Primary sensory map formations reflect unique needs and molecular cues specific to each sensory system

Introduction

Sensory organs are the windows of the brain to the environment, permitting behavioral interactions with conspecifics, prey, and predators. Finding an appropriate mate, being able to avoid being eaten, and finding food are essential features for survival and propagation. Sensory organs filter out the appropriate information for these tasks and relay it to the brain to elicit adequate motor responses^1–3. Sensory map features depend on the specific sensory modality and the relevant information to be extracted. For example, somatotopic maps project a topographic array of sensors to reflect the sensor distribution, density, and activity of the skin to the brain^4–6. Similarly, the retinotopic map projects distinct areas of the retina and the corresponding visual field as a two-dimensional (2D) map to the target brain area^7,8, whereas the cochlea map projects a unidimensional map of distinct frequencies to specific areas of the cochlear nuclei⁹ and auditory cortex^10,11. Beyond primary sensory maps, central map formation underlies binocular vision and depth perception^12,13. Likewise, the auditory space map is generated through binaural interactions^14–16 whereas the mechanosensory lateral line¹⁷ and the electrosensory space map¹⁸ are generated through integration of distributed sensors across the body. In contrast to these emerging centrally synthesized maps and continuous primary maps, discrete olfactory maps project unique properties of the odorant stimuli perceived by distributed olfactory neurosensory cells convergently onto specific glomeruli^10,19,20. A variation of the latter theme is the incomplete segregation of movement detection in the vestibular system, where angular movements always cause concomitant linear acceleration. This causes partial convergence of afferents from organs dedicated to either linear or angular acceleration perception^21,22. Even more difficult to understand are maps where a given stimulus and its intensity are differentially coded as the tastants for the yet-to-be-fully-defined taste map^23–26.

During the last century, specific properties of a given sensory map and basic rules how to form them, such as the chemoaffinity theory²⁷ and activity-mediated synaptic plasticity theory^28,29, have been worked out for some primary maps. Understanding the molecular cues that guide primary map development, the plasticity of primary map development mediated by activity to sharpen the map in neonates and adults^30–33, and the translation of primary sensory afferent map formation into cortical and midbrain maps for multisensory integration^2,6,34,35 will be the defining achievements of the 21st century. Toward this end, we provide here an overview of various primary sensory maps of mammals, characterized by continuous and discrete map properties³³. All primary maps require that a peripheral sense be wired to independently developing central target neurons by molecular cues in the target and matching cues expressed in the neurons as they navigate toward their target. Our aim is to turn primary sensory map formation into a neuronal pathfinding problem that combines with cell cycle exit to generate an embryonic primary map for each sense. Uncovering regulatory aspects of map formation across senses will facilitate sensory restoration badly needed for sensory repair of seniors in our rapidly aging societies.

Primary sensory maps compared

The six primary sensory maps of mammals have unique features and seemingly use distinct molecular cues, cell cycle exit, and activity combinations during development, regeneration, and plasticity. We will start with the best molecularly understood map formations followed by the less well understood map formations in the hindbrain, ending with the least understood map for taste that has recently seen dramatic revisions from past insights^24,36.

Molecular odorant map

Adult map organization

Since the cloning of genes encoding a family of odorant receptor (OR) nearly 30 years ago³⁷, the understanding of olfactory map formation has leapfrogged to be perhaps the best molecularly understood sensory map. The basic principle is that a given olfactory sensory neuron (OSN), coding for a given OR, projects its axon to a molecularly specified olfactory glomerulus in the olfactory bulb (OB), where it converges with axons of other OSNs coding for the same OR^20,38–40. Thus, OSNs coding for the same OR converge to the same glomerulus (Figure 1A). In the mouse, this results in a discrete expression of one of about 1100 ORs in a given OSN whose axon converges onto one or few of the roughly 3600 glomeruli. OR expression is not completely random but splits the olfactory epithelium into major divisions along the dorso-ventral axis, each with medio-lateral bands of randomly distributed OSNs that project to dorso-ventrally distinct sets of olfactory glomeruli^38,39,41. Specific odorant information is thus perceived by OSNs within certain zones that are, however, nearly randomly distributed within these zones. This is particularly obvious in mammals with a reduced complement of olfactory receptor genes that form glomeruli only in the ventral part of the OB⁴². Odor information is encoded in the odorant-specific glomeruli and not in the topology of OSNs in the olfactory epithelium. This organizational principle allows OSNs to be continuously replaced⁴³ without any change in the important central information storage^1,34. The brain learns and recognizes patterns of glomerular activity elicited by different odors⁴⁴.

