All experimental protocols and handling, use, and care of laboratory animals were conducted in compliance with the current normative standards of the European Union (Directive 2010/63/EU), the Spanish Government (Royal Decree 1201/2005 and 53/2013; Law 32/107) and had the approval of the University of Murcia Committee for Animal Experimental Ethics.

We used adult brains of a heterozygotic mice line developed at the Department of Biochemistry, School of Medicine, University of Murcia (López-Garcia et al., 2013). These mice express recombinant beta-galactosidase protein under control of the Azin2 promoter. After standard perfusion, dissection and embedding in agarose, vibratome 120 μm thick serial sections were obtained. Serial sections were obtained in either sagittal or horizontal planes. Floating sections were then reacted for beta-galactosidase and were finally washed, dehydrated, mounted on slides, and covered.

Digital microphotographs were acquired using Aperio CS2 technology (Leica Microsystems GmbH, Mannheim, Germany).

This database provides images from transgenic mice lines with EGFP as reporter for the expression of diverse genes. According to the recorded GENSAT procedures, the brain sections were processed for immunohistochemistry against EGFP, which normally leads to full labeling of the positive neurons including their soma and fibers. We searched for genes expressed in the trigeminal ganglion, whose axons reach the brain and form the ascending and descending trigeminal tracts, delimiting in this way the extent of the trigeminal sensory column. We selected two specimens, corresponding, respectively, to Calca and Avil reporter lines. We downloaded the selected images, and cropped them to show our region of interest.

We searched the AMBA and ADMBA for genes whose respective in situ hybridization (ISH) experiments included both sagittal and coronal section series at P56¹ as well as sagittal section series from stages P4, P14, and/or P28.² We screened these image series visually, selecting the genes with significant expression within subregions of Sp5. This analysis was initially carried out by two of the authors independently, whose preliminary results were discussed to reach a consensus. Our criteria to select genes was firstly that they displayed discrete expression patterns, with positive and negative regions visible along the longitudinal axis of Sp5; secondly, we checked that these patterns were coherently reproduced in both sagittal and coronal P56 stage series, as well as in the juvenile series (P4, P14 and/or P28). For some of the selected genes, part of the image series from these juvenile stages displayed generalized low ISH signal or high background, so that these image series were discarded from analysis.

As a result, we identified 12 genes (Table 1) that were differentially expressed in the trigeminal column, whose image series were analyzed, including the brightfield microphotographs or scanned images as well as their respective counterpart with color-coding of the expression level, both of them available for each brain section. In the brightfield images, the positive cells appear with a blue precipitate accordingly to standard ISH protocols. The sections corresponding to juvenile stages (P4, P14, and P28) are counterstained with HP Yellow. In the color-coded images from adult and juvenile stages, the expression intensity ranges from blue (low expression intensity), through green (medium intensity) to red (high intensity).

We downloaded the images that included the whole or part of the trigeminal column, cropping them to show the region of interest. In our figures, the images from parasagittal sections are oriented with the rostral end to the left, while those from coronal sections are details of the right side of the original image, with the midline to the left.

The list of the selected genes, together with the references of their respective experiments and downloaded images from the AMBA and ADMBA databases, is indicated in Supplementary Table 1.

In the mouse, the alar isthmocerebellar region forms rostrally the isthmus and caudodorsally the whole cerebellum, including the cerebellar cortex (CBX) and nuclei (CBN) (Sgaier et al., 2005) (Figures 1A, 2A,C, 3). The Pr5 is entirely formed by r2 and r3 in the mouse (Figures 1A,B, 2A,C, 3), with the ascending afferent trigeminal fibers stopping at the r1/r2 boundary (Oury et al., 2006). Therefore, in the mouse Pr5 abuts directly r1 structures, specifically the Kölliker-Fuse nucleus (KF) which is part of the parabrachial complex (PB) (Chatonnet et al., 2007); a neighboring landmark is the middle cerebellar peduncle (mcp) which courses through caudal r1 from ventral to dorsal (Franklin and Paxinos, 2008; Figures 1A,B, 2A, 3). Pr5 relates also rostrally with the ventral nucleus of the lateral lemniscus (VLL) within r1, separated from it by some fibers (Figure 2A).

The molecular mappings showed that the genes Baiap3 and Camk2a are expressed selectively in KF, while they show no significant expression in Pr5 (Figures 4A,B,D and data not shown). On the other hand, Irx2 is expressed by the Pr5 cells of r2, with no expression in KF (Figure 5A). These patterns thus delimit the r1/r2 boundary as the rostral end of the trigeminal column.

