Neurosensory development of the four brainstem-projecting sensory systems and their integration in the telencephalon

Introduction

Across vertebrates, the brainstem contains discrete nuclei which receive input from distinct peripheral sensory neurons. Neurons from these brainstem nuclei project to higher-order nuclei, permitting the relay of the specific sensory signal to unique areas of the telencephalon, where integration across sensory systems occurs. The telencephalon, or forebrain, is the last to form and develops as a bilateral expansion from a single end of the neuropore in the lancelet and the ascidians (Fritzsch and Martin, 2022). We will analyze four sensory systems (trigeminal, taste, vestibular, and auditory) from their peripheral nerve entry point within the hindbrain through their projection to the telencephalon. In addition, we will focus on the various genes expressed as well as the organizational principles for each system. For instance, while a topological organization exists for the somatosensory and the auditory projections to the telencephalon, the taste, and vestibular projections are not topologically organized. Moreover, higher-order interactions are provided to integrate the cohesive perception in the telencephalon that is documented chiefly in human organization.

Neurons of the peripheral nervous system (PNS) depend on specific genes for proper development (Kobayashi et al., 2019; Dvorakova et al., 2020; Xu et al., 2021). Most brainstem-projecting neurons differentiate by Neurog2 expression from epibranchial placodes (Fode et al., 1998; Dennis et al., 2019), whereas all other remaining placodally derived neurons depend on Neurog1 expression in mammals (Ma et al., 1998). There is a sequential expression of Neurog1/Neurog2 in the spinal cord of the dorsal root ganglia (DRGs; Ma et al., 1999; Meltzer et al., 2021). DRGs in the spinal cord depend on Sox10 followed by Neurog2 for large and mediums sized neurons, and by Neurog1 for small neurons and all depend on Pou4f1 (Huang et al., 2001; Meltzer et al., 2021). In contrast, epibranchial and dorsal placodes develop independently of Sox10 (Mao et al., 2013; Reiprich and Wegner, 2015). In mammals, the trigeminal, facial, glossopharyngeal, and vagus neurons depend on Neurog1, whereas the epibranchial neurons (geniculate, petrosal, and nodose) require Neurog2.

The central nervous system (CNS) consists of the brainstem and spinal cord. In the brainstem, the dorsal alar plate is innervated by placodally derived peripheral sensory neurons, which depend on Neurog1 or Neurog2 (Fode et al., 1998). On the other hand, the spinal cord receives DRG projections which depend on Neurog1/2 (Ma et al., 1999). These fibers extend ventrally to the basal plate in the spinal cord, including the muscle spindles (Lai et al., 2016; Meltzer et al., 2021). The CNS is divided dorsoventrally into the alar and basal plate, the former depending on choroid plexus formation along the dorsal part of the brainstem (Figure 1). Choroid plexus development requires Lmx1a/b expression, and without Lmx1a/b, neither the choroid plexus nor many of the dorsal alar plate structures form (Glover et al., 2018; Chizhikov et al., 2021). Downstream of Lmx1a/b is Gdf7, which is also necessary for roof plate formation (Lee et al., 2000; Mishima et al., 2009). The dorsoventral identity of neural progenitors within the developing dorsal brainstem and spinal cord are subdivided into eight or six distinct domains, respectively (Lai et al., 2016; Hernandez-Miranda et al., 2017; Hirsch et al., 2021). The most dorsal domain, dA1, is defined by an expression of Atoh1 across the brainstem and spinal cord. Ventral to dA1 is the dA2 domain (Neurog1/2 expression), followed by the dA3 domain (Ascl1, Phox2b, and Tlx3 expression) rostrally, which forms the solitary tract (Qian et al., 2001). Caudally to the more rostral dA2/3, we have two neuronal populations of the brainstem: one is presented as a duplication of Ptf1a expression (dA4/dB1), whereas another set of unique populations from dB2 (Atoh1, Phox2b, and Lbx1) expression that forms, among others, the lateral vestibular nucleus (LVN) in rhombomere r4 (Chen et al., 2012; Lunde et al., 2019). While some of these domain populations in the hindbrain are maintained through the spinal cord, such as Atoh1, some gene expression differences lead to the absence of two domains in the spinal cord that are present in the brainstem (Hernandez-Miranda et al., 2017). Although Atoh1 expression is continuous from the spinal cord to the cerebellum, separate nuclei are forming in this expression domain: the cerebellum (r0-1), the auditory nuclei (r2-5), and the pontine nuclei [r6-7 (Bermingham et al., 2001; Nichols and Bruce, 2006; Lowenstein et al., 2022)]. Further work is needed to detail the various nuclei and their specific migration to coalesce into distinct nuclei (r8-11; Rose et al., 2009; Watson et al., 2017a).

Second-order neuronal projections from the trigeminal, solitary tract, vestibular, and cochlear nuclei project in a relay to several additional nuclei before reaching the thalamus, whereas other second-order predictions are direct. Compared to the trigeminal projection that is common across vertebrates, information from taste, vestibular, and auditory nuclei is limited, and in addition, there is a unique loss of lateral line and electroreception in some species that are instead developing the auditory nuclei from frogs and amniotes (Grothe et al., 2004; Wullimann and Grothe, 2013; Fritzsch, 2021).

The bipartite of the telencephalon has long been seen as an exclusive input from the olfactory system of cyclostomes that is now shown to receive significant input from the trigeminal to the telencephalon in the lamprey (Suryanarayana et al., 2020, 2021), showing a basic telencephalon across vertebrates. Open questions concern the taste, vestibular, lateral line, electroreception, and auditory system in lampreys that show certain connections in gnathostomes (Striedter and Northcutt, 2019). It is unclear how the auditory system evolved beyond rudimentary insights that it is such a central system in mammals, frogs, reptiles, and birds (Briscoe and Ragsdale, 2019; Tosches and Laurent, 2019). Likewise, the taste system forms in all vertebrates but is unclear for higher-order connections to the telencephalon (Elliott et al., 2022a). The vestibular system has a basic understanding of the gnathostomes (Elliott et al., 2022a). Interestingly, there are multiple integrations of sensory systems. For example, auditory and visual input, taste and olfaction, and of vestibular and trigeminal (review in Shepherd, 2006; Majka et al., 2019; Chou et al., 2020; Mizoguchi et al., 2020; Rauschecker, 2021). We will provide the similarity of higher-order interactions of the four sensory systems and how they integrate multisensory modalities in the telencephalon with the four sensory systems.

Structure of the sensory systems

Somatosensory receptors receive touch, proprioception, pain, and thermoreception, which project via trigeminal neurons to nuclei in the brainstem. Two major brainstem nuclei (principle and descending tract) send second-order projections mostly to the contralateral thalamus, and from there, the projections end up in the somatosensory topological cortex. Tertiary connections integrate other cortical sensory system inputs to direct motor output.

Taste buds, the peripheral end organs of gustation, contain chemosensory transducing cells. These cells are in the five different sensory taste buds innervated by four distinct sensory neurons. Three epibranchial placodes develop from Neurog2-expressing ectoderm to provide innervation from the oral cavity, particularly the tongue, and end in the solitary tract (ST) from r3-r11. From there, neurons project to the parabrachial nucleus (PB) to spread directly or indirectly in the thalamus and the insular cortex. Interactions of the olfactory system with vision, auditory, and vestibular provide an integrative taste perspective.

Vestibular hair cells and vestibular neurons detect and send, respectively, linear, and angular acceleration information to the brainstem. Various organizations and sensory epithelia are dedicated to detecting this directional information in gnathostomes. Vestibular neurons depend on Neurog1 and provide innervation of hair cells at the periphery and end up in four distinct vestibular nuclei, as well as a part of the cerebellum (r1-r8) centrally. Within the cerebellum, vestibular neurons project a selected input to the uvula and nodulus. All spinal cord motoneurons receive an ipsilateral projection from the lateral vestibular nuclei (LVN), whereas three other vestibular neurons (superior, medial, and descending vestibular nuclei; SVN, MVN, and DVN) project to the three ocular motoneurons (III, IV, and VI). Bilateral outputs project the dorsal tegmental nucleus, the thalamus, and the parieto-insular vestibular cortex. Additional connections of auditory, visual, and taste information with vestibular input interact to drive higher-order spatial representations.

Auditory sensory neurons are unique neurons that evolved after the segregation of the basilar papilla developed. In mammals, auditory spiral ganglion neurons depend on Neurog1 to innervate Atoh1-dependent hair cells. In mammals, three cochlear nuclei form only in r2-r5. Like the hair cells at the periphery, these cochlear nuclei depend upon Atoh1 expression. Bilateral auditory nuclei interact with the superior olive nuclei (SOC) to go beyond the inferior colliculi (IC) to reach the medial geniculate body (MGB). The auditory cortex (aC) is the main target of the MGB, and auditory cortex neurons interact with those from vision, somatosensory, taste, and vestibular systems.

