Abstract
The three members of the Brn3 family of POU-domain transcription factors (Brn3a/Pou4f1, Brn3b/Pou4f2, and Brn3c/Pou4f3) are expressed in overlapping subsets of visual, auditory/vestibular, and somatosensory neurons. Using unmarked Brn3-null alleles and Brn3 conditional alleles in which gene loss is coupled to expression of an alkaline phosphatase reporter, together with sparse Cre-mediated recombination, we describe the following: (1) the overlapping patterns of Brn3 gene expression in somatosensory neurons; (2) the manner in which these patterns correlate with molecular markers, peripheral afferent arbor morphologies, and dorsal horn projections; and (3) the consequences for these neurons of deleting individual Brn3 genes in the mouse. We observe broad expression of Brn3a among DRG neurons, but subtype-restricted expression of Brn3b and Brn3c. We also observe a nearly complete loss of hair follicle-associated sensory endings among Brn3a−/− neurons. Together with earlier analyses of Brn3 gene expression patterns in the retina and inner ear, these experiments suggest a deep functional similarity among primary somatosensory neurons, spiral and vestibular ganglion neurons, and retinal ganglion cells. This work also demonstrates the utility of sparse genetically directed labeling for visualizing individual somatosensory afferent arbors and for defining cell-autonomous mutant phenotypes.
Introduction
Primary somatosensory neurons convey information from the skin and body interior about temperature, mechanical stimulation, tissue damage, and joint and muscle position (Lumpkin and Caterina, 2007; Basbaum et al., 2009; Proske and Gandevia, 2009). The diversity of somatosensation is mirrored in the diversity of peripheral nerve endings, which vary in complexity from free nociceptive endings in the skin to intricate structures such as Pacinian corpuscles and muscle spindles. Diversity is also seen in nerve conduction velocity and its anatomic counterparts, axon diameter and myelination. In the dorsal horn of the spinal cord, the central projections of the primary somatosensory neurons are organized in laminae, with distinct subtypes of somatosensory fibers projecting to only one or a few laminae (Sanderson Nydahl et al., 2004; Zylka et al., 2005; Bourane et al., 2009; Luo et al., 2009).
A growing collection of molecular markers and gene knock-out mouse lines has facilitated the identification and characterization of different subtypes of somatosensory neurons. The markers include TRPV (transient receptor potential vanilloid) channels, Mas-related G-protein-coupled receptors (Mrgprs), cytosolic proteins such as parvalbumin and neurofilament-200 (NF-200), and neuropeptides such as calcitonin gene-related peptide (CGRP) (Fundin et al., 1997a; Tominaga et al., 1998; Zylka et al., 2005; Liu et al., 2009; Luo et al., 2009). Of special interest are genes and proteins that control the development and diversification of somatosensory neurons—in particular, the neurotrophic factor receptors RET, Trk, and GFR-α (glial cell line-derived neurotrophic factor-α), and the transcription factors Sox10, Neurogenin-2, Islet-1, MafA, and the members of the Runx and Brn3 families (Eng et al., 2007; Marmigère and Ernfors, 2007; Inoue et al., 2008; Sun et al., 2008; Bourane et al., 2009; Lanier et al., 2009; Dykes et al., 2010). Sox10 controls the proliferation of neural progenitors in the DRG and TG, and Islet-1 and Brn3a control, directly or indirectly, the expression of a large number of genes in DRG and TG neurons, including Runx genes. The TrkA and TrkC neurotrophin receptors, which are controlled in part by Runx1 and Runx3, play critical roles in the differentiation and survival of nociceptors, mechanoreceptors, and proprioreceptors.
The present work focuses on the three members of the Brn3 family of transcription factors: Brn3a/Pou4f1, Brn3b/Pou4f2, and Brn3c/Pou4f3. Each Brn3 gene is expressed in distinct sets of neurons in each of three sensory organs—the retina, inner ear, and DRG/TG. We have recently developed Brn3a, Brn3b, and Brn3c conditional knock-out/alkaline phosphatase (AP) knock-in mouse lines that permit a genetic and morphologic analysis of individual neurons (Badea et al., 2009a; present work). We report here the use of these lines to define—for both WT and Brn3 mutant mice—the overlapping patterns of Brn3 gene expression among DRG neurons, the morphologies of their afferent arbors, and their central target fields in the dorsal horn of the spinal cord. The results imply that the Brn3 proteins contribute to sensory neuron diversity by participating in a combinatorial code of transcriptional regulation and that there are deep functional similarities in transcriptional circuits across diverse sensory systems.
Materials and Methods
Mouse lines.
The following lines were previously described: (1) Cre lines: Sox2Cre (Hayashi et al., 2002), Pax6αCre (Marquardt et al., 2001), ROSA26CreER (Badea et al., 2003), NFL-CreER (Rotolo et al., 2008), and R26rtTACreER (Badea et al., 2009b); (2) conventional knock-out lines: Brn3a (Xiang et al., 1996) Brn3b (Gan et al., 1996) and Brn3c (Xiang et al., 1997); and (3) conditional knock-in alleles: Brn3aCKOAP and Brn3bCKOAP (Badea et al., 2009a). The Brn3cCKOAP conditional allele was generated by homologous recombination in mouse embryonic stem cells using standard techniques. For the targeted allele, the following changes were made: a loxP site was inserted in the 5′ UTR 50 bp 5′ before the initiator ATG; three repeats of the SV40 early region transcription terminator were added to the 3′ UTR 600 bp 3′ of the Brn3c translation termination codon, followed by a second loxP site and the coding region of human placental AP. A positive selection cassette Phosphoglycerate-Kinase-Neomycin Resistance gene (PGK-Neo), flanked by frt sites, followed the AP coding region, and was subsequently removed by crossing to mice expressing Flp recombinase in the germline, as previously described (Badea et al., 2009a).
Sparse recombination.
For methods related to sparse Cre-mediated recombination, see Badea et al. (2003), Rotolo et al. (2008), Badea et al. (2009b), and Badea and Nathans (2011). For each of the three Brn3 genes, timed matings between Brn3+/−;R26rtTACreER/+ males and Brn3CKOAP/CKOAP females were set, conception date was determined by examining the copulation plug, and pregnant females were moved to cages with food pellets containing doxycyline (1.75 mg/g) at gestational day 3. At gestational day 9, 200 μg of 4-hydroxytamoxifen (4HT) in sunflower seed oil vehicle was delivered by intraperitoneal injection, and the doxycycline diet was continued until gestational day 11. P1–P4 pups were used for skin AP histochemistry (see Figs. 5, 6, 7), whereas adults were used for spinal cord AP histochemistry (see Fig. 7). For visualizing somatosensory afferents in Brn3CKOAP/+;NFL-CreER mice (see Fig. 5), females were injected intraperitoneally with 0–200 μg of 4HT at gestational day 14–17, and mice were analyzed at P1–P3. DRG immunostaining of P5 Brn3aCKOAP/+;R26CreER DRGs was performed using mice that were not exposed to 4HT, taking advantage of the background rate of Cre-mediated recombination in the absence of 4HT.
