source was adjusted to 20% and a cut-off of 20 s was applied to avoid tissue damage.

Mechanical sensitivity testing

Tape Response Assay

This test was performed as described by Ranade and colleagues (Ranade et al., 2014). Briefly, mice were allowed to put in a Plexiglas cylinder and allowed to acclimate for 5 min; a piece of 3 cm of tape was gently applied to the back of the mouse. Mice were then observed for 5 minutes and the total number of responses to the tape was counted. A response was scored when the mouse stopped moving and bit or scratched the piece of tape or showed a visible “wet dog shake” motion in an attempt to remove the foreign object on its back.

Von frey filaments test:

Mice were placed in plastic chambers on a wire mesh grid and stimulated with von Frey filaments

(Bioseb) using the up-down method (Chaplan et al., 1994) starting with 1g and ending with 2g filament as cutoff value.

Chemical sensitivity testing

100

Formalin test

Formalin solution was prepared at 2% in PBS 1X from a formalin stock (Fischer Scientific) (note that formalin stock corresponds to a 37% formaldehyde solution). Mice were placed individually into

Plexiglas chambers, allowed to habituate to the testing environment for 30 minutes. Following subcutaneous injection of 10 µl of formalin for a mouse weighing 25g in the left hind paw, the animals were immediately placed individually in observation chambers and then monitored for pain behavior

(shaking, licking and biting of the injected paw) for 60 min. The pain behavior cumulative time of the injected paw was counted at 5 minutes intervals. Time spent exhibiting these pain behaviors was recorded for the first phase (0-10 min) and the second phase (10-60 min).

Administration of PAM

NS11394 was dissolved in 5% Tween80/milliQ water (as adapted from (Munro et al., 2011)). Mice were administered per os with NS11394 (MedChem Express) 30 min prior to formalin injection just before being allowed to habituate to the Plexiglas cylinder. The dosing volumes were 1 and 3 mg/kg with an average mouse weighing 25g being administered 250 μl and later injected with 10 μl of formalin.

Carrageenan injection

20 μl of 1% λ -Carrageenan (Sigma-Aldrich, 22049-5G-F) in PBS1X was injected into the mouse left hind paw using a Hamilton syringe.

For the Carrageenan model, mechanical allodynia and hyperalgesia were assessed before and after injection using the up-down method starting with the 1g Von Frey hair filament and ending with the

2g filament as cutoff value. The uninjected right hind-paws serve as a control.

Unilateral peripheral mononeuropathy

For the chronic constriction of the sciatic nerve (CCI) model, unilateral peripheral mononeuropathy was induced in mice anaesthetized with Ketamine ( 40mg/kg ip) and Xylasine (5mg/kg ip) with three

101 monocryl (6/0, Ethicon) ligatures tied loosely (with about 1mm spacing) around the common sciatic nerve (Bennett and Xie, 1988). The nerve was constricted to a barely discernable degree, so that circulation through the epineurial vasculature was not interrupted.

For the chronic constriction model, mechanical allodynia and hyperalgesia were assessed before the surgery and every other 7 days post-surgery using the up-down method starting with the 1g Von Frey hair filament and ending with the 2g filament as cutoff value. The uninjured right hind-paws serve as a control.

Electrophysiological recording and calcium imaging

Transverse spinal cord slices with attached dorsal roots from juvenile (P24 to P45) bhlha9 KO and WT mice were prepared for whole-cell recording. Briefly, the animals were anesthetized using

Pentobarbital (200mg/kg), perfused with ice cold oxygenated sodium free artificial cerebro-spinal fluid

(ACSF; in mM: ClCholine 101; KCl 3.8; MgCl2 18.7; MgSO4 1.3; KH2PO4 1.2; HEPES 10; CaCl2 1; Glucose

1), and then beheaded. The vertebral column and surrounding muscles were quickly removed and immersed in ice cold oxygenated ACSF. Following laminectomy, the spinal cord was gently removed and its lumbar part was placed into a small 3% agarose block. Spinal slices (300μm thick) were cut using a Leica VTS1000 vibratome, and transferred in warm (31°C) ACSF (in mM: NaCl 130.5; KCl 2.4; CaCl2

2.4; NaHCO3 19.5; MgSO4 1.3; KH2PO4 1.2; HEPES 1.25; glucose 10; pH 7.4) equilibrated with 95%O2-

5%CO2 for at least one hour before starting patch clamp recordings. Spinal slices were placed in a recoding chamber bathed with warmed (31°C) ACSF Electrophysiological measurements were performed under the control of an Olympus BX51 microscope using a multiclamp 2B (Molecular devices). Patch pipettes (7-11 Ω) were filled with C-based pipette solution (in mM: Potassium D-

Gluconate 120; KCl 20; CaCl2 0.1; MgCl2 1.3; EGTA 1; HEPES 10; GTP 0.1; cAMP 0.2; Leupeptin 0.1;

Na2ATP 3; D-Manitol 77; pH 7.3). All drugs were purchased from Sigma. Synaptic analysis was performed using Spike 2 software and at least 3 minutes of recordings were used to quantify EPSC

102 parameters. Sag ratio was calculated as the difference between stady state (Vss) and peak voltage (Vp) variations during a hyperpolarizing current pulse (ratio = (Vss-Vp)/Vp) and IO rectification was calculated by dividing the slope of the current response curve at -60mV and -100mV (IOrect = [(dI/dV-

60) / (dI/dV-100)] ). For each cell, an average EPSC was calculated and the kitetics of these averaged EPSCs were compared between WT and KO animals.

Supplemental figure legends

Figure S1: Expression of bhlha9 in the nervous system is the highest in DRG neurons and its knockout does not affect maturation of DRG neurons, nor innervation of the skin or the spinal cord by C-

LTMRs. bhlha9 expression in DRG compared to cerebellum (cb), hippocampus (hp) and cortex (cx) (A)

Percentages of Ret, TrkA, TrkB and TrkC positive neurons in WT and bhlha9 KO DRGs are similar (B). In situ hybridization with a bhlha9 probe on transverse sections of thoracic bhlha9 KO DRG. bhlha9 probe in red (C), GINIP in blue (D) and IB4 in green (E), all three stainings in (F). Whole mount immunostaining of GINIP+ hair follicle-innervating neurons, known as C-LTMRs (white) shows similar innervation of back hairy skin in WT (G) and bhlha9 KO (H) animals. Immunostainings of PKCγ (red), CGRP (red), GINIP

(blue) and staining of IB4 (green) positive neurons on thoracic spinal cord slices show similar patterns of innervation of the dorsal horn in WT (I, K) and bhlha9 KO (J, L) animals.

Figure S2: bhlha9 KO and WT male and female mice display similar responses to most thermal sensitivity assays. bhlha9 KO and WT male (A, C) and female (B, D) mice show similar responses to noxious hot and cold stimulation. bhlha9 KO and WT animals behave similarly when given the choice between two different innocuous temperatures: 29°C-29°C (E-F, male and female respectively), 29°C-

24°C (G-H), 29°C-34°C (I, J) and 32°C-37°C (K, L).

Figure S3: bhlha9 KO and WT animals show grossly similar behaviors in response to mechanical stimulation under acute and injury-induced conditions. bhlha9 KO and WT male (A, B) and female (C,

103

D) mice show similar responses to innocuous mechanical stimulation in the tape test. bhlha9 KO and

WT male mice show grossly similar responses to mechanical (E) and noxious cold (F) stimulation following CCI (WT n=11, KO n=8), and similar responses to mechanical (G) and noxious hot (H) stimulation following carrageenan-induced inflammation (WT n=9, KO n=10).

Figure S4: bhlha9 KO and WT lamina IIi neurons show mostly similar passive and active properties.

Voltage responses recorded in lamina IIi neurons in WT (top) and bhlha9 KO (bottom) animals following the injection of a depolarizing or hyperpolarizing current pulse (A). Dashed line indicates the steady state used for sag ratio calculation. Current response recorded in lamina IIi neurons in WT (top) and bhlha9 KO (bottom) animals (B). Dashed lines indicate the slopes used to calculate the IO rectification.

Comparison of cell passive properties (membrane potential, Vm, (C); membrane resistance, Rm, (D); cell capacitance, Cm, (E) in WT and bhlha9 KO animals. Comparison of lamina IIi neurons spiking properties (F) and low threshold currents expression (potassium inward rectifier, (G), Ih current, (H)) in lamina IIi neurons in WT and bhlha9 KO animals.

104

105

106

107

108

Additional results of the study of bhlha9-KO mice

Subcutaneous formalin injection in the neck

Because C-LTMRs innervate exclusively hairy skin, a recurring question in the study of C-LTMRs is whether any behavioral experiment performed by stimulating the hindpaw glabrous skin is actually relevant. A negative answer would entirely disregard the role of C-LTMRs in pain modulation occurring at the level of the spinal cord, for example through tafa4 secretion and modulation of spinal networks’ excitability (Delfini et al., 2013; Kambrun et al., 2018). As such, applying a noxious stimulus to the plantar hindpaw is relevant to the study of C-LTMRs, a caress or a brush probably less so. However, LTMRs per definition are neurons activated by low intensity stimulation, but that does not exclude that they can be activated by noxious stimulations as well. For example, C-LTMRs respond perfectly well to pin-prick (Vallbo et al., 1999). And it is still unclear whether C-LTMRs could also be directly activated by formalin and play a dual role of formalin pain mediator and modulator, similarly to the hypersensitivity they seem to mediate following intra-muscular injection of hypertonic saline solution in human (Nagi et al., 2011). For that reason, I wanted to assess whether subcutaneous injection of formalin in the hairy skin would give rise to a similar phenotype as in glabrous skin. An already described model of hairy skin formalin test requires the injection of a small volume in the cheek of the mouse. But because I found this test quite gruesome, and also involving trigeminal neurons and not DRG neurons, I decided to inject the formalin in the back of the neck. I determined that an injection of 50 µl 5% formalin was necessary to obtain significant spontaneous scratching which lasted about 30 minutes in wild type mice. A monophasic response was always observed. Surprisingly, I did not record any difference in behavior between mutant and control animals (Figure 24). This could be explained by the fact that back skin is actually quite loose in the mouse and formalin may diffuse much more than in the cheek or the hindpaw. The skin may also be thicker and present a different sensitivity to the rest of hairy skin, considering mothers tend to grab pups by the scruff of the neck.

