AML1/Runx1 is important for the development of hindbrain cholinergic branchiovisceral motor neurons and selected cranial sensory neurons

Abstract

The mechanisms that regulate the acquisition of distinctive neuronal traits in the developing nervous system are poorly defined. It is shown here that the mammalian runt-related gene Runx1 is expressed in selected populations of postmitotic neurons of the embryonic central and peripheral nervous systems. These include cholinergic branchial and visceral motor neurons in the hindbrain, restricted populations of somatic motor neurons of the median and lateral motor columns in the spinal cord, as well as nociceptive and mechanoreceptor neurons in trigeminal and vestibulocochlear ganglia. In mouse embryos lacking Runx1 activity, hindbrain branchiovisceral motor neuron precursors of the cholinergic lineage are correctly specified but then fail to progress to a more differentiated state and undergo increased cell death, resulting in a neuronal loss in the mantle layer. In contrast, the development of cholinergic somatic motor neurons is unaffected. Runx1 inactivation also leads to a loss of selected sensory neurons in trigeminal and vestibulocochlear ganglia. These findings uncover previously unrecognized roles for Runx1 in the regulation of mammalian neuronal subtype development.

One of the critical events during nervous system development is the generation of distinct subclasses of neurons at precise locations and at defined times. In the developing mammalian CNS, cell cycle exit and activation of general neuronal traits act in concert. In addition, a coordinated acquisition of both general and subtype-specific neuronal features must be achieved to ensure the differentiation of different neuronal populations. Studies in the developing neural tube have identified several homeodomain (HD) transcriptional repressors that act in concert with specific basic helix–loop–helix transcription factors to regulate the differentiation and survival of individual neuronal subtypes (1–4). Much remains to be learned, however, about the identity of other proteins involved in regulating the differentiation of distinct neuronal subtypes in correct numbers and at different locations in the nervous system.

In this study, we have characterized the neural expression and function of the mammalian transcription factor AML1 (acute myeloid leukemia 1)/Cbfa2 (core binding factor α 2)/Runx1 (Runt-related transcription factor 1) (hereafter referred to as Runx1). The Runx protein family includes three members, Runx1–3, structurally and functionally related to the Drosophila protein Runt (5–9). In mice, individual Runx proteins act in mostly nonredundant manners to regulate a variety of cell differentiation events (6–8). In particular, Runx1 is required for fetal liver-derived hematopoiesis, and its human homolog is frequently targeted by chromosomal translocations that lead to acute myeloid leukemia (6, 7). In Drosophila, runt is involved in a number of developmental mechanisms, including neuronal development (5, 9). During insect embryonic neurogenesis, runt promotes the specification of a particular subset of CNS neurons, and its inactivation leads to a selective loss of those cells (5, 9). These findings, combined with the previous observation that murine Runx1 is expressed in embryonic neural tissues (10, 11), suggested that Runx1 may participate in mammalian neurogenesis. To date, this possibility had not been tested. Here we demonstrate for the first time that Runx1 activity is required for the development of selected populations of central and peripheral neurons, including cholinergic branchial and visceral motor neurons in the hindbrain and sensory neuron subtypes in trigeminal and vestibular ganglia. These results identify RUNX1 as a regulator of the development of particular motor and sensory neurons in the mammalian CNS and peripheral nervous system (PNS).

Materials and Methods

Embryological Analysis. Runx1^lacZ/+ mice were generated and genotyped as described (12). For staging of the embryos, the day of appearance of the vaginal plug was considered as embryonic day (E) 0.5. Embryos were dissected at various gestational stages; fixed in 2% paraformaldehyde, 10 mM sodium periodate, and 70 mM l-lysine; transferred to 30% sucrose; embedded in OCT compound (TissueTek); and cryostat sectioned (14 μm). β-Galactosidase (β-gal) activity was assessed as described (12), and counterstaining with eosin followed. Double-labeling immunofluorescence experiments (13) were performed with the following antibodies: mouse monoclonals against Nkx2.2 (1:15), Islet1 (1:75), Pax6 (1:15), HB9 (“Mnr2”, 1:5), Lim3 (1:5), or Lim1 + 2 (1:2) (obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA), β-gal (Promega; 1:250), Ki67 (BD Pharmingen; 1:100), or NeuN (Chemicon; 1:50); rabbit polyclonals against β-gal (Cappel; 1:5,000), calbindin D-28k (Chemicon; 1:250), tyrosine hydroxylase (TH) (Chemicon; 1:75), or retinaldehyde dehydrogenase 2 (RALDH2) (a kind gift of P. McCaffery, University of Massachusetts Medical School, Waltham, MA; 1:200); and goat polyclonal against choline acetyltransferase (ChAT) (Chemicon; 1:10). Double-labeling histochemical studies were performed by first subjecting the sections to incubation with rabbit antibodies against either Phox2b (1:700; ref. 14) or TrkA (1:200; ref. 15), followed by visualization with either the DAB or NovaRed substrate kits for peroxidase (Vector) and detection of β-gal activity.