Structure of the sensory systems and genes needed for their development

Somatosensory sensory inputs extend from the spinal cord to the trigeminal nuclei to provide somatosensory perception, giving rise to four brainstem PNS inputs (trigeminal, facial, glossopharyngeal, and vagal). These inputs reach rostrally from r2 and blend with the spinal cord past r11 caudally (Gray, 2013; Hans et al., 2020). The trigeminal nerve’s ophthalmic, maxillary, and mandibular branches project centrally in a ventral-to-dorsal orientation, ending with the small addition of other sensory neurons (Erzurumlu et al., 2010). The somatosensory input is comparable in the brainstem and the spinal cord: The spinal cord formed from Neurog2 first to develop the largest and intermediate DRGs, followed by the population of Neurog1-dependent small DRGs (Ma et al., 1999; Meltzer et al., 2021). In contrast, brainstem-derived PNS neurons are exclusively developed from Neurog1-expressing cells (Ma et al., 1998) to innervate a mix of placode and neural crest (trigeminal) or neural crest-derived facial, glossopharyngeal (superior ganglion), and vagal sensory neurons (jugular ganglion). Most of the largest neurons reach the motoneurons in the spinal cord, except for the input from the trigeminal neurons.

A unique set of mesencephalic trigeminal neurons (MesV or MTN) develop that reach the brainstem and branch into a distal projection to innervate the muscle spindles of the trigeminal motoneurons (Fritzsch and Northcutt, 1993; Lipovsek et al., 2017; Ter-Avetisyan et al., 2018). The MesV develops in the absence of Neurog1 (Ma et al., 1998; Marzban et al., 2019), independent of the neural crest (Lipovsek et al., 2017). The largest proprioceptor neurons of the spinal cord depend on Atoh1, but these continuations are interrupted at the hindbrain to generate brainstem neurons giving rise to auditory and other neurons, the cuneate and gracile to continue the lateral lemniscus (Kuypers and Tuerk, 1964; Bermingham et al., 2001; Lai et al., 2016).

Most trigeminal and proximal brainstem PNS neurons depend on Eya1/2, which is upstream of Sox2 expression (Xu et al., 2021; Zhang et al., 2021). In addition, Six1, Irx, Pax3, Dlx, Tbx, and Fox1 are expressed early. Sox2 functions as a pioneer factor that regulates the expression of various bHLH genes; the trigeminal neurons depend on Neurog1 expression (Ma et al., 1998; Kobayashi et al., 2019). Downstream of Neurog1 are Neurod1 and Pou4f1, which interact with Isl1 to regulate all trigeminal ganglion neuron development. The first PNS neurons depend on Tbx1/3, which overlaps with the ophthalmic and maxillary ganglions. Further subdivisions are driven by Oc2 (maxillary and mandibular ganglion) and Oc1 and Hmx2 (mandibular ganglion), forming discrete central projections to the CNS (Erzurumlu et al., 2010). An extensive set of distinct genes specify various sensory inputs and proteins fully described in the spinal cord input (Abraira and Ginty, 2013; Handler and Ginty, 2021) and the trigeminal ganglion as well as the MesV (Figures 2A–D). The expression of neurotrophin receptors differs between the spinal cord and trigeminal system for TrkA, TrkB, and TrkC. DRGs and trigeminal neurons depend on Neurog1 and mostly TrkA (Ma et al., 1999). Loss of TrkC leads to the absence of DRGs innervating the muscle spindles, but since about 50% of muscle spindles form in the absence of TrkC suggests many MesV neurons remain (Matsuo et al., 1999; Huang and Reichardt, 2003). Bdnf, which preferentially binds to TrkB, reduces the number of MesV neurons when lost but seems not to affect the muscle spindles, whereas Merkel cells and hair innervation require TrkC. The interaction between TrkC and possibly TrkB remains unclear for the remaining trigeminal neurons. In summary, our data show a different dependence of trigeminal and spinal cord concerning Neurog1/2 (spinal cord) compared to trigeminal, which depends on the placode and neural crest ganglions. A unique formation forms in the MesV that provide muscle spindle innervation but seems to develop independently of Neurog1, which diversifies into a set of neurons that depend on unique neurotrophins (Handler and Ginty, 2021; Meltzer et al., 2021).

Taste sensory information develops a unique step for taste buds in the oral cavity. Fungiform papillae depend on sonic hedgehog (Shh) and Wnt (Hall et al., 2003; Liu et al., 2007), while circumvallate papilla formations are regulated by FGF10 and its receptors Spry1-2 (Petersen et al., 2011). Sox2 expression and innervation are required for continued taste bud development (Okubo et al., 2006; Fan et al., 2019). Taste buds are not differentiated when Sox2 is conditionally deleted. The absence of Neurog2 results in a reduction of Sox2 expression (Okubo et al., 2006; Ohmoto et al., 2017) and causes taste bud formation to abort beyond a limited differentiation of small, single taste buds (Fan et al., 2019). Bitter-sensing taste buds seem to depend on Eya1, suggesting a unique expression of these sensory cells (Ohmoto et al., 2020). The differentiation of specific taste bud cells depends on numerous factors, only some of which have been identified (Pou2f3, Ascl1, and Eya1).

Neurons delaminate from the epithelium and migrate to achieve a final position near where three cranial nerves (facial, glossopharyngeal, and vagus) develop. Upstream of Neurog2, which defines the epibranchial placodes (Fode et al., 1998; O’Neill et al., 2012), is the expression of Eya1, Sox2, and Pax2. Downstream is the expression of Neurod1, Isl1, Pou4f1, Phox2b, and Foxg1 (Alsina, 2020), which are needed to differentiate into the distal part of epibranchial neurons. Phox2b is explicitly required for a visceral sensory neuron fate (D’Autréaux et al., 2011), whereas other genes are needed for differentiation, such as a set of genes, Coe1, Drg11, and Dcx. This set of genes is only turned on after the migrating cells have left the placode (Freter et al., 2012; Smith et al., 2015). Following neuronal differentiation, developing gustatory neurons express Ntrk2 (TrkB), required for the survival of all epibranchial neurons (Fritzsch et al., 1997). Ligands for the TrkB receptor, BDNF and NT-4, are expressed in the epibranchial neuron placodes and the taste buds (von Bartheld and Fritzsch, 2006; Huang and Krimm, 2010; Nosrat et al., 2012). Axons reach the correct target by making multiple decisions at several different locations along the path to the taste buds (Figures 3A–C). In summary, the development of both taste neurons and taste buds depends on a series of genes independent of one another: peripheral axons of neurons reach the taste epithelium. These two cell types become interdependent.

Vestibular neurons innervate vestibular hair cells located within the three canal cristae and two (mammals) or three (most gnathostomes; Fritzsch et al., 2002; Duncan and Cox, 2020) macula (Maklad and Fritzsch, 1999; Straka et al., 2014). Two vestibular neurons project to the vestibular hair cells (Desai et al., 2005): type I and type II, which innervate the two types of vestibular hair cells. The central projection of these neurons shows an incomplete and partial overlap of their target regions (Chagnaud et al., 2017; Elliott and Straka, 2022) that expands from the cerebellum rostrally to r8 caudally (Maklad and Fritzsch, 2003; Straka et al., 2014; Glover, 2020).

Vestibular neurons depend on certain common genes for specification and differentiation. For instance, no neurons develop without Neurog1 (Ma et al., 2000). Upstream of Neurog1 is the expression of the Brg1-based SWI/SNF chromatin-remodeling complex, which interacts with the neurosensory-specific transcriptional regulators of Eya1/Six1 (Xu et al., 2021). These regulator genes induce the expression of Sox2 (Dvorakova et al., 2020) to promote proneurosensory-lineage specification, such as upregulating Neurog1 expression (Reiprich and Wegner, 2015; Kageyama et al., 2019). Downstream of Neurog1 is Neurod1, which is important for neuronal differentiation; many neurons are lost without it. In addition, without Neurod1 ectopic hair cells develop (Jahan et al., 2010; Macova et al., 2019). Pou4f1, Isl1, and bHLHb5, among others, are also needed for normal vestibular or auditory neuron development (Filova et al., 2020). The initial induction of an otic placode depends on Foxi3 (Birol et al., 2016) and Fg3/7/10 (Urness et al., 2011). Any differentiation of neurons or hair cells from this placode requires the expression of Eya1 and Pax2/8 (Zou et al., 2004; Bouchard et al., 2010). A disorganized development occurs following the loss of several genes, including Lmx1a/b, Gata3, Dicer, and Shh, among others (see reviews Riccomagno et al., 2002; Kersigo et al., 2011; Duncan and Fritzsch, 2013; Elliott and Gordy, 2020; Chizhikov et al., 2021). The loss of distinct neurons following specific gene deletion remains unclear (Diaz and Glover, 2021).

Atoh1 is an essential gene for all mechanosensory hair cell development (Bermingham et al., 1999) beyond undifferentiated cells (Fritzsch et al., 2005). Vestibular hair cells differentiate into two types in amniotes: type I and type II (Elliott and Straka, 2022). The macula are divided into two distinct areas and are innervated by neurons that send central processes selectively to either the cerebellum or the brainstem, depending on the area of the macula that they innervate (Maklad and Fritzsch, 2003; Balmer and Trussell, 2019). Emx2 is only expressed in one of these distinct areas (Holley et al., 2010; Jiang et al., 2017) and requires a detailed analysis of central connections to the cerebellum or the brainstem (Figures 4A–C). Hair cells within the utricle and saccule show a polarity that provides 360 degrees of sensitivity. In contrast, hair cells within a particular canal crista are all polarized in the same direction (Fritzsch et al., 2002; Elliott and Gordy, 2020). In summary, the vestibular neurons and hair cells require specific genes for proper development.