109

Figure 24: Subcutaneous injection of 5% formalin in the back of the neck produces similar nocifensive behavior in WT and bhlha9 KO male and female mice.

RNA sequencing

In parallel to performing a wide array of behavioral tests on bhlha9 KO and WT male and female mice to assess potential consequences of the KO on motor function, stress, thermal and mechanical sensitivity in acute condition and following injury, I also carried out DRG dissections followed by RNA extraction of 7 pairs of lower Thoracic DRGs (T6 to T12) of 6 WT and 6 bhlha9 KO male mice, pooled 2 by 2 producing 3 high-quality RNA samples per genotype (Figure 25). The 2 samples of each genotype of the highest quality were sent for sequencing to the biotech GATC in Germany, and Andy Saurin in the institute performed the bioinformatic analysis of the data generated (Figure 25). The results were the following: bhlha9 KO samples were barely different from WT samples, in fact when comparing the 4 samples to each other, one bhlha9 KO sample was actually more similar to one WT sample than the other WT sample.

110

Figure 25: scheme of the different steps involved in the RNA sequencing experiment of whole thoracic DRGs of adult wild type and bhlha9 KO mice, and analysis of the results.

20 genes came out as either significantly upregulated or downregulated in bhlha9 KO animals compared to WT animals. However, for most of them it concerned either downregulated genes that already showed very low levels of expression in WT or the reverse. I tried to assess the actual profile of expression and potential ectopic expression / deregulation by performing in situ hybridization for each of these genes on fixed DRG slices and could confirm a very slight up or down regulation for few of them. It is worth noting that 25% of the genes that seemed deregulated were either upregulated in already high-expressing vascular cells, or downregulated in low-expressing glial cells.

In conclusion, when pooling the RNAs of whole DRGs of 2 bhlha9 KO animals and comparing them to a similarly pooled RNA sample from whole DRGs of 2 WT animals, no major difference in gene expression can be observed. Three lessons are to be drawn from this experience: it is more relevant not to pool RNA extracts from different animals, and better suited for an RNA sequencing experiment to have 3 samples of each genotype. But more importantly, considering that BHLHA9-positive C-LTMRs represent 9% of total DRG neurons, it would

111 probably be well-advised to sort at least small DRG neurons including C-LTMRs from medium and large ones, if not FACS them, and only sequence RNA from these sorted populations to avoid a possible “flattening” of any differences in gene expression.

Quantitative RT PCR combined with immunostainings / in situ hybridizations

To assess whether KO of bhlha9 perturbed the acquisition by C-LTMRs of the three molecular markers published so far, I performed immunostainings or in situ hybridizations on DRG slices and quantitative RT PCR for Vglut3, Tafa4 and Tyrosine Hydroxylase. Tafa4 seems slightly upregulated in bhlha9 KO DRGs when assessed by in situ hybridization, but this tendency is not significant in qRT-PCR results. However, intriguingly, TH is upregulated by 25% in bhlha9 KO DRGs (Figure 26 left panel). However, the role of TH in the function of C-LTMRs, if there is any, has not been determined yet.

It has been described in the literature that satellite cells present in the DRG around somatosensory neurons play a role in injury-induced pain (Takeda et al., 2009). In order to assess if somehow KO of bhlha9 affected the number or the size of glial cells I performed quantitative RT PCR on whole thoracic DRGs for the markers GFAP and IL-1b, and immunostainings for glutamine synthetase. GFAP and IL-1b seem overexpressed in bhlha9 KO DRG neurons, although they remain at extremely low levels which makes one wonder if this potential increase in expression has any biological relevance (Figure 26 right panel). Immunostainings did not show any difference between bhlha9 KO and WT animals.

Figure 26: quantitative RT PCR on whole thoracic bhlha9 KO and WT DRG RNA. Tyrosine hydroxylase seems upregulated, although its biological function in sensory neurons is still

112 unknown. Similarly, two markers of DRG glial cells, GFAP and IL-1b, seem upregulated in bhlha9 KO DRGs compared to male littermates, but levels of expression are globally extremely low.

Bhlha9 conditional KO mouse

In parallel to rederiving bhlha9 KO mice coming from Israel and generating and characterizing tafa4 transgenic mice, I decided to create a construct allowing the generation of mice conditionally knocked-out for bhlha9 (cKO). Indeed, generating a new mouse line can take up to 2 years and considering the hindpaw malformation observed by Schatz and colleagues on bhlha9 KO mice, it seemed highly relevant to generate a mouse where we could knock-out bhlha9 in DRG neurons without affecting limb development. The construct was relatively simple and involved inserting LoxP sites around the coding sequence of the bhlha9 monoexon so that prior to cre recombination, bhlha9 is still properly expressed (Figure 27). Once the construct generated, I helped screen for recombinant ES cells by Southern Blot, then started the process of back-crossing mice onto a C57BL/6J background for 5 generations (originally on a mixed C57BL/6J – 129/Sv) before crossing them with cre recombinase expressing mice.

113

Figure 27: scheme of the construct I designed and generated to obtain mice conditionally knocked out for bhlha9 in DRG neurons. The coding sequence is surrounded by loxP sites, in a way that, prior to cre recombination, bhlha9 is still properly expressed. Following cre-induced excision of the coding sequence, only cells expressing the Cre recombinase (here Nav1.8-expressing DRG neurons) will lack bhlha9 expression.

By the time I got mice carrying the floxed allele on a 95% pure C57BL/6J background, I had discovered that bhlha9 KO mice showed both a strong defect in the thermotaxis assay and an exacerbated response in the formalin-induced pain test. Hence, I decided to reiterate these tests in mice born from the crossing of female mice homozygous for the bhlha9 floxed allele and nav1.8-cre male mice homozygous for the bhlha9 floxed allele. Thus, pups from these litters would be either bhlha9 cKO or their littermate controls. The two phenotypes were not reproduced in mice lacking bhlha9 expression in nav1.8-expressing DRG neurons, which includes C-LTMRs (Figure 28).

114

Figure 28: Top panel, thermotaxis assay on bhlha9fl/fl and bhlha9fl/fl::nav1.8cre/+ animals. Bottom panels, timecourse (left) and total nocifensive response (right) of bhlha9fl/fl and bhlha9fl/fl::nav1.8cre/+ animals.

I confirmed that mice homozygous for the floxed allele and expressing cre recombinase from the nav1.8 allele were indeed conditional KO by performing quantitative RT PCR in RNA extracted from whole thoracic DRGs of WT animals, bhlha9fl/fl animals and bhlha9fl/fl::nav1.8cre/+ animals (Figure 29). The results show that: 1- expression of bhlha9 is not affected by the presence of the 2 loxP sites around the coding sequence, as the levels of expression of the gene are similar in WT and bhlha9fl/fl animals, 2- cre recombination happened in bhlha9fl/fl animals because these animals express much less bhlha9 than their littermate controls.

115

Figure 29: bhlha9 mRNA levels quantified by quantitative RT PCR on whole DRG from adult WT animal, bhlha9fl/fl and bhlha9fl/fl::nav1.8cre/+ animals.

Because nav1.8 itself is known to be involved in temperature sensation and in pain (Zimmermann et al., 2007), and the nav1.8-cre mouse used being a knock-in with only one allele of the gene, I rapidly decided to repeat the crossing process with Advillin-cre mice. These mice again did not show the same phenotypes as the total bhlha9 KO generated in Israel (Figure 30).

116

Figure 30: Top panel, thermotaxis assay on bhlha9fl/fl and bhlha9fl/fl::avilcre/+ animals. Bottom panels, timecourse (left) and total nocifensive response (right) of bhlha9fl/fl and bhlha9fl/fl::avilcre/+ animals.

I will discuss these results in the next part of my manuscript.

117

Discussion:

Bhlha9 total KO versus bhlha9 cKO DRG specific

The lack of a similar phenotype in terms of temperature sensation and inflammatory pain response in bhlha9 cKO mice compared to bhlha9 total KO mice was a surprise, especially considering that the phenotypes observed were strongly reminiscent of the ones observed in the few other studies where a gene enriched in C-LTMRs was Knocked-Out in the mouse (Tafa4, Cav3.2) (Delfini et al., 2013; Francois et al., 2015).

However, the conclusion we can draw from this result is that knock-out of bhlha9 specifically in DRG is not sufficient to recapitulate the sensory defects observed in the case of a total knock-out. The expression of bhlha9 elsewhere in the organism must be necessary for correct temperature and pain perception. We cannot exclude that the sensory defects exhibited by bhlha9 total KO animals are due to the combined loss of expression both in C-LTMRs and elsewhere. To answer that question, we would need to perform a selective inactivation of our gene in other tissues such as the brain or the mesenchymal tissues expressing Bhlha9. With such mouse models, one can identify the tissue(s) in which loss-of-function of Bhlha9 is responsible for the thermotaxis and formalin-evoked pain phenotypes observed in my global KO mouse.

We know that bhlha9 is expressed in the granular layer of the cerebellum, at lower levels than in the DRG (article figure S1, Allen Brain Atlas database). Hence, a conditional KO specific to this structure would give us a clue as to the role of bhlha9 in the cerebellum. This area of the brain has rarely been described as directly involved in temperature or pain perception. Indeed, the main functions of the cerebellum are body movement coordination and balance. However, cerebellar stroke, due to blockage of the vessels supplying blood to the cerebellum such as the posterior inferior cerebellar artery, can in some rare cases lead to a loss or an increase in pain and temperature sensation (Izumi et al., 1996; Ruscheweyh et al., 2014).