Histology.
Spinal cord and brain vibratome sections (typically 200 μm thickness) were fixed, AP stained, processed, and imaged as previously described (Badea et al., 2003, 2009a). For vibratome sections of early postnatal pups, eviscerated torsos were fixed overnight in PBS/4% paraformaldehyde at 4°C, and then decalcified for ∼4 d in 50 mm EDTA at 4°C. Skin flatmounts were prepared by pinning dissected skin (external surface downward) to a Sylguard surface with insect pins and fixing overnight in PBS/4% paraformaldehyde at 4°C. Retina sections were processed and immunostained as previously described (Badea et al., 2009a). For spinal cord and DRG immunostaining, adult or early postnatal mice were perfused intracardially with PBS/4% paraformaldehyde, and the vertebral column was dissected, decalcified, cryoprotected in Optimal Cutting Temperature Compound (Tissue-Tek-Sakura), and sectioned at 14 μm thickness on a cryostat. Complete spinal cords with attached DRGs from two to four mice of the same genotype were cut transversely into 4 segments of equal length, and these 8–16 segments were embedded together in a single block. For each immunostaining analysis cells were counted from six to eight sections at cervical, thoracic, and lumbar levels. No significant differences in patterns of dorsal horn lamination or the frequency of Brn3a-, Brn3b-, or Brn3c-expressing DRG neurons were noted across spinal cord levels with the markers used in this study. High-resolution images were captured on a Zeiss Imager.Z1 fitted with an Apotome for fluorescent imaging and Axiovision software. Skin afferent neurons were imaged with a black-and-white Axiocam camera using differential interference contrast/Nomarsky optics, and neuronal arbors were reconstructed using Neuromantic neuronal tracing freeware (Darren Myat, http://www.reading.ac.uk/neuromantic) and were exported to the Rotator visualization software using scripts written in Matlab (Mathworks).
Antibodies.
Rabbit polyclonal anti-Brn3a, anti-Brn3b, and anti-Brn3c antisera are described in Xiang et al. (1995). For double immunostaining with rabbit antibodies to Brn3 proteins and cytoplasmic markers, the anti-Brn3 immunostaining was performed first, the patterns of nuclear immunolabeling were captured, and then the anti-cytoplasmic marker immunostaining was performed and the labeling pattern compared with the earlier image. The sources of commercial antibodies are as follows: sheep anti-AP (American Research Products); rabbit anti-Neurofilament 200, anti-Peripherin, and anti-TrkA, and mouse monoclonal anti-Brn3a (MAB1585; Millipore); rabbit anti-parvalbumin (Swant); guinea pig anti-CGRP (Bachem); rabbit anti-PKCγ (Santa Cruz Biotechnology). Secondary antibodies were donkey antisera coupled with Alexa dyes (Invitrogen/Life Technologies). Isolectin B 4 (IB4) conjugates were from Invitrogen/Life Technologies.
Results
Central projections of Brn3c-expressing RGCs visualized by expression of alkaline phosphatase from the Brn3c locus
To eliminate Brn3c function via Cre-mediated recombination and to simultaneously visualize individual neurons expressing the recombined Brn3c allele, one loxP site was inserted in the Brn3c 5′ UTR, a second loxP site was inserted 3′ of the Brn3c transcription termination signal, and an AP reporter coding region was inserted distal to the 3′ loxP site (Fig. 1A). Expression of the AP reporter is activated by Cre-mediated deletion of the Brn3c coding region and 3′ UTR, bringing AP under the control of the Brn3c promoter. The same “conditional knockout with AP” (CKOAP) strategy was previously applied to the Brn3a and Brn3b genes (Badea et al., 2009a). When the conditional allele is placed over a WT allele (Brn3cCKOAP/+), Cre-mediated recombination generates phenotypically normal Brn3cAP/+ cells. When the conditional allele is placed over a conventional null allele (Brn3cCKOAP/−) (Xiang et al., 1997), Cre-mediated recombination generates phenotypically mutant Brn3cAP/− cells. In both cases, the AP reporter permits histochemical and immunocytochemical visualization of cell bodies and arbor morphologies. Figure 1B shows the colocalization of nuclear Brn3c and plasma membrane-anchored AP in Brn3cAP/+ RGCs (Fig. 1B, top), and the loss of Brn3c protein with the retention of AP expression in Brn3cAP/− RGCs (Fig. 1B, bottom).
Conditional Brn3c allele with an AP reporter reveals central targets of Brn3c-expressing RGCs. A, Gene targeting strategy. I, WT Brn3c gene; II, targeted Brn3c gene with a PGK-Neo cassette flanked by frt sites; III, Brn3cCKOAP allele with the Neo cassette excised by germline Flp recombination; and IV, Brn3cAP allele following Cre-mediated deletion of the Brn3c coding region. Filled black rectangles, 5′ UTR and coding region; open rectangle, 3′ UTR; red arrow labeled “pA,” additional polyadenylation sites added to the 3′ UTR; ATG, initiator methionine codon; AP, AP coding region and 3′ UTR. The line above the gene indicates the structure of the spliced Brn3c transcript. B, BamHI. The bar labeled “3′ probe” indicates the location of the Southern blot hybridization probe. Arrows flanking the 5′ loxP site show the locations of PCR primers. Right, Genotyping by Southern blot (top) and PCR (bottom). B, Cre-mediated deletion of the Brn3c coding region in Brn3cCKOAP/+ or Brn3cCKOAP/− retinas by Pax6αCre activates AP expression in a subset of RGCs and simultaneously eliminates Brn3c protein from RGCs in Brn3cCKOAP/− retinas as determined by immunostaining of adult retinas with anti-AP (green; plasma membrane) and anti-Brn3c (red; nuclei) antibodies. White arrowheads point to AP+ RGC somas; green fluorescent anti-mouse secondary antibodies also decorate intraretinal capillaries. INL, Inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. C, Central projections of Brn3a- versus Brn3c-expressing RGCs in coronal sections of adult brain histochemically stained for AP (purple). Retina-specific Cre-mediated recombination is conferred by Pax6αCre. Matched pairs of serial sections (e.g., a and f, b and g) proceed from anterior at the mid-LGN (a, f; approximately bregma −2.5) to posterior at the anterior colliculus (e, j; approximately bregma −3.5). Brn3aAP/+ RGC axons (a–e). Red arrows, medial terminal tract and nucleus; green arrow, lateral aspect of the accessory optic tract. Brn3cAP/+ RGC axons (f–l). k, The optic chiasm and optic tracts (bottom) are populated by Brn3c-expressing RGC axons, but the SCN (purple arrow) is not targeted by axons of Brn3c-expressing RGCs. l, The boxed LGN region in g is enlarged; the intergeniculate leaflet is seen as a central unstained region. Scale bars: B, 40 μm; Ca–j, 500 μm; Ck, 500 μm; Cl, 200 μm.