Another rostral component of the trigeminal system, the mesencephalic trigeminal nucleus can be observed across midbrain, isthmus and r1 in our AZIN2-lacZ sections (data not shown; see Martínez-de-la-Torre et al., 2018). We excluded this cell population from our analysis because it corresponds to primary sensory cells, therefore with a typology and category different from those of Pr5 and Sp5 (Hunter et al., 2001). An additional relation of the prepontine hindbrain region with the trigeminal system is represented by a small contribution to the motor trigeminal nucleus (5N) (Figures 1B, 3) as it can be deduced from the description in mid-gestational mouse embryos (Cordes, 2001).

This region is primarily characterized by the mass of the migrated pontine nuclei aggregated at the ventral surface of r3 and r4; they originate at the rhombic lip from rhombomeres r6-r8 (Di Meglio et al., 2013; Tomás-Roca et al., 2016; Figure 3).

Dorsally these rhombomeres participate in the formation of the cochlear nuclei, with r2 and r3 contributing principally to the ventral component (VC) and r4 plus r5 principally to the dorsal cochlear nucleus (DC) (Farago et al., 2006). In our material, we could not discern these differential contributions nor the precise limits between these subnuclei, since they homogenously express AZIN2-lacZ (Figures 1A,B). The cochlear nuclear complex abuts rostrally directly the cerebellum at the r1/r2 limit, so that it was used as an additional landmark for that boundary in our analyzed material (Figures 1A, 2A, 4B, 5C).

The 5N motor nucleus is formed principally across r2 and r3 (Schneider-Maunoury et al., 1997; Song et al., 2006; Figures 1B, 4B, 5C), similarly to Pr5 as commented above. Interposed between these two nuclei there lies the interfascicular trigeminal nucleus (IF5), formerly known as the tensor tympani part of the motor trigeminal nucleus (Fu et al., 2013; Figure 5C and data not shown).

The trigeminal nerve root (5n) is located at r2 in 10.0 dpc mice (Cordes, 2001), but lies displaced from the center of the neuromere, next to the r2/r3 boundary. We believe that due to its massive progressive growth it possibly invades the rostral part of r3 at later stages, with fibers encapsulating the r3 portion of Pr5 (Figures 1A, 2A,C). In this respect, it is known that the trigeminal ganglion continues its development and neurogenesis through gestational and postnatal stages up to even the adult stage (Lagares et al., 2007; Lanier et al., 2009) involving thus a progressive increase of afferent fibers of its entry root that may lead to the extension of the latter into r3.

After entering the hindbrain, many trigeminal primary afferents bifurcate into ascending and descending branches (Cajal, 1909) which are assumed to reach, respectively, Pr5 and Sp5 cells (Aström, 1953; Usunoff et al., 1997; Waite, 2004). Considering the aforementioned known rhombomeric location of Pr5 in the mouse, it can be deduced that in this species the so-called ascending branches follow their path into Pr5 within r2 and r3 (Oury et al., 2006; Rhinn et al., 2013; Figures 2B,D) while the descending branches would extend caudalwards through the rest of rhombomeres, forming the spinal trigeminal tract.

The fibers of the facial nerve (7n) extend radially through alar r4, down to their exit point located ventrolaterally within this rhombomeric domain (Di Bonito et al., 2017; Martínez-de-la-Torre et al., 2018; Figure 1B). In r4 there is also the entry root of the vestibulocochlear nerve (8n) as observed in early embryos (Cordes, 2001); in the adult brain this root appears in a dorsolateral position slightly separated from the 7n root (Franklin and Paxinos, 2008; Di Bonito et al., 2017; Martínez-de-la-Torre et al., 2018; Figures 3, 5C, 6E). According to the aforementioned mapping of Pr5 to r2 and r3, the rostralmost portion of Sp5 lies within r4 (Figures 1A,B, 2A,C, 3).

Due to the ventrally convex pontine flexure of the hindbrain, the r2, r3 and r4 rhombomeres appear skewed and wedge-shaped, narrowing from ventral to dorsal, as seen in sagittal and parasagittal sections (Figures 1A, 2A,C, 3). Their dorsalmost portions correspond to their respective parts of the alar cochlear nuclei and the choroidal roof (not shown), which are pushed backward by the large mass of the cerebellum. Their backward inclination causes that sections in the conventional coronal plane cut obliquely through these rhombomeres (e.g., Figures 4B, 5C,D, 6C–F).