Figure 1. Development of three distinct mammalian sensory maps.

Molecular cues (A, B) and spatio-temporal cues (C) are shown for the nearly non-spatial olfactory map (A), the two-dimensional (2D) retino-tectal map (B), and the unidimensional auditory map (C). (A) The olfactory map defines different olfactory receptor molecules in the dorsal and ventral zone of the olfactory epithelium. Receptor cells displaying distinct olfactory receptors (A–D) project their axons to the dorsal and ventral domain of the olfactory bulb where they converge and initiate olfactory glomeruli formation. Note that olfactory fibers sort before they reach the olfactory bulb and that some ventral zone receptors are expressed in the dorsal zone but afferents sort to the ventral domain. Different opposing gradients of receptors facilitate further the sorting of olfactory afferents. Within this limited topology, the distribution of specific olfactory receptor–expressing receptor cells is fairly random. (B) The retino-tectal system maps a 2D surface (the retina ganglion cells) onto another 2D surface (the midbrain roof or tectum opticum) via highly ordered optic nerve/tract fiber pathways. Within the midbrain, the presorted fibers are further guided by molecular gradients matching retinal gradients of ligand/receptor distributions. (C) The auditory map is unidimensional, projecting a species-specific frequency range from the mammalian hearing organ, the organ of Corti via orderly distributed spiral ganglion neurons (SGNs), and their fibers in the auditory (cochlear) nerve onto the ventral cochlear nucleus complex. Both SGNs and cochlear nucleus neurons show a matching temporal progression of cell cycle exit followed by matching differentiation that could be assisted by spatio-temporal expression changes of receptors and ligands (shown here are the putative Wnt/Fzd combinations) that further support the fiber sorting. Note that this map projects a single frequency of an inner hair cell of the organ of Corti via a set of SGNs onto longitudinal columns of cochlear nucleus neurons in a cell-to-band projection and thus is not a point-to-point map as the olfactory and visual map. Moreover, afferents innervating multiple outer hair cells (OHCs) generate a band-to-band projection centrally. A, anterior; D, dorsal; L, lateral; M, medial; N, nasal; P, posterior; T, temporal; V, ventral. Modified after 12,41,53–58.

Development

The main and accessory (vomeronasal) olfactory epithelium develops from the olfactory placode that also gives rise to a set of gonadotrophin-releasing hormone (GnRH)-positive neurons migrating into the hypothalamus^45–48. This allows the olfactory system to interconnect with the retina in some species⁴⁹. A sequence of basic helix-loop-helix (bHLH) genes, in combination with other transcription factors, guides the transformation of the olfactory placode cells into OSNs^50,51. Dorso-ventral zones of ORs are expressed in selective OSNs and each OSN projects specific odorant molecule information to a given glomerulus⁵². Matching gradients of OR expression define the pathfinding properties of OSNs to select a given band of glomeruli and a specific glomerulus within that band (Figure 1 and Figure 2). Misexpression of a given OR in another set of OSNs results in misdirection to a different glomerulus. This supports the idea that both the selection of a given OR and the level of gene expression bestow on an OSN an identity that allows the OSN growth cone to navigate to a specific glomerulus. The G protein–coupled ORs define expression levels of adenylate cyclases such as Ac3³³. Knockouts of Ac3 lead to disorganized OR projections. This appears to be related to Ac3-mediated activation of downstream guidance cues via cAMP/CREB/PKA, such as neuropilin 1 (Nrp1). Gradients of Nrp1 code for anterior-posterior patterning⁵³ in combination with matching expression of semaphorins^59–61. However, detailed tests question the proposed model of Nrp1 guidance by showing more complicated outcomes inconsistent with the simple Nrp1 gradient model⁵⁴. Since G-coupled ORs are found on the growth cone of OSNs, those ORs could locally interact with the environment to guide confined responses via the cAMP/PKA intracellular signal cascades. Though clearly important, a gradation of G protein/cAMP alone is not the only cue, and Robo/Slit is used for larger-scale dorso-ventral patterning^40,62. In addition, two different classes of OSNs have been identified and their axons sort out as they extend toward the OB, leading to a complete segregation of axons of dorsal but not ventral OSNs⁴⁰. This fiber sorting (Figure 1A and Figure 2A) happens prior to and even in the absence of OBs, establishing a topographic order of OSN axons as they approach the OB⁵³. Further refinement of the olfactory mapping is achieved through differential expression and activity-regulated levels of ephrinA ligands and Eph-A5 receptors as well as the molecularly related Kirrel2/3 (Figure 1A). In a given glomerulus, there is an opposing gradient of either the Kirrel2/3 pair dorsal or ephrinA/EphA5 ventral. This expression defines a dorsal and a ventral domain of glomeruli (Figure 1A) matching to the dorso-ventral zones of OR expressing OSNs in the olfactory epithelium. Thus, although the dorso-ventral patterning of bands of OSNs to project to bands of olfactory glomeruli seems to be settled, the details of antero-posterior patterning remain less clear and seemingly are less precise⁵⁴.