Based on a recent study of the molecular architecture of the mouse nervous system (Zeisel et al., 2018), we now know that bhlha9 is expressed at very low levels in a number of neuronal populations including neurons of the cortex. KO of bhlha9 in the brain with an Emx1-cre mouse for example would be interesting to analyze.

118

During my PhD, I assessed bhlha9 expression in the spinal cord by in situ hybridization at very early stages of development (E13 and E15) and found none, and later on in adult and again found none. This was confirmed by a thorough search of available online databases. However, I never checked for expression during the developmental period where bhlha9 starts to appear in DRG: P7 to P20, which is also a crucial period for spinal cord networks maturation where transient expression of certain genes such as vglut3 is necessary for proper pain modulation later in life (Peirs et al., 2015). This potential transient expression of bhlha9 in the spinal cord would need to be assessed, and if it exists, then spinal cord-specific conditional KO of bhlha9 is necessary, through the use of hoxb8-cre mice for example.

Finally, it is important to keep in mind that BHLHA9 was first described for its role in limb bud development by Schatz and colleagues, and 25% of bhlha9 KO mice present a syndactyly of the forepaw. It is interesting to note that bhlha9 is expressed in both the forepaw and the hindpaw during embryogenesis but only the forepaw shows a malformation in adult, with an incomplete penetrance. Furthermore, male and female bhlha9 KO mice present this congenital malformation but only male KO mice show sensory deficits. But we cannot exclude that a lack of expression of bhlha9 in the developing limb bud affects innervation of the skin of the paw, even when no actual malformation is visible. To rule out this hypothesis, we would need to inactivate bhlha9 specifically in the developing limb bud.

Sexual dimorphism of the sensory phenotypes observed in bhlha9 KO male mice

It is intriguing that only male but not female bhlha9 KO mice show temperature perception and inflammatory pain defects. As I mentioned in the discussion of my article, a study published in 1983 (Batchelder et al., 1983) looked at huddling behavior in male and female mice in response to changes in housing temperature. What they noticed is that while both male and female huddle at low temperatures, males huddle less at warmer temperatures whereas females maintain high levels of huddling. One possible explanation for this could be that while males need to huddle to survive at low temperatures, when the housing environment is more favorable, their naturally territorial and dominant behavior may overcome the need for huddling. The instinct to huddle in female mice however does not seem to only be temperature-motivated and as such their thermosensitive nervous system may

119 have evolved differently. This is a possible explanation for the fact that female bhlha9 KO mice do not present any temperature perception defect whereas males do.

In the case of formalin-induced inflammatory pain response however, the sexual dimorphism presented by male and female bhlha9 KO mice seems different. Indeed, when analyzing mouse behavior in the thermotaxis assay, bhlha9 KO males show a strong defect and bhlha9 KO females behave exactly the same way as their WT littermates. Conversely, in the formalin assay, female pain response seems to follow the same tendency as male, when comparing WT of both sexes and bhlha9 KO of both sexes respectively. Indeed, in both male and female bhlha9 KO there is a tendency for a longer first phase and a higher peak of pain behavior at 25 minutes post-injection, compared to their WT littermates. This could suggest that bhlha9’s mechanism of action in inflammatory pain is similar in male and female mice but either its role is less important in female, or they have a compensatory mechanism that males lack.

The analysis of the time course of the male pain response suggests two potential phenomena: 1- a stronger activation of peripheral fibers during the first phase, which can lead to a stronger sensitization of central networks and so a stronger second phase, and/or 2- a lesser inhibitory tone in bhlha9 KO male mice central networks which leads to more pain both during the first and the second phase.

Pertaining to the first hypothesis, the fact that DRG-specific conditional knock-out of bhlha9 does not seem to affect first nor second phase pain response suggests that the increased formalin pain response of total KO male mice is not due to a peripheral component.

The increased inflammatory pain exhibited by bhlha9 KO male mice is insensitive to positive modulation of GABAA receptors (PAM), whereas it reduces pain by more than 70% in WT animals. Thus, the role of bhlha9 in inflammatory pain involves the ionotropic GABAergic system which supports my second hypothesis. Another argument in favor of my hypothesis comes from electrophysiological recording on spinal cord slices. Indeed, using whole cell patch clamp, we were able to show that under naïve conditions, the average amplitude of excitatory synaptic events was significantly higher in bhlha9 KO spinal cord slices than in WT littermates.

It would be very interesting to assess the effect of PAM on formalin-induced inflammatory pain in bhlha9 KO female mice also. Even though they are not in significantly more pain than WT females, if PAM administration does not produce an analgesic effect in bhlha9 KO females,

120 it means that bhlha9 does play a similar role in relation to the ionotropic GABAergic system in females and in males. But this could also have one major implication: that the involvement of the ionotropic GABAergic system is much less important in formalin-induced inflammatory pain in female mice than in male mice.

It is well known that chronic pain affects predominantly women (Mogil, 2012). However, very few studies so far have aimed at deciphering the neuronal differences between male and female, especially in the field of pain research. One recent study investigated by RNA sequencing and quantitative RT PCR potential differences in gene expression between male and female mice DRG neurons, immune cells or dorsal horn immune cells, at steady-state and after spinal nerve ligation (Lopes et al., 2017). No significant difference was found at steady- state in any of these populations that could explain sexual dimorphism in pain. However, the authors found an intriguing difference in the immune cells that infiltrate the DRG in response to injury. More precisely, they found that there was a high increase in the frequency of B cells in male mice and a prevalence of T cells in female mice. Thus, in the peripheral component of pain, the difference between male and female seems to reveal itself after injury. No such experiment was performed on dorsal horn populations of neurons. But even if such populations were to be similar in male and female, the complex architecture of the synaptic networks of the dorsal horn would need to be finely analyzed before we could assert that there is no sex difference in the integration of the pain signal at this level. Indeed, subtle differences in GABA signaling have already been described in the periaqueductal gray (PAG) area in the brain (Tonsfeldt et al., 2016). In this study, the authors show that ionotropic GABAergic signaling in the ventrolateral PAG, a structure of the descending pain modulation pathway, is affected differently by persistent inflammation in male and female rats. Practically speaking, morphine induces greater antinociception in females than males, in the case of CFA- induced inflammation. And this antinociception is reversed by positive modulation of GABAA receptors in females but not in males. Similar sex differences in GABAA modulation of formalin pain at the level of the spinal cord and/or the brain could explain the sexual dimorphism we find in bhlha9 KO male and female mice.

In this study, I have discovered that BHLHA9 is a transcription factor with a very restricted pattern of expression in the nervous system. Its expression appears in C-LTMRs only after birth making it an interesting genetic tool to study them. Male mice lacking BHLHA9 show a wide-

121 ranging thermotaxis defect compared to littermate controls, and compared to female mice. Moreover, bhlha9 KO male mice present an exacerbated inflammatory pain in the formalin test paradigm. Intriguingly, this exacerbated pain is insensitive to the normally analgesic effect of GABAA receptor positive modulation. In conclusion, I uncovered BHLHA9 as a new molecule required for correct perception of temperature and for inflammatory pain processing, in a sexually dimorphic manner.

The study of C-LTMRs

Even though it turns out that expression of BHLHA9 in C-LTMRs may not actually be the main component at the origin of the sensory defects presented by bhlha9 total KO mice, the main goal of my PhD was originally to study C-LTMRs themselves and gain knowledge as to their developmental specification and their function in adult. We designed the first two mice that we generated with that objective in mind (bhlha9 KO-KI and Tafa4 transgenic mice) and I spent quite a lot of time envisioning the experiments that would allow us to characterize better these neurons. I will now discuss this aspect of my PhD work.

I mentioned in my introduction the many functions of C-LTMRs, demonstrated or still up to debate. These range from a role in affective and affiliative light touch to pain modulation, but also temperature perception and possibly mechanical itch.

How can we study specifically C-LTMRs ? The ultimate key to the study of C-LTMRs, which we have been missing until now, is a genetic tool that would allow us to probe specifically their function. In order to design one, it is clear by now that we need 1- an excellent understanding of the genes expressed by this population of neurons during development and in adult, 2- intersectional genetics.

Intersectional genetics is a genetic tool that allows a more precise restriction of the expression of a transgene of interest, such as a reporter, by combining the use of the cre-lox system and the FRT-flippase system. This transgene is located downstream of two stop cassettes flanked by either loxP sites or FRT sites, and these cassettes will be excised in cells expressing the corresponding recombinases, cre or flippase respectively. Hence, the transgene will be expressed only in cells expressing both recombinases. Moreover, the mice lines currently

122 available for intersectional genetics containing a transgene of interest preceded by the two stop cassettes have been generated in the Tau locus (mapt), which allows inherent restriction to the nervous system.

The list of genes that we know to be quite specific to C-LTMRs in DRG in adult is rather short: Tyrosine Hydroxylase, the transporter of glutamate Vglut3, Tafa4 and now bhlha9. Any of these genes shows a dynamic pattern of expression during development in the DRG and/or is expressed elsewhere in the organism in adult.

And so, it will be necessary to cross mice expressing cre and flippase under the control of two genes with a relevant pattern of expression to doubly restrict to C-LTMRs the expression of any reporter/cassette expressing a gene of interest such as DTR, inhibitory or excitatory DREADDs. Crossing a mouse expressing a recombinase under the control of any of the 4 genes mentioned above with a mouse expressing the other recombinase in DRG (Advillin, nav1.8) could be a good approach. But it still has its limitations: Advillin for example is also expressed in sympathetic neurons which share the expression of TH with DRG neurons, it is also expressed in DRG from E14 onward and any gene expressed at the same time with a dynamic pattern of expression could lead to expression of the transgene of interest in other populations. As a consequence, the first step after generating such a mouse will be to check that only C-LTMRs have expressed at one time both of the genes controlling the expression of the recombinases. This can be done with the use of a reporter such as GFP, such a line is available at the Jackson Laboratory. Any other “double-stop” lines have also been generated by Martyn Goulding but are not yet available at the Jackson Laboratory.