In earlier work, we used Brn3aCKOAP and Brn3bCKOAP alleles to determine the central projections of RGCs expressing these two genes. A similar analysis with Brn3cCKOAP, using the retina-specific Pax6αCre transgene (Marquardt et al., 2001), shows projections to the lateral geniculate nucleus and superior colliculus (Fig. 1Cf–j,l). In contrast to Brn3a- and Brn3b-expressing RGCs, Brn3c-expressing RGCs do not project to the medial terminal nucleus nor do they contribute to the lateral terminal tract, both of which are involved in eye movement control as part of the accessory optic tract (Fig. 1C, compare a–e, f–j). Also in contrast to Brn3b-expressing RGCs, Brn3c-expressing RGCs bypass the suprachiasmatic nucleus (SCN) (Fig. 1Ck). An analysis of the dendritic morphologies of both Brn3cAP/+ and Brn3cAP/− RGCs is reported in Badea and Nathans (2011); loss of Brn3c appears to have little or no effect on the survival or morphology of these RGCs (see also Wang et al., 2002).
The effect of mutations in each Brn3 gene on the expression of other Brn3 family members in RGCs
A high degree of functional similarity among the members of the Brn3 family is suggested by the following: (1) the near identity of the DNA binding domains of Brn3a, Brn3b, and Brn3c; (2) the ability of the Brn3a coding region to substitute for the Brn3b coding region to permit RGC survival (Pan et al., 2005); and (3) the partial redundancy of Brn3b and Brn3c in mediating RGC survival and intraretinal axon guidance (Wang et al., 2002). The extensive overlap in Brn3a, Brn3b, and Brn3c expression in RGCs (Xiang et al., 1995; Badea et al., 2009a) and the evidence that Brn3a negatively regulates its own transcription in DRG neurons (Trieu et al., 2003) led us to ask whether there might be cross-regulation among Brn3 family members in those neurons in which two or more Brn3 genes are expressed. As an initial test of this idea, we have immunolocalized each Brn3 protein together with AP in RGCs of genotypes Brn3aAP/+, Brn3aAP/−, Brn3bAP/+, Brn3bAP/−, Brn3cAP/+, and Brn3cAP/−. For this analysis, the Pax6αCre transgene was used to generate retina-specific Cre-mediated recombination at ∼E9.5 (Marquardt et al., 2001). For the Brn3-null mutant cells (i.e., Brn3AP/−), the AP reporter identifies those cells that were programmed to express the particular Brn3 gene but from which that gene's coding region has been deleted. Data for Brn3a and Brn3b-null RGCs are described by Badea et al. (2009a); Figure 2, A and B, presents the complete dataset for all three Brn3 genes.
Expression of Brn3a, Brn3b, and Brn3c in Brn3aAP/+, Brn3aAP/−, Brn3bAP/+, Brn3bAP/−, Brn3cAP/+, and Brn3cAP/− RGCs and DRG neurons determined by anti-AP and anti-Brn3 double immunostaining. A, Adult retina sections stained for cell surface AP (green) and nuclear Brn3a, Brn3b, or Brn3c (red). Note that green fluorescent anti-mouse secondary antibodies also decorate intraretinal capillaries. INL, Inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Scale bar, 40 μm. B, Quantification of the patterns of Brn3 immunostaining (labels at left) in adult WT and KO RGCs of the indicated genotypes (labels at top). Pie charts quantify the fraction of AP+ cells that were positive (blue) or negative (red) for the indicated transcription factor (TF). The number of RGCs that were counted in each category is indicated. Data for Brn3a and Brn3b expression in Brn3aAP/+ and Brn3aAP/− RGCs were described by Badea et al. (2009a) and are included here for completeness. Definitions of red and blue pie chart colors apply to B, D, and F. C, P5 DRGs following mosaic Cre-mediated recombination in Brn3aCKOAP/+;R26CreER, Brn3bCKOAP/+;R26CreER or Brn3cCKOAP/+;R26CreER mice stained for cell surface AP (red) and nuclear Brn3a, Brn3b, or Brn3c (green). Scale bar, 40 μm. D, Quantification of data from C; color code as in B. E, P0–P1 DRGs following mosaic Cre-mediated recombination in Brn3aCKOAP/+;R26rtTACreER, Brn3aCKOAP/−;R26rtTACreER, Brn3bCKOAP/+;R26rtTACreER, Brn3bCKOAP/−;R26rtTACreER, Brn3cCKOAP/+;R26rtTACreER, and Brn3cCKOAP/−;R26rtTACreER mice stained for cell surface AP (green) and nuclear Brn3a, Brn3b, or Brn3c (red). Scale bar, 40 μm. F, Quantification of data from E; color code as in B.
As expected, all AP+ RGCs from Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ retinas also exhibit nuclear immunoreactivity for the corresponding Brn3 family member, and all AP+ RGCs from Brn3aAP/−, Brn3bAP/−, and Brn3cAP/− retinas lack nuclear immunoreactivity for the corresponding Brn3 family member (Fig. 2A,B). With respect to cross-regulation, loss of Brn3b has little effect on expression of Brn3a or Brn3c, and loss of Brn3c has little effect on expression of Brn3a or Brn3b. However, among Brn3aAP/− RGCs there is a statistically significant increase in the fraction of cells expressing Brn3c (p = 0.0013) and a modest decrease that does not rise to statistical significance in the fraction of cells expressing Brn3b (p = 0.20). These observations suggest that Brn3a may normally suppress Brn3c expression in a subset of RGCs.
Expression patterns of Brn3 family members in DRG neurons
We also determined the patterns of coexpression of Brn3 family members in DRG neurons. Figure 2, C and D, shows the patterns of colocalization in Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ DRG neurons at P5 from Brn3aCKOAP/+;R26CreER, Brn3bCKOAP/+;R26CreER, or Brn3cCKOAP/+;R26CreER mice that had undergone sparse Cre-mediated recombination. We note that at P5 as many as 10% of AP+ DRG neurons are not detectably immunostained for the transcription factor that corresponds to the AP knock-in allele (e.g., 10% of Brn3bAP/+ neurons are not detectably stained with anti-Brn3b antibodies) (Fig. 2D). We ascribe this apparent discrepancy to the higher sensitivity of AP compared with Brn3 immunostaining and the wide variation in Brn3 protein levels per cell. In comparing the patterns of Brn3 expression, RGCs and DRG neurons show several intriguing similarities: (1) there are far fewer Brn3c- than Brn3a- or Brn3b-expressing neurons; (2) many neurons express both Brn3a and Brn3b; and (3) among Brn3c-expressing neurons, a larger fraction coexpress Brn3a than Brn3b (Fig. 2A–D). The principal difference between RGC and DRG expression patterns is that, among DRG neurons, Brn3a is expressed in nearly all neurons and Brn3b and Brn3c are rarely coexpressed.