Auditory neurons innervate auditory hair cells located within the basilar papilla or cochlea. As with the vestibular system, there are two types of auditory neurons (type I and type II) that innervate two types of hair cells: inner hair cells (IHCs) and outer hair cells (OHCs) (Grabner and Moser, 2018; Jean et al., 2019; Moser, 2020). In mammals, these are arranged in a highly stereotyped organization with one row of IHCs and three rows of OHCs. The central projection of auditory neurons shows a highly organized tonotopic organization in auditory nuclei (Fritzsch et al., 2019).

As with vestibular neurons, auditory neurons also require the expression of Neurog1 as these neurons never develop in the absence of its expression (Ma et al., 2000). Neurog1 interacts with transcriptional regulators to promote the development of auditory neurons, the spiral ganglion neurons (SGN). In addition to Neurog1, these neurons depend on Gata3, Pax2, and Lmx1a/b: without any of these three transcription regulators, all SGNs are lost (Bouchard et al., 2010; Duncan and Fritzsch, 2013; Chizhikov et al., 2021). Downstream of Neurog1 is Neurod1 which regulates the differentiation of sensory neurons and their projections to the organ of Corti and the ventral cochlear nuclei (Jahan et al., 2010; Macova et al., 2019). In addition, SGNs require the expression of Pou4f1, Isl1, and bHLHb5, among others (Filova et al., 2020).

Eya1, followed by Sox2, is needed to initiate hair cell development (Dvorakova et al., 2020; Xu et al., 2021). As with vestibular hair cells, auditory hair cells also critically depend upon Atoh1 expression (Bermingham et al., 1999). Downstream of Atoh1, hair cell differentiation requires Pou4f1 and Gfi1 (Xiang et al., 2003; Hertzano et al., 2004; Eng et al., 2007). Cochlear hair cells also need Pax2, Gata3, Lmx1a/b, and Dicer: without each case, there are no hair cells formed in the cochlear sac (Bouchard et al., 2010; Kersigo et al., 2011; Duncan and Fritzsch, 2013; Chizhikov et al., 2021). Unique genes are needed for the specific differentiation of IHCs (Nakano et al., 2020) and OHCs (Lorenzen et al., 2015; Chessum et al., 2018; Wiwatpanit et al., 2018). For instance, Emx2 is needed to differentiate OHCs (Holley et al., 2010; Kindt et al., 2021). The total length of hair cells in the cochlea depends on Neurog1, Neurod1, Foxg1, Prox1, and n-Myc (Matei et al., 2005; Pauley et al., 2006; Fritzsch et al., 2010; Kopecky et al., 2011; Filova et al., 2020). Hair cells must be polarized with stereocilia aligned in a stair-step organization to function optimally. Specific genes are expressed that establish the distinct polarity of both vestibular and auditory hair cells (Montcouquiol et al., 2003). This polarity develops through steps, resulting in a kinocilium on one end (for vestibular hair cells; cochlear hair cells lack a kinocilium) connected with stereocilia that form a staircase (Fritzsch and Elliott, 2017; Elliott and Straka, 2022). After the polarity is established, Pcdh15 and Chd23 interact as tip links between taller and shorter stereocilia to respond to proper stimulation. Hair cells require a channel, Tmc1/2, to open, allowing an influx of K⁺ ions (Pan et al., 2013; Shibata et al., 2016; Erives and Fritzsch, 2020).

Somatosensory—Developing distinct mechanosensory innervation

Nuclei dedicated for nociception (pain), thermosensation (temperature), pruriception (itch), mechanosensation (cutaneous/touch), and proprioception (limb and body position) are found in the brainstem and spinal cord (Erzurumlu et al., 2010; Lai et al., 2016; Meltzer et al., 2021). An extensive set of different neuronal categorizations is dependent on their size (largest and fast conducting Aα fibers, a variety of medium-sized conducting Aβ and Aδ fibers, and the thinnest unmyelinated C fibers) and provides proprioception (Ia, Ib, and II) that innervates the muscle spindle and tendon organs. The four mechanosensory cells are Merkel cells, Meissner, Ruffini, and Pacinian corpuscles (Abraira and Ginty, 2013; Handler and Ginty, 2021). Most of the innervation is driven by low-threshold mechanoreceptors (LTMR) for detecting mechanical stimuli, but there is also specific innervation for high-threshold mechanoreceptors (HTMR), which provide noxious input by thin, unmyelinated fibers (Meltzer et al., 2021).

Merkel cells depend on the local upregulation of Atoh1 (Maricich et al., 2009) and Pou4f3 for normal development. Like many other sensory innervations, these cells depend on Piezo2 for mechanosensory cation channels, including the Aβ-LTMR innervation. Merkel cells are local next to the keratinocyte cells and show multiple indentations with desmosomes to interact with Merkel cells. An exciting feature is the presence of Piezo2 in the afferent fiber and the Merkel cells, suggesting an interaction between the complex. Synapses form in Merkel cells, releasing them upon proper stimulations (Handler and Ginty, 2021). Compared to the TrkC-dependent Merkel cells that respond to forces of about 10 Hz, some fibers react to specific frequencies around 100 Hz (Meissner cells, which depend on TrkB). In comparison, Pacinian corpuscles are most sensitive to about 200–700 Hz and require Ret for proper innervation (Handler and Ginty, 2021). In summary, a wide variety of mechanosensory inputs depend on Piezo2 to drive obviously in Merkel cells, a unique bHLH gene-dependent formation.

Central nuclei differ between the brainstem and the spinal cord

A large set of expressed genes and proteins regulate the different central projections in the spinal cord (Lai et al., 2016; Handler and Ginty, 2021; Loutit et al., 2021) and the brainstem (Hernandez-Miranda et al., 2017; Hirsch et al., 2021). Whereas the spinal cord receives mechanosensory fibers from proprioceptors (muscle spindles), Aβ, Aδ, and C fibers, the trigeminal system in the hindbrain receives only Aβ, Aδ, and C from the trigeminal neurons to project the principle and the descending trigeminal branches that extend to C4. A separate projection from proprioceptors reaches the trigeminal motoneurons through bifurcations (Matsuo et al., 1999; Ter-Avetisyan et al., 2018) and provides information from the dental gums along the descending track (Lazarov, 2002). Trigeminal motoneurons are a part of the branchiomotor nuclei that belong to the facial, ambiguous, and dorsal motor nuclei (DMNs) that express Phox2a/b and Lmx1b (Figures 1, 2; Fritzsch et al., 2017). At least four dorsal nuclei receive trigeminal innervation, defined by the Ascl1 gene (dA3, dA4, dB1, and dB3). In contrast, only three nuclei receive somatosensory information in the spinal cord (dl3-5; Figures 2A–D). A continuation of Neurog1/2 (dl2) and Atoh1 (dl1) defines the following expression in the spinal cord to regulate. Interestingly, this continued expression of Atoh1 from the spinal cord to the cerebellum (dl1, dA1) shows a gap for Neurog1 (dl2, dA2 to r7 and expression of r1) to add the cerebellum Purkinje cells (Hawkes, 2012; Obana et al., 2018). In the brainstem, the dorsal Atoh1 expressing column gives rise to cerebellar granule cells (r1), the cochlear nuclei (r2-5), the pontine nuclei (r6-7), the pre-cerebellar nuclei (r7-8) and the external cuneate, the cuneate and the gracile (r9-11; review in Farago et al., 2006; Rose et al., 2009; Gray, 2013; Watson et al., 2017a; Arimura et al., 2019; Hirsch et al., 2021; Pop et al., 2021), that project to the cerebellum (pontine, external cuneate, inferior olivary nucleus; ION). Some areas expand from dA4 to expand from the ION, and depend on Ptf1a (Yamada et al., 2014; Iskusnykh et al., 2016; Hirsch et al., 2021), a unique contribution to the brainstem (Hernandez-Miranda et al., 2017).

Rodents best describe the detailed terminals of trigeminal neurons (Erzurumlu and Gaspar, 2020). Central projection of these neurons depends on Hoxa2, Lmx1b, Epha4/7, Slit2, Robo1/2, Neuropilin1, and Sema3a to define the maxillary branch that innervates the unique whiskers for normal central innervation of r2 (mandibular) and r3 (maxillary). Five rows of the maxillary projection were formed to develop a ‘barrelette.’ Robo deletions can cause a duplication from both ipsilateral and contralateral fibers to reach the somatosensory cortex (Iwasato and Erzurumlu, 2018). Retinoic acid (RA) influences the patterning of Hoxa2 (Glover et al., 2006). Lmx1b acts upstream of Fgf8 and limits the rostral distribution of Hoxa2 to r2 (Glover et al., 2018). Fate mapping shows the principal trigeminal nucleus (PrV) is derived from r2 and r3 and the spinal cord (caudalis; SpV) to present an inverted projection that starts in the ophthalmic region to the more dorsal mandibular fibers in a topographical projection of the face. Lower jaws and lips project from the mandibular ganglia to r2, whereas the maxillary fibers project to r3 to innervate the ‘barrelette’ (Erzurumlu et al., 2010). Second order trigeminal neurons give rise to the contralateral lemniscal of SpV, whereas the PrV projects to the thalamus to innervate the VPM (Erzurumlu et al., 2010; Iwasato and Erzurumlu, 2018). Mutants have disorganized central projections that lack the ‘barrelette’ formation (Drg11, Nmdar1, and Lmx1b; reviewed in Erzurumlu and Gaspar, 2020).