As regards our molecular characterization, Irx2 is expressed in the r2 portion of Pr5, with some disperse cells within r3, as can be observed in adult and juvenile sections (Figures 5A–H and Table 1). Therefore, this gene marks a rhombomeric subdivision of Pr5. Additionally, in some sections Baiap3 appears expressed in scarce cells in r2 as compared to r3 (data not shown).

In the adult Kcng4 is expressed in the r2 and r3 portions of the trigeminal column, that is, the territory corresponding to Pr5 (Figures 6A–D,G,H), showing additionally some disperse positive cells within Sp5 at r4 (Figures 6A,B,E,F). This pattern appeared also in juvenile stages, although with an apparently regionalized pattern within Pr5, so that the ventral portion of r2 lacked expression of this gene (Figures 6G,H). Besides their different density, the cells marked within r4 showed a larger size as compared to those in r3, at least as observed in adult parasagittal and coronal sections (Figures 6A–F).

Similarly, Calb2 is significantly expressed in the r2 and r3 portions of Pr5, together with some positive cells in r4 (Figures 10J–L). Calb2, Irx2 and Kcng4 display also expression in more caudal hindbrain regions, as commented below.

This region is characterized, among other features, by the abducens motor nucleus (6N) and the facial nerve genus in r5, and the facial ascending fibers (asc7) plus the migrated facial motor nucleus (7N) in r6 (Figures 1B, 3; Cordes, 2001; Tomás-Roca et al., 2016; Martínez-de-la-Torre et al., 2018). 7N bulges somewhat from r6 into the adjoining r5 and r7 segments, deforming the interrhombomeric boundaries rostrally and caudally (Figure 3) as deduced from the fate mapping based on Hox gene expression (Tomás-Roca et al., 2016). The rostral limit of r5 lies just caudal to the fiber bundles of 7n and 8n in r4, as observed in sagittal and coronal sections of adult brains with transgenic lineage tracings (Di Bonito et al., 2017; Watson et al., 2017).

The parts of Sp5 corresponding to rhombomeres r5 and r6 strongly express Mafb (Figures 7A–C,F,G) and Fn1 (Figures 8A–D,G,H), with a higher intensity than the rest of rhombomeres. There appears moderate expression of both genes rostrally at r4 and caudally through r7 and r8, while the caudal rest of rhombomeric trigeminal portions contains sparse positive cells (Figures 7A–E,H,I, 8A–F,I,J). Fn1 displays a negative zone in the dorsalmost part of Sp5 extending from r4 to r8, probably corresponding to the dorsomedial subdivision of the trigeminal spinal nucleus (DMSp5) (Figures 8G,H).

The cryptic segments within this region (lacking visible boundary constrictions) were first discovered through fate mapping in avian chimeras (Cambronero and Puelles, 2000). They were subsequently corroborated by analysis of step-like changes in Hox expression patterns (Marín et al., 2008; Tomás-Roca et al., 2016). The medullary region is characterized by longitudinal plurisegmental structures such as the inferior olive (IO), found ventrally from r8 to r11, the complex formed by the hypoglossal nucleus (12N) close to the midline, plus the migrated vagal preganglionic motor nucleus (10N) from r9 to r11, and the ambiguous branchiomotor nucleus (Amb) extending caudalwards to the 7N in a deep intermediate position from r7 to r10 (Figure 3). The medulla comprises as well sizeable portions of the trigeminal, viscerosensory (solitary), and vestibular columns in the alar plate, and raphe nuclei at or next to the ventral midline (Alonso et al., 2013). The caudalmost rhombomere, r11, abuts the first myelomere or spinal cord segment (my1) occupied ventromedially by the pyramidal decussation (pyx). The rostrocaudal axis of the medulla oblongata follows more or less a straight horizontal direction. However, its caudal end, together with the upper spinal cord, forms the cervical flexure. Its pronounced curvature (together with possible variations in the angle of the sectioning plane) causes that coronal sections through r11 or even r10 cross obliquely the hindbrain/spinal cord (r11/my1) boundary, including thus the pyx in the same sections than the caudalmost rhombomeres.