A comparison of Brn3 gene expression patterns in Brn3aAP/+ versus Brn3aAP/−, Brn3bAP/+ versus Brn3bAP/− and Brn3cAP/+ versus Brn3cAP/− DRGs at P0–P1 indicates that there is little cross-regulation among Brn3 family members; that is, loss of one Brn3 family member does not lead to repression or induction of other family members (Fig. 2E,F). In comparing P0/P1 versus P5 DRG neurons (Fig. 2C–F), we observe a progressive increase in the fraction of Brn3aAP/+ neurons that also express Brn3b (7 of 81 neurons at P0/P1 versus 16 of 42 neurons at P5; p = 0.0002), indicative of ongoing postnatal refinement within this subset of DRG neurons.
To relate the expression of Brn3 family members to functional DRG neuron subtypes during early postnatal development, we performed double labeling for a series of molecular markers and for AP in P5 Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ DRGs (Fig. 3A–C). The same markers were also analyzed in adult WT DRGs using anti-Brn3a, anti-Brn3b, and anti-Brn3c antibodies (Fig. 3D; and data not shown). The DRG markers consist of NF-200 (mechanoreceptors and proprioreceptors with large axon diameters), parvalbumin (proprioceptors), TrkA (small and medium diameter unmyelinated nociceptors and mechanoreceptors), CGRP (peptidergic nociceptors), IB4 (nonpeptidergic nociceptors), and peripherin (nociceptors and mechanoreceptors). TrkA is expressed in and required for the development of both peptidergic nociceptors (Marmigère and Ernfors, 2007) and a variety of mechanoreceptors (Fundin et al., 1997b; Cronk et al., 2002; Sedý et al., 2004).
Correlation among Brn3a, Brn3b, and Brn3c expression and DRG neuron subtype as defined by molecular markers. A, B, Double immunostaining for AP (red) and the indicated molecular markers (green) in DRGs at P5 following mosaic Cre-mediated recombination in Brn3aCKOAP/+;R26CreER, Brn3bCKOAP/+;R26CreER, and Brn3cCKOAP/+;R26CreER mice. Scale bar, 40 μm. C, Quantification of data from A and B. D, Quantification of adult WT DRGs using anti-Brn3a, anti-Brn3b, and anti-Brn3c immunostaining with the same series of markers as in A–C. Color code refers to C and D.
As summarized in Figure 3, C and D, the patterns of labeling at P5 and in the adult are distinctive for each Brn3 family member. Within the diverse set of Brn3a-expressing DRG neurons, all six markers are represented, both at P5 and in the adult. A seventh marker, tyrosine hydroxylase, which is present in 10–15% of adult mouse DRG neurons (Brumovsky et al., 2006), showed no colabeling at P5 with Brn3bAP/+ and Brn3cAP/+ DRG neurons. Among 124 Brn3aAP/+ neurons and 50 tyrosine hydroxylase-expressing neurons, only one neuron was possibly double-positive. At P5, Brn3b-expressing DRG neurons express only NF-200, peripherin, and TrkA at appreciable frequency, and Brn3c-expressing neurons express only TrkA and CGRP at appreciable frequency.
In comparing P5 and adult DRGs, the main differences are as follows: (1) an increase among all three classes of adult Brn3-expressing DRG neurons in the proportion that express NF-200, an effect that is most dramatic among Brn3c-expressing neurons; (2) a conversion of all or nearly all Brn3c-expressing neurons from peripherin negative to peripherin positive; (3) an increase in the percentage of Brn3c-expressing neurons that express CGRP from ∼55% to ∼85%; and (4) a decline in the proportion of Brn3a- and Brn3b-expressing neurons that express TrkA, with expression of TrkA persisting in a majority of Brn3c-expressing neurons. These changes between P5 and adulthood suggest that Brn3c-expressing neurons mature relatively late. Together, this analysis indicates that Brn3a-expressing DRG neurons encompass many, but not all, of the major classes of somatosensory neurons, Brn3b-expressing DRG neurons likely correspond to mechanoreceptors, and Brn3c-expressing DRG neurons likely correspond to peptidergic nociceptors.
Distinctive laminar targets in the dorsal horn of the spinal cord for Brn3a-, Brn3b-, and Brn3c-expressing DRG neurons
The correlations described in the preceding paragraph imply that Brn3a-expressing DRG neurons should project to many laminae in the dorsal horn of the spinal cord, whereas Brn3b- and Brn3c-expressing DRG neurons should project to only one or a few laminae. To test these predictions, the projections of Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ DRG neurons in the adult dorsal horn were visualized using AP immunostaining together with immunostaining for CGRP, IB4, or PKCγ, markers that define, respectively, peptidergic nociceptive laminae I and IIo, a nonpeptidergic nociceptive zone between laminae IIo and IIi, and mechanoreceptive lamina IIi (Zylka et al., 2005; Neumann et al., 2008). The central projections of Brn3a-expressing DRGs were distributed throughout the dorsal horn (Figs. 4A,B, 7A), consistent with the broad expression of Brn3a in DRG neurons. In contrast, the central projections of Brn3b- and Brn3c-expressing DRG neurons were narrowly targeted to the outer half of the PKCγ lamina and to the CGRP lamina, respectively (Figs. 4C–H, 7A). These data are consistent with an assignment of Brn3b-expressing DRG neurons as likely mechanoreceptors, based on the assignment of the PKCγ lamina as a target area for non-noxious stimuli (Neumann et al., 2008), and Brn3c-expressing DRG neurons as likely peptidergic nociceptors.
Axons of Brn3a-, Brn3b-, and Brn3c-expressing DRG neurons target distinct territories in the dorsal horn of the spinal cord. A–H, Double immunostaining for AP (red) and the indicated molecular markers (green) in the adult dorsal horn following Cre-mediated recombination in Brn3aCKOAP/+;Sox2Cre, Brn3bCKOAP/+;Sox2Cre and Brn3cCKOAP/+;Sox2Cre mice. AP staining at the dorsal edge of the spinal cord is derived from DRG fibers that have not yet entered the dorsal horn or are passing between spinal segments. A, B, Brn3aAP fibers project broadly within the dorsal horn. C–E, Brn3bAP fibers project to laminae ventral to the narrow zones of CGRP localization and IB4 staining, and overlapping the dorsal ∼50% of the broad zone of PKCγ localization. F–H, Brn3cAP fibers are largely overlapping with the distribution of CGRP in the dorsal-most laminae. Scale bar, 100 μm.
