The transcription factor Hmx1 and growth factor receptor activities control sympathetic neurons diversification

Abstract

The sympathetic nervous system relies on distinct populations of neurons that use noradrenaline or acetylcholine as neurotransmitter. We show that fating of the sympathetic lineage at early stages results in hybrid precursors from which, genetic cell-lineage tracing reveals, all types progressively emerge by principal mechanisms of maintenance, repression and induction of phenotypes. The homeobox transcription factor HMX1 represses Tlx3 and Ret, induces TrkA and maintains tyrosine hydroxylase (Th) expression in precursors, thus driving segregation of the noradrenergic sympathetic fate. Cholinergic sympathetic neurons develop through cross-regulatory interactions between TRKC and RET in precursors, which lead to Hmx1 repression and sustained Tlx3 expression, thereby resulting in failure of TrkA induction and loss of maintenance of Th expression. Our results provide direct evidence for a model in which diversification of noradrenergic and cholinergic sympathetic neurons is based on a principle of cross-repressive functions in which the specific cell fates are directed by an active suppression of the expression of transcription factors and receptors that direct the alternative fate.

Introduction

The paravertebral sympathetic nervous system is organized as a chain of ganglia along the rostro-caudal axis of the vertebra receiving preganglionic innervation from central nervous system neurons located in the spinal cord. Most sympathetic neurons differentiate into noradrenergic neurons, which innervate internal organs regulating their function, for instance gut motility, salivation, piloerection, pupil diameter and vasoconstriction. Some sympathetic neurons instead differentiate into cholinergic neurons. Sympathetic cholinergic neurotransmission is believed to play unique roles compared to the noradrenergic, innervating the periosteum (Asmus et al, 2000), and is particularly important during thermoregulation by innervating sweat glands and stimulating perspiration. In mammals, apart from rodents, cholinergic sympathetic vasodilator nerves also innervate muscle vasculature (Morris et al, 1998; Ernsberger and Rohrer, 1999), which actively redistributes oxygen and nutrients to the working muscle (Anderson et al, 2006).

A sympathetic fate defined by a regulatory network of transcription factors is induced by bone morphogenetic proteins (BMPs) expressed in the dorsal aorta acting on migrating neural crest cells (NCCs; Schneider et al, 1999). The paired-like homeodomain transcription factor PHOX2B and the bHLH protein HAND2 are required for defining a noradrenergic sympathetic fate (Hirsch et al, 1998; Lo et al, 1998; Pattyn et al, 1999; Howard, 2005; Lucas et al, 2006). PHOX and HAND factors drive expression of the noradrenaline biosynthesis enzymes tyrosine hydroxylase (TH) and dopamine-β-hydroxylase (DBH) in sympathetic precursors. Thus, development of the sympathetic lineage is governed by an early gene-regulatory network likely defining early aspects of sympathetic neuron precursors (Apostolova and Dechant, 2009). In later development, diversification of sympathetic neurons into the distinct functional types is believed to be governed by extrinsic soluble signals (Habecker and Landis, 1994; Asmus et al, 2001). However, the transcriptional regulators and their mechanistic interactions with environmental signals during diversification of sympathetic neurons into the distinct types remain largely unknown.

The H6 homeobox gene Hmx1 (Nkx5-3) caught our attention as a possible candidate for regulation of cell type specification in the neural crest, as initiation of its expression coincides with neurogenesis and becomes restricted to only a few neural crest-derived neurons later in development (Yoshiura et al, 1998; Adameyko et al, 2009; Munroe et al, 2009). A mutation in Hmx1 causes the autosomal recessive Oculo-Auricular syndrome in humans, characterized by malformation of the eyes and external ears (Schorderet et al, 2008), and Dumbo mice containing a mutation in Hmx1 substantially recapitulate the human condition (Munroe et al, 2009). Here, we report that induction of the sympathetic fate results in a hybrid precursor population that diversifies into distinct neuronal types by initiation of Hmx1 expression in a mechanism involving the activity by different classes of growth factor receptors.

Results

Hierarchical development of sympathetic subtypes

While expression of noradrenergic traits in developing sympathetic precursors is well documented, direct quantitative measurements of segregation of noradrenergic and cholinergic fates in early sympathetic precursors remain to be described (Goridis and Rohrer, 2002; Howard, 2005; Apostolova and Dechant, 2009). To understand if mechanisms operating during sympathetic neuron consolidation into distinct types encompass the extinction of phenotypes already existing in precursors, we first examined noradrenergic and cholinergic associated markers in E12.5 paravertebral thoracic level 9–13 sympathetic ganglia (SG) neurons. The enzymes in noradrenalin synthesis TH, DBH, vesicular monoamine transporter (VMAT2), the enzyme for acetylcholine biosynthesis (choline acetyltransferase (ChAT)), the vesicular acetylcholine transporter (VAChT), the glial cell line-derived factor (GDNF) family ligand receptor RET, and the neurotrophin-3 (NT3) receptor, TRKC, as well as peripherin (PRPH) were abundantly present (Figure 1A). All ISL1⁺ precursors contained expression of these markers at E12.5 (Figure 1A). Expression of TrkC in the precursors started to be downregulated at E14.5 and was largely mutually exclusive with RET and TRKA in sympathetic neurons at E15.5 (Supplementary Figure S1).