Sp5 extends from r4 to r11 as commented above (Figures 1A–C, 2A,C). Therefore, the descending trigeminal primary afferents, from their entry point in r2 (or probably r2 plus the upper r3, as commented above) extend along the whole length of this nucleus as the spinal trigeminal tract, which continues into the spinal dorsal column and marginal stratum of the dorsal horn at least along the first 2 spinal segments. The dorsal horn of this upper portion of the spinal cord is accordingly described as part of Sp5C, considering also that their laminated cytoarchitecture is almost identical (Usunoff et al., 1997; Ten Donkelaar, 2020). The trigeminal afferents coming from the original branches of the trigeminal nerve are ordered dorsoventrally in the ascending and descending trigeminal tracts, so that dorsally there appear the mandibular fibers (sp5md) and ventrally the ophthalmic fibers (sp5ot) (Figures 1A,B, 2A,C) with the maxillary fibers in an intermediate position (data not shown) (Waite, 2004). These three bundles of fibers remarkably display different expression levels of the AZIN2-lacZ transgene (compare spt5md and sp5ot in Figure 1A) as previously described in coronal brain sections of these mice (Martínez-de-la-Torre et al., 2018).

Concerning the classical regionalization of Sp5, the boundary between the interpolar and the caudal subnuclei apparently lies at the r9/r10 transition, as was deduced from the Hox-based mapping of rhombomeres in mice (Tomás-Roca et al., 2016). This boundary is easily recognizable histologically attending either to their cytoarchitecture or to their pattern of afferent fiber terminals. A clearcut transition occurs from the non-laminar interpolar Sp5 to the laminar caudal Sp5 (Figures 2A,C; Olszewski, 1950; Waite, 2004). On the other hand, the boundary between the interpolar and oral subnuclei of Sp5 is uncertain on a morphological basis, so that in the literature it has been located tentatively at different positions (see section “Discussion”).

The dorsal horn of the first 2 myelomeres or cervical segments (C1 and C2), is functionally a continuation of the caudal subnucleus of Sp5, as commented above; it receives the terminals of the descending primary trigeminal afferents, and has a largely similar cyto- and myeloarchitectural aspect. In our molecular mappings we noted a continuous pattern from r10/r11 to my1 concerning the expressed genes and their laminar pattern, so that in our figures showing gene expression we have drawn the contour of Sp5 as including the dorsal horn of my1. The eventual continuation of this pattern into my2 could not be ascertained in the studied material, since seemingly the analyzed sagittal or coronal brain sections included only down to my1.

Among the genes we mapped, Baiap3 (Figures 4A–C) and Tac1 (Figures 9A–C) display specific expression in the region formed by r10, r11, and my1, corresponding to the caudal subnucleus of Sp5 as commented above. This labeled region limits rostrally with r9 that remains negative, corresponding to the caudal end of the interpolar Sp5. Both genes appear expressed in the superficial zone of the caudal Sp5 subnucleus, corresponding to the marginal and gelatinous layers (laminae I and II), as previously reported for substance P (product of the Tac1 gene) (Del Fiacco and Cuello, 1980; Ribeiro-da-Silva and Hökfelt, 2000). There appear also disperse Tac1 positive cells from r2 to r9 within the rest of the trigeminal column (Figures 9A–C and data not shown). Baiap3 and Tac1, as well as other genes commented below, display an uniform pattern along the rostrocaudal axis of the region formed by r10, r11 and my1, so that they cannot be used as markers for the delimitation of the r10/r11 and r11/my1 boundaries. Therefore, in the image series showing their expression we have traced these intersegmental boundaries tentatively, following the results of Hox gene expression (Tomás-Roca et al., 2016; see section “Discussion”).

Tac2 is expressed only in the gelatinous layer of the caudal subnucleus of Sp5 (Figures 9D–F). It displays a regionalized pattern within this subnucleus, with substantially fewer positive cells in r10 than in r11 or my1, as observed across juvenile and adult stages (Figures 9D,F; also compare Figure 9E with Figure 9B, respectively, showing Tac2 and Tac1 in r10). Therefore, from the analyzed genes Tac2 is the only one displaying a rostrocaudal regionalized pattern within the caudal Sp5, as compared with the homogeneous rostrocaudal expression of the aforementioned Baiap3 and Tac1, as well as Calb1, Calb2, and Camk2a as commented below.

Calb2 signal also appears within Sp5 (the r10 and r11 parts are positive, while r9 is negative) in the P14, P28, and adult brains. This marker is also expressed at these stages in discrete neuronal populations in r2 and r3, and in a less intense manner in disperse cells from r4 to r8 (Figures 10J–R and data not shown). In its turn, Calb1 is expressed from r9 to r11 across all stages (Figures 10A–I). Although Calb1 labels these three rhombomeres crossing the interpolar-caudal limit (r9/r10), the labeling in r9 is nuclear and homogeneous, whereas the expression pattern is laminar in r10 and r11, thus displaying the interrhombomeric limit (Figures 10A–C). Camk2a displays a similar pattern concerning r9, r10 and r11 (Figures 4D–G). On the other hand, a vertical palisade of Calb1-negative cells was distinguished along the aforementioned r9/r10 limit, enhancing its visibility and pointing to a specific cell population at this precise location (Figures 10A–C and Supplementary Figure 1).