Cell Counting. To count the numbers of Hoechst-stained nuclei, as well as the numbers of β-gal⁺, NeuN⁺, and TrkA⁺ cells in cranial ganglia from Runx1^lacZ/+ or Runx1^lacZ/rd embryos, >35 serial sections for each genotype (n = 4 embryos per genotype) were used. To count the numbers of Phox2b⁺ and β-gal⁺ cells in the hindbrain of the same embryos, >30 sections for each genotype (n = 3 embryos per genotype) were used. The numbers of Phox2b⁺ cells were counted in three separate domains of the hindbrain defined as follows: (i) ventromedial (a 35- × 40-μm area extending from the ventral ventricular zone (VZ) to the lateral mantle layer adjacent to the VZ), (ii) lateral (a 70- × 70-μm area comprising the mantle layer lateral to the ventromedial domain), or (iii) dorsolateral (a 30- × 50-μm area located dorsal to the lateral domain and extending above the dorsoventral boundary). The numbers of β-gal⁺ cells were counted in the ventromedial and lateral quadrants but not in the dorsolateral quadrant where β-gal is not expressed. Values are shown as mean ± SD.

Results

Expression of Runx1 in Selected Cranial Ganglion Sensory Neurons. To determine the neural cell types in which Runx1 is expressed, we used Runx1^lacZ/+ mice in which the β-gal gene was recombined into the Runx1 locus to produce a fusion protein of the N-terminal 242 aa of RUNX1 (containing a nuclear localization sequence) and β-gal (12). The expression of this nuclear fusion protein in heterozygous animals faithfully reproduces the expression of Runx1 transcripts (10–12). In agreement with previous in situ hybridization studies (10), we found that Runx1 was expressed in neither the CNS nor the PNS at E9.5 even though, as reported (10, 12), its expression was detected in the wall of the dorsal aorta (Fig. 1A and data not shown). Expression of β-gal in the PNS was first observed in a few trigeminal ganglion cells at approximately E10.5 (Fig. 1 B and C) and became robust in trigeminal ganglia by E12 (Fig. 1E). Expression was also observed in the vestibular (dorsolateral), but not cochlear (ventromedial), portion of vestibulocochlear ganglia at E12 (Fig. 1 F and G). In mice, vestibular neurons differentiate earlier (E10–E12) than cochlear neurons (E12–E14) (16, 17). These combined findings suggest that Runx1 expression is correlated with differentiated neurons and not neuronal progenitors. In agreement with this possibility, essentially all β-gal⁺ cells in vestibular ganglia also expressed calbindin D-28k, a marker of bipolar mechanoreceptor neurons (16) (Fig. 1G). Interestingly, not all calbindin D-28k⁺ cells were β-gal⁺, suggesting that Runx1 is not expressed in all vestibular neurons. Similarly, essentially all of the β-gal⁺ cells in trigeminal ganglia expressed the TrkA neurotrophin receptor, a marker of postmitotic nociceptive and thermoceptive neurons (18) (Fig. 1H). No β-gal expression was observed in proliferating cells expressing the mitotic cell marker protein Ki67 (data not shown). Together, these findings show that Runx1 is expressed in certain specific populations of postmitotic trigeminal and vestibular ganglion neurons.

Fig. 1.