What to do with a “C-LTMR-specific” mouse? Understand their circuitry The various ways to study C-LTMRs are not quite endless but almost. As I mentioned, they seem to play a role in temperature perception, social behavior and pain modulation. Thus, once we have a genetic tool specific to these neurons, their function can be assessed in as many contexts.

It seems obvious that we need to determine what their role is in pain modulation and how it works, and for that we need to understand the pathway involved from the first neurons to

123 receive input from them in the spinal cord, which are multiple (Abraira et al., 2017), to the rest of the network involved in the processing of pain modulated by C-LTMRs, all the way to the brain regions receiving this information, and how it is affected when C-LTMRs are defective. Discovering this spinal network will be helpful not only in the context of pain but also to understand how these neurons can transduce innocuous sensations such as cooling and caress, and whether the spinal network involved is the same. In order to do this, we need to perform tracing experiments, for example by viral infection. This is a complex experiment considering that C-LTMRs are among the most difficult DRG neurons to infect (Albisetti et al., 2017). Albisetti and colleagues performed retrograde infection of peripheral sensory neurons from the spinal cord but they mention that even with an optimized version of the rabies virus infection, almost no C-LTMRs were infected. The other option is flank injection of the rabies virus in hairy skin which should allow trans-synaptic tracing of next-order spinal neurons as described by Bourane and colleagues in 2015 (Bourane et al., 2015), a rare study mentioning C-LTMR infection by viruses. However, the authors do not mention the efficiency of infection.

When considering the various functions of C-LTMRs, one wonders how they can transduce such different innocuous sensations as cooling and caress if the next-order neurons receiving the input are the same in both cases and at least one of the target area in the brain is also the same (the insular cortex). One thing to take into account is that the actual molecules at the membrane of the neurons that allow peripheral detection of such stimuli are probably different and can lead to a different firing pattern in terms of the electrophysiological response induced. One recent paper compares how C-LTMRs respond to thermal and mechanical stimulation and shows that in the case of cooling, afterdischarges are often observable, but not in the case of slow caress (Ackerley et al., 2018). This could be one of the mechanisms underpinning the transduction by C-LTMRs of such two different sensations, resulting in a globally different output transmitted by the spinal cord to the brain.

Pertaining to the spinal network relevant for pain modulation by C-LTMRs, it may actually be different from the one considered in the case of steady-state stimulation of C-LTMRs by cooling or caress. Indeed, when put in the context of the gate control theory (Melzack and Wall, 1965), the role of C-LTMRs in pain modulation could be directly related to the plasticity occurring in the spinal cord in terms of inhibitory tone and circuitry. Following injury, we know that a certain amount of plasticity is at play in the spinal cord in terms of glial activation but

124 also in terms of reorganization of the inhibitory network. For example, following CFA injection, the balance between the glycinergic and the GABAergic inhibitory tone is reversed in lamina II (Takazawa et al., 2017). The idea is that following injury, certain peripheral neurons such as LTMRs “gain access” to pain circuits that they normally would not be able to activate thanks to primary afferent depolarization by GABA for example (Witschi et al., 2011). This new access to the pain pathway would be due to what is called disinhibition: the synapses between the central afferences of C-LTMRs and the pain pathway already existed but were under inhibitory control, and due to a decrease in inhibition, C-LTMR stimulation could now lead to the activation of a pain pathway.

Being able to map functional synapses between C-LTMRs and next-order spinal neurons will allow us to understand C-LTMR connectivity at the level of the spinal cord in physiology and pathology. But this would be an incomplete picture of their function.

Understand the extent of their role in physiology and pathology

Studies in human on the role of C-LTMRs, or CT-afferents in that case, have outlined different functions for this population of LTMRs in social and affective touch and in injury-induced pain modulation (Krahe et al., 2018; Shaikh et al., 2015). However, because light touch activates also myelinated LTMRs, correlations are observed between “CT-optimal touch” (light slow touch, around 3 cm/sec) and hedonic ratings or changes in heartbeat or attentional engagement in infants (Fairhurst et al., 2014). For example, the “Midas touch” effect proposes that light touch contributes to positive feelings and reinforces prosocial behavior. We also know that emotions affect skin temperature and hedonic ratings of touch, and the reverse is also true in that skin temperature affects how we perceive light touch. Pregnant women lightly stroking their belly tends to appease both the mother-to-be and the baby, and according to midwives it could actually reduce post-partum depression. Similarly, maternal stroking of the infant in the first weeks of life is correlated with lower postnatal depression rates (Murgatroyd et al., 2015). One possible mechanism underlying the positive effects of gentle touch could be an increase in the release of the “feel-good” hormone oxytocin and/or a decrease in cortisol blood levels. Conversely, although the ASD spectrum is wide, children suffering from autistic symptoms tend to avoid social interactions and interpersonal touch but sometimes show an

125 increased fondness for other tactile stimulations by repetitively stroking certain surfaces or objects with a particular texture. In this latter example of a possible involvement of CT- afferents in everyday life events, one cannot help but note that interpersonal touch tends to involve stimulation of hairy skin (arm, shoulder…) whereas stroking something involves stimulating the glabrous skin of the palm and the fingers. Moreover, CT-optimal touch is the most pleasant one in normal conditions, but following injury, for example capsaicin injection, it becomes the most unpleasant one. This suggests that CT stimulation-induced allodynia, which may be the case in some forms of ASD and chronic pain, results from perturbed processing of CT input. In conclusion, human studies have opened avenues into investigating the many different ways in which CT-afferents are not only necessary for normal social interactions, but could actually be considered as a tool to treat or at least alleviate symptoms of stress, depression. But C-LTMR-focused animal research is needed to confirm and further our knowledge with the use of appropriate genetic tools.

What would be the read-out of the emotional aspect of C-LTMR function in mouse?

Mice are animal models which allow us to investigate many aspects of sensory research, but the study of the emotional aspect of light touch, such as caress, is limited. Tests such as conditioned place preference (CPP) now allow us to discriminate and study emotional pain independently from evoked mechanical pain (usually assessed by Von Frey stimulations). In terms of read-out of the emotional aspect of innocuous stimulation, Vrontou and colleagues showed in their study of MrgprB4-positive light touch-sensitive C-fibers that CPP is a test particularly relevant to assess whether the activation of a subpopulation of sensory neurons is associated with positive reinforcement. As such, it seems particularly important to include this kind of test in the study of C-LTMRs, be it in a physiological or in a pathological context.

For now, it seems like CPP is the only test that could provide a direct read-out of the emotional aspect of C-LTMR activation. A second, more indirect, option to decipher the effect of C-LTMR stimulation on brain activity in mouse is to couple their stimulation with functional Magnetic Resonance Imaging (fMRI). Although this area of C-LTMR study is still ongoing in the mouse, we know that CTs activate a certain number of brain areas in human. One structure often mentioned in relation to CT activation is the posterior insula (Olausson et al., 2002), which has

126 been associated with the emotional aspect of pain (Davis, 2000) and is also activated by cooling (Craig et al., 2000). Similarly to myelinated mechanoreceptors projecting in a somatotopically organized manner onto the somatosensory cortex areas S1 and S2, CT- induced activation of the posterior insula seems to be somatotopically organized: a 3cm/sec light stimulation of the arm detected more anteriorly in the insula than a stimulation of the thigh (Bjornsdotter et al., 2009). This lends support to the hypothesis that the posterior insula is a primary region of CT input processing in the brain, which may provide a read-out independent of myelinated LTMR activation, these discriminative touch neurons projecting more importantly to the S1 and S2 somatosensory cortex.

Are CTs in human and C-LTMRs in mouse the same neurons?

One more aspect of affective touch research we need to take into account is that human CTs and mouse C-LTMRs may present differences in terms of molecular content, function, target areas in the brain… Indeed, 1- mice are much hairier than human, 2- Aδ-LTMRs have never been recorded in hairy skin which, if they actually do not exist in human hairy skin, suggests a reorganization of the functions subserved by the remaining Aβ and C-LTMRs. Finally, piezo2 is a mechanosensitive channel broadly expressed in mouse and human DRG neurons including in C-LTMRs. Lou and colleagues infer in 2013 that piezo2 is necessary for C-LTMR mechanosensitivity (Lou et al., 2013). Recent findings show that people lacking the expression of piezo2 are still fully capable of detecting CT-optimal light touch, although they perceive it as pricking but non-painful (Chesler et al., 2016). Conversely, mice conditionally knocked-out for piezo2 in sensory neurons do not feel gentle touch anymore (Szczot et al., 2018). Such a discrepancy raises the question of possible differences that may have appeared along the evolution between rodents, maybe cats, C-LTMRs and human CTs.