Figure 1.

Sympathetic precursors of hybrid phenotype diversify into distinct neurotransmitter subtypes. (A) E12.5 sympathetic precursors display a mixed phenotype, expressing genes associated with both noradrenergic and cholinergic sympathetic neuron phenotypes. Double immunostaining for indicated proteins on E12.5 sympathetic ganglia (SG) sections. Section for VMAT2 is the same as VAChT, with red channel converted to green. (B) Association of markers with noradrenergic and cholinergic sympathetic neurons at P5. Note co-expression of noradrenergic (TH, VMAT2, DBH, NPY and TRKA, upper panels) and cholinergic (RET, VAChT, PRPH, VIP and SST, lower panels) markers. (C–F) Diversification of the sympathetic lineage as revealed by Ret, TrkA, VAChT and Vmat2 expression during development. (C, D) Double immunostaining for RET/TRKA (C) and VAChT/VMAT2 (D) on SG sections from indicated ages. Note the onset of TrkA expression at E14.5 and the concomitant downregulation of Ret (C, inset) and VAChT (D). Sporadic hybrid neurons expressing both Ret and TrkA are observed until P5 (C, arrowheads). (E, F) Quantification of (C) and (D). Results are represented as mean±s.e.m. n=10–26 ganglia from at least two animals for each stage. (G, H) E11.5 RET⁺ SG neurons contribute to the TRKA⁺, RET⁺ and hybrid RET⁺/TRKA⁺ neurons at E18.5. (G) Double immunostaining for RET and TRKA on SG sections at E18.5. Note TOMATO⁺ cells (TOM) in both the RET⁺ and TRKA⁺ lineages. (H) Quantification of (G). Scale bar in all images represents 50 μm.

Analysis of P5 SG revealed that VAChT⁺ neurons segregated with RET, PRPH, Vasoactive Intestinal Peptide (VIP) and Somatostatin (SST), while TrkA expression was initiated in VMAT2⁺ neurons (Figure 1B). These neurons also contained DBH and some, neuropeptide Y (NPY) (Figure 1B). Quantification of the number of neurons expressing Vmat2, TrkA, VAChT and Ret throughout development was performed to resolve the exact developmental sequence of sympathetic neuron diversification (Figures 1C–F). While RET was rapidly extinguished in most neurons between E14.5 and E15.5, TRKA⁺ neurons first appeared at E14.5 and increased progressively until E18.5. Approximately half of the TRKA⁺ neurons at E14.5 were RET⁺/TRKA⁺ hybrid after which RET was rapidly segregated, and only 2.2±0.6% of the TRKA⁺ neurons remained as a hybrid population at P5 (Figure 1E). Also Vmat2 and VAChT, expressed in nearly all precursors at early stages, rapidly segregated between E14.5 and E15.5 (Figures 1D and F). Hence, already at E18.5 a noradrenergic TRKA⁺/VMAT2⁺/RET⁻ and cholinergic VAChT⁺/RET⁺/PRPH⁺ population were nearly completely segregated by a mechanism involving extinction of TRKC in most segregated neurons, and in noradrenergic neurons maintenance of TH, VMAT2 and DBH, induction of TRKA combined with extinction of ChAT, VAChT and PRPH while cholinergic neurons emerged through maintenance of ChAT, VAChT, RET and PRPH and extinction of TH, VMAT2 and DBH.

To directly prove that both noradrenergic and cholinergic sympathetic neurons arise from a common pool of RET⁺ precursors, genetic-based cell lineage tracing experiments were performed using mice heterozygous for an inducible CRE (CreERT2) in the Ret locus (Ret^ERT2) (Luo et al, 2009) crossed to the Gt(ROSA)26^{tm9(CAG-tdTomato)} strain (R26^TOMATO). Tamoxifen was administered at E11.5 and the animals were sacrificed at E18.5 for analysis. Analysis of TOMATO⁺ cells expressing TrkA, Ret or both, revealed 66.4, 25.7 and 7.9% of traced cells, respectively (Figures 1G and H). This shows that early RET⁺ sympathetic precursor cells are the cellular origin of both noradrenergic and cholinergic sympathetic neurons. Taken together, the above results suggest a defined organization of segregation which is intimately associated with expression of TrkA and Ret from an early precursor population.