Visualizing afferent arbors of Brn3a-, Brn3b-, and Brn3c-expressing DRGs
Among somatosensory neurons, the integration of physiological, morphological, and molecular properties is most advanced in the context of pain, temperature, and itch sensation mediated by TRP channels and the Mrgprd class of GPCRs (Zylka et al., 2005; Lumpkin and Caterina, 2007; Dussor et al., 2008; Basbaum et al., 2009; Cavanaugh et al., 2009; Liu et al., 2009; Rau et al., 2009). Largely missing, thus far, has been a morphologic analysis of somatosensory afferents beyond the structures of individual sensory terminals. In particular, there is very little information regarding the size and geometry of the afferent arbors of individual somatosensory neurons. The paucity of data reflects several factors: (1) somatosensory afferents typically travel long distances from their cell bodies, thus hampering their visualization following injection of the soma with intracellular tracers; (2) immunostaining for somatosensory proteins or staining with nonspecific methods such as silver impregnation generally reveals a complex meshwork of afferent processes from which individual arbors cannot be reconstructed; and (3) many of the tissues within which sensory afferents reside—skin, muscle, the viscera, and connective tissue—are relatively thick and refractile, making it difficult to immunostain and image them in whole-mount preparations. This situation stands in marked contrast to the sophisticated morphologic analyses that have been conducted in the retina, where a relatively thin two-dimensional tissue offers ready access to physiologic recordings, cell filling, and immunochemical or histochemical staining in both flatmount and sectioned preparations (Masland, 2001; Dacey et al., 2003; Badea and Nathans, 2004).
To our knowledge, only one published study has used a genetically directed reporter to visualize the complete arbors of individual somatosensory afferents, that of Liu et al. (2007), in which skin afferents expressed an AP knockin at the MrgprB4 locus. AP histochemistry lends itself to visualizing skin afferents because of the high sensitivity of the AP reporter, the efficient diffusion of the low-molecular-weight AP substrate into relatively thick tissue samples, and the insolubility of the nitro-blue tetrazolium reaction product in the tissue clearing solvent benzyl benzoate:benzyl alcohol.
Individual afferent arbors of Brn3a-, Brn3b-, and Brn3c-expressing DRG neurons were visualized following extremely sparse Cre-mediated recombination with either of two CreER lines. Most of the experiments shown in Figure 5 used an IRES-CreER cassette knocked into the 3′ UTR of the gene coding for the neurofilament light chain (NFL-CreER) (Rotolo et al., 2008). For Figures 5U, 6, and 7, CreER was expressed from the ROSA26 locus under the control of the reverse Tet transactivator (rtTA), thus providing pharmacologic control over both transcription and nuclear translocation of Cre recombinase (R26rtTACreER) (Badea et al., 2009b). As described below, NFL-CreER appears to be expressed in many types of DRG neurons as judged by the diversity of labeled Brn3a-expressing afferents in skin and muscle and the labeling of central Brn3a-expressing projections throughout the dorsal horn of the spinal cord. This conclusion is also supported by the high degree of similarity in afferent arbor patterns observed between the NFL-CreER and R26rtTACreER datasets; the latter presumably derives from ubiquitous expression of CreER.
Individual Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ somatosensory arbors visualized histochemically following sparse Cre-mediated recombination. A–N, Brn3aCKOAP/+ afferents. A, 200 μm transverse section of P1 spinal cord and DRGs. In addition to DRG cell bodies and their processes, several large multipolar interneurons are labeled within the spinal cord. B–E, Innervation of muscle and tendons: hemisected P1 foot with the palmar surface at the bottom (B); adult diaphragm (C) and adult esophagus (D); and isolated muscle spindle in subdermal muscle at P1 (E). F–N, P1 abdominal skin flatmounts showing individual sensory arbors (F–K, M, N); tangential section of P1 skin showing a guard hair with a single sensory ending (green arrow) (L). F–N, Blue arrows indicate individual afferent fibers. F, A single elaborate follicle associated ending. M, A large arbor with a mixture of follicle-associated and nonassociated endings. Scale bars: A, C, 500 μm; B, D, E, G, N, 200 μm. F–L are at the same scale, and M and N are at the same scale. O–V, Brn3bCKOAP/+ afferents. O–T, P1 abdominal skin flatmounts. Blue arrows indicate individual afferent fibers. U, Innervation of glabrous skin in the P1 foot in cross section. V, Brn3b-expressing fibers in a P1 Brn3bAP/+ cochlea (following germline recombination) innervate sensory hair cells in the organ of Corti. Scale bars: O, 500 μm; P–U, 200 μm. W–Y, Brn3cCKOAP/+ afferents. I–K, Brn3cAP/+ afferents in P1–P3 abdominal skin flatmounts from Brn3cCKOAP/+;NFL-CreER mice. Red arrows show low-level read-through AP expression from the un-recombined Brn3cCKOAP allele in Merkel cells adjacent to each guard hair. Blue arrows indicate individual afferent fibers. Scale bar, 200 μm.
Three-dimensional reconstructions of afferent arbors from Brn3a-, Brn3b-, and Brn3c- expressing DRG neurons in P1–P3 skin. A–L, Each dendritic arbor reconstruction is assigned a letter (A–L), and two projections—one in the plane of the skin (top) and one perpendicular to that plane (bottom)—are shown. For the territories encompassing arbors A and C and arbors E and I, the two neighboring arbors are shown in yellow and green. Outlines of guard hair follicles are shown in gray, and the point of contact between each follicle and the epidermis is marked by a pink horizontal line in the transverse projection. Stratification levels of individual subdermal muscle fibers are represented by dark red bars in the transverse projections. The most proximal point of each afferent fiber reconstruction is marked by a black arrow. Small C-shaped hair follicle-associated structures are present in arbors D, F, and K, and are highlighted in blue; more elaborate hair follicle associated structures are present in E, G, and H. Distinct morphological types representative of Brn3aAP/+, Brn3bAP/+, and Brn3cAP/+ arbors are outlined in green, red, and blue, respectively. All morphologies were observed among Brn3aAP/+ arbors. Types A and B were also observed among Brn3cAP/+ arbors, types C–H were also observed among Brn3bAP/+ arbors, and types I–L were unique to Brn3aAP/+ arbors. The specific examples shown are derived from Brn3aCKOAP/+;R26rtTACreER/+ (A–F, H–L) and Brn3bCKOAP/+;R26rtTACreER/+ (G) P1 skins. Scale bar, 100 μm.