Runx1 expression in selected cranial ganglion sensory neurons. (A–F) Histochemical detection of β-gal activity (blue) in near transverse sections through E9.5 (A), E10.5 (B and C), and E12 (E and F) Runx1^lacZ/+ embryos. (D) Schematic representation of the planes of the sections shown in E and F.(G) Combined double-label immunofluorescence analysis of β-gal (green) and calbindin D-28k (red) expression in vestibular ganglia. (H) Combined double-label analysis of β-gal activity (blue) and TrkA immunoreactivity (brown) in trigeminal ganglia at E12.5. (I–L) Expression of either NeuN (I and J) or Islet1 (K and L) in trigeminal ganglia of Runx1^lacZ/+ (I and K) or Runx1^lacZ/rd (J and L) embryos at E10.5. Fv, fourth ventricle; mye, myelencephalon; Rp, Rathke's pouch; tnr, trigeminal nerve rootlets; V, trigeminal ganglion; VIII, vestibulocochlear ganglion; vz, ventricular zone. [Scale bars = 1nm(D), 130 μm(E and F), and 30 μm(H).]

Involvement of Runx1 in the Development of Selected Trigeminal and Vestibular Sensory Neurons. Runx1^lacZ/+ mice were crossed to mice heterozygous for a disrupted Runx1 allele lacking coding sequences for the DNA-binding Runt domain (Runx1^rd/+ mice) (8). Similar to Runx1^rd/rd embryos, doubly heterozygous Runx1^lacZ/rd embryos lack Runx1 activity and die at ≈E12.5 because of impaired fetal liver-derived hematopoiesis (12). At E10.5, Runx1^lacZ/rd and Runx1^lacZ/+ embryos showed no differences in the number of trigeminal ganglion neurons expressing the general neuronal marker protein NeuN (Fig. 1 I and J) or the HD protein Islet1, which is expressed in essentially all ganglionic neurons at this stage (19) (Fig. 1 K and L). Further, we observed no differences in either TrkA or β-gal expression in mutant or control ganglia at E10.5 (data not shown). These results suggest that Runx1 inactivation does not perturb the differentiation of cranial sensory neurons at E10.5.

In contrast, at E11.5 we observed a significant decrease in both Islet1 (Fig. 2 A and B) and NeuN (Fig. 2D) expression in the vestibular ganglia of Runx1^lacZ/rd embryos compared with Runx1^lacZ/+ littermates. These changes were correlated with a considerable reduction in the number of β-gal⁺ cells in the same ganglia (Fig. 2C). Runx1^lacZ/rd embryos also displayed a significant decrease in the number of β-gal⁺ cells in trigeminal ganglia (Fig. 2E), and this decrease was correlated with a reduced number of TrkA⁺ cells (Fig. 2F). These combined findings strongly suggest that Runx1 inactivation causes a selected loss of the vestibular and trigeminal ganglion neurons in which Runx1 would normally be expressed. In agreement with this, Runx1-deficient embryos displayed a proportional decrease in both the total number of trigeminal ganglion cells (Fig. 2G) and the expression of the general neuronal marker NeuN (Fig. 2H). In contrast, no changes in Ki67 expression were observed in mutant trigeminal ganglia (Fig. 2 I and J), suggesting that Runx1 inactivation does not perturb the proliferation of ganglionic progenitors or their transition into sensory neurons.

Fig. 2.

Loss of cranial ganglion sensory neurons in E11.5 Runx1-deficient embryos. (A and B) Expression of Islet1 in vestibular ganglia of Runx1^lacZ/+ (A) or Runx1^lacZ/rd (B) embryos. (C–H) Cell counts of β-gal⁺ (C and E), NeuN⁺ (D and H), TrkA⁺ (F), or Hoechst⁺ (G) cells in equivalent transverse sections through vestibular (C and D) or trigeminal (E–H) ganglia of either Runx1^lacZ/+ or Runx1^lacZ/rd embryos. Data are shown as mean ± SD; *, P < 0.01. (I and J) Expression of Ki67 in trigeminal ganglia of either Runx1^lacZ/+ (I) or Runx1^lacZ/rd (J) embryos. The dotted lines define the area of the trigeminal ganglia; mitotic cells are also present in the tissue surrounding the ganglia. (K and L) Analysis of TUNEL⁺ cells in trigeminal ganglia of Runx1^lacZ/+ (K) or Runx1^lacZ/rd (L) embryos. (M and N) Phox2b expression in geniculate ganglia of Runx1^lacZ/+ (M) or Runx1^lacZ/rd (N) embryos. (Scale bar = 30 μm.)