Acquisition of Hmx1 expression defines the major population of noradrenergic neurons

An Hmx1 mouse allele containing LacZ fused to the start codon of Hmx1 (Hmx1^LcZ mice, Supplementary Figure S2) was used for β-galactosidase immunohistochemistry to quantify Hmx1 expressing neurons during sympathetic neuron differentiation. HMX1 first appeared in very few cells in the ganglia at E13.5, which increased in numbers until E18.5 and remained at a similar level at P5 (Figure 2A). HMX1 was strongly associated with TrkA expressing neurons throughout development. However, prior to TrkA expression, that is, at early stages such as E13.5–E14.5, Hmx1 was initiated in RET⁺ precursors. Ret was then rapidly extinguished in HMX1⁺ neurons, since already at E15.5, Ret and Hmx1 expression were largely mutually exclusive (Figure 2B). Hmx1 expression clearly preceded both the upregulation of TrkA and the downregulation of Ret (Figures 2A, insets and C). Consistently, expression of Hmx1 was initiated in TRKC⁺ precursors at E13.5 fated to the noradrenergic lineage of sympathetic neurons (Supplementary Figure S3). In agreement with this, Hmx1 was expressed in VMAT2⁺ neurons but was never present in RET⁺/VAChT⁺ neurons at E15.5 (Supplementary Figure S4A and B) and P5 (Supplementary Figure S5K). A proposed hierarchical organization of the development of sympathetic subtypes is shown in Figure 2D.

Figure 2.

Expression of the homeobox transcription factor Hmx1 defines the major population of noradrenergic neurons. (A) Triple immunostaining for Hmx1-βgal, RET and TRKA on SG section of indicated ages in Hmx1^LcZ/+ or Hmx1^LcZ/+;RET^CFP/+mice. Note the onset of Hmx1 expression, revealed by βgal expression, at E13.5 in RET⁺ neurons (inset). At E14.5, βgal⁺/TRKA⁺/RET⁻, βgal⁺/TRKA⁻/RET⁺ and βgal⁺/RET⁺/TRKA⁺ (inset) neurons are observed. At E15.5, most neurons co-express βgal and TrkA (inset) while RET⁺ neurons are βgal⁻. At E18.5 and P5, nearly all neurons in the ganglion are either βgal⁺/TRKA⁺ or only RET⁺ (inset). Scale bar in all images represents 50 μm. (B) Quantification of (A). Results are represented as mean±s.e.m. n=6–26 ganglia from at least 2 animals for each stage. (C) Graph showing proportion of RET⁺, βgal⁺, TRKA⁺ and RET⁺/TRKA⁺ neurons in ISL1⁺ neurons throughout development. Note Hmx1 preceding TrkA expression by about 1 day. (D) Schematic representation of the different main lineages with indicated percent of their contribution to all neurons at each stage during the course of sympathetic neuron differentiation in mouse. Hmx1 expression predefines a noradrenergic fate.

Hmx1 consolidates the sympathetic noradrenergic fate

The strict relation of HMX1 to the TRKA⁺/VMAT2⁺/RET⁻ neuronal population led us to generate an Hmx1 conditional null mutant allele lacking the homeobox domain and the 3′ untranslated region (UTR) (Figure 3A; Supplementary Figure S5A and G). The conditional allele was crossed to Wnt1-Cre mice to delete Hmx1 in neural crest-derived cells, including sympathetic neurons (Hmx1^fl/fl;Wnt1-Cre mice). In situ hybridization for Hmx1 confirmed the efficient deletion of Hmx1 in sympathetic neurons in the Hmx1^fl/fl; Wnt1-Cre mouse strain (Supplementary Figure S5H).

Figure 3.

Consolidation of noradrenergic and suppression of cholinergic traits by Hmx1. (A) Schematic illustration of the Hmx1 allele in conditional mutant mice. Exon 2 encoding amino acids 124–332 of Hmx1, including the homeobox domain, was flanked by LoxP sites. (B, C) Immunostaining for TRKA, RET, TH, VIP and SST on SG sections from control and Hmx1^fl/fl;Wnt1-Cre animals at P0 (B) and for RET, TRKA and TH at E15.5 (C). (B) In the absence of HMX1, the noradrenergic genes TrkA and Th are markedly decreased, while cholinergic-associated genes including Ret, Vip and Sst fail to be repressed in P0 mutant animals. Phox2a/b expression is unchanged at P0 (C). Double staining for RET and TRKA at P0 reveals that remaining TRKA⁺ neurons fail to repress Ret (B, arrowheads). (C) Ret downregulation and TrkA induction fail as early as E15.5 in the absence of HMX1. Furthermore, TH expression levels are decreased in Hmx1^fl/fl;Wnt1-Cre animals at E15.5. Scale bar in all images represents 50 μm.