In humans, we can define between the PrV, which forms a dorsomedial division, and a ventrolateral division, likely belonging to r2 and r3, respectively. The dorsomedial provides the dorsal trigeminothalamic tract (DTTT; Figure 1A) to reach the ipsilateral VPM, whereas the ventrolateral division, the ventral and dorsal trigeminothalamic tracts (VTTT, DTTT) that crosses to innervate, like other trigeminal lemniscus fibers, the VPM next to the VPL in humans (Suer, 2021) and monkeys (Andrew et al., 2020). It requires a detailed analysis in mice for the ipsilateral projections from r2. A discrete projection from the PrV projects predominantly ipsilateral fibers. Atoh1 is expressed from the cerebellum to the spinal cord but shows a regular formation to generate granular cells, auditory nuclei, pontine, and ION, and receives the external cuneate, cuneate and gracile between r1-r11 (Figure 2). At least four Ascl1 positive genes (dA3, dA4, dB1, and dB3) are driving, in part, the development of trigeminal nuclei. In summary, the spinal and trigeminal central projections are contralateral and are, in part, like the medial lemniscus in the SpV, cuneate, and gracile.

The thalamus receives the topological somatosensory inputs

In mice, fibers reach the VPM to innervate as ‘barreloids’ the thalamic nucleus (Erzurumlu et al., 2010) from r3 projections, while the spinal fibers enter as the contralateral medial lemniscus (SpV, gracile, and cuneate) to innervate the VPL (Iwasato and Erzurumlu, 2018). Interesting bilateral information is described in gums and teeth that show faster ipsilateral projections in humans and monkeys (Andrew et al., 2020; Hihara et al., 2020). We have a somatosensory, topological innervation (Figures 1A, 5A) to project to the mammalian cortex with a primary and secondary innervation (Liao et al., 2020; Qi et al., 2020; Zilles and Palomero-Gallagher, 2021). Based on the earliest description of Brodmann’s work, they defined cortical organizations that were further distinct, showing the ‘homunculus’ by Penfield. In addition, lampreys and gnathostomes show a similar direct trigeminal projection that reaches, with VPM/VPL, the cortex (Briscoe and Ragsdale, 2019; Suryanarayana et al., 2020).

Mice, and other rodents, have a unique organization, the barrel field, a significant innervation that can be manipulated by loss and gain of input (Erzurumlu and Gaspar, 2020). The cortical organization is affected by changing the pattern of central projections. For example, sensory receptors of whiskers differentiate after birth and can change the pattern from the brainstem (barrelettes), thalamus (barreloids), and the cortex (barrel field). Further, manipulations have studied the effect of plasticity on morphological and physiological plasticity. Fgf8 defines the rostromedial neocortical primordia that organize somatosensory cortical maps. Calcium signal drives the thalamus and barrel field before sensory input is present, and without calcium, it results in hyperexcitability to provide projections to the barrel field organization (Antón-Bolaños et al., 2019). The organization is controlled by specific transcription factors with dendrites that suggest the self-organization of a proto-map assembly to drive thalamocortical sensory circuits (Steiner et al., 2020). Cross-modal plasticity can substitute input modality during development, which has important implications for plasticity and evolution after surgical replacement of part of the cortex (Pallas, 2001; Kral and Pallas, 2011). In summary, the developing somatosensory input generates the topological input that all mammals and other vertebrates provide. The role of innervation depends on specific inputs to create the barrel field in mice and can be used as cortical reorganizations.

Taste buds innervate four sensory neurons from five-sensory receptors

Four types of taste bud cells have 30 to 150 tall columnar cells in the orofacial innervation. The largest population (about 50% of cells) are Type I cells, glia-like cells. Type II and Type III cells are receptor cells for the 5 basic taste qualities (sweet, salty, sour, bitter, and umami). Type IV cells are considered immature cells that do not extend to the taste and differentiate into other cell types over the next 20 days (Perea-Martinez et al., 2013; Ohmoto et al., 2020). Type II, III, and IV contribute to 19, 15, and 14% of taste bud cells, respectively (Yang et al., 2020).

Type II taste cells are chemosensors that express G protein-coupled taste receptors (Tas1Rs, Tas2Rs) to detect sugars, amino acids, and bitter compounds. Sugars bind to the dimeric receptor Tas1R2/Tas1R3 while umami activates the dimeric receptor Tas1R1/Tas1R3. It is unknown whether Tas2Rs form dimers in bitterness (Kinnamon and Finger, 2019; Ohmoto et al., 2020; Roper, 2020). Salt/NaCl taste responds to a subset of Type II cells to release neurotransmitters (Bigiani, 2021). Type III taste cells form conventional synapses with sensory afferent fibers and respond to sour tastes (Roper, 2020). These cells respond to a combination of (1) intracellular acidification and (2) proton (H⁺) influx through apical OTOP1 channels (Teng et al., 2019; Zhang Y. et al., 2020).

Six neurotransmitters have been identified in taste buds: ATP, 5-HT, GABA, ACh, glutamate, and noradrenaline (see review Roper and Chaudhari, 2017; Roper, 2021), with receptors expressed on the postsynaptic sensory afferent terminals. Individual taste cells and gustatory afferent nerve fibers respond either to single tastes (sweet, salty, sour, and umami) or, more broadly, to several different taste qualities (Wooding and Ramirez, 2020) that are more compatible with cross-fiber or combined tastes (Roper, 2020). A subset of neurons innervates taste buds that express the mechanosensitive channel Piezo1, which may play a role in the tactile responsiveness of taste buds (Dvoryanchikov et al., 2017). In summary, three sensory cells (sweet, bitter, and umami) are sensory Type II cells, whereas sour cells are Type III cells that synapse onto neurons. Salt sensation is unclear to date.

The solitary tract depends primarily on three genes

Central projections of three different nerves for taste perception (geniculate, nodose, and petrosal) show initially entirely overlapping terminal fields (Figures 3A–E); each nerve occupies a discrete, overlapping territory, suggesting a weak oral topography (Hill and May, 2007; Lundy and Norgren, 2015). A ‘homotopy’ is more pronounced in lampreys and lower gnathostomes than in mammals (Finger, 1978; Fritzsch and Northcutt, 1993; Daghfous et al., 2020). It raises the question of the functionality of constant input (Finger, 2008) and suggests further investigation (Lundy and Norgren, 2015).

Centrally, taste information ends up in the nucleus of the solitary tract (ST or NST; Rose et al., 2009; Gray, 2013; Hernandez-Miranda et al., 2017). The development of the ST depends on a combination of bHLH genes (Ascl1, Neurog2, and Olig3) that outline the dA3 brainstem through the spinal cord (Figures 3A,B). The genes, Tlx3, Lmx1b, and Phox2b are co-expressed in the neurons of the ST but show a different longitudinal expression that is prominent for Phox2b in the dA3 (r1-r11) that end before the DRGs (Qian et al., 2001; Hirsch et al., 2021). Phox2b is essential for the viscerosensory fate that determines the dA3 that convert into dB3-like development (D’Autréaux et al., 2011). In contrast, Tlx3 expressing neurons project from the cerebellum and continue from r4 through the spinal cord (Figure 3; Qian et al., 2001). Lmx1b is expressed throughout the cerebellum and spinal cord (Mishima et al., 2009). A fusion of the left and right ST forms a prominent input that reaches r8-11 and forms the area postrema that expands ventrally to the gracile and overlaps with the hyoglossus (XII; Gray, 2013).

A subpopulation, which expresses the dA3 markers Phox2b/Lmx1b (Figures 3C–E), is localized to the ST by projecting to the parabrachial nucleus (PB) and reaching out to the pre-locus coeruleus (pLC) complex in the prepontine hindbrain (r0-1; Watson et al., 2017b; Gasparini et al., 2021) to subdivide PB as depending on Lmx1b and Phox2b. A large regulation provides homeostatic functions, hunger, thirst, sodium appetite, taste aversion, and cardiorespiratory influence (Karthik et al., 2022). Other ST reaches ventrally and caudally to reach different cervical and thoracic spinal cord segments to innervate pre-motor neurons (Hirsch et al., 2021), among many other connections (Lundy and Norgren, 2015). In summary, taste buds send taste information through three cranial nerves (VII, IX, and X) to the hindbrain that develops specific gene expression (dA3) to innervate the solitary tract (r4-11) that retains a rough orthotopic organization.