Loss of Brn3a but not Brn3b or Brn3c causes a loss of dorsal horn projections and hair follicle-associated sensory endings in the skin. A, AP histochemistry of vibratome sectioned spinal cords from adult Brn3aCKOAP/+;R26rtTACreER, Brn3aCKOAP/−;R26rtTACreER, Brn3bCKOAP/+;R26rtTACreER, Brn3bCKOAP/−;R26rtTACreER, Brn3cCKOAP/+;R26rtTACreER, and Brn3cCKOAP/−;R26rtTACreER/+ mice following Cre-mediated recombination at E9.5. Red arrow points to the loss of Brn3aCKOAP/− fibers in the dorsal horn. B, AP histochemistry of flatmounted abdominal skins from Brn3aCKOAP/+;R26rtTACreER and Brn3aCKOAP/−;R26rtTACreER P1 mice. Red arrowheads point to AP+ hair follicle-associated C-shaped sensory endings; these endings are common among Brn3aAP/+ arbors (left) but are extremely rare among Brn3aAP/− arbors (right). The right panel shows a region of Brn3aCKOAP/−;R26rtTACreER skin with a somewhat higher density of AP+ sensory arbors relative to the Brn3aCKOAP/+;R26rtTACreER skin in the left panel to emphasize the point that that the low density of C-shaped endings does not arise from a lower density of sensory arbors. C, Flatmount images of an isolated Brn3aAP/− sensory arbor (two focal planes at left) and its reconstruction (Fig. 6, right; as), with relatively simple branching and small area, similar to the example shown in Figure 6C. D, Box plots quantifying the density of AP+ follicle-associated C-shaped sensory endings in the skin of Brn3aCKOAP/+;R26rtTACreER and Brn3aCKOAP/−;R26rtTACreER P1 mice. Regions with similar overall densities of labeled arbors were chosen for comparison. The central red mark is the median, the blue edges of the box are the 25th and 75th percentiles, and the whiskers extend to the most extreme data points. For each genotype, eight fields of 2.6 × 2.6 mm were scored. p = 1.56 × 10−4. Scale bars: A, 300 μm; B, C, 100 μm.
In Figure 5, Brn3aCKOAP/+;NFL-CreER, Brn3bCKOAP/+;NFL-CreER, and Brn3cCKOAP/+;NFL-CreER fetuses were exposed to 0–200 μg of 4HT during late gestation and then abdominal skin from these mice at P0–P3 was processed for AP histochemistry. Neonatal rather than adult skin was analyzed because the pigment content, large size, and high density of adult hair follicles interfere with imaging of labeled afferents. In the mouse, many mechanoreceptors that innervate the skin appear to be physiologically and anatomically mature by the early postnatal period (Woodbury et al., 2001; Woodbury and Koerber, 2007). Figure 5A illustrates the specificity of AP expression in a 200 μm vibratome section through a P1 Brn3aCKOAP/+;NFL-CreER mouse. AP activity is restricted to DRG cell bodies, afferent fibers in the dorsal roots, projections throughout the dorsal horn of the spinal cord, and occasional multipolar neurons within the spinal cord. Figure 5B–E illustrates the diversity of nondermal afferents labeled in Brn3aCKOAP/+;NFL-CreER mice. Labeled afferents in the P1 foot, adult esophagus, and P1 abdominal wall terminate in enlarged structures, presumably tendon organs (foot) or muscle spindles (esophagus and abdominal wall) (Fig. 5B,D,E). In the adult diaphragm, the termini of the highly branched afferents are not obviously enlarged (Fig. 5C).
In flatmounts of Brn3aCKOAP/+;NFL-CreER P1 skin, AP-stained arbors vary in size from single dense C-shaped endings associated with hair follicles (likely lanceolate endings) (Fig. 5F,L) to arbors with >100 branches and a diameter of ∼1 mm (Fig. 5M,N). A range of intermediate sizes is also seen (Fig. 5G–K), as well as individual arbors with both lanceolate and nonlanceolate endings (Fig. 5M). Similar analyses of Brn3bCKOAP/+; NF-LCreER P1 skin show a more limited range of labeled arbor types, with few very large arbors and many arbors with single lanceolate endings or with a mixture of lanceolate and nonlanceolate endings (Fig. 5O–T), as well as sensory endings in the glabrous skin of the footpad (Fig. 5U). Brn3b is also expressed in all or nearly all spiral ganglion neurons, as judged by the density of fibers targeting the organ of Corti in a whole-mount Brn3bAP/+ P1 cochlea (Fig. 5V). Brn3cCKOAP/+;NFL-CreER P1 skin flatmounts show an even further reduction in the morphologic diversity of labeled sensory arbors, with almost all labeled afferents exhibiting a coarsely branched structure devoid of follicle-associated endings (Fig. 5W–Y). In Brn3cCKOAP skin, AP is expressed at low level in a semicircle of cells—presumably Merkel cells—surrounding each guard hair (Fig. 5W–Y, red arrows). These AP-expressing cells are not contiguous with labeled afferents, and they are readily distinguished from the more compact semicircular follicle-associated C-shaped endings (Fig. 5O–T). AP expression in Merkel cells appears to derive from low-level read-through transcription into the AP coding region at the un-recombined Brn3cCKOAP locus; a similar pattern of low-level read-through is seen for Brn3aCKOAP and Brn3bCKOAP loci in RGCs.
Three-dimensional reconstruction and morphologic classification of afferent arbors in the skin
To more rigorously characterize the morphologies of individual sensory arbors, we have reconstructed representative examples from sparsely labeled Brn3aCKOAP/+;R26rtTACreER/+ and Brn3bCKOAP/+;R26rtTACreER/+ abdominal skin at P1 (Fig. 6). The colored outlines in Figure 6 indicate the patterns of Brn3a, Brn3b, and/or Brn3c expression among the various morphologic classes. Arbors from DRG neurons that express Brn3c, as well as a subset that express Brn3a, cover a large area, branch sparsely, stratify either broadly within the dermis (Fig. 6A) or narrowly near the dermal-epidermal boundary (Fig. 6B), and do not make contacts with hair follicles. Arbors from DRG neurons that express Brn3b, as well as a subset that express Brn3a, branch more densely and, in most cases, make one or more C-shaped contacts with hair follicles (Fig. 6C–H). Arbors from DRG neurons that express Brn3a exclusively are diverse, including examples that cover a large area with numerous dense branches, stratify either narrowly near the dermal–epidermal boundary (Fig. 6J) or more deeply (Fig. 6I,L), and make few or no contacts with hair follicles. Also in the Brn3a-only group are arbors that contact hair follicles at more than one depth (Fig. 6K). The great diversity of dermal and nondermal sensory afferents that express Brn3a is consistent with the nearly ubiquitous expression of Brn3a in DRG neurons (Figs. 2, 3) and the large number of dorsal horn laminae targeted by these neurons (Figs. 4, 5A, 7A).