Expression of TrkA was markedly absent in the mutant at E15.5, the critical stage for acquisition of TrkA expression (46.5±2.5% positive neurons in wild-type and 9.9±1.0% positive neurons Hmx1^fl/fl;Wnt1-Cre mice) and was not recovered by P0 (88.1±1.9 wild-type and 28.3±2.0% Hmx1^fl/fl;Wnt1-Cre mice) (Figure 3B and C). Analysis of RET and TRKA double-stained neurons from P0 Hmx1^fl/fl; Wnt1-Cre mice showed that most or all of the few neurons expressing TrkA independent of Hmx1 were hybrid RET⁺/TRKA⁺ neurons (Figure 3B). While some markers, including DBH and VMAT2 (Supplementary Figure S5J), were not regulated by HMX1 at P0, HMX1 was found to be critical for the expression of the noradrenergic neurotransmitter phenotype, since the number of TH⁺ neurons was also dramatically reduced at P0, from 93.3±0.8% in wild-type mice to 27.9±2.7% in Hmx1^fl/fl;Wnt1-Cre mice (Figure 3B). Interestingly, Th was expressed at E15.5 in similar number of cells in Hmx1^fl/fl;Wnt1-Cre mice as in control (Hmx1^fl/fl) mice, although at lower levels (Figure 3C). This shows that HMX1 is necessary around E15 to consolidate a noradrenergic phenotype in precursors that segregate into noradrenergic sympathetic neurons by mechanisms of induction and maintenance of noradrenergic lineage-specific genes TrkA and Th, respectively. Although neuronal numbers were increased in Hmx1^−/− mice at P0 (Supplementary Figure S5I), arrector pili muscle and cutaneous blood vessel innervation by PGP9.5⁺ and TH⁺ noradrenergic fibres was markedly reduced at P19 (Supplementary Figure S7), consistent with a requirement of NGF signalling via TRKA for a full innervation of target tissues. Analysis of RET⁺ neurons in Hmx1^fl/fl; Wnt1-Cre mice revealed a complete failure of Ret suppression, as most or all of the sympathetic neurons expressed Ret at P0 (Figure 3B). Similar results were obtained when analysing E15.5 Hmx1^fl/fl;Wnt1-Cre mice (Figure 3B and C). In addition to Ret, Vip and Sst, normally expressed in the RET⁺ cholinergic sympathetic neurons, also failed to be repressed in the noradrenergic sympathetic neurons (Figure 3B), while ChAT and VAChT were unaffected (Supplementary Figure S5J). In Hmx1^LcZ/fl;Wnt1-Cre mice, which allow monitoring of cells that should normally express Hmx1, the absence of HMX1 led to Ret expression in β-gal⁺ cells while, in control animals, Ret and β-gal expression was mutually exclusive (Supplementary Figure S5K). Thus, this conclusively shows a requirement for HMX1 for suppression of cholinergic traits in noradrenergic sympathetic neurons.

Th expression is governed by PHOX2B and HAND2 in sympathetic precursors, consistently, in Phox2b and Hand2 null mutant mice, Th expression is absent in sympathetic precursors (Pattyn et al, 1999; Morikawa et al, 2007). Hmx1^fl/fl;Wnt1-Cre mice did not display any deficits of Phox2a/b expression (Figure 3C), indicating that early expression of Th in sympathetic precursors is governed by these transcription factors independently of HMX1. Hence, although the previously identified gene-regulatory network is sufficient for inducing a sympathetic lineage with noradrenergic traits, consolidation of this phenotype and repression of other fates within the major population of noradrenergic sympathetic neurons requires HMX1.

TrkC biases segregation into cholinergic sympathetic neurons

TrkC is expressed in early sympathetic precursors (Ernfors et al, 1992). We find that Hmx1 is initiated within the TRKC⁺ precursors at E13.5 (Supplementary Figure S3), but is largely mutually exclusive with HMX1 at E15.5 through an extinction of TrkC expression in neurons attaining HMX1 (Figure 4A and B), consistent with the findings that TrkC is expressed in precursors, since TRKC⁺ neurons were rarely positive for RET or TRKA at E15.5 (Supplementary Figure S1). TrkC^−/− and TrkC^−/−;Bax^−/− mice were analysed to address the role of TRKC during diversification of sympathetic neurons into distinct types. TrkC^−/−;Bax^−/− mice were used to ascertain analysis of survival-independent functions of TRKC (Patel et al, 2000). Lack of TRKC led to a marked de-repression of Hmx1 expression (Figure 4C). Consistently, mice lacking TRKC displayed a large reduction in RET⁺ neurons at E15.5 (TrkC^−/−;Bax^−/−) and E18.5 (TrkC^−/−) (Figure 4D, F and G). The reduction in RET⁺ neurons was paralleled by an increase in TRKA⁺ neurons (Figure 4D, F and G). This switch in receptor expression represented a change in fate, since ChAT⁺ neurons decreased while TH⁺ neurons increased (Figure 4E and G) to a similar extent as RET⁺ and TRKA⁺ neurons. Hence, a de-repression of Hmx1 in TrkC^−/−;Bax^−/− mice leads to induction of TrkA and repression of Ret that participate in the failure of diversification of cholinergic sympathetic neurons in the absence of TrkC.

Figure 4.