Thalamus provides the input to the insula

Solitary tract neurons send output projections to a variety of different targets, including the PB as one of the heaviest targets to receive input from the isthmus (r0) and rhombomere 1 (r1; Watson et al., 2017b). The PB neurons relay via gustatory and general visceral input to the “gustatory/visceral” thalamic region that also receives input from the ST directly (Figure 3A). The thalamic gustatory relay is referred to as ventral posterior thalamic nucleus parvocellular (VPPC) and is defined as a ‘taste relay’ (Herbert et al., 1990; Lundy and Norgren, 2015). This unique, predominantly ipsilateral connection breaks down into a complex interaction from the VPPC that receives fibers from the ipsi- and, to a lesser extent, the contralateral PB. The largest PB injections reach the medial septum, olfactory tubercle, and receive the gustatory cortex (Saper and Loewy, 1980; Lundy and Norgren, 2015). The orthotopic organization in the ST is not preserved in higher-order nuclei, making the need for any orthotopic primary map even fuzzier (Lundy and Norgren, 2015). PB neurons project axons indirectly to the insular cortex via VPPC (reviewed in Lundy and Norgren, 2015; Gasparini et al., 2021).

In primates, including humans, the central gustatory pathway bypasses the PB, projects directly to the VPPC, and reaches the granular region of the insular cortex (Lorenzo, 2021). Highly conserved second-order neurons project taste information to the cortex (Lundy and Norgren, 2015; Vendrell-Llopis and Yaksi, 2015; Staszko and Boughter, 2020). The major terminal field in the insula region interacts with other inputs, in addition to taste (Fain, 2019; Staszko et al., 2020). In contrast to the clear-cut neuroanatomical pathways for taste, a uniform interpretation of taste coding has yet to emerge in the insula cortex. Imaging studies have conflicting reports (Chen et al., 2011, 2021), and many studies report a broad distribution of taste neurons throughout the insula cortex with no spatial organization (Staszko et al., 2020), including fMRI studies in human subjects (Canna et al., 2019; Avery et al., 2020). Gustatory responses were elicited from the anterior tongue and posterior tongue, and pharyngeal responses were located caudal to the anterior tongue (Hanamori et al., 1998). The reciprocal connection between the cortex, PB, and ST indicates descending projection from forebrain areas converging the brainstem gustatory neurons, the PB, and the ST into an integrated interaction. In summary (Figure 3A), gustatory PB neurons reach the taste cortex without or with a relay with the thalamus and have reciprocal connections with the gustatory and visceral cortex directly from the PB.

Vestibular nuclei develop from two areas in the hindbrain

Vestibular neurons innervate the semicircular canals, and otolithic organs at the periphery and project centrally to the vestibular nuclei (Straka et al., 2014). Mammals have four defined nuclei: the descending, lateral, medial, and superior vestibular nuclei (DVN, LVN, MVN, and SVN). Nuclei that form the SVN are from r1-3, LVN mostly from r4, DVN from r5-8, and MVN from r2-8 (Chagnaud et al., 2017; Glover, 2020; Elliott and Straka, 2022). Vestibular afferents project partially overlapping and separate within the vestibular nuclei (Figures 3A–D). There is no topologically organized central or peripheral organization compared to the trigeminal system, which is more akin to the taste input (Fritzsch et al., 2019). Within the overlap of different central projections, one can distinguish them into incomplete projections in partial central information. For example, the cerebellum receives only fibers from the saccule to the uvula, whereas the posterior canal is prominent to the nodulus and has a limited expansion to the uvula (Maklad and Fritzsch, 2003; Balmer and Trussell, 2019), like the remaining vestibular organization (Elliott and Straka, 2022). Central projections depend on the expression of specific genes. For example, in the absence of Neurod1, Gata3, or Fzd3, there is an overlap of cochlear and vestibular central projections (reviewed in Duncan and Fritzsch, 2013; Duncan et al., 2019; Filova et al., 2020; Stoner et al., 2021). Lmx1a/b double null mice cross afferents to the contralateral side after the choroid plexus is eliminated (Chizhikov et al., 2021).

Vestibular nuclei develop from at least two longitudinal central nuclei progenitor populations: dB1 and dB2 (Hernandez-Miranda et al., 2017; Hirsch et al., 2021). dB1 expresses Ascl1 and Ptf1 from the spinal cord to the cerebellum (r0-r11, spinal cord). Additional genes expressed are Lbx1, Pax2, and Lhx1/5 (Diaz and Glover, 2021). In contrast, a unique gene expression is present in dB2, Phox2b, and Lbx1, from r1 to r6. A later manifestation of Atoh1 is reported (Hirsch et al., 2021). Gene expression of Hoxb1 overlaps with r4 and depends on Phox2b and Lbx1 that define the LVN, reaching the ipsilateral spinal motor neurons (Chen et al., 2012). Further, developing mice (E13-15; Lunde et al., 2019) show a nearly equal expression of Esrrg, Maf, and Pou3f1 in LVN/MVN. In contrast, Lhx1/5, Evx2, and Foxp2 are expressed in the nearby MVN in r5 that project in a different direction with additional expression of several other genes (Lunde et al., 2019). In addition, a reduced expression level in Onecut1-3 is higher in the MVN compared to the LVN (Glover, 2020). Additional expression is possible from other origins (Figures 4A,D). In particular, dB4 is positive for bHLH genes Neurog1 and Neurog2 (Diaz and Glover, 2021). Gene expression changes by migration later, following in part the information from neurons developing spinal and brainstem neurons (Lai et al., 2016; Hernandez-Miranda et al., 2017). Further work is needed to define the contribution of at least two earliest gene expressions (dB1, dB2) that originate along the longitudinal gene expression from r1-r8 (Diaz and Glover, 2021; Hirsch et al., 2021). In summary, at least two areas are defined to add distinct vestibular nuclei (Figures 3A,C,D) that indicate other origins, making it a complex migration from different neuronal populations (reviewed in Diaz and Glover, 2021).

Neurons from the thalamus project to the insula to provide vestibular information

Semicircular canal organs converge with otolithic organs to generate an end-organ-specific central organization that is incompletely presented in overlapping and distinct functional inputs (Straka et al., 2014; Elliott and Straka, 2022). Considerable overlap of vestibular end-organs precludes defining a classical sensory map as the organizational principle (Chagnaud et al., 2017). Three major 2^o order output fibers are determined from the vestibular nuclei:

1. The LVN gives rise to the ipsilateral projection to the spinal cord motoneurons via the lateral vestibulospinal tract (LVST).

2. The MVN, SVN, and DVN project bilaterally to the ocular motoneurons (III, IV, and VI).

3. Additional fibers project to the lateral mammillary and dorsal tegmental nucleus to extend via the anterodorsal and ventral posterior lateral thalamus (AND, VPL) to the parietal-insula vestibular (PIVC) and the ventral intraparietal cortex (VIP).

We concentrate on the thalamocortical interaction (Figure 4A), providing a limited discussion on the two motor outputs of the LVN to the spinal cord, as all other vestibular fibers project to the ocular motoneurons (see review Fritzsch et al., 2017; Glover, 2020; Elliott and Straka, 2022). Within this connection is the cerebellum that receives and emits fine control of movements and storage of velocity and activates visual-vestibular integration using the nucleus propositus hypoglossal and contacts with the parabrachial nucleus to deal with cardiovascular and respiratory control (Cullen et al., 2021; Dieterich and Brandt, 2021; Elliott and Straka, 2022). The vestibular-ocular reflex (VOR) provides active and passive head-body motion, whereas the optokinetic reflex (OKN) provides visuomotor responses (Figure 3A).

Vestibular projections are bilateral, primarily from the vestibular nuclei to the thalamus and cortex. Certain areas may project ipsilaterally and contralaterally to the cortex. It appears that certain stimuli project via ipsilateral connections to the extent of the PIVC in humans. However, other stimuli project bilaterally or even a dominant contralateral input that reaches across three different levels to innervate the cortex (Dieterich et al., 2017), showing an asymmetric interaction from vestibular nuclei (VN) followed by a dominant contralateral input that crosses back except for the thalamus (Dieterich and Brandt, 2018). At least three midline crossings are located: (1) at the level of the vestibular nuclei; (2) at the pontine level above the vestibular nuclei; (3) at the mesencephalic level (Dieterich and Brandt, 2021). Vestibular nuclei project via a relay to pass through the nucleus propositus hypoglossal (NPH) and reach the supragenual nucleus (SGNu) that plays a role in the head direction network (HDC; Cullen et al., 2021; Taube and Yoder, 2021). From here, projections expand to the anterior vestibular-thalamic pathway (ADN) in the thalamus, and, from there, it projects to the limbic and entorhinal cortex. In addition, vestibular projections project bilaterally to the ventral posterolateral nucleus (VPL) that innervates the parietal insula of the vestibular cortical area (PIVC). In addition to the parietal-insula vestibular area (PIVC), vital information is sent to the visual temporal Sylvian area (VTS) in the retroinsular cortex, the superior temporal gyrus (STG), the inferior parietal lobule (IPL), the anterior cingulum, the hippocampus, and area 6a. Posterolateral thalamic nuclei project bilaterally via VOPL, VPM, and VL (Dieterich et al., 2017) to end up in the PIVC, but some fibers from the brainstem can project directly to the PIVC. Overall, fiber tracking data in humans shows that both the VPL (crossed and uncrossed projections) and VP (uncrossed projections) thalamus serve as vestibular relay stations for brainstem projections to (or from) the vestibular cortex.