The r9 portion of Sp5 appears labeled specifically by the expression of Pde1c (Figures 11A–D) and Zbtb16 (Figures 11E–N) genes, in which cases the rest of the trigeminal column remains negative. This rhombomeric portion is also differentially labeled by the expression of Camk2a and Calb1, which show a homogenous moderate expression in contrast with the scattered cells at r8 and the laminar pattern of r10 and r11 (Figures 4D–G, 10A–I). The r9 part of Sp5 is also characterized by the absence of expression of Calb2 as compared with adjacent rhombomeres (Figures 10J–R). Moreover, this portion of Sp5 expresses homogenously Kcng4 in contrast to the scattered positive cells seen in r8 and the absence of expression in r10 and r11 (Figures 6A,B,G,H).

In our screen, we did not identify any specific molecular markers for the rostral rhombomeres of the medulla oblongata, r7 and r8. However, the r8 portion of Sp5 can be distinguished by a pattern of scattered intensely labeled Calb1-positive cells, which differentiates this segmental module from the pattern in r7 (with much less density and intensity of the labeling) and r9 (with a more homogenous expression than r8) (Figures 10A–G). In its turn, Calb2 showed an heterogenous pattern in the rostral medulla (r7 and r8) that also extended into r4, r5 and r6, but did not allow to distinguish a rhombomeric periodic pattern after examination of the image series (Figures 10J–L and data not shown).

Considering the set of r7, r8 and r9 Sp5 domains (that is, the medullary part of Sp5, after excluding the caudal subnucleus), it is jointly characterized by Irx2 expression, present as a shared pattern of scattered positive cells, in contrast with the absence of expression in the rest of Sp5 (the retropontine region -r5 and r6- and the caudal subnucleus -r10 and r11) (Figures 5E–H).

The subdivision of Sp5 into oral, interpolar and caudal subnuclei was first described in primates on a cytoarchitectural basis, according to cell size and morphology, as well as the density of the neuronal populations (Olszewski, 1950). This subdivision was extrapolated to other mammal species as well as sauropsides (Molenaar, 1978; Arends and Zeigler, 1989) and established as the basis for anatomical, pharmacological or physiological studies of this structure. Differences between these Sp5 subnuclei concerning expression of molecular markers, patterns of incoming trigeminal primary afferents, or connectivity with other brain regions have been reported (Waite, 2004).

Some of the boundaries of the classic Sp5 subdivisions can be identified with acceptable precision, and correlated with the rhombomeric map. The oral subnucleus abuts rostrally Pr5, which lies within r2 and r3 (Oury et al., 2006). The limit between Pr5 and the oral Sp5 subnucleus accordingly coincides with the r3/r4 limit. The caudal Sp5 subnucleus is continuous with the dorsal horn of the spinal cord, mapping thus down to the r11/my1 limit and extending into the upper two spinal cord segments (Usunoff et al., 1997; Ten Donkelaar, 2020). Other authors use a nomenclature that differentiates, as two components of the sensory trigeminal column, the caudal subnucleus (within the medulla oblongata) and the upper cervical dorsal horn, continuous at the hindbrain/spinal cord junction (Jacquin et al., 1986; Noma et al., 2008; Faunes and Wild, 2017). In its turn, the abrupt structural transition between the non-laminar interpolar Sp5 portion and the laminar caudal subnucleus allows the limit between them to be defined, which corresponds to the r9/r10 limit as identified by the Hox gene pattern (Tomás-Roca et al., 2016).

However, a problem arises when we try to differentiate between the oral and interpolar subnuclei of Sp5. They display subtle cytoarchitectural differences, but these apparently do not render a clearcut boundary, so that the precise morphological location of the oral/interpolar limit varies across the literature (Phelan and Falls, 1989). Besides, there are also differences between these two subnuclei concerning molecular markers or connectivity (Erzurumlu and Killackey, 1979; Veinante et al., 2000; Waite, 2004; Ashwell et al., 2006). None of these studies showed a distinct boundary between them. In rodents it is conventionally assumed that the oral/interpolar limit lies at the level of the caudal pole of the facial motor nucleus (Phelan and Falls, 1989; Waite, 2004) so that following the rhombomeric map it would correspond to the r6/r7 limit. Accordingly, the oral and interpolar Sp5 subnuclei would consist, respectively, of the r4-r6 and r7-r9 portions of the plurineuromeric trigeminal column.