A failure of Hmx1 repression and development of cholinergic neurons in mice lacking TrkC. (A) Triple immunostaining for TRKC, Hmx1-βgal and ISL1 on SG section of E15.5 Hmx1^LcZ/+ embryos. Note βgal⁺/TRKC⁺ neurons (arrowheads), βgal⁺/TRKC⁻ (double arrowheads), βgal⁻/TRKC⁺ (arrows) and βgal⁻/TRKC⁻ (asterisk) neurons in the E15.5 SG. Scale bar represents 20 μm. (B) Quantification of (A). (C) In situ hybridization for (mRNA) Hmx1 on SG section of a P0 TrkC^−/−;Bax^−/− animal. Note increased expression of Hmx1 mRNA in the absence of TRKC. Scale bar represents 50 μm. (D) Triple immunostaining for TRKA, RET and ISL1 on SG sections of E15.5 WT and TrkC^−/−;Bax^−/− embryos. Note the decreased and increased proportion of RET⁺ neurons and TRKA⁺ neurons, respectively, in the absence of TRKC. Scale bar represents 20 μm. (E, F) Double immunostaining for TH and ChAT (E) and for TRKA and RET (F) on SG sections of E18.5 WT and TrkC^−/− embryos. Note the decreased number of ChAT⁺ (E) and RET⁺ (F) neurons. Scale bar represents 20 μm. (G) Quantification of (C–F). Results are represented as mean±s.e.m. n=9–26 ganglia from at least 2 animals for each stage. Statistical significance was determined using Student’s t-test. *P<0.05, **P<0.01, ***P<0.0001.

A Ret-dependent segregation of sympathetic neurotransmitter fates

In Ret^−/− mice, ChAT and VAChT fail to be expressed at E16 (Burau et al, 2004). Hence, the marked loss of RET⁺ and ChAT⁺ neurons and increase in TRKA⁺ and TH⁺ neurons in mice lacking TRKC suggest that TRKC may bias a cholinergic fate in precursors via its regulation of Ret. In this case, Hmx1 expression is also expected to be de-repressed in Ret^−/− mice. Analysis of β-gal expression from the Hmx1 locus in Ret^CFP/CFP;Hmx1^LcZ/+ mice revealed a near three-fold increase in Hmx1 expressing neurons at E14.5 (Figure 5B; Supplementary Figure S6A) and an around 1.4-fold increase at E15.5 (Figure 5A and B) as compared to control Ret^CFP/+;Hmx1^LcZ/+ mice. To address whether the increases in HMX1⁺ cells represented only precocious expression, or also expression in neurons which normally do not express it, we analysed co-localization of RET with TRKC or HMX1 in control and RET-deficient mice. For this purpose, we used the Ret^CFP allele that expresses CFP from the null Ret locus and therefore can be used to identify cells that normally should have expressed Ret. Control Ret^CFP/+;Hmx1^LcZ/+ and Ret null Ret^CFP/CFP;Hmx1^LcZ/+ mice were analysed at E15.5. In control mice, Hmx1 and Ret expression was largely mutually exclusive. In contrast, in the absence of RET, a large number of neurons that normally should express Ret had initiated ectopic Hmx1 expression (Figure 5M). Therefore, an absence of RET leads to both precocious and ectopic expression of Hmx1 in some precursor cells. TrkA was also precociously expressed at E14.5 and the increase in TRKA⁺ neuronal numbers was similar to HMX1⁺ neuron increase at both E14.5 and E15.5, with a full co-localization of HMX1 and TRKA in the neurons (Figure 5C and D; Supplementary Figure S6A). Absence of RET signalling not only led to an increase in noradrenergic phenotypes, but also to a partial or near complete loss of cholinergic phenotypes, including VIP, SST, ChAT and VAChT (Figure 5G–L; Supplementary Figure S6B).

Figure 5.

A critical role of RET for suppression of noradrenergic and maintenance of a cholinergic phenotype. (A–L) Expression of Hmx1-βgal (A, B), TrkA (C, D), TrkC (E, F), Vip (G, H), Sst (I, J) and VAChT (K, L) on SG sections from E15.5 control and Ret mutant mice (i.e., Ret^CFP/CFP) mice, as indicated. (B, D, F, H, J and L) Quantifications of (A, C, E, G, I and K). Results are represented as mean±s.e.m. n=10–20 ganglia from at least 2 animals for each stage. Scale bar represents 50 μm. Statistical analysis was performed using Student’s t-test. **P<0.01, ***P<0.0001. (M, N) Change of fate of RET⁺ neurons determined by identification of cells that normally should express Ret using a Ret^CFP knock-in allele and staining with an anti-GFP antibody that detects CFP in E15.5 RET^CFP embryos. (M) Neurons normally expressing Ret fail to suppress Hmx1 in E15.5 mice as seen by double immunostaining for CFP and Hmx1-βgal on SG sections from E15.5 Ret^CFP/+ and Ret^CFP/CFP embryos. Note expression of Hmx1 and CFP being almost mutually exclusive in the control while CFP⁺ neurons express βgal in Ret^CFP/CFP mice. Quantifications are presented on the right. Numbers above bars indicate number of cells positive for the marker. n=10–20 ganglia. Scale bar represents 10 μm. (N) Note almost mutually exclusive expression of CFP and TrkC in control (Ret^CFP/+) mice and a failure of suppression of TrkC in cells which normally express Ret in Ret mutant (Ret^CFP/CFP) mice. Quantifications are presented on the right. Numbers above bars indicate number of cells positive for the marker. n=10–20 ganglia. Scale bar represents 10 μm.