To examine these phenotypes further, we asked whether the observed neuronal losses were correlated with increased numbers of dying cells. Terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL) labeling showed a significant increase in cell death in cranial ganglia of Runx1-deficient embryos (Fig. 2 K and L). To determine the specificity of these effects, we examined geniculate ganglia, where Runx1 is not expressed at E11.5 (data not shown). The expression of the paired-like HD protein Phox2b, which is specifically expressed in neurons of geniculate but not vestibulocochlear ganglia (14), was unchanged when Runx1 was inactivated (Fig. 2 M and N). Moreover, no increased numbers of TUNEL⁺ cells were observed in the geniculate ganglia of Runx1-deficient embryos (data not shown). These combined results show for the first time that RUNX1 is important for the postmitotic development of selected populations of cranial sensory neurons. They do not suggest that Runx1-deficient cells adopt an alternative neuronal fate(s) because such an event would be expected to leave the total number of neurons unchanged, a possibility that is not consistent with our demonstration of reduced neuronal cell numbers, as well as decreased total cell numbers, in the mutant ganglia. Instead, our combined findings suggest that Runx1 inactivation causes a loss of cranial sensory neurons in which this gene would have normally been expressed. This may be caused by an arrest of the postmitotic development of these cells, resulting in their elimination through apoptosis, or by compromised prosurvival mechanisms involving RUNX1.

Expression of Runx1 in Cholinergic Branchial and Visceral Motor Neurons of the Hindbrain. Runx1 expression was first detected in the CNS at approximately E10.5 (Fig. 3B). At this stage, β-gal activity was absent in the forebrain and midbrain (data not shown) but present in the caudal hindbrain where most β-gal⁺ cells were found in the ventral region of the mantle layer lateral to the VZ, with a few others located along columns ascending toward the dorsoventral boundary (Fig. 3 B–F). At E11.5, β-gal expression displayed an almost converse pattern characterized by a limited expression in the ventral domain and a gradually more pronounced expression along symmetrical trajectories that terminated at the outer edge of the mantle layer at the dorsoventral boundary (Fig. 3 G and I). This pattern persisted until the brainstem/cervical spinal cord junction (data not shown). These observations suggested that Runx1 might be expressed in hindbrain branchiomotor (bm) and visceromotor (vm) neurons. These cells derive from ventral progenitors that express the Nkx2.2 HD protein. bm/vm neurons initially appear in the ventral mantle layer and then migrate to more dorsolateral positions from where they innervate either muscles derived from the branchial arches or parasympathetic targets. bm/vm neurons express Phox2b as well as the Islet1, but not Islet2, HB9, or Lim3 HD proteins (2, 14, 20). In contrast, somatic motor (sm) neurons of the hindbrain derive from progenitor cells that express Pax6 and are located dorsal to Nkx2.2⁺ cells. These neurons occupy more ventral locations than bm/vm neurons and are characterized by the expression of Islet1, Islet2, HB9, and Lim3, but not Phox2b (2, 14, 20).

Fig. 3.

Runx1 expression in hindbrain branchial and visceral motor neurons. (A) Schematic representation of the plane of the sections shown in G–L.(B and G) Expression of β-gal in the hindbrain of either E10.5 (B) or E11.5 (G) Runx1^lacZ/+ embryos. (C and H) Combined double-label immunofluorescence analysis of β-gal and Nkx2.2 at either E10.5 (C) or E11.5 (H). (D–F) Double-labeling analysis of β-gal and Islet1 expression in the hindbrain at E10.5. (F) Combined β-gal and Islet1 staining. (I–L) Analysis of β-gal (blue) and Phox2b (pale red) expression performed either separately on adjacent sections (I and J) or together on the same section (K and L). The dotted line indicates the approximate location of the dorsoventral boundary. (L) Higher magnification view of the area boxed in K; most if not all Runx1⁺ cells are also Phox2b⁺ (arrowhead pointing to darkly stained cells), but a number of Phox2b⁺ cells do not express Runx1 (arrow pointing to pale red stained cells). (M–R) Double-labeling analysis of β-gal and HB9 expression at either E10.5 (M–O) or E11.5 (P–R). [Scale bars = 1mm(A), 130 μm(B and G), 50 μm(C–F and M–O), and 35 μm(H–L).]