127

Abrahamsen, B., Zhao, J., Asante, C.O., Cendan, C.M., Marsh, S., Martinez-Barbera, J.P., Nassar, M.A., Dickenson, A.H., and Wood, J.N. (2008). The cell and molecular basis of mechanical, cold, and inflammatory pain. Science 321, 702-705. Abraira, V.E., and Ginty, D.D. (2013). The sensory neurons of touch. Neuron 79, 618-639. Abraira, V.E., Kuehn, E.D., Chirila, A.M., Springel, M.W., Toliver, A.A., Zimmerman, A.L., Orefice, L.L., Boyle, K.A., Bai, L., Song, B.J., et al. (2017). The Cellular and Synaptic Architecture of the Mechanosensory Dorsal Horn. Cell 168, 295-310 e219. Ackerley, R., Backlund Wasling, H., Liljencrantz, J., Olausson, H., Johnson, R.D., and Wessberg, J. (2014). Human C-tactile afferents are tuned to the temperature of a skin-stroking caress. J Neurosci 34, 2879- 2883. Ackerley, R., Wiklund Fernstrom, K., Backlund Wasling, H., Watkins, R.H., Johnson, R.D., Vallbo, A., and Wessberg, J. (2018). Differential effects of radiant and mechanically applied thermal stimuli on human C-tactile afferent firing patterns. J Neurophysiol 120, 1885-1892. Albisetti, G.W., Ghanem, A., Foster, E., Conzelmann, K.K., Zeilhofer, H.U., and Wildner, H. (2017). Identification of Two Classes of Somatosensory Neurons That Display Resistance to Retrograde Infection by Rabies Virus. J Neurosci 37, 10358-10371. Batchelder, P., Kinney, R.O., Demlow, L., and Lynch, C.B. (1983). Effects of temperature and social interactions on huddling behavior in Mus musculus. Physiol Behav 31, 97-102. Beaudry, H., Daou, I., Ase, A.R., Ribeiro-da-Silva, A., and Seguela, P. (2017). Distinct behavioral responses evoked by selective optogenetic stimulation of the major TRPV1+ and MrgD+ subsets of C- fibers. Pain 158, 2329-2339. Bennett, G.J., and Xie, Y.K. (1988). A peripheral mononeuropathy in rat that produces disorders of pain sensation like those seen in man. Pain 33, 87-107. Bessou, P., Burgess, P.R., Perl, E.R., and Taylor, C.B. (1971). Dynamic properties of mechanoreceptors with unmyelinated (C) fibers. J Neurophysiol 34, 116-131. Bjornsdotter, M., Loken, L., Olausson, H., Vallbo, A., and Wessberg, J. (2009). Somatotopic organization of gentle touch processing in the posterior insular cortex. J Neurosci 29, 9314-9320. Bourane, S., Duan, B., Koch, S.C., Dalet, A., Britz, O., Garcia-Campmany, L., Kim, E., Cheng, L., Ghosh, A., Ma, Q., et al. (2015). Gate control of mechanical itch by a subpopulation of spinal cord interneurons. Science 350, 550-554. Brenner, D.S., Golden, J.P., and Gereau, R.W. (2012). A Novel Behavioral Assay for Measuring Cold Sensation in Mice. Plos One 7. Britz, O., Zhang, J., Grossmann, K.S., Dyck, J., Kim, J.C., Dymecki, S., Gosgnach, S., and Goulding, M. (2015). A genetically defined asymmetry underlies the inhibitory control of flexor-extensor locomotor movements. Elife 4. Brown, A.G., and Iggo, A. (1967). A quantitative study of cutaneous receptors and afferent fibres in the cat and rabbit. J Physiol 193, 707-733. Brumovsky, P., Villar, M.J., and Hokfelt, T. (2006). Tyrosine hydroxylase is expressed in a subpopulation of small dorsal root ganglion neurons in the adult mouse. Exp Neurol 200, 153-165. Brumovsky, P.R. (2016). Dorsal root ganglion neurons and tyrosine hydroxylase--an intriguing association with implications for sensation and pain. Pain 157, 314-320. Cavanaugh, D.J., Lee, H., Lo, L., Shields, S.D., Zylka, M.J., Basbaum, A.I., and Anderson, D.J. (2009). Distinct subsets of unmyelinated primary sensory fibers mediate behavioral responses to noxious thermal and mechanical stimuli. Proc Natl Acad Sci U S A 106, 9075-9080. Chaplan, S.R., Bach, F.W., Pogrel, J.W., Chung, J.M., and Yaksh, T.L. (1994). Quantitative assessment of tactile allodynia in the rat paw. J Neurosci Methods 53, 55-63. Cheng, L., Duan, B., Huang, T., Zhang, Y., Chen, Y., Britz, O., Garcia-Campmany, L., Ren, X., Vong, L., Lowell, B.B., et al. (2017). Identification of spinal circuits involved in touch-evoked dynamic mechanical pain. Nat Neurosci 20, 804-814. Chesler, A.T., Szczot, M., Bharucha-Goebel, D., Ceko, M., Donkervoort, S., Laubacher, C., Hayes, L.H., Alter, K., Zampieri, C., Stanley, C., et al. (2016). The Role of PIEZO2 in Human Mechanosensation. N Engl J Med 375, 1355-1364. 128

Cole, J., Bushnell, M.C., McGlone, F., Elam, M., Lamarre, Y., Vallbo, A., and Olausson, H. (2006). Unmyelinated tactile afferents underpin detection of low-force monofilaments. Muscle Nerve 34, 105- 107. Cole, J.D., and Sedgwick, E.M. (1992). The perceptions of force and of movement in a man without large myelinated sensory afferents below the neck. J Physiol 449, 503-515. Correa, C.R., and Calixto, J.B. (1993). Evidence for participation of B1 and B2 kinin receptors in formalin- induced nociceptive response in the mouse. Br J Pharmacol 110, 193-198. Craig, A.D., Chen, K., Bandy, D., and Reiman, E.M. (2000). Thermosensory activation of insular cortex. Nat Neurosci 3, 184-190. Crowley, C., Spencer, S.D., Nishimura, M.C., Chen, K.S., Pitts-Meek, S., Armanini, M.P., Ling, L.H., McMahon, S.B., Shelton, D.L., Levinson, A.D., et al. (1994). Mice lacking nerve growth factor display perinatal loss of sensory and sympathetic neurons yet develop basal forebrain cholinergic neurons. Cell 76, 1001-1011. Croy, I., Luong, A., Triscoli, C., Hofmann, E., Olausson, H., and Sailer, U. (2016). Interpersonal stroking touch is targeted to C tactile afferent activation. Behav Brain Res 297, 37-40. Davis, K.D. (2000). The neural circuitry of pain as explored with functional MRI. Neurol Res 22, 313- 317. Delfini, M.C., Mantilleri, A., Gaillard, S., Hao, J., Reynders, A., Malapert, P., Alonso, S., Francois, A., Barrere, C., Seal, R., et al. (2013). TAFA4, a chemokine-like protein, modulates injury-induced mechanical and chemical pain hypersensitivity in mice. Cell Rep 5, 378-388. Dong, X., Han, S., Zylka, M.J., Simon, M.I., and Anderson, D.J. (2001). A diverse family of GPCRs expressed in specific subsets of nociceptive sensory neurons. Cell 106, 619-632. Douglas, W.W., and Ritchie, J.M. (1957). Nonmedullated fibres in the saphenous nerve which signal touch. J Physiol 139, 385-399. Douglas, W.W., Ritchie, J.M., and Straub, R.W. (1960). The role of nonmyelinated fibres in signalling cooling of the skin. J Physiol 150, 266-283. Draxler, P., Honsek, S.D., Forsthuber, L., Hadschieff, V., and Sandkuhler, J. (2014). VGluT3(+) primary afferents play distinct roles in mechanical and cold hypersensitivity depending on pain etiology. J Neurosci 34, 12015-12028. Duan, B., Cheng, L., Bourane, S., Britz, O., Padilla, C., Garcia-Campmany, L., Krashes, M., Knowlton, W., Velasquez, T., Ren, X., et al. (2014). Identification of spinal circuits transmitting and gating mechanical pain. Cell 159, 1417-1432. Dykes, I.M., Tempest, L., Lee, S.I., and Turner, E.E. (2011). Brn3a and Islet1 act epistatically to regulate the gene expression program of sensory differentiation. J Neurosci 31, 9789-9799. Fairhurst, M.T., Loken, L., and Grossmann, T. (2014). Physiological and behavioral responses reveal 9- month-old infants' sensitivity to pleasant touch. Psychol Sci 25, 1124-1131. Fernandez, J.M., Davalos, A., Ferrer, I., Cervera, C., Codina, A., and Ochoa, J.L. (1994). Acute sensory neuronopathy: report of a child with remarkable clinical recovery. Neurology 44, 762-764. Francois, A., Schuetter, N., Laffray, S., Sanguesa, J., Pizzoccaro, A., Dubel, S., Mantilleri, A., Nargeot, J., Noel, J., Wood, J.N., et al. (2015). The Low-Threshold Calcium Channel Cav3.2 Determines Low- Threshold Mechanoreceptor Function. Cell Rep. Franklin, K.B., and Abbott, F.V. (1993). Pentobarbital, diazepam, and ethanol abolish the interphase diminution of pain in the formalin test: evidence for pain modulation by GABAA receptors. Pharmacol Biochem Behav 46, 661-666. Fromy, B., Josset-Lamaugarny, A., Aimond, G., Pagnon-Minot, A., Marics, I., Tattersall, G.J., Moqrich, A., and Sigaudo-Roussel, D. (2018). Disruption of TRPV3 Impairs Heat-Evoked Vasodilation and Thermoregulation: A Critical Role of CGRP. J Invest Dermatol 138, 688-696. Fukuoka, M., Miyachi, Y., and Ikoma, A. (2013). Mechanically evoked itch in humans. Pain 154, 897- 904. Gaillard, S., Lo Re, L., Mantilleri, A., Hepp, R., Urien, L., Malapert, P., Alonso, S., Deage, M., Kambrun, C., Landry, M., et al. (2014). GINIP, a Galphai-interacting protein, functions as a key modulator of peripheral GABAB receptor-mediated analgesia. Neuron 84, 123-136. 129