A small increase in TRKC⁺ neurons was observed in the absence of RET (Figure 5E and F). In Ret^CFP/CFP null mutant mice, large numbers of cells that normally should express Ret displayed TrkC expression (Figure 5N), likely reflecting a delayed commitment of some precursors at E15.5 in the absence of RET. Combined, these data suggest that in the absence of RET, following a short delay in specification of TRKC⁺ precursors, Hmx1 is de-repressed and cholinergic markers including VIP, SST, ChAT and VAChT fail to be maintained, these neurons attain ectopic TrkA expression and ultimately change fate from cholinergic to noradrenergic lineage of sympathetic neurons.

By microarray experiments and confirmative RT–PCR, expression of T-cell leukaemia homeobox 3 (Tlx3) has been shown to increase in cultures of E12 chick sympathetic chain grown in the presence of pro-cholinergic differentiation factors (Apostolova et al, 2007). Immunohistochemical staining for TLX3 revealed its expression in nearly all precursors at E13.5 and became largely mutually exclusive with HMX1 already at E15.5 by a downregulation in many neurons (Figure 6A). Most TLX3⁺ cells were VAChT⁺ at E15.5, although some still expressed TH, while at P60 TLX3 was confined exclusively to VAChT⁺ neurons (Figure 6B). Tlx3 failed to be repressed in Hmx1^fl/fl;Wnt1-Cre mice at E15.5 and P0 and all TLX3⁺ cells also stained for RET (Figures 6C and D). A direct role of HMX1 in repression of Tlx3 in the cholinergic lineage of sympathetic neurons was established by analysis of Hmx1^LcZ/fl; Wnt1-Cre mice at P5 in which β-galactosidase is expressed in Hmx1 null cells that normally should have expressed Hmx1. In contrast to control (Hmx1^lcZ/fl) mice, Tlx3 failed to be repressed in HMX1-deficient neurons fated to the noradrenergic lineage, as revealed by co-localization of β-galactosidase and TLX3 in Hmx1^LcZ/fl;Wnt1-Cre mice (Figure 6E). These results show that Tlx3, initially expressed in all precursors, becomes restricted to cholinergic neurons via a mechanism that involves its repression by HMX1 in neurons fated to the noradrenergic lineage.

Figure 6.

TLX3 segregates in the RET⁺/VAChT⁺ neurons and its expression is regulated by HMX1. (A) Double immunostaining for TLX3 and Hmx1-βgal on SG section in Hmx1^LcZ/+ mice of indicated ages. At E13.5, note emergence of Hmx1 expression in TLX3⁺ neurons and downregulation of Tlx3 expression in these neurons (inset). At E15.5, E18.5 and P5, expression of Hmx1 and Tlx3 becomes almost mutually exclusive (insets) although rare, TLX3⁺/HMX1⁺ neurons are present even at late stages (P5, arrowhead). (B) Triple immunostaining for TLX3, VAChT and TH on SG section in wild-type mice of indicated ages. Note expression of Tlx3 in some TH⁺ neurons at E15.5 (arrowhead) but complete segregation in VAChT⁺ neurons at later embryonic stages. (C, D) Double immunostaining for TLX3 and RET on SG sections from control and Hmx1^fl/fl;Wnt1-Cre animals at E15.5 (C) and for TLX3 at P0 (D). Note de-repression of Tlx3 and Ret expression in neurons lacking HMX1 at both stages. (E) Double immunostaining for TLX3 and Hmx1-βgal on SG section from control (Hmx1^LcZ/+;Wnt1-Cre) and mutant (Hmx1^LcZ/fl;Wnt1-Cre) animals at P5. Note Hmx1 expression mutually exclusive with Tlx3 in control (upper panel) but not in mutant SG (lower panel) showing that in the absence of HMX1, many of the neurons that normally express Hmx1 (i.e., Hmx1-βgal⁺ neurons) acquire Tlx3 expression. Scale bar represents 50 μm. (F) Schematic representation of gene-regulatory interactions determining noradrenergic and cholinergic sympathetic subtypes. (G) Schematic representation showing the temporal activation of the genetic program defining the noradrenergic and cholinergic differentiation programs. Arrows show positive regulation and red lines show negative regulation.

Discussion

Here, we delineate the hierarchical cascade of sympathetic neuron diversification and identify how a transcriptional regulatory mechanism and neurotrophic tyrosine kinase receptors coordinate specification of the sympathetic lineage in the trunk. Although a transcriptional gene-regulatory network fating the sympathetic lineage and noradrenergic phenotype in precursors has previously been described, our study demonstrates that these early precursors display a mixed noradrenergic and cholinergic phenotype. The hybrid precursors with mixed phenotype segregate into distinct sympathetic lineages in a process starting around E13.5. We found that initiation of Hmx1 expression in precursors plays a critical role for the expression of TrkA and for the consolidation of a noradrenergic fate, while TRKC and RET determine a cholinergic phenotype. Together, our results unveil an intricate developmental process during specification of sympathetic neurons in which tyrosine kinase receptors and transcriptional activators, in cross-regulatory interactions, determine the ultimate function of sympathetic neuronal subtypes. A schematic representation depicting the gene-regulatory interactions determining sympathetic subtypes is shown in Figure 6F and G.