Double-labeling studies showed that β-gal expression did not overlap with, but was immediately lateral to, the expression of Nkx2.2 at both E10.5 (Fig. 3C) and E11.5 (Fig. 3H), suggesting that Runx1 is expressed in postmitotic motor neurons derived from those progenitors. In agreement with this, essentially all β-gal⁺ cells in the hindbrain mantle layer expressed the general motor neuron marker protein Islet1 (Fig. 3 D–F) and the bm/vm neuron marker Phox2b (Fig. 3 I–L) at both E10.5 and E11.5. Importantly, not all Phox2b⁺ cells expressed β-gal, suggesting that Runx1 is expressed in a subset of hindbrain bm/vm neurons. In contrast, β-gal expression did not overlap with that of sm neuron markers like HB9 (Fig. 3 M–R) or Lim3 (data not shown) at either E10.5 or E11.5. Together, these results demonstrate that Runx1 is expressed in selected hindbrain bm/vm neurons.

To identify the specific bm/vm neuron subtypes in which Runx1 is expressed, double-labeling studies were performed by using antibodies against either ChAT, a marker of cholinergic neurons, or TH, a general marker of noradrenergic and dopaminergic neurons (21). Virtually all of the β-gal⁺ cells in the hindbrain mantle layer also expressed ChAT, indicating that they corresponded to cholinergic bm/vm neurons (Fig. 4 A–D, arrow in A). In contrast, β-gal was not expressed in ventral ChAT⁺ cells likely corresponding to sm neurons based on their location, ventral projections, and expression of sm neuron markers (Fig. 4 B and C, and data not shown). In addition, no evident overlap of β-gal and TH expression was detected, further suggesting that Runx1 is specifically expressed in cholinergic bm/vm neurons (Fig. 4 E–I). Consistent with these observations, at later gestational stages β-gal expression was observed in hindbrain motor nuclei that contain cholinergic bm/vm neurons, including the bm nucleus of the facial nerve, the vm dorsal motor nucleus of the vagal nerve, and the nucleus ambiguus (data not shown). Together, these findings show that in the hindbrain Runx1 is selectively expressed in postmitotic cholinergic bm/vm neurons.

Fig. 4.

Expression of Runx1 in selected hindbrain cholinergic motor neurons. (A–I) Double-label immunofluorescence analysis of β-gal, ChAT, and TH expression in the hindbrain of E11.5 Runx1^lacZ/+ embryos. (C and D) Combined β-gal and ChAT staining. (D) Higher magnification view of the area boxed in C.(G–I) Combined β-gal and TH staining. (H and I) Higher magnification views of the areas boxed in G.

Restricted Expression of Runx1 in Spinal Cord Motor Neurons. The second site of robust Runx1 expression in the CNS was at the brachial (C5–T1) level of the spinal cord, where two separate populations of β-gal⁺ cells were observed at E11.5. One group was located medioventrally (Fig. 6B, arrows, which is published as supporting information on the PNAS web site) and comprised cells that also expressed Islet1, HB9 (Fig. 6 C–J), and Lim3 (data not shown), indicating that they correspond to spinal cord motor neurons. Based on these immunological properties, their cell body position, and the fact that they did not express the RALDH2 protein (Fig. 6 K–N), these Runx1⁺ motor neurons likely correspond to medial constituents of the median motor column (MMC) (1, 2, 22). It is unlikely that these cells correspond to lateral MMC neurons because the latter are found only at thoracic levels (22) where Runx1 is not expressed (ref. 10 and data not shown). The second population of β-gal⁺ cells was located dorsolaterally (Fig. 6B, arrowheads) and was positive for both Islet1 and RALDH2 expression (Fig. 6 C–F and K–N), suggesting that these particular Runx1⁺ cells comprise lateral motor column (LMC) neurons (1, 22). Further, because we observed no overlapping immunoreactivity of β-gal and Lim1 + 2, a marker of lateral LMC neurons (1, 22) (Fig. 6P), it is likely that these cells correspond to medial LMC neurons. These findings show that Runx1 is expressed in restricted types of spinal motor neurons.

