The role of the transcription factor Rbpj in the development of dorsal root ganglia

Background The dorsal root ganglion (DRG) is composed of well-characterized populations of sensory neurons and glia derived from a common pool of neural crest stem cells (NCCs), and is a good system to study the mechanisms of neurogenesis and gliogenesis. Notch signaling is known to play important roles in DRG development, but the full scope of Notch functions in mammalian DRG development remains poorly understood. Results In the present study, we used Wnt1-Cre to conditionally inactivate the transcription factor Rbpj, a critical integrator of activation signals from all Notch receptors, in NCCs and their derived cells. Deletion of Rbpj caused the up-regulation of NeuroD1 and precocious neurogenesis in DRG early development but led to an eventual deficit of sensory neurons at later stages, due to reduced cell proliferation and abnormal cell death. In addition, gliogenesis was delayed initially, but a near-complete loss of glia was observed finally in Rbpj-deficient DRG. Furthermore, we found P75 and Sox10, which are normally expressed exclusively in neuronal and glial progenitors of the DRG after the NCCs have completed their migration, were co-expressed in many cells of the DRG of Rbpj conditional knock-out mice. Conclusions Our data indicate that Rbpj-mediated canonical Notch signaling inhibits DRG neuronal differentiation, possibly by regulating NeuroD1 expression, and is required for DRG gliogenesis in vivo.


Background
The nervous system is made up of a wide variety of neuronal and glial cell types. How these different cell types are generated from multipotent progenitors during development is a fundamental and largely unanswered question in neuroscience. The dorsal root ganglion (DRG), which consists of several well-characterized types of sensory neurons and glial cells, is an attractive model system to investigate the molecular processes underlying cellular differentiation in the nervous system [1]. Sensory neurons and glial cells in the DRG derive from the neural crest, a tissue that exists transiently during mammalian embryogenesis at the border between the ectoderm and the neural plate [2]. Between embryonic day (E)8.5 and E10.0 in the mouse, neural crest stem cells (NCCs) begin to exit the neural tube, and those that migrate along the ventral pathway between the neural tube and the dermamyotome form the DRG [3]. Between E9. 25 and E13.5 in the mouse, NCCs first give rise to large neurons that express neurotrophic tyrosine receptor kinase (Trk)C, and then to medium-sized TrkB + and small-sized TrkA + sensory neurons [4,5]. NCC-derived glial cells are also generated during this period, including satellite cells and Schwann cells, although these begin to differentiate about 1.5 days after sensory neurons [4,6].
Many genes are involved in the generation of sensory neurons and glia from multipotent NCCs. Among the various cell-surface proteins known to be expressed by NCCs, the low affinity neurotrophin receptor P75 has been widely used to identify and isolate NCCs [7]. In addition, P75 interacts with TrkC, the high-affinity receptor for Neurotrophin-3, to promote neuronal differentiation of NCCs in vitro [8]. The high-mobility group transcription factor SRY (sex determining region Y) box 10 (Sox10) is expressed in pre-migratory and migratory NCCs and plays a role in maintaining the multipotency of NCCs [9]. Expression of Sox10 turns off in daughter cells committed to neuronal fates, but persists in glia-restricted progenitors and differentiated glia [9,10]. The specification of DRG sensory neuron lineages is also controlled by several transcription factors. For example, the basic helix-loop-helix (bHLH) transcription factors Neurogenin-1 (Ngn1) and Neurogenin-2 (Ngn2) promote sensory fates, as opposed to autonomic ones [1,5,11,12], and are required for the initiation of neurogenesis [5]. Neurogenic differentiation 1 (NeuroD1) is thought to act downstream of the neurogenins in the regulation of neuronal differentiation [13,14]. As terminal differentiation progresses, sensory neuron subtypes with distinct modalities acquire specific patterns of Trk expression, uniquely expressing either TrkA, TrkB, or TrkC [1,15].