Loss of Brn3a but not Brn3b or Brn3c causes a loss of DRG neuron subtypes
Previous genetic analyses of Brn3a function in the somatosensory system have been restricted to the prenatal period because Brn3a−/− mice die shortly after birth. Here we present a postnatal analysis of Brn3a function, based on the availability of the Brn3aCKOAP allele, which permits individual, histochemically marked Brn3aAP/− neurons to be produced by sparse Cre-mediated recombination at any time during development in Brn3aCKOAP/−;R26rtTACreER/+ embryos. Since the vast majority of Brn3a-expressing neurons do not undergo Cre-mediated recombination with the doxycycline and 4HT regimen used here (see Materials and Methods), the resulting mice can be studied at any postnatal age. Moreover, the phenotypes exhibited by individual Brn3aAP/− DRG neurons in the Brn3aCKOAP/−;R26rtTACreER/+ background are presumably cell autonomous since almost all of the neighboring neurons are phenotypically WT. This last point is of special relevance because the massive neuronal loss observed in Brn3a−/− DRGs would be expected to produce cell-nonautonomous effects arising from reduced competition with neighboring DRG neurons for target innervation, aberrant axonal pathfinding by neighboring mutant neurons, or a large local concentration of apoptotic cells. While cell-nonautonomous effects are of interest in their own right, experiments conducted in Brn3a−/− mice cannot distinguish autonomous from nonautonomous effects.
As an initial step in defining the cell-autonomous phenotypes associated with loss of Brn3a, Brn3b, and Brn3c, we compared the territories targeted by AP-expressing DRG neurons in the dorsal horn of adult Brn3aCKOAP/+;R26rtTACreER versus Brn3aCKOAP/−;R26rtTACreER, Brn3bCKOAP/+;R26rtTACreER versus Brn3bCKOAP/−;R26rtTACreER, and Brn3cCKOAP/+;R26rtTACreER versus Brn3cCKOAP/−;R26rtTACreER/+ spinal cords under conditions of sparse Cre-mediated recombination (Fig. 7A). While no differences were seen between Brn3bAP/+ and Brn3bAP/− projections or between Brn3cAP/+ and Brn3cAP/− projections, Brn3aAP/+ and Brn3aAP/− projections differed markedly, with the latter showing a large zone of hypoinnervation (Fig. 7A, red arrowhead). Similarly, in early postnatal skin, a comparison between Brn3bAP/+ and Brn3bAP/− arbors or between Brn3cAP/+ and Brn3cAP/− arbors showed no apparent differences in density or in the types of morphologies, whereas a comparison between Brn3aAP/+ and Brn3aAP/− skin showed a reduced density of AP+ arbors in the Brn3aAP/− sample (data not shown). Strikingly, the morphologies of Brn3aAP/− arbors represent only a subset of the Brn3aAP/+ arbor morphologies. In particular, the Brn3aAP/− skin is nearly devoid of AP+ C-shaped endings that contact hair follicles, a difference that is apparent even when comparing regions in which the overall density of AP+ arbors in the Brn3aCKOAP/−;R26rtTACreER skin is comparable or greater to that in the Brn3aCKOAP/+;R26rtTACreER control skin (Fig. 7B,D). Many of the AP+ arbor morphologies that remain in Brn3aCKOAP/−;R26rtTACreER skins conform to types observed in control Brn3aCKOAP/+;R26rtTACreER skins, as seen, for example, in comparing Figures 6C and 7C. Thus, loss of Brn3a leads to a selective loss and/or fate switch of hair follicle-associated terminals.
Discussion
The experiments described here, together with those reported in Badea et al. (2009a) and Badea and Nathans (2011), use a set of three Brn3-null alleles and three AP-expressing Brn3 conditional alleles, together with sparse Cre-mediated recombination, to define the overlapping patterns of Brn3 gene expression in RGCs and somatosensory neurons, the manner in which the patterns of gene expression correlate with morphologically distinct neuronal types, and the consequences for these neurons of deleting each Brn3 gene. Among RGCs, we observe that Brn3 gene expression correlates with the patterns of dendritic arborization and lamination in the inner plexiform layer and the patterns of axonal projections to retinorecipient targets in the brain. Among somatosensory neurons, we observe subtype-specific expression of Brn3b and Brn3c. We also demonstrate the general utility of sparse, genetically directed reporter expression and histochemical labeling for visualizing individual somatosensory afferent arbors, an approach that should aid in classifying somatosensory neurons and determining the sizes and structures of their receptive fields.
Brn3 gene expression and somatosensory neuron identity
The essential role of Brn3a in the development and survival of somatosensory neurons was appreciated in the initial analyses of Brn3a−/− mice (McEvilly et al., 1996; Xiang et al., 1996). Subsequent work has refined this picture by characterizing the downregulation of a variety of somatosensory cell-type-specific molecular markers, including Trk receptors, in the DRG and TG of Brn3a−/− mice (Huang et al., 1999; Ichikawa et al., 2002, 2004, 2005; Ma et al., 2003), by demonstrating defects in pathfinding of somatosensory afferents (Eng et al., 2001) and by showing that, either directly or indirectly, Brn3a represses genes associated with neurogenesis and induces genes associated with terminal differentiation, including Runx1 and Runx3 (Lanier et al., 2009; Dykes et al., 2010). The nearly complete loss of hair follicle-associated endings observed here among Brn3aAP/− afferents is consistent with the idea that Brn3a controls a program of Runx and Trk receptor expression that is required for mechanosensory neuron development and/or survival (Dykes et al., 2010).
Before the present work, the relationship between Brn3b and Brn3c expression and somatosensory neuronal subtypes was largely unexplored. Although the immunohistochemical profiles of DRG cell bodies, morphologies of afferent arbors in the skin, and locations of target laminae in the dorsal horn suggest that Brn3b is expressed principally in hair follicle-associated mechanoreceptors and that Brn3c is expressed principally in peptidergic nociceptors, a definitive assignment of sensory types must await an electrophysiologic characterization of the stimulus–response characteristics of these neurons. The persistence of each of these cell types upon elimination of the associated Brn3 protein suggests that, unlike Brn3a, Brn3b and Brn3c may play a relatively subtle role in the somatosensory system, perhaps conferring aspects of molecular identity, but not determining subtype-specific morphology, central target specificity, or survival.
Genetically directed sparse labeling as a tool for characterizing neuronal morphology in mice
Our analyses of Brn3 expression and function have relied on the use of sparse Cre-mediated Brn3 deletion coupled to activation of an AP reporter. This strategy offers several experimental advantages over the use of conventional null alleles or conditional alleles that do not generate marked cells. First, it permits mosaic mice to survive beyond the age at which homozygous null mutants die. For Brn3a−/− mice, death occurs in the neonatal period, a severe limitation for the analysis of retinal development, much of which occurs postnatally. Second, the sparse recombination strategy limits the phenotypic analysis to cell-autonomous effects, since the vast majority of the cells surrounding the rare AP/− cell are phenotypically WT. As noted in the Results section, in Brn3a−/− embryos the large numbers of dying DRG and TG neurons and/or the reduced competition for peripheral and central targets among the surviving DRG and TG neurons could produce cell-nonautonomous effects. A similar argument applies to the Brn3b−/− retina, where ∼70% of RGCs eventually die. Third, by using pharmacologic control of Cre recombinase—in the present study, a dual doxycycline and 4HT strategy (Badea et al., 2009b)—the rare recombination events that lead to loss of gene function can be precisely timed. Finally, the sparse recombination method permits single-cell morphologic analysis by revealing a Golgi-like image of neuronal processes, either in AP/+ or AP/− cells, the latter being null for the protein of interest but expressing the AP reporter by virtue of continued activity of the Brn3 promoter. With respect to sparse visualization of somatosensory afferents and central projections, the combination of Brn3aCKOAP and NFL-CreER might be generally useful as a tool for surveying these structures in the context of surgical, toxicologic, genetic, or other experimental perturbations.