Comparison of the brachial spinal cord of E11.5 Runx1^lacZ/+ and Runx1^lacZ/rd littermates revealed no significant differences in the number of cells expressing β-gal, showing that Runx1 inactivation causes neither reduced β-gal expression nor a change in the number of β-gal⁺ cells (Fig. 6O). HB9 and RALDH2 expression also was not affected by Runx1 inactivation (data not shown). Further, we failed to detect any obvious differences in Lim1 + 2 immunoreactivity (Fig. 6 P and Q). Moreover, the number of TUNEL⁺ cells in the spinal cord of these embryos was essentially the same (Fig. 6 R and S). These combined findings suggest that Runx1 activity is not required for the development of the MMC and LMC neuron subtypes in which it is expressed at E11.5. Moreover, its inactivation does not appear to cause changes in cell fate choices in the spinal cord. Although it remains possible that Runx1 may play important roles during later stages of spinal motor neuron development, the early embryonic death of Runx1-deficient embryos has thus far precluded the analysis of this possibility.

Involvement of Runx1 in the Development of Hindbrain Cholinergic Branchial and Visceral Motor Neurons. At E10.5, no differences in the expression of β-gal (Fig. 5 A and B), Phox2b (which is expressed in both bm/vm progenitors and neurons at this stage) (Fig. 5 C and D), and Nkx2.2 (Fig. 5 E and F) were observed in the hindbrain of Runx1-deficient and control littermates, suggesting that Runx1 is not important for early phases of bm/vm neuron differentiation.

Fig. 5.

Loss of cholinergic branchiovisceral motor neurons in the hindbrain of Runx1-deficient embryos. (A–F) Expression of β-gal (A and B), Phox2b (C and D), or Nkx2.2 (E and F) in the hindbrain of either Runx1^lacZ/+ (A, C, and E) or Runx1^lacZ/rd (B, D, and F) embryos at E10.5. (G) Schematic representation of the planes of the sections shown in the indicated panels. (H–K) Expression of Phox2b (H and I) or β-gal (J and K) in the hindbrain of either Runx1^lacZ/+ (H and J) or Runx1^lacZ/rd (I and K) embryos at E11.5. Notice the decreased β-gal expression in both the hindbrain mantle layer and the vestibular ganglia (VIII) of Runx1^lacZ/rd embryos. (L) Quantitative analysis of Phox2b expression in the ventromedial (VMQ), lateral (LQ), or dorsolateral (DLQ) quadrants of the hindbrain of Runx1^lacZ/+ (bars 1, 3, and 5) or Runx1^lacZ/rd (bars 2, 4, and 6) embryos. (M) Quantitative analysis of β-gal expression in the VMQ and LQ of the hindbrain of Runx1^lacZ/+ (bars 1 and 3) or Runx1^lacZ/rd (bars 2 and 4) embryos. (L and M) Results of cell counts in each quadrant are depicted as mean ± SD; *, P < 0.01. (N–Q) Analysis of the hindbrain of Runx1^lacZ/+ (N and P) or Runx1^lacZ/rd (O and Q) embryos for expression of β-gal (N and O) or ChAT (P and Q). (P) Arrow points to ChAT⁺ cells that are lost in Runx1-deficient embryos; arrowhead points to ChAT⁺ cells that are not affected by Runx1 inactivation. (R–U) Analysis of TUNEL⁺ cells in either the hindbrain (R and S) or the forebrain (T and U) of Runx1^lacZ/+ (R and T) or Runx1^lacZ/rd (S and U) embryos. The areas stained in R and S correspond to the boxed regions in N and O. (U) * indicates intraventricular blood cells. tv, telencephalic vesicle; VIII, vestibular ganglion; vz, ventricular zone. [Scale bars = 1mm(G), 90 μm(H–K), 80 μm(N–Q), and 25 μm(T and U).]