Moreover, the ventral intraparietal areas (VIP) play a role in the vestibular circuits, where they interact with somatosensory and visual information to interact with the optokinetic (reviewed in Dieterich and Brandt, 2021). In summary, consistent with the interactions from the brainstem, there is a bilateral projection from the vestibular nuclei to the cortex, except for the lack of crossing fibers in the thalamus. The cortical areas innervated are the parietal insula, limbic and entorhinal cortex, and contacts the premotor cortical areas.

Auditory nuclei depend on Atoh1

The expression of Atoh1 in r2-5 (Wang et al., 2005) is needed to develop the anterior-ventral, posterior-ventral, and dorsal cochlear nuclei (AVCN, PVCN, and DCN; Figures 4A–C). Within dA1 is the expression of the bHLH gene Olig3 and Lhx2/9, Barhl1/2, and Pou4f1 to differentiate cells into neurons (Hernandez-Miranda et al., 2017). Atoh1 expression depends on the choroid plexus formation to initiate Lmx1a/b to distinguish cochlear nuclei (Chizhikov et al., 2021). In addition, there is a loss of Wnt3a in Lmx1a/b double null mice (Elliott et al., 2021b). Furthermore, specific migration reaches out to add to dA1 and at least dA4 that depends on Ptf1a (Iskusnykh et al., 2016; Lipovsek et al., 2017; Hirsch et al., 2021). Cochlear nuclei are active before hair cells (Marrs and Spirou, 2012). Previous work showed that the SGN central projection typically develops in the absence of Atoh1 and, thus, lacks hair cells and cochlear nuclei (Elliott et al., 2017). Future differentiation will generate a large set of cochlear nuclei that must be traced from the earliest gene expression to establish into bushy, globular, spherical, stellate, and pyramidal neurons (Malmierca, 2015; Oertel and Cao, 2020). In addition, unipolar brush cells, granule cells, and fusiform cells develop in the DCN; each is uniquely positive for calbindin, calretinin, and parvalbumin (reviewed in Caspary and Llano, 2018; Elliott et al., 2022b).

In addition to the cochlear nuclei (Figures 5A–C), several neurons migrate more ventrally to develop mostly from r4 to generate the medial olivocochlear nuclei (MOCs; Farago et al., 2006; Di Bonito et al., 2017). Different projections innervate ipsi- and contralateral MOCs (Marrs et al., 2013; Lipovsek and Wingate, 2018; Kandler et al., 2020; Grothe, 2021). For example, the development of detailed innervation shows an exuberant innervation of the MNTB followed by the development of the calyx of Held with time (Holcomb et al., 2013) that may decline in density formation with age (Elliott et al., 2022b). SGN fibers end up in a tonotopic organization that follows a base to apex progression to the MOCs. Both time and intensity are used to identify the unique frequency for future interactions (Grothe, 2021; Kandler et al., 2021). Innervation shows a bilateral projection from each cochlear nuclei to reach the lateral lemniscus to extend to the inferior colliculi (Macova et al., 2019; Syka, 2020). In summary (Figure 5), Atoh1 is developing the three cochlear nuclei and designing the MOCs into distinct innervation to create massive calyx formations.

Inferior colliculi and medial geniculate bodies project to the auditory cortex

Ascending projections to the inferior colliculi (IC) are bilateral projections either directly from the cochlear nucleus or indirectly from the cochlear nuclei (AVCN, PVCN, and DCN) through the MOCs and the lateral lemniscus, all while maintaining the tonotopic organization (Malmierca, 2015; Macova et al., 2019; Syka, 2020). The IC projects from the unilateral medial geniculate ganglion (MGB) that have limited crossing fibers in the IC (Grothe, 2021). Limited data (Gurung and Fritzsch, 2004; Razak et al., 2009) shows the earliest projection that has a prominent expression of Neurod1 to the IC to project to the auditory nucleus (Malone et al., 2021). The central part of the IC is organized in a tonotopic organization that responds to both direct and indirect fibers (Malmierca, 2015; Budinger, 2020).

The MGB consists of several subareas, the major parts being the ventral (MGV), medial (MGM), and dorsal (MGD) divisions (Budinger, 2020). Its main ascending input, from the ventral nucleus of the medial geniculate body, provides connections that comprise only a fraction of the inputs to the auditory cortex (A1; ∼18%), where these ascending signals are integrated with feedback connections from ipsilateral (∼70%) and contralateral (∼10%) cortex (Malone et al., 2021). The MGV constitutes the tonotopic lemniscal part of the MGB (Figure 5A). The MGD and MGM are usually assigned to the largely non-tonotopically organized, non-lemniscal position (Malmierca, 2015; Budinger, 2020). MGB and IC receive direct input from the auditory cortex to influence information processing (Budinger, 2020; King, 2020). In summary, a bilateral relay from the IC is mostly unilateral to the MGB, which expands bilaterally in the auditory cortex.

Integration of sensory inputs

Within the cortex, inputs from the various sensory systems interact at a tertiary level with inputs from other systems to provide an integrated sensory experience. Cortical integration of somatosensory, taste, vestibular, and auditory systems with other sensory and motor systems will be discussed below. Unfortunately, much of the detailed description is presented in humans, with a few exceptions from other mammals. Moreover, the detailed description will require additional analysis to evaluate the distinct genes needed for the integrated perspective.

Somatosensory integration

Cortical somatosensory projections interact tertiarily with those from other sensory systems to form a cohesive perception (Hans et al., 2020). For example, multisensory integration depends on auditory, visual, olfactory, and somatosensory interaction (Meehan et al., 2017). Somatosensory neurons show integration properties like those of visual stimuli. Neural responses account for the perceived direction of the stimuli across all stimulus conditions tested, highlighting the somatosensory system, a vector average mechanism to compute tactile motion direction that bears striking similarities to its visual counterpart (Pei et al., 2011; Goodman and Bensmaia, 2021). In addition, the vibrissa information from the primary motor cortex (M1) modulates sensory processing in the primary somatosensory barrel cortex to provide interneurons in vivo that of somatostatin interneurons decreased during whisking, suggesting the firing rates during whisking depended on M1 activity (Figure 6A). A circuitry exists by which inputs from the motor cortex influence activity in the somatosensory cortex (Lee et al., 2013).

A secondary somatosensory area (S2) and the ventral parietal area (PV) are organized mirror-symmetrically relative to the primary somatosensory area in humans. Bilateral spatiotemporal integration in S2/PV takes place over a sizeable cortical area and presents the properties of spatiotemporal integration in anterior parietal regions. Importantly, S2/PV showed significant temporal integration while spatial integration was reduced (Zhu et al., 2007; Caldwell et al., 2019).

A tactile and visual interactive response to activation of various brain regions shows the effects of visuotactile multisensory integration on the amount of brain activation in the somatosensory cortical regions. Tactile stimulation-induced cortical activations occur in the left primary sensory-motor cortex (SM1) and secondary somatosensory cortex (S2). In the visuotactile stimulation task, cortical activations can happen in both primary S1 and secondary S2 and can reach the posterior parietal cortex by activation (Kwon et al., 2017). A visual and somatosensory input is reliable, reducing reliance on the vestibular system. Postural training can alter sensory organization after a visual feedback-vestibular activation training protocol, suggesting a possible sensory reweighting through vestibular adaptation and/or habituation occurs (Appiah-Kubi and Wright, 2019) that helps indicate with the somatosensory input, in particular the tendon of the limbs in humans (Peterka and Loughlin, 2004). Multisensory work in rodents suggested a segregation of mechanically driven, temperature-selective and gustatory responses that provide stimuli in the oral somatosensory cortex (Clemens et al., 2018). Furthermore, the mid-dorsal insula and the primary and secondary somatosensory cortex suggest that it involves a specific set of brain regions of the oral-sensory integration processes (Wistehube et al., 2018).

Beyond a hierarchical model of multisensory integration, a unimodal input converges onto higher multisensory areas that integrate multimodal information and guide decision-making and behavior. The superior temporal sulcus, lateral occipital–temporal cortex, posterior parietal cortex, and ventrolateral frontal cortex have been implicated in higher cortical multisensory processing (Meehan et al., 2017; Zhou et al., 2019). The involvement of primary sensory areas in multisensory processing indicates a multisensory integration that may involve more distributed neural networks beyond classic hierarchical multisensory-specific areas, including primary sensory cortices of the somatosensory input, including the associated somatosensory information (Brunert and Rothermel, 2020).

Somatosensory-auditory integrations

In the mouse, primary (S1) and secondary (S2) somatosensory cortices pair sound with whisker stimulation, modulating tactile responses in both S1 and S2. After local mutually inhibitory S2 circuit can spectrally select tactile versus auditory inputs suggesting a multisensory integration that provides a key role for S2 in combining auditory and tactile information (Zhang M. et al., 2020), Magnetoencephalography can be used to study auditory stimulation by cortical activity in the primary somatosensory cortex. Somatosensory–acoustic interaction reaches area 3b from the thalamus during the early stages of synaptic transmission (Sugiyama et al., 2018).