Nevertheless, the classic subdivision into oral, interpolar and caudal parts turned out to be insufficient to interpret the rostrocaudal organization of Sp5 in several studies. In his own seminal work proposing this subdivision, the oral subnucleus is described as containing a rostral subdivision with different cytoarchitecture (Olszewski, 1950). A detailed cyto- and myeloarchitectural study of the interpolar subnucleus shows significant differences between successive transversal sections of this structure (Phelan and Falls, 1989), so that these rostrocaudal differences may be explained by an internal segmental organization within this subnucleus. In birds the analysis of the pattern of trigeminocerebellar connections needed the further subdivision of the oral and interpolar subnuclei into two rostrocaudal subunits each (Arends and Zeigler, 1989). In the cat, the whole trigeminal column (Pr5 plus Sp5) was subdivided into 12 rostrocaudal units according to the peripheral origin of their primary trigeminal afferents (Marfurt, 1981). Interestingly the caudal subnucleus -including the upper cervical dorsal horn- can be subdivided into four rostrocaudal regions innervating, respectively, oral/perioral, snout, periocular, and periauricular receptive fields (Panneton et al., 2017). Studies about the central processing of pain have introduced an additional trigeminal rostrocaudal subdivision, identified as the interpolar/caudal transition zone, which specifically receives nociceptive afferents (Ren and Dubner, 2011).

Therefore, considering these multiple rostrocaudal subdivisions of Sp5, plus the modern concept of a neuromeric hindbrain (Tomás-Roca et al., 2016; Martínez-de-la-Torre et al., 2018), a rhombomere-based framework needs to be developed that should facilitate the finding and characterization of trigeminal functional units, which apparently remain partly hidden if only the tripartite classic schema is assumed. An advantage of the rhombomeric model is that it would be underpinned by modern embryological and molecular criteria. Such a model would be also a basis for interspecies comparison, taking into account the conservation of the rhombomeric (neuromeric) pattern across vertebrates.

From the different proneuromeric regions of the hindbrain (prepontine, pontine and retropontine regions, and medulla oblongata; Puelles et al., 2013, 2018) the latter appears as the more extense, including as well more rhombomeres. The medulla oblongata has the peculiarity of lacking the typical interrhombomeric limits of the more rostral hindbrain. Such limits are visible as constrictions at early stages, and consist of specific neuroepithelial cell populations that act as barriers to clonal cell dispersion, and as a source of morphogens for adjacent rhombomeres, among other cellular and molecular characteristics (Pujades, 2020). Albeit the absence of these overt limits in the medulla oblongata, five rhombomeres (r7-r11) were proposed in this region, separated by cryptic limits antimeric to neighboring intersomitic boundaries, through experiments with quail-chick chimeras (Cambronero and Puelles, 2000). Ulterior analysis showed that r7-r11 are differentially characterized by the nested expression of Hox genes of the paralog groups 4–8, very much similarly as Hox genes of the paralog groups 1–3 delimit the overt rhombomeres r1-r6 (Marín et al., 2008; Tomás-Roca et al., 2016). Initially these 5 medullary rhombomeres were identified as “pseudorhombomeres” (Cambronero and Puelles, 2000), but later the descriptor “cryptorhombomeres” was thought to be more appropriate, since these units are not false neuromeres, but only hidden ones, only separated by given molecular properties.

Concerning the trigeminal column, the medulla oblongata contains the most conspicuous internal limit of this structure, which separates the interpolar and caudal subnuclei at the r9/r10 limit (Tomás-Roca et al., 2016; Martínez-de-la-Torre et al., 2018; Puelles et al., 2018). Morphologically this limit is visible as an interface between the homogenous nuclear cytoarchitecture of the interpolar subnucleus and the laminar pattern of the caudal subnucleus (Olszewski, 1950). This change correlates with the expression pattern of diverse genes (present data) including Hoxa6 and -b6 whose domains end rostrally at this limit, as we observed at perinatal stages (Tomás-Roca et al., 2016). Physiologically, this limit is highly relevant considering the involvement of the caudal subnucleus in the central processing of craniofacial pain (Bereiter et al., 2000).