The function of Notch signaling in DRG development, as revealed by in vitro studies and in vivo chick studies, is consistent with its two fundamental roles in the development of the nervous system: maintaining undifferentiated progenitors by inhibiting neuronal differentiation, and promoting glial determination [16,17]. Studies of chick development show that Notch signaling is essential for the maintenance of NCCs and prevents neuronal differentiation in the DRG and sympathetic ganglion [18]. Furthermore, transient activation of Notch signaling in NCCs in vitro prevents neuronal differentiation and promotes glial differentiation [19,20]. On the other hand, results from mutant mice with Notch signaling pathway deletions are not consistent. Delta-like-1 (a Notch ligand) null mice had aberrantly fused and/or reduced DRG and sympathetic ganglia, suggesting that Notch signaling is also essential for the proper migration of NCCs [21]. Hes1 and Hes5 (Notch signaling effectors) double-null mutant mice maintain the expression of the specific marker for glial cell precursors, brain fatty acid binding protein (BFABP), in the DRG, suggesting that Notch signaling is not involved in the generation of glial cell precursors from NCCs [22,23]. By over-expressing or inactivating Notch signaling specifically in Schwann cell precursors in vivo, Notch signaling has been shown to drive the differentiation of immature Schwann cells [23]. Thus, manipulating individual components of Notch signaling in the mouse yields varying results and the full scope of Notch functions in mammalian DRG development remains to be elucidated.
The presence of four Notch receptors and at least five Notch ligands, all of which are partially functionally redundant, has made it difficult to investigate the physiological function of Notch signaling in mammals [16]. Recombination signal binding protein for immunoglobulin kappa J region (Rbpj) can interact with the intracellular domains of all four Notch receptors and is required to mediate their transcriptional effects [24,25]. Therefore, deletion of Rbpj would be expected to completely abolish canonical Notch signaling. Taylor et al. [26] conditionally knocked out Rbpj expression in NCCs and NCC-derived cells and observed a modest reduction in sensory neurons and profound defects in gliogenesis in the DRG of Wnt1-Cre;Rbpj flox/flox (Rbpj conditional knock-out (CKO)) mice. However, the whole range of DRG phenotypes, especially the defects of neurogenesis, in Rbpj CKO mice has not yet been described, and the mechanisms underlying the phenotype remain unclear.
In the present study, we focused on early developmental events in the DRG of Rbpj CKO mice. We found that Ngn1 and Ngn2 expression was unchanged in the absence of Rbpj, but NeuroD1 was up-regulated and precocious neurogenesis occurred in the DRG. The elevated rate of neuronal differentiation at early time points was followed by a reduction in cell proliferation and abnormal cell death. In addition, the initiation of BFABP expression in glial progenitors was delayed, and this expression was lost at later stages. In wild-type mice, P75 and Sox10 are co-expressed in NCCs during early DRG development, but are subsequently expressed in two distinct populations as development progresses [27]. In contrast, this separation did not occur in Rbpjdeficient DRG, as evidenced by the presence of P75/ Sox10 co-expressing cells at later developmental stages, suggesting that the multipotency of NCCs was abnormally maintained, thus arresting their development. These data provide further insights into the physiological functions of Notch signaling in the development of DRG.