At E11.5, Runx1^lacZ/rd embryos continued to exhibit no statistically significant difference in Phox2b expression in a ventromedial quadrant of the hindbrain encompassing the ventral VZ and the mantle layer immediately lateral to the VZ and containing mostly neuron progenitors and premigratory precursors (Fig. 5 H, I, and L). Moreover, we observed essentially equivalent numbers of Nkx2.2⁺ cells in this ventromedial quadrant in Runx1^lacZ/+ (219 ± 47 per section; >15 sections; n = 3 embryos) and Runx1^lacZ/rd (227 ± 46 per section; >15 sections; n = 3 embryos) littermates. These combined observations strongly suggest that the inactivation of Runx1 does not perturb the initial generation of those Phox2b⁺ neurons in which Runx1 is expressed.

In contrast, E11.5 Runx1^lacZ/rd embryos displayed a significant reduction in the number of Phox2b⁺ cells in a lateral quadrant of the mantle layer that contains more developmentally mature bm/vm neurons that have migrated from their place of birth (Fig. 5 H, I, and L). The decrease in Phox2b expression was specifically observed in this lateral region of the hindbrain, where Runx1 and Phox2b expression overlaps (Fig. 3). In contrast, we detected no differences in Phox2b expression in a more dorsolateral quadrant where Runx1 is not expressed (Fig. 5 H, I, and L). We therefore examined whether this specific loss of Phox2b⁺ cells was correlated with a loss of Runx1⁺ bm/vm neurons. Cell counting studies in the ventromedial (where small numbers of β-gal⁺ cells are found) and lateral (where the majority of Runx1⁺ cells are located) quadrants of the hindbrain of E11.5 embryos revealed a significant decrease in the number of β-gal⁺ cells in Runx1-deficient embryos (Fig. 5 J, K, and M). These findings suggest that Runx1 inactivation causes a specific loss of Phox2b⁺/Runx1⁺ bm/vm neurons. To test this possibility further, we compared ChAT expression in mutant and control embryos. A significant decrease in both ChAT (Fig. 5 P and Q, arrow) and β-gal (Fig. 5 N and O) expression was detected in the lateral quadrant of the hindbrain, consistent with a loss of Runx1⁺/ChAT⁺ bm/vm neurons. In contrast, no significant changes in ChAT expression were detected at more ventral locations of the hindbrain (Fig. 5 P and Q, arrowhead), where ChAT⁺ cells do not express Runx1 (Fig. 4).

The decreased expression of β-gal in the lateral quadrant of the hindbrain of Runx1-deficient embryos was not correlated with ectopic β-gal expression in ventral territories occupied by sm neurons (Fig. 5 J, K, N, and O). Similarly, neither an ectopic expression of Phox2b (Fig. 5 H and I) nor an increased expression of ChAT (Fig. 5 P and Q) were observed in the same ventral regions. These combined observations strongly suggest that Runx1 inactivation causes neither a rerouting of bm/vm neurons to ventral domains nor a conversion to an alternative neuron cell fate. Instead, Runx1^lacZ/rd embryos displayed an increase in the number of TUNEL⁺ cells near the boundary between the ventromedial and lateral quadrants (Fig. 5 R and S), suggesting a correlation between increased rate of cell death and the loss of β-gal⁺ cells. This increase in TUNEL⁺ cells was specific to this particular region because we did not detect changes in TUNEL staining in either the spinal cord (Fig. 6 R and S) or the forebrain (Fig. 5 T and U). Together, these results show that Runx1 plays an important role in the postmitotic development and/or survival of hindbrain cholinergic bm/vm neurons.

Discussion

In this study, we have demonstrated that Runx1 function is important for the postmitotic development of specific neuronal subtypes in the CNS and PNS. In both the hindbrain and spinal cord, Runx1 is expressed in selected postmitotic neurons of the mantle layer and not in neural progenitors of the VZ. Within the mantle layer, Runx1 expression is restricted to cholinergic bm/vm neurons in the hindbrain and certain MMC and LMC motor neurons in the spinal cord. Similarly, in both trigeminal and vestibular ganglia, Runx1 expression is correlated with phases of active neuronal differentiation and coincides with nociceptive/thermoceptive and mechanoreceptor neurons, respectively. Apart from these cells, Runx1 expression was observed in only a few other neuron types, including nociceptive, but not proprioceptive, sensory neurons in dorsal root ganglia and restricted neuron subtypes in the retina (refs. 10 and 12 and data not shown).