The caudal subnucleus (Sp5C) shows apparently an uniform cyto- and genoarchitecture through its r10 and r11 portions, which also extend with little variation into my1, according to previous literature (Olszewski, 1950; Bereiter et al., 2000; Waite, 2004) now corroborated by our data on gene expression. Nevertheless, we observed that the expression of Tac2 shows a gradient-like pattern with decreasing positive cells from my1 into r10. At perinatal stages the r11 part of the caudal subnucleus is positive for Hoxb7 while the r10 part is negative. The domain with highest expression of Hoxb8 reaches up to the r11/my1 limit, with additional weaker expression within r11 (Tomás-Roca et al., 2016). Taken jointly, these data suggest the existence of subtle molecular differences starting with differential Hox gene properties within the medullary and spinal portions of Sp5C, which may lead to functional properties that would need to be explored.

Within the medulla oblongata, the rhombomere that was singled out more directly by our approach is r9, considering its specific expression of Ped1c and Zbtb16, as well as its distinctive pattern in relation to Calb1, Calb2, Camk2a, and Kcng4 expression, as commented in Results. This rhombomere represents the rostralmost expression domain of paralog Hox5 genes (Tomás-Roca et al., 2016). It may correspond to the part of the interpolar subnucleus that is reactive to orofacial nociceptive stimuli defined as the so-called interpolar/caudal transition zone (Ren and Dubner, 2011).

It would remain to be assessed the possible individuality at molecular level of the r7 and r8 portions of Sp5, which are quite similar structurally and molecularly, as commented in Results. Another case of a pair of rhombomeric domains with relative molecular similarity is that of r5 and r6, which share specific Mafb and Fn1 expression. However, the r6 portion of Sp5 has a specific trigeminocerebellar projection in the chick (Díaz and Puelles, 2003) so that further research could yield molecular and functional differences between r5 and r6, as well as between r7 and r8. On the other hand, at perinatal stages these pairs of rhombomeres are internally differentiated by Hox expression, with Hoxd3 and Hox4 paralogs expressed up to the r5/r6 and r7/r8 limits, respectively (Tomás-Roca et al., 2016).

Therefore the medullary portion (r7-r11) of the trigeminal column consists of successive rostrocaudal units that largely fit the rhombomeric map, an organization also observed in the vagal motor nucleus (Cambronero and Puelles, 2000; Marín et al., 2008), the inferior olive (Hidalgo-Sánchez et al., 2012), the reticular formation (Gray, 2013) or the raphe nuclei (Alonso et al., 2013) all of them longitudinal structures that span several rhombomeres of the medulla oblongata. Its apparent early morphologic homogeneity or uniformity contrasts with the overt segmentation of the rostral half of the hindbrain (Parker and Krumlauf, 2020; Pujades, 2020; Krumlauf and Wilkinson, 2021). However, according to current evidence the medulla oblongata emerges as a complex region formed by multiple neuromeric units with differential molecular profiles (Nieuwenhuys and Puelles, 2016; Puelles et al., 2018; Watson et al., 2019).

Some of the molecular markers we report (Calb1, Calb2, Camk2a, Tac1, Tac2) have been previously described in relation to the regionalization of the trigeminal column, while others (Irx2, Mafb, Fn1, Pde1c, Zbtb16, Kcng4, and Baiap3) are novel in this respect.

In the rat, Calb1 and Calb2, which code, respectively, for the calcium binding proteins calbindin and calretinin, are expressed in the caudal subnucleus, principally in its gelatinous layer or lamina II. Both are also expressed in disperse cells in the rest of the trigeminal column (that is, the Pr5, and the oral and interpolar Sp5 subnuclei) although the description of this pattern varies across the literature, probably due to different sensitivity or specificity of antibodies (Arai et al., 1991; Bennett-Clarke et al., 1992; Rogers and Résibois, 1992; Ashwell et al., 2006). A significant novelty of our study is the identification of a segment-like portion of the interpolar subnucleus (r9) that is positive for Calb1 and negative for Calb2. This had not been noticed previously possibly because the scarce use of parasagittal sections in classical neuromorphological analyses.

Another calcium-related gene, Camk2a (calcium/calmodulin-dependent kinase II α) is functionally involved in nociceptive pathways, as deduced from the study of mutant mice (Zeitz et al., 2004). It was described as expressed in neurons of the marginal and gelatinous layers of the spinal dorsal horn, as well as in unmyelinated neurons of the trigeminal ganglion. We show now that its expression extends to the superficial layers of the caudal Sp5 subnucleus in r10 and r11, with a pattern similar to that described in the spinal cord, in addition to the expression in the non-laminar segment-like portion of the interpolar Sp5 in r9. This latter expression would support the correspondence of the r9 portion with the “caudal/interpolar transition zone,” which is involved in pain processing, as commented above.