Normal induction and early migration of NCCs in Rbpj CKO embryos
In order to determine at what time point Rbpj becomes inactivated in NCCs and their derivatives in Rbpj CKO mice, we performed X-gal staining of Wnt1-Cre;Rosa26 reporter (R) embryos that express β-galactosidase (β-gal) in the Cre expression domain. The X-gal product did not appear in the roof plate of the neural tube or in the presumptive DRG until E9.5 (data not shown). At this stage, nearly all P75 + NCCs were also positively labeled with βgal antibody in both Rbpj CKO (Wnt1-Cre;Rbpj flox/flox ; Rosa26R) and control (Wnt1-Cre;Rosa26R) embryos ( Figure 1A-F), demonstrating that Rbpj could be deleted in NCCs and their derivatives from E9.5 on. P75 and Sox10 are co-expressed in pre-migratory and migrating NCCs [27]. The double staining of P75 and Sox10 showed there was no obvious difference in the distribution and number of P75/Sox10-double labeled cells between CKO and wild-type embryos at E9.5 ( Figure 1G-L). This result is consistent with a previous report [26] and suggests that deleting Rbpj from E9.5 does not affect the induction and initial migration of NCCs.

Precocious neurogenesis in Rbpj-deficient DRG
We next examined neurogenesis in Rbpj-deficient DRG by looking at the expression of neurogenins in these mice. Ngn1 and Ngn2 are required for the specification of DRG sensory neurons [5]. In E9.5 wild-type embryos, Ngn1 and Ngn2 were expressed in a cluster of migratory NCCs (Figure 2A,G). The expression domain of Ngn1 was enlarged at E10.0 and E10.5 ( Figure 2C,E), whereas Ngn2 expression was unchanged at E10.0 and had disappeared by E10.5 ( Figure 2I,K), a stage at which differentiating sensory neurons are condensed into ganglia [28]. We did not observe any obvious difference in Ngn1 and Ngn2 expression between wild-type and CKO embryos from E9.5 to E10.5 (Figure 2), suggesting that the specification of sensory neurons may not be affected in the absence of Rbpj.
Detectable defects in neurogenesis in Rbpj CKO mice were first observed at E10.5, a stage when nearly all primary sensory neurons co-express the POU homeodomain transcription factor Brn3a and the LIM homeodomain transcription factor Islet1 [29], and the onset of their expression coincides largely with the commitment of NCCs to neuronal fates [28]. We found that the numbers of both Islet1 + ( Figure 3A,D,Y) and Brn3a + ( Figure 4A,B) cells were dramatically increased and that of Sox10 + cells ( Figure 3B,E,Y) was decreased in Rbpj CKO mice at E10.5 compared with wild types. Furthermore, many Islet1-expressing cells were abnormally      Figure 3A,D), an area normally occupied by Sox10 + cells in the wild-type DRG (arrowheads in Figure 3B,E). The overall increase in the number of neurons in the E10.5 CKO DRG was confirmed by increased expression of the pan-neuronal marker SCG10 ( Figure 4C,D,E). To explore the mechanisms underlying precocious neurogenesis in the CKO DRG, we examined the expression of the proneural gene Neu-roD1. In situ hybridization showed a great increase in NeuroD1 expression in the DRG of CKO embryos relative to wild-type controls at E10.5 (Figure 4E,F,G), and co-immunostaining revealed that almost all NeuroD1 + neurons were also positive for Islet1 ( Figure 3G-L). Taken together, our data indicate that, in the absence of Rbpj, DRG progenitors precociously differentiate into sensory neurons, possibly due to up-regulation of NeuroD1.

Rbpj-deficient DRG
The increase in the number of sensory neurons in the DRG of Rbpj CKO mice was not obvious at E11.5 (Figure 3M-R), and by E12.5 the number was dramatically reduced relative to wild-type littermates ( Figure 3S-X), consistent with a previous report [26]. We reasoned that the precocious neurogenesis caused by the deletion of Rbpj led to premature depletion of the progenitor pool, which in turn resulted in an overall reduction in sensory neuron production. To test this possibility, we pulse labeled the E10.5 to E11.5 DRG with bromodeoxyuridine (BrdU) and analyzed the rates of proliferation of progenitor cells 2 hours later. There was no significant difference in BrdU incorporation between wild-type and CKO mice at E10.5, but the number of BrdU-labeled cells was greatly reduced in the CKO DRG at E11.5 ( Figure 5A-D,I).
To determine whether elevated rates of cell death may also have contributed to the overall reduction of sensory neurons, we performed terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining. No difference in TUNEL staining between the wild-type and CKO DRG was observed at E10.5, but the number of TUNEL + cells was increased in CKO mice relative to controls at E11.5 ( Figure 5E-H,J). In order to explore which kinds of cells are dying in the E11.5 DRG of CKO mice, Caspase3, an apoptosis marker, was coimmunostained with Sox10 ( Figure 6A-G), P75 ( Figure  6H-N) or Islet1 ( Figure 6X-V). Like TUNEL staining, many Caspase3 + cells were present in CKO mice, whereas they were rarely observed in wild-type controls at E11.5 ( Figure 6). Co-immunostaining showed Cas-pase3 was co-localized with Sox10 ( Figure 6G), P75 ( Figure 6N) or Islet1 ( Figure 6V), showing that abnormal cell death occurs in both progenitor cells and early differentiating neurons. Taken together, these results suggest that reduced cell proliferation and increased cell death contribute to the reduction of sensory neurons in Rbpj CKO mice.
As mentioned above, sensory neurons in the DRG can be divided into three groups based on their expression of the three Trk genes [1]. To determine whether Rbpj deletion has distinct effects on the development of different types of sensory neurons, we examined the expression of TrkA, TrkB and TrkC. We found that the number of TrkC + neurons in Rpbj CKO DRG was greatly increased (Figure 4 L,M), but the numbers of TrkA + and TrkB + neurons were not changed at E10.5 ( Figure 4H-K). The significant increase of TrkC + neurons in Rpbj CKO DRG no longer existed at E11.5 (Figure 7G,H,Q), and TrkA + neurons in Rpbj CKO DRG were reduced in number compared with wild-type mice ( Figure 7C,D,Q). At postnatal day 0, the number of TrkC + neurons in Rpbj CKO DRG was comparable to that of wild-type mice ( Figure 7O,P,R), but TrkA + and TrkB + neurons were greatly reduced in CKO mice (Figure 7K-N,R). During DRG morphogenesis, large TrkC + neurons are the first to be generated [1]; thus, the selective increase in TrkC + neurons supports the idea that deletion of Rbpj leads to enhanced neurogenesis during the first wave of NCC differentiation, but this results in a depletion of the neural progenitor pool, which in turn leads to the reduction of consequent generation of TrkA + and TrkB + neurons. Of course, abnormal cell death also contributes to the reduction of these two types of sensory neurons.