Combinatorial expression of the Brn3 family in the visual, somatosensory, and auditory/vestibular systems
Figure 8 summarizes the patterns of Brn3a, Brn3b, and Brn3c expression in the WT inner ear, retina, and DRG/TG. The Venn diagrams in Figure 8A illustrate both the relative numbers of neurons expressing each Brn3 gene and the extent of overlap of Brn3 gene expression among these neurons. In most cases, the Brn3 genes are expressed in the projection neurons that communicate information from the primary sensory structures to the brain or spinal cord (Fig. 8B–D), the two exceptions being the expression of Brn3c in vestibular and auditory sensory hair cells (Fig. 8C) and in Merkel cells (Fig. 5W–Y). The general pattern of expression in projection neurons applies regardless of whether the Brn3-expressing cells are or are not primary sensory neurons or whether they are derived from surface ectoderm (spiral and vestibular ganglia), the neuroectoderm (DRG/TG), or neural tube (retina). Conceptualized in this manner, the logic of Brn3 expression suggests an evolutionary equivalence of RGCs, primary somatosensory neurons, and auditory and vestibular ganglion neurons.
Combinatorial expression of Brn3a, Brn3b, and Brn3c in projection neurons in the visual, auditory/vestibular, and somatosensory systems. A, Venn diagrams summarizing the relative numbers of expressing neurons and the patterns of overlapping expression among Brn3 family members in the retina, inner ear, and DRG. B, Central targets of RGCs that express different combinations of Brn3a, Brn3b, and Brn3c. For illustrative purposes only a subset of the many accessory optic areas is shown. For example, the central targets that control pupil constriction are not shown; these receive input from Brn3b-expressing RGCs and are functionally impaired in Brn3b−/− mice (Badea et al., 2009a). C, In the auditory and vestibular systems, sensory hair cells express Brn3c (Xiang et al., 1997), and most or all spiral and vestibular ganglion cells express both Brn3a and Brn3b (Huang et al., 2001). D, In the somatosensory system, Brn3a is expressed by all or nearly all classes of sensory neurons, Brn3b is expressed by sensory neurons that have follicle-associated C-shaped endings, and Brn3c is expressed by sensory neurons that have endings not associated with hair follicles.
How do the patterns of Brn3 gene expression relate to the evolution of increasingly complex sensory structures? It is reasonable to suppose that the primordial architecture of all sensory neurons initially resembled that seen in present-day nociceptors: the neuron that was the primary receiver of sensory information also projected directly to a central target. Following this logic, we imagine that the two outer layers of neurons in the trilayered vertebrate retina represent a later addition of specialized photoreceptors and interneurons. Consistent with this view, present-day vertebrate retinas contain a minor class of intrinsically photosensitize RGCs (ipRGCs). The ipRGCs mediate non-image-forming visual functions such as circadian entrainment and pupil constriction (Güler et al., 2008), and they may represent the last relics of a primordial monolayer retina. Analogous arguments apply to the auditory and vestibular systems: primordial mechanosensory neurons may have projected directly to their central targets, and only later acquired a specialized epithelial partner—hair cells in the auditory/vestibular system and a variety of specialized epithelial cells in mechanosensory end organs in the skin, muscle, tendons, and viscera—to enhance their sensitivity. In this regard, it is striking that Brn3c is not only expressed in auditory hair cells (Xiang et al., 1997) but is also expressed in Merkel cells (Fig. 5W–Y, red arrows). This serendipitous finding is consistent with a growing body of evidence demonstrating that Merkel cells—specialized skin epithelial cells that synapse onto mechanosensory afferents—share molecular, morphologic, and functional similarities with auditory hair cells (Lumpkin and Caterina, 2007). As Brn3c is only required in the final stages of inner ear sensory hair cell differentiation (Xiang, 1998), these data suggest that Brn3c may play a general role in promoting the mechanosensory differentiation of specialized epithelial cells.
A further inference from the suggested evolutionary equivalence of RGCs, primary somatosensory neurons, and auditory and vestibular ganglion neurons is that there may be a corresponding equivalence in the logic of information processing at their initial CNS targets. Thus, the separation of somatosensory submodalities by targeting of DRG axons to distinct laminae in the dorsal horn of the spinal cord may be analogous to the separation of visual information streams produced by targeting of RGC axons to different retino-recipient regions in the diencephalon, or to the separation of auditory and vestibular inputs by targeting of spiral and vestibular ganglion axons to the cochlear and vestibular nuclei in the brainstem.
Functions analogous to those of mammalian Brn3 genes are seen in the single Brn3 homologs in Caenorhabditis elegans (Unc86) and Drosophila melanogaster (Acj6). Unc86 is required both for the production and differentiation of primary touch-sensitive neurons and the correct functioning of the chemosensory AIZ interneurons (Duggan et al., 1998; Sze and Ruvkun, 2003). Acj6 is required for specifying the identity of primary olfactory receptor neurons, at the level of olfactory receptor gene choice and at the level of axon targeting to the appropriate glomerulus (Komiyama et al., 2004; Bai et al., 2009). Intriguingly, differential splicing of the Acj6 gene generates multiple isoforms with distinct activities in olfactory receptor neuron specification (Bai and Carlson, 2010). Initial steps have been taken to identify transcriptional targets of Unc86, Acj6, and the mammalian Brn3s (Erkman et al., 2000; Eng et al., 2004, 2007; Mu et al., 2004; Lanier et al., 2009; Dykes et al., 2010). It will be of great interest to compare these targets both between species and across sensory modalities.
Footnotes
This work was supported by the Howard Hughes Medical Institute and the National Institutes of Health. We thank Rocky Cheung, Sumit Kumar, and Yanshu Wang for help with histology and immunostaining; and Xinzhong Dong, David Ginty, Qin Liu, and Ting Guo for helpful discussions and/or comments on this manuscript.
- Correspondence should be addressed to either of the following: Dr. Tudor Constantin Badea, Building 6, National Eye Institute, NIH, Bethesda, MD 20892, badeatc{at}mail.nih.gov; or Dr. Jeremy Nathans, 805 PCTB, 725 North Wolfe Street, Johns Hopkins University School of Medicine, Baltimore, MD 21205, jnathans{at}jhmi.edu