Tac1 and Tac2 encode, respectively, the neurotransmitters substance P and neurokinin B. Both genes, or their protein products, have been described in the superficial laminae of the dorsal horn and caudal trigeminal subnucleus (Del Fiacco and Cuello, 1980; Ribeiro-da-Silva and Hökfelt, 2000; Mar et al., 2012). We have found that their expression patterns display segment-related differences, with Tac1 reaching up to the r10 trigeminal portion in a homogenous pattern, and Tac2 showing less positive cells in r10, as reported in Results.

Irx2, a member of the Iroquois family of transcription factors, has expression along the hindbrain in 10 dpc mouse embryos (Bosse et al., 1997; Cohen et al., 2000) so that the pattern we describe in the trigeminal column would be derived from that early expression. Interestingly Irx2 differentiates the r2 from the r3 portion of Pr5, as well as a medullary portion of Sp5 (r7-r9) that would correspond to the interpolar subnucleus as described in the classical nomenclature. Given the known role of the Iroquois gene family in pattern formation and differentiation (Gómez-Skarmeta and Modolell, 2002) it would be expected the involvement of Irx2 in trigeminal development. It would exist also the possibility of a role in the differentiated mature neurons where it is expressed.

Mafb is selectively expressed in r5 and r6 of the early developing hindbrain as well as in several brainstem structures such as a part of the cochlear nuclei (Cordes and Barsh, 1994; Eichmann et al., 1997). Mafb is necessary for the normal development of these rhombomeres through the regulation of the developmental genes Krox20 and Fgf3, and several Hox genes (Frohman et al., 1993; Manzanares et al., 1999, 2002; Giudicelli et al., 2003), all of which are involved in hindbrain segmentation. Our results show that Mafb maintains its expression in the trigeminal derivatives within r5 and r6 in the adult brain, suggesting additional roles in these neuronal populations.

Fn1 encodes the extracellular matrix component fibronectin. Its mRNA expression appears in telencephalic migrating neurons while in the adult brain it is restricted to the subiculum (Sheppard et al., 1995; Kashima et al., 2019). Here, we describe that it is additionally expressed at least within given segmental portions of the trigeminal column. Its possible function in these neuronal populations would need to be tested.

Pde1c encodes the calmodulin-dependent cyclic nucleotide phosphodiesterase and is expressed in olfactory sensory neurons (Yan et al., 1995), in migrating neurons in the cerebellum and cerebral cortex (Gong et al., 2003), and in lamina I neurons of the dorsal horn involved in nociception, principally at lumbar levels (Torsney et al., 2006). Interestingly, our results show a specific expression of this gene in the r9 part of Sp5, delimiting the caudal/interpolar transition zone, involved in pain signaling.

Zbtb16, also known as promyelocyte leukemia zinc finger (PLZF), was identified in an unique case of acute promyelocytic leukemia (APL), where it is fused to the retinoic acid receptor cx (RARor) (Chen et al., 1993). Zbtb16 is characterized by a dynamic pattern of expression across multiple regions of the neural tube (Avantaggiato et al., 1995; Cook et al., 1995). PLZF is a transcriptional regulator of Hox genes during hindbrain development (Ivins et al., 2003). We have described a novel feature of the expression pattern of this gene, namely its specific location in a single rhombomeric domain (r9), similarly to Pde1c.

Kcng4 encodes a subunit of a potassium voltage-gated channel that has been identified as potentially linked to migraine (Lafrenière and Rouleau, 2012). Although this brain disorder has been extensively linked to the trigeminal system, and mainly to the caudal Sp5 subnucleus (Ashina et al., 2019; Messlinger and Russo, 2019), the pathophysiology of chronic migraine is not fully understood. Our work shows that the migraine-related gene Kcng4 is actually expressed from r2 to r9, which correspond to the principal, oral and interpolar subnuclei, which also have recently been proposed to play a role in migraine (Youn, 2018).

Baiap3 is a member of the mammalian uncoordinated 13 (Munc13) protein family of synaptic regulators of neurotransmitter exocytosis (Shiratsuchi et al., 1998; Koch et al., 2000). In the human brain, its mRNA is expressed in several regions including the cortex, amygdala, hypothalamus and periaqueductal gray (Shiratsuchi et al., 1998; Lauridsen et al., 2011; Wojcik et al., 2013). Knockout mice and given polymorphisms for this gene lead to or are correlated with anxiety behavior and substance abuse (Wojcik et al., 2013). We have described a strong and specific expression in the caudal Sp5 trigeminal nucleus, so that it might have a role in the nociceptive function of this region.