Severe gliogenesis defects in Rbpj-deficient DRG
Taylor et al. [26] showed that severe gliogenesis defects are present in Rbpj-deficient DRG. Consistent with their findings, we also found a complete loss of BFABP (a glia-specific marker) expression in the Rbpj-deficient DRG at E16.5 (Additional file 1). At E11.0, a half day after the occurrence of precocious neurogenesis in Rbpj CKO mice, BFABP expression is initiated in gliarestricted progenitors of the wild-type DRG ( Figure 8A) [30]. In contrast, BFABP expression was not detected in the DRG of Rbpj CKO mice at this stage ( Figure 8B). By E12.5, only a few BFABP-labeled cells were observed in Rbpj CKO ganglia, whereas many BFABP + cells were distributed throughout wild-type DRG ( Figure 8E,F). Thus, BFABP expression in Rbpj CKO DRG was delayed and drastically reduced as development progressed. P75 and Sox10 are co-expressed in multipotent NCCs at E8.5 to E9.5, but these genes gradually become expressed in distinct DRG progenitor pools as the DRG condenses; after E10.5, P75-expressing cells commit to the neuronal lineage while Sox10-expressing cells become glia [27]. To explore the mechanism underlying defective gliogenesis, we performed double immunolabeling of P75 and Sox10 in CKO mice. At E10.5, there was no apparent difference in P75 expression between wild-type and Rbpj CKO DRG, and similar numbers of P75/Sox10 co-labeled cells were observed in both genotypes ( Figure 8G-L,Y). At E11.5 and E12.5, few P75/ Sox10 co-labeled cells were observed in wild-type mice, revealing the segregation of P75 and Sox10 expression that occurs as DRG development progresses ( Figure 8M, O,Q,W,Y). In contrast, P75 expression was increased in Rbpj CKO mice at E11.5 and E12.5, and many P75/ Sox10 co-labeled cells were present, particularly at the DRG periphery ( Figure 8N,P,R,X,Y). Note that P75 and Sox10 localize to the cytoplasm and nucleus, respectively, and co-labeled cells are easily identified when imaged at high magnification ( Figure 8W,X). The presence of P75/Sox10-co-labeled cells suggests that the restriction and bifurcation of NCC fates from E10.5 fails to occur in Rbpj CKO DRG.

Discussion
In the present study, we specifically inactivated the critical transcription factor downstream of all four Notch receptors, Rbpj, in NCCs and their derivatives. Up-regulation of NeuroD1 and precocious neurogenesis were observed in the DRG of Rbpj CKO mice, followed by reduced proliferation and abnormal cell death. These phenotypes were not reported in a previous examination of Rbpj CKO (Wnt1-Cre;Rbpj flox/flox ) mice [26]. We confirmed previous findings revealing a near-complete loss of glia in Rbpj-deficient DRG [26], and further found that a large number of P75/Sox10 co-expressing NCCs were abnormally maintained in Rbpj CKO mice, suggesting that defective NCC development contributes to the loss of glia in these mice.