Gascon, E., Gaillard, S., Malapert, P., Liu, Y., Rodat-Despoix, L., Samokhvalov, I.M., Delmas, P., Helmbacher, F., Maina, F., and Moqrich, A. (2010). Hepatocyte growth factor-Met signaling is required for Runx1 extinction and peptidergic differentiation in primary nociceptive neurons. J Neurosci 30, 12414-12423. Gorokhova, S., Gaillard, S., Urien, L., Malapert, P., Legha, W., Baronian, G., Desvignes, J.P., Alonso, S., and Moqrich, A. (2014). Uncoupling of molecular maturation from peripheral target innervation in nociceptors expressing a chimeric TrkA/TrkC receptor. PLoS Genet 10, e1004081. Guan, Y., Liu, Q., Tang, Z., Raja, S.N., Anderson, D.J., and Dong, X. (2010). Mas-related G-protein- coupled receptors inhibit pathological pain in mice. Proc Natl Acad Sci U S A 107, 15933-15938. Hahn, J.F. (1971). Thermal-mechanical stimulus interactions in low-threshold C-fiber mechanoreceptors of cat. Exp Neurol 33, 607-617. Haring, M., Zeisel, A., Hochgerner, H., Rinwa, P., Jakobsson, J.E.T., Lonnerberg, P., La Manno, G., Sharma, N., Borgius, L., Kiehn, O., et al. (2018). Neuronal atlas of the dorsal horn defines its architecture and links sensory input to transcriptional cell types. Nat Neurosci 21, 869-880. Hasegawa, H., Abbott, S., Han, B.X., Qi, Y., and Wang, F. (2007). Analyzing somatosensory axon projections with the sensory neuron-specific Advillin gene. J Neurosci 27, 14404-14414. Hunskaar, S., and Hole, K. (1987). The formalin test in mice: dissociation between inflammatory and non-inflammatory pain. Pain 30, 103-114. Iggo, A. (1960). Cutaneous mechanoreceptors with afferent C fibres. J Physiol 152, 337-353. Iggo, A., and Kornhuber, H.H. (1977). A quantitative study of C-mechanoreceptors in hairy skin of the cat. J Physiol 271, 549-565. Izumi, M., Terao, S., Sobue, G., Koshimura, J., Takatsu, S., Yokoi, Y., Takahashi, M., and Mitsuma, T. (1996). [Clinical features of anterior inferior cerebellar artery territory infarcts--a study of ten patients]. No To Shinkei 49, 152-156. Johansen, J.P., Fields, H.L., and Manning, B.H. (2001). The affective component of pain in rodents: direct evidence for a contribution of the anterior cingulate cortex. Proc Natl Acad Sci U S A 98, 8077- 8082. Johansson, R.S., Trulsson, M., Olsson, K.A., and Abbs, J.H. (1988). Mechanoreceptive afferent activity in the infraorbital nerve in man during speech and chewing movements. Exp Brain Res 72, 209-214. Kambrun, C., Roca-Lapirot, O., Salio, C., Landry, M., Moqrich, A., and Le Feuvre, Y. (2018). TAFA4 Reverses Mechanical Allodynia through Activation of GABAergic Transmission and Microglial Process Retraction. Cell Rep 22, 2886-2897. Kim, J., Lo, L., Dormand, E., and Anderson, D.J. (2003). SOX10 maintains multipotency and inhibits neuronal differentiation of neural crest stem cells. Neuron 38, 17-31. Krahe, C., von Mohr, M., Gentsch, A., Guy, L., Vari, C., Nolte, T., and Fotopoulou, A. (2018). Sensitivity to CT-optimal, Affective Touch Depends on Adult Attachment Style. Sci Rep 8, 14544. Kramer, I., Sigrist, M., de Nooij, J.C., Taniuchi, I., Jessell, T.M., and Arber, S. (2006). A role for Runx transcription factor signaling in dorsal root ganglion sensory neuron diversification. Neuron 49, 379- 393. Kumazawa, T., and Perl, E.R. (1977a). Primate cutaneous receptors with unmyelinated (C) fibres and their projection to the substantia gelatinosa. J Physiol (Paris) 73, 287-304. Kumazawa, T., and Perl, E.R. (1977b). Primate cutaneous sensory units with unmyelinated (C) afferent fibers. J Neurophysiol 40, 1325-1338. Lagerstrom, M.C., Rogoz, K., Abrahamsen, B., Persson, E., Reinius, B., Nordenankar, K., Olund, C., Smith, C., Mendez, J.A., Chen, Z.F., et al. (2010). VGLUT2-dependent sensory neurons in the TRPV1 population regulate pain and itch. Neuron 68, 529-542. Li, L., Rutlin, M., Abraira, V.E., Cassidy, C., Kus, L., Gong, S., Jankowski, M.P., Luo, W., Heintz, N., Koerber, H.R., et al. (2011). The functional organization of cutaneous low-threshold mechanosensory neurons. Cell 147, 1615-1627. Li, Y., Stam, F.J., Aimone, J.B., Goulding, M., Callaway, E.M., and Gage, F.H. (2013). Molecular layer perforant path-associated cells contribute to feed-forward inhibition in the adult dentate gyrus. Proc Natl Acad Sci U S A 110, 9106-9111. 130

Liljencrantz, J., and Olausson, H. (2014). Tactile C fibers and their contributions to pleasant sensations and to tactile allodynia. Front Behav Neurosci 8, 37. Lindeberg, J., Usoskin, D., Bengtsson, H., Gustafsson, A., Kylberg, A., Soderstrom, S., and Ebendal, T. (2004). Transgenic expression of Cre recombinase from the tyrosine hydroxylase locus. Genesis 40, 67- 73. Liu, Q., Tang, Z., Surdenikova, L., Kim, S., Patel, K.N., Kim, A., Ru, F., Guan, Y., Weng, H.J., Geng, Y., et al. (2009). Sensory neuron-specific GPCR Mrgprs are itch receptors mediating chloroquine-induced pruritus. Cell 139, 1353-1365. Liu, Q., Vrontou, S., Rice, F.L., Zylka, M.J., Dong, X., and Anderson, D.J. (2007). Molecular genetic visualization of a rare subset of unmyelinated sensory neurons that may detect gentle touch. Nat Neurosci 10, 946-948. Liu, Y., Latremoliere, A., Li, X., Zhang, Z., Chen, M., Wang, X., Fang, C., Zhu, J., Alexandre, C., Gao, Z., et al. (2018). Touch and tactile neuropathic pain sensitivity are set by corticospinal projections. Nature 561, 547-550. Loken, L.S., Evert, M., and Wessberg, J. (2011). Pleasantness of touch in human glabrous and hairy skin: order effects on affective ratings. Brain Res 1417, 9-15. Lopes, D.M., Malek, N., Edye, M., Jager, S.B., McMurray, S., McMahon, S.B., and Denk, F. (2017). Sex differences in peripheral not central immune responses to pain-inducing injury. Sci Rep 7, 16460. Lou, S., Duan, B., Vong, L., Lowell, B.B., and Ma, Q. (2013). Runx1 controls terminal morphology and mechanosensitivity of VGLUT3-expressing C-mechanoreceptors. J Neurosci 33, 870-882. Lou, S., Pan, X., Huang, T., Duan, B., Yang, F.C., Yang, J., Xiong, M., Liu, Y., and Ma, Q. (2015). Incoherent feed-forward regulatory loops control segregation of C-mechanoreceptors, nociceptors, and pruriceptors. J Neurosci 35, 5317-5329. Low, P.A., Burke, W.J., and McLeod, J.G. (1978). Congenital sensory neuropathy with selective loss of small myelinated fibers. Ann Neurol 3, 179-182. Lu, Y., and Perl, E.R. (2003). A specific inhibitory pathway between substantia gelatinosa neurons receiving direct C-fiber input. J Neurosci 23, 8752-8758. Luo, W., Enomoto, H., Rice, F.L., Milbrandt, J., and Ginty, D.D. (2009). Molecular identification of rapidly adapting mechanoreceptors and their developmental dependence on ret signaling. Neuron 64, 841- 856. Ma, Q., Fode, C., Guillemot, F., and Anderson, D.J. (1999). Neurogenin1 and neurogenin2 control two distinct waves of neurogenesis in developing dorsal root ganglia. Genes Dev 13, 1717-1728. Malmberg, A.B., Brandon, E.P., Idzerda, R.L., Liu, H., McKnight, G.S., and Basbaum, A.I. (1997). Diminished inflammation and nociceptive pain with preservation of neuropathic pain in mice with a targeted mutation of the type I regulatory subunit of cAMP-dependent protein kinase. J Neurosci 17, 7462-7470. Marics, I., Malapert, P., Reynders, A., Gaillard, S., and Moqrich, A. (2014). Acute heat-evoked temperature sensation is impaired but not abolished in mice lacking TRPV1 and TRPV3 channels. PLoS One 9, e99828. Marmigere, F., Montelius, A., Wegner, M., Groner, Y., Reichardt, L.F., and Ernfors, P. (2006). The Runx1/AML1 transcription factor selectively regulates development and survival of TrkA nociceptive sensory neurons. Nat Neurosci 9, 180-187. Matthies, B.K., and Franklin, K.B. (1992). Formalin pain is expressed in decerebrate rats but not attenuated by morphine. Pain 51, 199-206. McCoy, E.S., Taylor-Blake, B., Street, S.E., Pribisko, A.L., Zheng, J., and Zylka, M.J. (2013). Peptidergic CGRPalpha primary sensory neurons encode heat and itch and tonically suppress sensitivity to cold. Neuron 78, 138-151. McGlone, F., Olausson, H., Boyle, J.A., Jones-Gotman, M., Dancer, C., Guest, S., and Essick, G. (2012). Touching and feeling: differences in pleasant touch processing between glabrous and hairy skin in humans. Eur J Neurosci 35, 1782-1788.