Induction of the sympathetic fate and its relation to diversification and consolidation of neurotransmitter phenotypes

An important aspect of these results is the relation of identified factors to the previously much studied gene-regulatory network inducing and defining the sympathetic lineage during development. The sympathetic lineage is induced in NCCs migrating in the ventral migratory pathway. BMPs are expressed in the dorsal aorta, and as NCCs reach the vicinity of the dorsal aorta, BMPs induce expression of a complex transcriptional regulatory unit including Mash1 (Guillemot et al, 1993; Hirsch et al, 1998; Lo et al, 1998), basic helix-loop-helix (bHLH), DNA binding protein Hand2 and Phox2a and Phox2b (Morin et al, 1997; Pattyn et al, 1997, 1999; Lo et al, 1999; Stanke et al, 1999). Transcription of Th and Dbh, encoding enzymes in noradrenalin synthesis, is regulated by Phox2 proteins and enhanced by HAND2 (Xu et al, 2003), which is also critical for the noradrenergic fate and proliferation in vivo (Goridis and Rohrer, 2002; Howard, 2005; Apostolova and Dechant, 2009). It is interesting that we find a failure of maintenance of Th expression in newborn Hmx1^fl/fl;Wnt1-Cre mice without any alterations in Phox2a/b expression. Our results show that PHOX2A/B and HAND2 are sufficient for Th expression independently of HMX1 at early stages, while a defining aspect during diversification is a change in requirements. During diversification, Hmx1 expression becomes critical for the maintenance of the noradrenergic phenotype as defined by expression of the rate-limiting enzyme in noradrenalin synthesis, TH.

One critical underlying mechanism that may participate in the change of requirements might be the extinction of TRKC activity. We find that TRKC is expressed in all hybrid noradrenergic/cholinergic precursors but is largely downregulated as these diversify into distinct types. In TrkC^−/− mice, Hmx1-mRNA levels are markedly elevated, showing that TRKC is suppressing Hmx1 expression. Therefore, reduced levels of TRKC activity in hybrid precursors might be a requirement for onset of Hmx1 expression. BMP2 has been shown to robustly upregulate TrkC expression in cultured E15 sympathetic neurons (Zhang et al, 1998). Hence, BMP signalling may not only be critical for inducing the sympathetic lineage at early stages, but a temporal window of BMP with declining levels starting around E13 could also initiate diversification in the sympathetic lineage, via an insufficient sustainment of TrkC expression and activity. Although most targets eventually receive some sympathetic innervation, proximal axon extension requires the TRKC ligand, NT3 (ElShamy et al, 1996; Francis et al, 1999; Kuruvilla et al, 2004). While NT3 does not mediate sympathetic precursor survival, the 50% loss of sympathetic neurons in Nt3^−/− mice coincides at a later stage in development with the period of excessive neuronal loss in Ngf^−/− mice (Wyatt et al, 1997; Francis et al, 1999; Kuruvilla et al, 2004). At these late embryonic stages, we find that TrkC is largely downregulated in sympathetic neurons consistent with the interpretation that a deficit in early axon growth leads to a subsequent impairment in attaining target-derived NGF that is required for neuronal survival (Glebova and Ginty, 2005). Here, we find that in addition to affecting proximal axon growth, TRKC plays an important role in sympathetic neuron diversification by regulating the neurotransmitter fate and in its absence, a switch in fate from cholinergic to noradrenergic neurons is observed.

Specification of noradrenergic neurons

Onset of Hmx1 expression in sympathetic precursors is critical for several, but not all, defining aspects of development of noradrenergic sympathetic neurons. First, while PHOX2AB/HAND2 is required for initiation of Th expression in hybrid precursors, HMX1 is necessary for its maintenance in noradrenergic sympathetic neurons as these diversify from sympathetic precursors. This requirement is clearly an independent mechanism from Phox2ab/Hand2 induction, since Dbh expression, which also requires PHOX2AB/HAND2, is not affected by Hmx1 deficiency. Furthermore, the onset of Hmx1 expression is critical for expression of TrkA in noradrenergic sympathetic neurons. TrkA expression is defining for these neurons, as signalling through this receptor plays fundamental roles in development of the noradrenergic sympathetic nervous system. NGF promotes extensive neurite growth and hypertrophy of sympathetic neurons (Levi-Montalcini, 1987), an activity critical for final sympathetic target innervation but not proximal growth in vivo (Glebova and Ginty, 2004). TRKA signalling is also critical for survival of sympathetic neurons and in its absence essentially no noradrenergic sympathetic neurons remain, as revealed in mice lacking TRKA or NGF (TrkA^−/− and Ngf^−/− mice) (Crowley et al, 1994; Smeyne et al, 1994; Fagan et al, 1996b). NGF signalling via TRKA leads to robust upregulation of Th expression in cultured sympathetic neurons (Otten et al, 1978; Brodski et al, 2000). A dependence on Hmx1 expression for maintenance of Th expression in the noradrenergic sympathetic lineage and for consolidation of the noradrenergic phenotype may therefore be a result of direct and/or indirect activity of HMX1 via its regulatory activities on TrkA expression. Thus, it is possible that HMX1-induced TrkA expression allows for TRKA signalling, which maintains Th expression in the noradrenergic sympathetic neuronal subtypes. The deficits of TrkA expression and specification of the noradrenergic phenotype of Hmx1^fl/fl;Wnt1-Cre mice resulted in a marked reduction of innervation of the arrector pili muscle and subcutaneous blood vessels, consistent with the critical role of TrkA for target innervation. Our results on cell counts revealed an unexpected increase in trunk sympathetic neurons in the Hmx1^fl/fl;Wnt1-Cre mice. One possibility is that a remaining low level of TrkA is sufficient for neuronal survival. This is consistent with that despite that both heterozygous mice for Ngf and TrkA display reduced levels of gene expression, Ngf^+/− mice but not TrkA^+/− mice display excessive neuronal loss (Brennan et al, 1999; Ghasemlou et al, 2004). Hence, TrkA does not appear to be limiting during development. Another possibility is that in the absence of HMX1, the altered phenotype (i.e., for instance massive increase in Ret expression) allows for TRKA-independent survival.