The role of Rbpj in DRG neurogenesis
We observed precocious neurogenesis in Rbpj-deficient DRG, a phenotype consistent with the known role of canonical Notch signaling in maintaining the undifferentiated state of neural progenitors by inhibiting the expression of genes involved in neuronal differentiation [16]. This finding is at odds with those of a recent study that reported there was no evidence of premature neuronal differentiation in Rbpj-deficient DRG, a conclusion based on Tuj1 and Peripherin expression patterns [26]. In the present study, we analyzed Islet1 and Brn3a expression to determine the timing and extent of neurogenesis in the DRG, and found that an elevated number of cells expressed these sensory neuronal markers between E10.5 and E11.5. Consistent with these observations, we found that BrdU incorporation in Rbpj-deficient DRG was less prevalent at this stage of development, indicating that NCC progenitor cells were prematurely differentiating into neurons. After E11.5, however, Islet1 and Brn3a staining suggested that the number of sensory neurons in Rbpj-deficient DRG was reduced. Therefore, we conclude that arresting Notch signaling in DRG NCCs removes critical inhibition, allowing them to differentiate into neurons at an exuberant rate during the early stages of DRG development, leading to depletion of the progenitor pool and an overall deficit of sensory neurons. Furthermore, the sensory neuron deficit in Rbpj-deficient DRG from E12.5 was also partly due to increased apoptosis, which may have occurred as a consequence of the uncoordinated development of the DRG.
To explore the mechanism underlying precocious neurogenesis, we examined the expression of several genes involved in the specification and differentiation of sensory neurons. Previous studies have shown that Notch signaling acts through the Hes genes, transcriptional corepressors activated by Notch signaling that inhibit the expression of proneural genes, such as Ngn1 and Ngn2. This repression, in turn, prevents the activation of neurogenic differentiation genes such as NeuroD [17,31]. However, we detected no difference in the expression of Ngn1 and Ngn2 in the DRG between Rbpj CKO and wild-type mice, but observed a great up-regulation of NeuroD1 in Rbpj-deficient DRG. These results suggest that Rbpj normally inhibits neuronal differentiation of NCCs in the DRG by repressing NeuroD1 expression via a novel mechanism, which is independent of Ngn1 and Ngn2. On the other hand, evidence from studies of retina development shows that NeuroD1 governs the neuron versus glial fate decision by promoting neurogenesis and suppressing gliogensis [32,33], and it is unclear whether up-regulation of NeuroD1 is also involved in defective gliogenesis in CKO mice (see below).
Mechanisms underlying severe gliogenesis defects in Rbpj CKO mice Premature neuronal differentiation, reduced cell proliferation, and abnormal cell death in progenitor cells were all observed in Rbpj-deficient DRG. However, these phenotypes cannot fully account for the severity of the observed defects in gliogenesis. Consistent with previous findings [26], we found that the initiation of gliaspecific expression of BFABP was delayed in Rbpj CKO mice, and expression in the relatively small population of cells was lost at later developmental stages. In wildtype NCCs, P75 and Sox10 are co-expressed early in DRG development, but as development progresses P75 becomes restricted to neuronal precursors while Sox10 becomes restricted to glial precursors [27]. Interestingly, we found that this separation did not occur in Rbpj-deficient NCCs, and P75/Sox10 co-expressing cells were still observed at E12.5. These results suggest that, in the absence of Rbpj, the subpopulation of NCCs that normally become restricted to glial fates fails to differentiate and instead maintains pluripotency, thus retaining the potential to differentiate into neurons. Taylor et al. [26] showed that these NCCs also maintain glial fate potential and can be induced to differentiate normally in vitro by stimulation with the gliogenic factor Neuregulin1-β1. Furthermore, BFABP is known to play a critical role in gliogenesis [34,35], Interestingly, the BFABP promoter contains an Rbpj binding site that is essential for BFABP transcription in radial glial cells [36], and thus it is likely that Rbpj promotes glial differentiation in the DRG by directly activating the transcription of BFABP.
The reduction in Sox10 expression was one of the earliest defects that we observed in Rbpj CKO mice, suggesting that Notch signaling is necessary for the maintenance of Sox10 expression. Because Sox10 maintains multipotency, inhibits neuronal differentiation of NCCs, and serves as a key regulator for peripheral glial development [9,37], we speculated that the reduction in Sox10 expression might contribute to defects in both gliogenesis and neurogenesis. Consistently, DRG phenotypes observed in Sox10 null mice [27,37] are very similar to those of Rbpj CKO mice. In addition, a study on the enteric nervous system proposed that, by suppressing proneural genes such as Mash1, Notch signaling might be required for continuous Sox10 expression and the maintenance of enteric neural crest progenitors [38].
In light of the up-regulation of NeuroD1 and the reduction of Sox10, we propose that Rbpj might maintain Sox10 expression by repressing NeuroD1. However, further studies are required to determine the relationships among them, as well as their combined effects on differentiation processes in precursor cells of the DRG.