Somatosensory-vestibular interactions

Vestibular stimulation can speed tactile detection and indicates vestibular facilitation of somatosensory processing (Pfeiffer et al., 2018). Specifically, somatosensory- vestibular interactions affect the perceived timing of tactile stimuli (Moro and Harris, 2018). Caloric vestibular stimulation can result in somatosensory stimulation (Bretas et al., 2020) that can be combined with vestibular and postural training to improve balance in healthy young adults (Appiah-Kubi and Wright, 2019; Ferrè and Haggard, 2020). Area 3a extends from the somatosensory cortex and reaches the superior parietal lobule of the parietal-insula vestibular area (PIVC), and the extrastriate regions (Shaikh et al., 2020).

Somatosensory-taste interactions

Cross-modal correspondences occur between tastes and the somatosensory attributions of food and beverage products (Spence and Ngo, 2012). Beyond cross-modal integration occurs with somatosensory with sweet tastes (Spence, 2022). Results show that flavor perception must affect somatosensory stimuli as components of flavor perception and taste modulators (Green and Nachtigal, 2012). In mice, ascending projections from trigeminal neurons reach the parabrachial nucleus (PB). Trigeminal fibers use TRPV1 and TRPA1 to play a role in thermosensation and pain communication with parabrachial taste neurons (Li et al., 2021; Rhyu et al., 2021). In summary, mammals’ somatosensory cortical regions (S1, S2, and PV) have integrated with the motor cortex and visual, auditory, vestibular, taste, and olfactory cortical areas.

Conclusion of somatosensory inputs

Trigeminal input (Figures 2A, 6A) spans the spinal cord through the hindbrain to organize a topological input (leg > arm > face). Fibers reach the medial lemniscus, which projects to the VPL/VPM to innervate the principal (S1) and secondary somatosensory fibers (S2, PV). A tertiary interaction is evident in the motor cortex and has visual, auditory, vestibular, olfactory, and taste interactions to generate a cohesive perception. Further work is needed to detail the various sensory inputs to coordinate with the trigeminal primary (S1) and secondary (S2, PV) into a cohesive interaction.

Taste integration

The solitary tract (ST) innervates, directly and indirectly, the parabrachial nuclei (PB, Figure 3A), from which primarily ipsilateral fibers extend the thalamus in mice (Lundy and Norgren, 2015; Gasparini et al., 2021). Neurons also directly reach the thalamus from ST to reach the gustatory and taste relay, referred to as ventral posterior thalamic nucleus parvocellular (VPPC). The unique ipsilateral connection and, to a lesser extent, the contralateral PB reach the insular cortex, which develops a reciprocal connection with gustatory and visceral afferents. The insula functions from sensory processing to representing feelings and emotions, autonomical and motor control, risk prediction and decision-making, bodily- and self-awareness, and complex social functions like empathy (Gogolla, 2017). Moreover, the insula subserves various human functions ranging from sensory and affective processing to high-level cognition (Uddin et al., 2017).

Beyond function, the loss of the right insula affects ipsilateral taste recognition, whereas the left-hemispheric stroke patients’ taste recognition on the right side of the tongue suggests that taste information from both sides of the tongue passes through the left insula (Pritchard et al., 1999). The insular cortex may be organized into a hedonic or “viscerotropic” map rather than one ordered according to taste quality. In contrast to a spatial map ordered by taste quality, the insular cortex interacts with other sensory inputs, integrating taste, somatosensation, and olfaction into a cohesive experience of food intake (Shepherd, 2006; Schier and Spector, 2019). Once it reaches the mouth, the taste, texture, and retronasal smell of the food deliver sensations of olfaction and taste (Figure 5B). The cortical presentation of both spatial and temporal divisions of taste coding changes with experience. Cortical taste representation is not organized in a spatially discrete map but instead is plastic and spatially dispersed that uses temporal information to encode many types of ingested stimuli (Staszko et al., 2020).

In contrast to visual, auditory, and somatosensory cortices that are topographically organized, the spatial organization of taste responses in the gustatory insular cortex responses shows no sign of spatial clustering or topography. Thus, the idea that taste qualities in the insula are sparse and distributed is analogous to the representation of odorants in the piriform cortex (Chen et al., 2021). In humans, recent studies do not support the topographic model of taste quality representation, but rather taste quality is a distributed pattern within gustatory regions of the insula (Avery, 2021). Several indications suggest that information about taste stimuli is more profound when paired with odors than when presented alone (Lorenzo, 2021). These data show that a lack of direct projections from the olfactory bulb or piriform cortex may be conveyed via the gustatory cortex, amygdala, or lateral hypothalamus, suggesting a convergence of information of olfaction and taste (Lorenzo, 2021).

Four functionally distinct regions have been identified in the human insula (Figure 5B): (1) a sensorimotor region located in the mid-posterior insula; (2) a central olfactory-gustatory region; (3) a socio-emotional region in the anterior-ventral insula; and (4) a cognitive anterior-dorsal region (Uddin et al., 2017). Trigeminal and olfactory input converges and integrates within the insular region (Hummel et al., 2009) which is expanded for spicy stimulations in humans (Han et al., 2021; Spence et al., 2021). Somatosensory manifestations represent a large proportion of responses elicited by electrical stimulation of the insular cortex in humans. Likewise, input from the auditory system and the parieto-insular vestibular cortex reach the insular cortex. The posterior granular insula in monkeys integrates vestibular inputs and proprioceptive, visual motion, and auditory activities comparable to humans (Evrard, 2019). However, an interaction between auditory, speech influence, and vestibular integration with taste input is unclear (Uddin et al., 2017). A poly-modal integration likely contributes to “higher” vestibular functions, including self-motion perception in a body- (proprioceptive) and the world- (audio-visual) centered referential system that could be organized into a dorsoventral gradient (Evrard, 2019).

Taste-auditory integrations

Tasting in the presence of auditory stimuli of various frequencies or silence suggests that low frequencies increase red wine’s perception and aromatic intensity (Crisinel and Spence, 2009; Burzynska et al., 2019). Moreover, cross-modal interactions can affect the background music on the drink’s taste (Hauck and Hecht, 2019).

Taste-vestibular interactions

Vestibular stimulation can speed tactile detection and indicates vestibular facilitation of somatosensory processing (Pfeiffer et al., 2018). Caloric vestibular stimulation can result in somatosensory stimulation (Bretas et al., 2020) combined with vestibular and postural training (Appiah-Kubi and Wright, 2019; Ferrè and Haggard, 2020).

Taste-somatosensory interactions

Cross-modal correspondences between taste and the somatosensory attributions of food and beverage products (Spence and Ngo, 2012). Beyond cross-modal integration with somatosensory in particular, sweet flavors (Spence, 2022) that connect to drinking have unclear interactions (Green and Nachtigal, 2012).

Conclusion of taste inputs

The taste bud projection (Figures 3A, 6B) from the three visceral ganglia reaches r4-11 to form the solitary tract. Fibers reach from the ST to the PB and go bilaterally to the VPPC to innervate the insula. To generate a cohesive perception, a tertiary interaction is limited to somatosensory, olfactory, auditory, and vestibular interactions.

Vestibular integration

In monkeys, the PIVC seems to be the dominant multi-modal vestibular cortical area, which is closely connected to the other regions, the vestibular nucleus, and to the opposite hemisphere (Figure 4A). Neurons are focused on VPL and ADN but have additional projections from MVN and SVN (Dieterich and Brandt, 2018). The thalamus is multimodal and encodes vestibular signals to project somatosensory, proprioceptive, visual, and auditory sensory information (Figure 4A). From here, the VPL is important as multisensory inputs to have a perceptual perception of head and body motion that are directly innervating the parietal-insula of the vestibular cortical area (PIVC) and the ventral intraparietal areas (VIP; Cullen, 2019). A strong connection exists with 3a and 2v that interacts with multiple connections to play a role in the frontal eye field. Notable inputs are the medial superior temporal area that plays a role in visual motion that interacts with vestibular and visual stimulation (Cullen, 2019). Without any doubt, the visual-vestibular interactions are directly and indirectly involved in optokinetic stimulation and caloric irrigation in normal and partial deletion of various regions (Dieterich and Brandt, 2019; Cullen et al., 2021). Two areas, the lateral entorhinal and postrhinal cortex contain center-bearing and center-distance cells (Figure 4C), which discharge based on the animal’s orientation and distance to the center of its environment (Taube and Yoder, 2021). Going beyond the visual/vestibular interactions, we provide an overview of somatosensory, taste, and auditory integration. Because this vestibular cortical network is so widely distributed, it could impact multiple neurocognitive functions in health and disease. The most impact is dependent on Caloric Vestibular Stimulation (CVS) or Galvanic Vestibular Stimulation (GVS) to cause an integrated perspective (Ferrè and Haggard, 2020).

Vestibular-somatosensory interactions

Following the peripersonal space can be used to define upper limbs for vestibular signals (Pfeiffer et al., 2018). For example, vestibular stimulation can speed tactile detection and indicates vestibular facilitation of somatosensory processing, suggesting that vestibular inputs dynamically interact with multisensory input (Pfeiffer et al., 2018). Caloric vestibular stimulation can result in somatosensory activity (Bretas et al., 2020) that can be activated after somatosensory reduction and can combine vestibular and postural training to improve balance in healthy young adults and likely in those with deficits (Appiah-Kubi and Wright, 2019; Ferrè and Haggard, 2020).