Figure 2. Distribution of sensory maps and the development of hindbrain sensory maps.

(A) Schematic presentation of the main features of the six cranial senses projected onto an embryonic mouse brain. (Pale yellow, left) Distributed olfactory sensory neurons of the olfactory epithelia coalesce their axons before reaching a specific olfactory glomerulus in the olfactory bulb (OB). (Pale lavender) Axons of retina ganglion neurons leave the eye orderly to project via the optic nerve to the optic chiasm (OC). Crossed contralateral axons form the orderly optic tract that distributes axons within the midbrain using matching gradients of several factors. (Gray) The trigeminal ganglion has three distinct branches and matching sensory neuron populations that reach different areas of the face. The central axons form in a temporal progression resulting in an inverted presentation of the face. (Pale pink) Taste buds of the tongue and pharynx are innervated by three cranial nerves that form a somewhat orotopic central projection to the solitary tract. (Light blue) The five vestibular sensory organs are innervated by somewhat orderly distributed sensory neurons that project via the vestibular nerve. Within the brain, vestibular afferents from different ear organs are partially segregated and partially overlapping in the various vestibular nuclei as well as the posterior lobes of the cerebellum. (Pale green) The organ of Corti of the cochlea is innervated by a temporally generated longitudinal array of spiral ganglion neurons that project in an orderly organization to dorso-ventral distinct regions of the cochlear nucleus complex, projecting a one-dimensional frequency array along the cochlea onto a matching frequency array of afferents in the cochlear nuclei. (B) (Left) In the axolotl, there is a timing factor of afferent ingrowth such that the most ventral trigeminal projection reaches the hindbrain first (V at stage 32) whereas the most dorsal projection from the electroreceptive (lateral line) ampullary organs reaches the most dorsal part of the hindbrain last (ELL, stage 38). The inner ear vestibular ganglia (VG, stage 34) and mechanosensory lateral line ganglia (LL, stage 36) are reaching the alar plate between those extremes. (Right) In the mouse, the dorso-ventral patterning of the hindbrain is driven by countergradients of Wnt/BMP and Shh to regulate expression of transcription factors defining various nuclei. How these gradients define the positon of central nuclei and afferents is not completely clear. A temporal gradient of afferent development and projection development has thus far been demonstrated only for the spiral ganglion, taste and trigeminal system where the first neurons to form are the first to project to the most ventral part of their respective tract. Note that the auditory nuclei show an apparent inversion such that the most ventral projection from the basal spiral ganglion ends up in the more dorso-medial part of the cochlear nuclei because of the morphogenetic changes in cochlear nucleus neuron position. A, anterior; AC, anterior crista; Ascl1/Mash1, achaete-scute family basic helix-loop-helix transcription factor 1; Atoh7, atonal basic helix-loop-helix transcription factor 7; AVCN, antero-ventral cochlear nucleus; BMP, bone morphogenic protein; C, cochlea; CB, cerebellum; CN V, VII, IX, X, cranial nerve V, VII, IX, X; CP, choroid plexus of IV ventricle; D, dorsal; DCN, dorsal cochlear nucleus; dV, descending trigeminal tract; ELL, electroreceptive (ampullary organ) lateral line; GG, geniculate ganglion; HC, horizontal crista; L, lateral; LL, (mechanosensory) lateral line; M, medial; N, nasal; Neurog1/2, Neurogenin 1/2; NG, nodose ganglion; OB, olfactory bulb; OC, optic chiasm; OE, olfactory epithelium; P, posterior; PC, posterior crista; PG, petrosal ganglion; pV, principal trigeminal nucleus; r1, rhombomere 1; r2, rhombomere 2; S, saccule; SG, spiral ganglion; Shh, sonic hedgehog; ST, solitary tract; T, temporal; TIx3, T-cell leukemia homeobox 3; U, utricle; V, ventral; V1, ophthalmic branch of trigeminal nerve; V2, maxillary branch; V3, mandibular branch; VG, vestibular ganglion; VN, vestibular nucleus complex. Modified after 5,12,17,33,36,54,56,63–66.