131

McNamara, C.R., Mandel-Brehm, J., Bautista, D.M., Siemens, J., Deranian, K.L., Zhao, M., Hayward, N.J., Chong, J.A., Julius, D., Moran, M.M., et al. (2007). TRPA1 mediates formalin-induced pain. Proc Natl Acad Sci U S A 104, 13525-13530. Melzack, R., and Wall, P.D. (1965). Pain mechanisms: a new theory. Science 150, 971-979. Mogil, J.S. (2012). Sex differences in pain and pain inhibition: multiple explanations of a controversial phenomenon. Nat Rev Neurosci 13, 859-866. Moqrich, A., Earley, T.J., Watson, J., Andahazy, M., Backus, C., Martin-Zanca, D., Wright, D.E., Reichardt, L.F., and Patapoutian, A. (2004). Expressing TrkC from the TrkA locus causes a subset of dorsal root ganglia neurons to switch fate. Nat Neurosci 7, 812-818. Moqrich, A., Hwang, S.W., Earley, T.J., Petrus, M.J., Murray, A.N., Spencer, K.S., Andahazy, M., Story, G.M., and Patapoutian, A. (2005). Impaired thermosensation in mice lacking TRPV3, a heat and camphor sensor in the skin. Science 307, 1468-1472. Morrison, I., Loken, L.S., Minde, J., Wessberg, J., Perini, I., Nennesmo, I., and Olausson, H. (2011). Reduced C-afferent fibre density affects perceived pleasantness and empathy for touch. Brain 134, 1116-1126. Munro, G., Erichsen, H.K., Rae, M.G., and Mirza, N.R. (2011). A question of balance--positive versus negative allosteric modulation of GABA(A) receptor subtypes as a driver of analgesic efficacy in rat models of inflammatory and neuropathic pain. Neuropharmacology 61, 121-132. Murgatroyd, C., Quinn, J.P., Sharp, H.M., Pickles, A., and Hill, J. (2015). Effects of prenatal and postnatal depression, and maternal stroking, at the glucocorticoid receptor gene. Transl Psychiatry 5, e560. Nagi, S.S., and Mahns, D.A. (2013). C-tactile fibers contribute to cutaneous allodynia after eccentric exercise. J Pain 14, 538-548. Nagi, S.S., Rubin, T.K., Chelvanayagam, D.K., Macefield, V.G., and Mahns, D.A. (2011). Allodynia mediated by C-tactile afferents in human hairy skin. J Physiol 589, 4065-4075. Nordin, M. (1990). Low-threshold mechanoreceptive and nociceptive units with unmyelinated (C) fibres in the human supraorbital nerve. J Physiol 426, 229-240. Olausson, H., Lamarre, Y., Backlund, H., Morin, C., Wallin, B.G., Starck, G., Ekholm, S., Strigo, I., Worsley, K., Vallbo, A.B., et al. (2002). Unmyelinated tactile afferents signal touch and project to insular cortex. Nat Neurosci 5, 900-904. Olson, W., Abdus-Saboor, I., Cui, L., Burdge, J., Raabe, T., Ma, M., and Luo, W. (2017). Sparse genetic tracing reveals regionally specific functional organization of mammalian nociceptors. Elife 6. Olson, W., Dong, P., Fleming, M., and Luo, W. (2016). The specification and wiring of mammalian cutaneous low-threshold mechanoreceptors. Wiley Interdiscip Rev Dev Biol 5, 389-404. Patel, T.D., Jackman, A., Rice, F.L., Kucera, J., and Snider, W.D. (2000). Development of sensory neurons in the absence of NGF/TrkA signaling in vivo. Neuron 25, 345-357. Peirs, C., Williams, S.P., Zhao, X., Walsh, C.E., Gedeon, J.Y., Cagle, N.E., Goldring, A.C., Hioki, H., Liu, Z., Marell, P.S., et al. (2015). Dorsal Horn Circuits for Persistent Mechanical Pain. Neuron 87, 797-812. Ranade, S.S., Woo, S.H., Dubin, A.E., Moshourab, R.A., Wetzel, C., Petrus, M., Mathur, J., Begay, V., Coste, B., Mainquist, J., et al. (2014). Piezo2 is the major transducer of mechanical forces for touch sensation in mice. Nature 516, 121-125. Rexed, B. (1952). The cytoarchitectonic organization of the spinal cord in the cat. J Comp Neurol 96, 414-495. Rexed, B. (1954). A cytoarchitectonic atlas of the spinal cord in the cat. J Comp Neurol 100, 297-379. Reynders, A., Mantilleri, A., Malapert, P., Rialle, S., Nidelet, S., Laffray, S., Beurrier, C., Bourinet, E., and Moqrich, A. (2015). Transcriptional Profiling of Cutaneous MRGPRD Free Nerve Endings and C-LTMRs. Cell Rep. Roberts, W.J., and Elardo, S.M. (1985). Sympathetic activation of A-delta nociceptors. Somatosens Res 3, 33-44. Rogoz, K., Andersen, H.H., Lagerstrom, M.C., and Kullander, K. (2014). Multimodal use of calcitonin gene-related peptide and substance P in itch and acute pain uncovered by the elimination of vesicular glutamate transporter 2 from transient receptor potential cation channel subfamily V member 1 neurons. J Neurosci 34, 14055-14068. 132

Roybon, L., Mastracci, T.L., Ribeiro, D., Sussel, L., Brundin, P., and Li, J.Y. (2010). GABAergic differentiation induced by Mash1 is compromised by the bHLH proteins Neurogenin2, NeuroD1, and NeuroD2. Cereb Cortex 20, 1234-1244. Ruscheweyh, R., Kuhnel, M., Filippopulos, F., Blum, B., Eggert, T., and Straube, A. (2014). Altered experimental pain perception after cerebellar infarction. Pain 155, 1303-1312. Rutlin, M., Ho, C.Y., Abraira, V.E., Cassidy, C., Bai, L., Woodbury, C.J., and Ginty, D.D. (2014). The cellular and molecular basis of direction selectivity of Adelta-LTMRs. Cell 159, 1640-1651. Saito, N., and Shirai, Y. (2002). Protein kinase C gamma (PKC gamma): function of neuron specific isotype. J Biochem 132, 683-687. Samour, M.S., Nagi, S.S., and Mahns, D.A. (2015). Cav3.2-expressing low-threshold C fibres in human hairy skin contribute to cold allodynia--a non-TRPV1- and non-TRPM8-dependent phenomenon. Pain 156, 1566-1575. Schatz, O., Langer, E., and Ben-Arie, N. (2014). Gene dosage of the transcription factor Fingerin (bHLHA9) affects digit development and links syndactyly to ectrodactyly. Hum Mol Genet 23, 5394- 5401. Schmidt, R., Schmelz, M., Forster, C., Ringkamp, M., Torebjork, E., and Handwerker, H. (1995). Novel classes of responsive and unresponsive C nociceptors in human skin. J Neurosci 15, 333-341. Seal, R.P., Wang, X., Guan, Y., Raja, S.N., Woodbury, C.J., Basbaum, A.I., and Edwards, R.H. (2009). Injury-induced mechanical hypersensitivity requires C-low threshold mechanoreceptors. Nature 462, 651-655. Seltzer, Z., Dubner, R., and Shir, Y. (1990). A novel behavioral model of neuropathic pain disorders produced in rats by partial sciatic nerve injury. Pain 43, 205-218. Shaikh, S., Nagi, S.S., McGlone, F., and Mahns, D.A. (2015). Psychophysical Investigations into the Role of Low-Threshold C Fibres in Non-Painful Affective Processing and Pain Modulation. PLoS One 10, e0138299. Shields, S.D., Ahn, H.S., Yang, Y., Han, C., Seal, R.P., Wood, J.N., Waxman, S.G., and Dib-Hajj, S.D. (2012). Nav1.8 expression is not restricted to nociceptors in mouse peripheral nervous system. Pain 153, 2017- 2030. Shields, S.D., Cavanaugh, D.J., Lee, H., Anderson, D.J., and Basbaum, A.I. (2010). Pain behavior in the formalin test persists after ablation of the great majority of C-fiber nociceptors. Pain 151, 422-429. Simone, D.A., and Kajander, K.C. (1997). Responses of cutaneous A-fiber nociceptors to noxious cold. J Neurophysiol 77, 2049-2060. Sinclair, D.C. (1955). Cutaneous sensation and the doctrine of specific energy. Brain 78, 584-614. Smith, A.K., O'Hara, C.L., and Stucky, C.L. (2013). Mechanical sensitization of cutaneous sensory fibers in the spared nerve injury mouse model. Mol Pain 9, 61. Stam, F.J., Hendricks, T.J., Zhang, J., Geiman, E.J., Francius, C., Labosky, P.A., Clotman, F., and Goulding, M. (2012). Renshaw cell interneuron specialization is controlled by a temporally restricted transcription factor program. Development 139, 179-190. Szczot, M., Liljencrantz, J., Ghitani, N., Barik, A., Lam, R., Thompson, J.H., Bharucha-Goebel, D., Saade, D., Necaise, A., Donkervoort, S., et al. (2018). PIEZO2 mediates injury-induced tactile pain in mice and humans. Sci Transl Med 10. Takazawa, T., Choudhury, P., Tong, C.K., Conway, C.M., Scherrer, G., Flood, P.D., Mukai, J., and MacDermott, A.B. (2017). Inhibition Mediated by Glycinergic and GABAergic Receptors on Excitatory Neurons in Mouse Superficial Dorsal Horn Is Location-Specific but Modified by Inflammation. J Neurosci 37, 2336-2348. Takeda, M., Takahashi, M., and Matsumoto, S. (2009). Contribution of the activation of satellite glia in sensory ganglia to pathological pain. Neurosci Biobehav Rev 33, 784-792. Tan, L.L., Pelzer, P., Heinl, C., Tang, W., Gangadharan, V., Flor, H., Sprengel, R., Kuner, T., and Kuner, R. (2017). A pathway from midcingulate cortex to posterior insula gates nociceptive hypersensitivity. Nat Neurosci 20, 1591-1601. Todd, A.J. (2010). Neuronal circuitry for pain processing in the dorsal horn. Nat Rev Neurosci 11, 823- 836. 133