Vestibular-taste interactions

Pairing a novel taste with an offensive vestibular stimulation result in conditioned taste aversions in rats and humans. Vestibular system involvement in gustatory conditioning was compared in sham-lesioned or labyrinth ectomized rats with control rats for oral rejection reactions (Ossenkopp et al., 2003). Previous work showed that vestibular input could influence several aspects of consumer behavior to affect food taste (Biswas et al., 2019). Robust taste aversions can be solicited in rats and mice affected after the bilateral loss of labyrinth function, suggesting a direct interaction of the magnet on the vestibular system in Het mutant mice (Cote, 2020).

Vestibular-auditory interactions

Rhythm can be biased by passive motion, which suggests that vestibular input may play a vital role in the effect of movement on auditory rhythm processing (Phillips-Silver and Trainor, 2008; Phillips-Silver et al., 2020). Recent evidence suggests that simultaneous auditory–vestibular training facilitates short-term auditory plasticity, producing stronger oscillator connections in the acoustic network (Tichko et al., 2021). Most importantly, a combined vestibular and cochlear prosthesis may restore hearing and balance in patients who have lost both (Phillips et al., 2020). A significant interaction between different sensory modalities during stimulation with a combined vestibular and cochlear prosthesis can help challenge stimulation strategies to simultaneously restore auditory and vestibular function after such an implant (Phillips et al., 2020).

Conclusion of vestibular inputs

Vestibular information (Figures 4A, 6C) is sent from the periphery to four vestibular nuclei (spanning r1-r8) along vestibular sensory neurons. Ascending projections are bilateral to innervate the ventral posterolateral nucleus (VPL) and anterior vestibular-thalamic pathway (ADN) to the parietal insula of the vestibular cortical area (PIVC) and the ventral intraparietal regions (VIP). Higher-order integration beyond the demonstrated visual-vestibular interaction is limited to somatosensory, taste, and, importantly, auditory-vestibular interactions.

Auditory integration

The auditory cortex (A1) is a central hub located at a pivotal position within the auditory system, receiving a tonotopic input from the medial geniculate body (MGB) and playing a role in the sensation, perception (Figure 5A), and interpretation of the acoustic environment (Malone et al., 2021). ‘Receptive fields’ (RF) end up a particular stimulus dimension: the choice of acoustic stimulus dimensions to explore are related to the basic perceptual attributes such as spectral pitch (frequency), loudness (intensity), periodicity, or virtual pitch (amplitude modulations, harmonic series), timbre (spectral envelope), and sound location (interaural time, level differences, and spectral shape). In humans, hemispheric asymmetries between the left and suitable auditory cortex form in gross anatomical features and cortical microcircuitry (Budinger, 2020; Rauschecker, 2021).

There are two major classes of neurons in the neocortex: principal cells and interneurons. The auditory cortex is organized into six horizontal layers, another defining feature of the neocortex. It has anatomical and functional vertical columns and has intense interhemispheric connections between the auditory cortices of both hemispheres, which cross the midline via the (caudal body of the) corpus callosum (Budinger, 2020; Malone et al., 2021). Primary sensory cortices like A1, S1 (somatosensory), and V1(visual) are not purely unimodal but also process other non-matched sensory and non-sensory information. Projections from V1 and, to a lesser degree, from S1 toward A1 mainly arise from subgranular layers and provide feedforward organizations (Budinger, 2020).

The auditory corticobulbar system includes projections from the nuclei of the medial geniculate body (MGB), the inferior colliculi (IC), the lateral lemniscus (NLL), superior olivary complex (SOC), and the cochlear nuclear complex (CN) including the medial olivary efferents that project out to the outer hair cells (OHCs; Simmons et al., 2011; King, 2020). The prominent corticothalamic, corticocollicular, colliculofugal, and olivocochlear connections descend from auditory pathways that include long-range connections (Malmierca, 2015; Budinger, 2020; King, 2020). Overall, the feedback loop might modulate auditory response properties of neurons in the midbrain and hindbrain, adapting their sensitivity to sound frequency, intensity, and location (Budinger, 2020; King, 2020).

As previously mentioned, the auditory cortex interacts with multiple other sensory projections as tertiary fibers (Figure 5D). For example, auditory soundscape or visual landscape can influence the perception in a real, multisensory environment (Pheasant et al., 2010; Zhao et al., 2018). An interesting McGurke effect showed an interaction between visual and auditory stimuli, highlighting a higher interaction of otherwise discrete stimulations (Spence and Soto-Faraco, 2010). Most interestingly, multisensory interactions are audiovisual speech perception, in which visual speech substantially enhances auditory speech processes (Plass et al., 2020) that be more robust in audio-visual speech responses than what would have been expected from the summation of the audio and visual speech responses, suggesting that multisensory integration occurs (O’Sullivan et al., 2021).

Auditory-somatosensory interactions

Typical, auditory-somatosensory (AS) humans can multisensory stimulus pairs yield significant reaction time facilitation relative to their unisensory counterparts that exceeded probability (Murray et al., 2005). Multisensory interactions across space suggest perceptual-cognitive phenomena of sensory-cognitive processing (Foxe et al., 2002). Further, integration of auditory and somatosensory information in speech processing can be used in a bimodal perceptual task, suggesting somatosensory information on sound categorization can affect adults more than children (Trudeau-Fisette et al., 2019). This results in somatosensory cross-modal reorganization of the auditory cortex in adults with early stage, mild-moderate age-related hearing loss [ARHL (Cardon and Sharma, 2018)]. Pairing sound with whisker stimulation modulates tactile responses in both S1 and S2, with the most prominent modulation being robust inhibition in S2 (Zhang M. et al., 2020). Mutually inhibitory activation from the S2 circuit can spectrally select tactile versus auditory inputs in mice (Zhang M. et al., 2020).

Auditory-taste interactions

In contrast to vision and somatosensory interaction, very few examples exist supporting some level of interaction between the auditory and taste systems (Keast and Breslin, 2003; Zampini and Spence, 2012). The strongest is provided through a cross interaction between taste and auditory pitch (Holt-Hansen, 1968; Spence, 2011). Most recent evidence suggests an attenuation of taste neophobia induced by taste familiarity and is auditory context-dependent in mice (Grau-Perales and Gallo, 2020). Context dependency involved dopaminergic activity mediated by D1 receptors, which might be responsible for adequately acquiring safe taste recognition memory (Grau-Perales et al., 2021). In humans, associations between auditory attributes and a number of the commonly agreed basic tastes are now recently added as ‘saltiness’ (Wang et al., 2021), among other interactions of taste with auditory interactions (Spence et al., 2021).

Auditory-vestibular interactions

Since rhythm can be biased by passive motion, it suggests that vestibular input may play a vital role in the effect of movement on auditory rhythm processing (Phillips-Silver and Trainor, 2008; Phillips-Silver et al., 2020). Recent evidence suggests that simultaneous auditory-vestibular training facilitates short-term auditory plasticity, producing stronger oscillator connections in the auditory network (Tichko et al., 2021). Most importantly, it can be combined in a vestibular and cochlear prosthesis to restore hearing and balance in patients who have lost both (Phillips et al., 2020). A significant interaction between different sensory modalities during stimulation with a combined vestibular and cochlear prosthesis can help challenge stimulation strategies to simultaneously restore auditory and vestibular function after such an implant (Phillips et al., 2020).

Conclusion on auditory integration

Cochlear spiral ganglia (Figures 5A, 6D) are unique compared to the nearby vestibular ganglia and project tonotopically to innervate the cochlear nuclei that form from r2-5. A notable development migrates from the cochlear nuclei (r4) to receive order projections. Projections are bilateral to innervate via IC and MGB to innervate the A1. Higher-order auditory interactions are limited to somatosensory, taste, and, importantly, the auditory-vestibular interactions.

A combined perspective of the four brainstem sensory systems

The peripheral sensory innervation depends on unique regional cell types (Merkel cells, free endings, taste buds, and mechanosensory hair cells). Trigeminal, vestibular, auditory and taste neurons depend on the bHLH genes, Neurog1/2. Central organization depends on Atoh1 (dA1) for auditory, Phox2b/Tbx3 (dA3) for taste, Phox2b/Ptf1a (dB1/2) for vestibular and Ascl1/Neurog1 (dB3/4) for trigeminal to integrate through migration into the distinct brainstem organization. Output is prominent mainly to the contralateral thalamus (trigeminal), ipsilateral via an intermediate (PB, taste) to the thalamus, bilaterally through the midbrain intermediate to the thalamus in the vestibular, and bilaterally from the auditory a secondary (SOC, LL) to the tertiary (IC) and the thalamus (MGB). From here, a topological organization projects to the somatosensory (S1) and the auditory telencephalon (A1), whereas the taste (insula) and vestibular (VPL, PIVC) are not organized topologically. Higher-order interactions are demonstrated between trigeminal and auditory, between taste and vestibular, between the auditory and the vestibular, and between taste and trigeminal to integrate the cohesive perception of the telencephalon mostly documented in human.

Data availability statement

The original contributions presented in this study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Author contributions

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

Acknowledgments

We thank members for their comments on this manuscript. Grants to EY and BF supported this work from the National Institutes of Health (DC016099, DC015252, DC015135, AG060504, and AG051443).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References