Retinotopic map

Somatosensory map

Vestibular maps for linear and angular acceleration detection

Tonotopic map

Primary taste maps challenge past taste concepts

Overview of brainstem maps

Olfactory and retinotopic maps differ from brainstem maps as the former either involve the only mammalian sensory neuron/cell (the OSN) with its own axon that is continuously replaced or deal with a region of the brain transformed into the retina, generating the “optic nerve” out of an intracerebral tract. In addition, both of the above-described maps provide a point-to-point connection that either projects one surface (the retina) onto another surface (the midbrain⁸) in two dimensions or ensures that a given odor binding to distributed OSNs converges on the same glomerulus (olfactory map³³). No such point-to-point map is obvious in the hindbrain (Figure 1C and Figure 2A) where a given peripheral connection (such as a specific area of the facial skin or the cochlea) is innervated by a neuron residing in a given ganglion with a distinct molecular (Neurog1 versus Neurog2^177,122) and developmental (epibranchial placode, otic placode, trigeminal placode, and neural crest^46,73,164) origin. Instead of a point-to-point connection, each of the hindbrain targeting sensory neurons forms an extended longitudinal track along the alar plate of the hindbrain (Figure 1C and Figure 2A). As a first approximation, the hindbrain alar plate can be regarded as a highly transformed part of the spinal cord that has developed rhombomere-specific nuclei, which receive hindbrain-specific innervation^{109,63,178,179}. Within each of the longitudinal hindbrain tracks, rhombomere-specific nuclei can be identified^{5,17,178,179–182}, and each has its own higher-order projection. How the well-known cortical maps such as for somatosensation^5,6 are exactly derived from the organizational principle of primary afferents^102,103 is only in the case of the auditory and somatosensory system partially clarified^5,10,11. An emerging principle of hindbrain and visual map formation, likely not shared with the olfactory map because of its continuous replacement, is the role of cell cycle exit.

Summary

Primary sensory maps mirror the unique properties of a given sensory modality. Maps can reflect (a) local receptor density and activity (somatosensory), (b) convergence of distributed receptors (olfactory), (c) continuous one-dimensional (tonotopic) or (d) 2D (retinotopic and somatotopic) maps, or (e) convergence and segregation of information gathered by distinct sensory organs (vestibulotopic and orotopic) maps. An emerging principle of several maps is the role of cell cycle exit that allows distinct inputs to interact specifically with matching cell cycle–exited second-order neurons. This is particularly obvious in the temporal progression of afferent projections from various sensory systems in amphibians and bony fish, the temporal and maturational progression of spiral ganglion, and cochlear nucleus cell cycle exit, and it plays a role in RGN type specifications. Afferent fiber sorting prior to the target is an obvious common feature in all sensory systems and may reflect both fiber–fiber molecular interactions and cell cycle exit. As a consequence of fiber sorting, a crude topology of processes arises before a target is innervated and even in the absence of a target as demonstrated in the olfactory and auditory systems. Once the specific map has been established by various non-activity-related means, a common feature is that activity sharpens the map. To the best of our knowledge, there is only one developing sensory system currently known where a single deletion in a viable mutant results in near-random distribution of peripheral and central processes that cannot be corrected by physiological activity. Such mutants can test the limits of activity-mediated refinement of distorted primary maps.

Combinations of classic embryologic manipulations, such as transplantations, rotations, or partial deletions, have been extremely helpful to formulate basic principles such as the chemoaffinity theory. Whereas the topographical information coded in such diffusible gradients may be uniform across all sensory maps, the molecular nature of specific guidance cues used certainly is not. However, it is noteworthy that several maps have a dorso-ventral axis that could reflect known countergradients of diffusible molecules needed to define different dorso-ventral nuclei, such as Bmp4, Wnt3, and Shh. Going forward, combining heterochronic and heterotopic transplantations with molecular perturbation of map formation and with the evaluation of the role of activity to sharpen such distorted maps will reveal how best to use such information to enhance sensory organ replacements for functional recovery to cure anosmia, blindness, vestibular, and auditory dysfunction. Whole face or tongue transplants could also benefit from an understanding of such detailed map formation. Clearly, cortical maps will plastically respond to peripheral manipulations, but meaningful integration of various sensory information requires that each primary map be appropriately organized to allow a multisensory cortical or subcortical integration of relevant information extracted out of primary maps.

Abbreviations

2D, two-dimensional; bHLH, basic helix-loop-helix; MesV, mesencephalic trigeminal nucleus; Nrp, natriuretic peptide receptor; OB, olfactory bulb; OR, odorant receptor; OSN, olfactory sensory neuron; RGN, retinal ganglion neuron