Todd, A.J. (2017). Identifying functional populations among the interneurons in laminae I-III of the spinal dorsal horn. Mol Pain 13, 1744806917693003. Todd, A.J., and McKenzie, J. (1989). GABA-immunoreactive neurons in the dorsal horn of the rat spinal cord. Neuroscience 31, 799-806. Todd, A.J., and Spike, R.C. (1993). The localization of classical transmitters and neuropeptides within neurons in laminae I-III of the mammalian spinal dorsal horn. Prog Neurobiol 41, 609-645. Tonsfeldt, K.J., Suchland, K.L., Beeson, K.A., Lowe, J.D., Li, M.H., and Ingram, S.L. (2016). Sex Differences in GABAA Signaling in the Periaqueductal Gray Induced by Persistent Inflammation. J Neurosci 36, 1669-1681. Ueda, H. (2006). Molecular mechanisms of neuropathic pain-phenotypic switch and initiation mechanisms. Pharmacol Ther 109, 57-77. Urien, L., Gaillard, S., Lo Re, L., Malapert, P., Bohic, M., Reynders, A., and Moqrich, A. (2017). Genetic ablation of GINIP-expressing primary sensory neurons strongly impairs Formalin-evoked pain. Sci Rep 7, 43493. Usoskin, D., Furlan, A., Islam, S., Abdo, H., Lonnerberg, P., Lou, D., Hjerling-Leffler, J., Haeggstrom, J., Kharchenko, O., Kharchenko, P.V., et al. (2015). Unbiased classification of sensory neuron types by large-scale single-cell RNA sequencing. Nat Neurosci 18, 145-153. Vallbo, A., Olausson, H., Wessberg, J., and Norrsell, U. (1993). A system of unmyelinated afferents for innocuous mechanoreception in the human skin. Brain Res 628, 301-304. Vallbo, A.B. (2018). Microneurography: how it started and how it works. J Neurophysiol 120, 1415- 1427. Vallbo, A.B., Olausson, H., and Wessberg, J. (1999). Unmyelinated afferents constitute a second system coding tactile stimuli of the human hairy skin. J Neurophysiol 81, 2753-2763. Vrontou, S., Wong, A.M., Rau, K.K., Koerber, H.R., and Anderson, D.J. (2013). Genetic identification of C fibres that detect massage-like stroking of hairy skin in vivo. Nature 493, 669-673. Walcher, J., Ojeda-Alonso, J., Haseleu, J., Oosthuizen, M.K., Rowe, A.H., Bennett, N.C., and Lewin, G.R. (2018). Specialized mechanoreceptor systems in rodent glabrous skin. J Physiol 596, 4995-5016. Wall, P.D. (1978). The gate control theory of pain mechanisms. A re-examination and re-statement. Brain 101, 1-18. Wei, F., Wang, G.D., Zhang, C., Shokat, K.M., Wang, H., Tsien, J.Z., Liauw, J., and Zhuo, M. (2006). Forebrain overexpression of CaMKII abolishes cingulate long term depression and reduces mechanical allodynia and thermal hyperalgesia. Mol Pain 2, 21. Wende, H., Lechner, S.G., Cheret, C., Bourane, S., Kolanczyk, M.E., Pattyn, A., Reuter, K., Munier, F.L., Carroll, P., Lewin, G.R., et al. (2012). The transcription factor c-Maf controls touch receptor development and function. Science 335, 1373-1376. Witschi, R., Punnakkal, P., Paul, J., Walczak, J.S., Cervero, F., Fritschy, J.M., Kuner, R., Keist, R., Rudolph, U., and Zeilhofer, H.U. (2011). Presynaptic alpha2-GABAA receptors in primary afferent depolarization and spinal pain control. J Neurosci 31, 8134-8142. Yoshihara, Y. (2002). Visualizing selective neural pathways with WGA transgene: combination of neuroanatomy with gene technology. Neurosci Res 44, 133-140. Zeilhofer, H.U., Wildner, H., and Yevenes, G.E. (2012). Fast synaptic inhibition in spinal sensory processing and pain control. Physiol Rev 92, 193-235. Zeisel, A., Hochgerner, H., Lonnerberg, P., Johnsson, A., Memic, F., van der Zwan, J., Haring, M., Braun, E., Borm, L.E., La Manno, G., et al. (2018). Molecular Architecture of the Mouse Nervous System. Cell 174, 999-1014 e1022. Zhou, Q., and Anderson, D.J. (2002). The bHLH transcription factors OLIG2 and OLIG1 couple neuronal and glial subtype specification. Cell 109, 61-73. Zimmermann, K., Hein, A., Hager, U., Kaczmarek, J.S., Turnquist, B.P., Clapham, D.E., and Reeh, P.W. (2009). Phenotyping sensory nerve endings in vitro in the mouse. Nat Protoc 4, 174-196. Zimmermann, K., Leffler, A., Babes, A., Cendan, C.M., Carr, R.W., Kobayashi, J., Nau, C., Wood, J.N., and Reeh, P.W. (2007). Sensory neuron sodium channel Nav1.8 is essential for pain at low temperatures. Nature 447, 855-858. 134

Zotterman, Y. (1939). Touch, pain and tickling: an electro-physiological investigation on cutaneous sensory nerves. J Physiol 95, 1-28. Zylka, M.J., Dong, X., Southwell, A.L., and Anderson, D.J. (2003). Atypical expansion in mice of the sensory neuron-specific Mrg G protein-coupled receptor family. Proc Natl Acad Sci U S A 100, 10043- 10048.

135

List of abreviations :

°C : degrees centigrade A-MH : A-mechno-heat (receptors) AMPA : α-amino-3-hydroxy-5-méthylisoazol-4-propionate, ionotropic glutamate receptor AP : action potential ASD : autism spectrum disorder BAC : bacterial artificial chromosome bhlha9 : basic helix-loop-helix family member a9 Cav3.2 : cacna1h Calcium channel, voltage-dependent, T type, alpha 1H subunit CCI : chronic constriction injury cDNA : complementary desoxyrobinucleic acid CFA : complete Freund’s adjuvant CGRP : calcitonin-gene related peptide cKO : conditional knock-out C-LTMR : C-low threshold mechanoreceptor cm : centimeter CNQX : 6-cyano-7-nitroquinoxaline-2,3-dione, competitive AMPA receptor antagonist CPP : conditioned place preference CT : C-tactile (afferents) CTB : cholera toxin subunit B DAT1 : dopamine active transporter 1 DNA : desoxyrobinucleic acid DREADD : designer receptor exclusively activated by designer drugs DRG : dorsal root ganglia DTR : diphteria toxin receptor DTX : mice injected with diphteria toxin EPSC : excitatory postsynaptic potential ES : embryonic stem (cell) FACS : fluorescence-activated cell sorting fl : flox/floxed, flanked by loxP sites FRT : flippase recognition site

136 g : gram GABA : gamma-amino butyric acid

GABAA receptor: ionotropic gamma-amino butyric acid receptor

GABAB receptor: metabotropic gamma-amino butyric acid receptor GCT : gate control theory GFAP : glial fibrillary acidic protein GFP : green fluorescent protein Ginip : Gαi-interacting protein h : hour hDTR : human diphteria toxin receptor HTMR : high threshold mechanoreceptor IB4 : isolectin B4 IL-1b : interleukin 1 beta ip : intraperitoneal IPSC: inhibitory postsynaptic potential IRES : internal ribosome entry site IT : intrathecal KI : knock-in kg : kilogram KO : knock-out Lbx1 : ladybird homeobox 1 L-DOPA: levo-3,4-dihydroxyphenylalanin LTMR : low threshold mechanoreceptor mGluR5 : metabotropic glutamate receptor 5 MOR : μ opioid receptor MrgprD : Mas-related g-protein coupled receptor member D MrgprB4 : Mas-related g-protein coupled receptor member B4 μA: microAmpere μl: microiter μm: micrometer mg : milligram mm : millimeter mN : milliNewton 137 msec : millisecond min : minute Nav1.8 : scn10a subunit alpha sodium channel type 10 NPY : neuropeptide Y NS11394 : 3'-[5-(1-hydroxy-1-methyl-ethyl )-benzoimidazol-1-yl]-biphenyl-2 carbonitrile pA : picoAmpere PAF : paraformaldehyde PAG : periaqueductal gray (area)

PAM : positive allosteric modulator (of GABAA receptors) PBS : Phosphate-buffered saline PCR : polymerase chain reaction PKCγ: gamma isoform of protein kinase C PSN : peripheral sensory neuron PSDC : post-synaptic dorsal column q RT PCR: quantitative reverse transcription polymerase chain reaction RNA: ribonucleic acid RORβ : RAR-related orphan receptor beta Runx1 : Runt-related transcription factor 1 SCT : spinocervical tract s/sec : second S1 : somatosensory cortex area 1 SC : spinal cord SEM : standard error of the mean SG : substantia gelatinosa SHFLD : split-hand/foot malformation and long-bone deficiency SNI : spared nerve injury Tg : transgenic TH : tyrosine hydroxylase TrpC3 : transient receptor potential channel subfamily C member 3 TrpM8 : transient receptor potential cation channel subfamily M member 8 TrpV1 : transient receptor potential vanilloid 1 TTA-A2 : T-type calcium channel antagonist A2 Vglut3 : vesicular glutamate transporter 3 138

WGA : wheat germ agglutinin WT : wild type

139

Annexe:

During my PhD, another project was ongoing in team focusing on the study of C-LTMRs. The strategy was to perform a specific ablation of GINIP-positive neurons and perform behavioural tests on these mice. To perform this ablation, mice genetically modified to express the human Receptor of the Diphteria Toxin from the ginip locus were injected with Diphteria Toxin at 3 weeks of age.

The GINIP-positive population of DRG neurons is made up of two subpopulations: C-LTMRs and mechanonociceptors that not only express GINIP but, conversely to C-LTMRs, bind IB4. As such, any sensory defect presented by mice lacking GINIP-positive neurons could be due to the ablation of C-LTMRs and/or of mechanonociceptors. Except if the test performed does not involve the activation of one or the other population of neurons. That is the case of the tape test, which involves light innocuous mechanical stimulation by way of the application of a 3cm long piece of tape on the back hairy skin of mice. This innocuous mechanical stimulation should not lead to stimulation of mechanonociceptors only activated by high intensity mechanical stimulation. It would however activate all three subtypes of LTMRs, including C- LTMRs specifically ablated in ginipDTR/+ mice.

I had the opportunity of performing this experiment and found that ablation of Ginip-positive neurons leads to a significant decrease in light mechanical sensitivity as these mice had a higher threshold of sensitivity in the tape test, compared to littermate animals. From the results of this experiment we could conclude that C-LTMRs are required for detection of light mechanical stimulation such as the one induced by putting a piece of tape on back hairy skin. We also confirmed that this test is relevant for the study of C-LTMRs.

The results of the study of the genetic ablation of Ginip-positive neurons were the object of a publication in which I am an author.

140

141

142

143

144

145

146

147

148

149

150

151

152