Our studies have shown further that Runx1 expression only becomes detectable in hindbrain bm/vm neurons at approximately E10.5, even though these neurons already begin to be generated at roughly E9.5 (20, 21). At E11.5, when hindbrain cholinergic bm/vm neuron production has essentially ceased (20, 21), only small numbers of Runx1⁺ cells are found in the most ventral aspect of the hindbrain and most of them are located dorsolaterally, likely as a result of the migration of the earlierborn cells. These combined observations show that Runx1 expression is activated after the initial generation of specific bm/vm neurons and suggest that this gene is involved in the developmental maturation of these cells. In agreement with this possibility, we have shown that although Runx1 inactivation does not cause a detectable loss of hindbrain bm/vm neurons at E10.5, a significant loss of bm/vm neurons in which Runx1 is normally expressed is observed in the hindbrain mantle layer at E11.5. This neuronal loss is specific because a number of dorsolateral Phox2b⁺ neurons that do not express Runx1 are not affected by the disruption of Runx1 function. Moreover, ChAT⁺/Runx1^– neurons located in the ventral hindbrain, likely corresponding to sm neurons, are also spared. In addition, no detectable differences in the numbers of neurons expressing TH are observed among Runx1-deficient and control embryos (data not shown), suggesting further that Runx1 inactivation perturbs only the development of those bm/vm neurons in which it is expressed. A similar situation was observed in trigeminal and vestibular ganglia, where disruption of Runx1 function also causes no detectable effect at E10.5 but results in a significant loss of sensory neurons at E11.5. Importantly, no loss of sensory neurons was observed in geniculate ganglia, which are adjacent to vestibulocochlear ganglia but do not express Runx1 at E11.5.

The observed decrease in Runx1⁺ bm/vm neurons in the lateral hindbrain does not appear to be the consequence of their abnormal migration to ventral regions because we detected neither an ectopic expression of β-gal and Phox2b nor an increased number of ChAT⁺ neurons in ventral territories where sm neurons are located. Further, we failed to detect an increase of either TH⁺ bm/vm neurons or ventral sm neurons in the hindbrain of Runx1-deficient embryos. Similarly, we found no evidence that Runx1-deficient cells in trigeminal and vestibular ganglia adopt alternative neuronal fates and instead observed that the expression of a number of different neuronal markers is reduced in the mutant ganglia. Together, these results strongly suggest that in both the neural tube and cranial ganglia Runx1 is involved in mechanisms important for the postmitotic development of the restricted neuronal populations in which it is expressed. They suggest further that Runx1-inactivation is not correlated with changes in neuronal fates but rather with a reduced development/survival of those cells.

Runx1 may be directly involved in promoting cell survival. This possibility is in agreement with our finding that in both cranial ganglia and hindbrain, the neuronal losses associated with Runx1-inactivation are correlated with increased cell death. These effects are specific and not simply a consequence of the hematopoietic deficiency of Runx1-deficient embryos, because no differences in TUNEL labeling were observed in other regions of the neural tube where either Runx1 expression is not detected, like the forebrain, or Runx1 is expressed but its inactivation does not cause a detectable phenotype, like the spinal cord. A prosurvival role for Runx1 would also be consistent with previous studies showing that its overexpression in T hybridoma cells results in an up-regulation of the antiapoptotic gene Bcl-2 and renders those cells resistant to apoptosis mediated by the T cell receptor (23). Alternatively, or in addition, Runx1 may be involved in regulating the expression of genes that define the particular phenotypic traits of those neuron subtypes, and its inactivation may cause a developmental arrest eventually resulting in the elimination of those cells through apoptosis. An involvement of Runx1 in cranial sensory neuron differentiation is consistent with recent studies showing that Runx1 and the related gene Runx3 are expressed in separate dorsal root ganglion neuron subtypes. Runx1 is expressed in nociceptive neurons whereas Runx3 is found in proprioceptive neurons (12, 15, 24). Importantly, Runx3 inactivation causes a perturbation of the development of dorsal root ganglion proprioceptive, but not nociceptive, neurons (15, 24). Future studies aimed at further elucidating the involvement of Runx1 in cell survival and/or differentiation events in selected sensory and motor neurons will provide important information that will help clarify mechanisms underlying the development of neuronal subtypes in the mammalian CNS and PNS.