Conclusions
We conditionally knocked out the transcription factor Rbpj, a master integrator of activation signals from all Notch receptors, in NCCs, and examined the role of Rbpj in early events of DRG development. Premature neuronal differentiation, reduced cell proliferation, and increased apoptosis in both progenitor cells and early differentiating neurons were all observed in Rbpj-deficient DRG. The up-regulation of NeuroD1 in the absence of Rbpj may lead to premature neuronal differentiation, and abnormal maintenance of stem cell potential by NCCs may contribute to the profound defects in gliogenesis as well as in neurogenesis in Rbpj CKO mice.

Mouse breeding and genotyping
Wnt1-Cre, Rbpj flox/flox and Rosa26 reporter mice were generated and genotyped as previously described [39][40][41]. To inactivate Rbpj expression in the neural crest, we crossed Wnt1-Cre mice with Rbpj flox/flox mice to obtain Wnt1-Cre;Rbp flox/+ progeny. Then Wnt1-Cre; Rbp flox/+ mice were crossed with each other to obtain Wnt1-Cre; Rbp flox/flox progeny. The morning of the day on which the vaginal plug appeared was designated as E0.5. In each set of experiments, at least three CKO embryos or pups in each experimental group and an equal number of littermate control mice (for example, wild-type, Rbpj flox/+ or Wnt1-Cre;Rbpj flox/+ ) were used. Animal care procedures were reviewed and approved by the Animal Studies Committee at the Tongji University School of Medicine, Shanghai, China.

BrdU labeling and TUNEL staining
For BrdU labeling, a single BrdU pulse (60 μg/g of body weight) was delivered intraperitoneally to timed-pregnant females at E10.5 and E11.5, and embryos were fixed 2 hours later. Sections were processed for BrdU and BFABP or Sox10 double immunostaining as described above. TUNEL staining was performed according to the In Situ Cell Death Detection Kit instructions (Roche, Indianapolis, USA).

Statistical analysis
For cell counts in E10.0, E10.5 or E11.5 DRG, we selected embryos with 26 to 28 somites as E10.0, and those with 36 to 38 somites as E10.5 [45]. In order to obtain transverse sections with similar angles, wild-type and mutant embryos at the desired stages with similar size were cut along the black line as shown in Figure  2N, and then the upper parts of embryos were positioned vertically (relative to the spinal neural tube) on the frozen pedestal prior to sectioning. Because DRG development shows big differences along the anteroposterior axis, counting data for comparison were collected from the sections at the thoracic level, which was determined by the appearance of the heart and liver. Six sets of consecutive 10-μm thick sections were made in each embryo, and cell counts were done in one set of sections that had been processed for in situ hybridization or immunostaining (see above). Approximately five sections on one slide were counted. Similarly, cell counts in post natal day 0 DRG were done on one set of consecutive 20-μm thick cryostat sections through lumbar DRGs. Cell counting was conducted by a co-author who did not know the genotyping data. For each set of comparisons, at least three CKO mice and three littermate controls (for example, wild-type, Rbpj flox/+ or Wnt1-Cre; Rbpj flox/+ ) were included. All data were analyzed using OriginPro7.5 [46] software and are presented as mean ± standard error of the mean. Comparisons were made using an unpaired Student's t-test and statistical significance was set at P < 0.05.

Additional material
Additional file 1: Reduced number of neurons and near-complete loss of glia in Rbpj-deficient DRG at E16.5. (A-D) Tuj1 (A,B) and BFABP (C,D) immunostaining of transverse sections through wild-type and Rbpjdeficient DRG at E16.5 with Hoechst counterstaining. Note that BFABP expression is present in the spinal cord, but not in the DRG (arrow) of Rbpj CKO mice. VH, spinal ventral horn. Scale bars: 100 μm.