HMX1 is not only important for induction of TrkA and maintenance of Th, but also for repression of phenotypic characteristics associated with the alternative, cholinergic, fate. In Hmx1^fl/fl;Wnt1-Cre mice Ret, Vip and Sst are broadly expressed in neurons throughout the ganglion, representing a failure of suppression of their expression in the noradrenergic lineage, due to Ret failing to be repressed in neurons that normally should have expressed Hmx1 (as revealed using Hmx1^LcZ/fl;Wnt1-Cre mice). The failure of suppression of Ret in Hmx1^fl/fl;Wnt1-Cre mice explains the persistent expression of Vip and Sst in these mice, since both these genes are induced by RET signalling. Hence, HMX1 defines a noradrenergic fate by sustaining the noradrenergic sympathetic phenotype already present in precursors and by eliminating gene expression associated with the alternative cholinergic phenotype.

Specification of cholinergic neurons

Cholinergic sympathetic innervation of sweat glands and periosteum in rodents have been shown to emerge postnatally from noradrenergic innervation as a result of a change in neurotransmitter phenotype from noradrenergic to cholinergic by soluble signals (Schotzinger et al, 1994; Francis and Landis, 1999). Our results show that TRKC directly affects the fate of sympathetic precursors by promoting development along the cholinergic sympathetic fate. It is unclear if this role of TRKC is ligand dependent or ligand independent. An increased cell death reported in Nt3^−/− but not in TrkC^−/− mice (Fagan et al, 1996a; Tessarollo et al, 1997) and no additional loss in Nt3^−/−/Ngf^−/− compound mutant mice as compared to Ngf^−/− mice (Francis et al, 1999) indicate that the early role of NT3 for proximal axon growth that later results in an NGF-deprivation induced death caused by failure of target innervation is mediated via TRKA interactions. This conclusion concur with that NT3 can signal via both TRKC and TRKA in sympathetic neurons (Davies et al, 1995). Although previous studies have not addressed deficits in specification of sympathetic neuronal types in Nt3^−/− mice, TH and DBH expression has been examined prior to normal cell death (i.e., at E15). TH and DBH in the superior cervical ganglion of Nt3^−/− and TrkA^−/− mice but not Ngf^−/− mice display deficits of expression, suggesting that NT3 can act via TRKA to regulate their expression (Andres et al, 2008). Thus, much of the previously known functions of NT3 during sympathetic neuron development may include interactions and signalling via TRKA in the TRKA⁺ noradrenergic lineage. Alternative splicing generates transcripts that encode both TRKC receptors with and without the catalytic tyrosine kinase domain. Quantitative PCR data show that transcripts lacking the tyrosine kinase domain are expressed at much greater levels than transcripts encoding catalytic full-length TRKC receptors in developing sympathetic neurons (Wyatt et al, 1997), supporting a ligand-independent activity of TRKC during sympathetic neuron diversification. In contrast to this conclusion, NT3 has been shown to induce ChAT and Vip expression in cultured E12 sympathetic neurons in a mechanism likely not involving TRKA since unlike NT3, NGF fails to induce ChAT and VIP (Brodski et al, 2000, 2002). This indicates that sympathetic precursors express sufficient amounts of catalytically active TRKC receptors for a functional response. It is therefore unclear whether the role of TRKC for sympathetic diversification in the present study is ligand dependent or ligand independent. If it does involve a ligand interaction, then Nt3^−/− mice are expected to display a much more complex phenotype than TrkC^−/− mice as NT3 may also interact with other receptors, such as TRKA during sympathetic neuron development.

The roles of TRKC during cholinergic sympathetic neuron differentiation described in the present study may partly involve its repressive activity on Hmx1 expression in sympathetic precursors, because HMX1 represses Ret, and hence suppression of Hmx1 expression by TRKC activity is compatible with continuous expression of Ret. Mice with null mutations in Ret, or one of its ligand, artemin or the artemin co-receptor GFRα3 display short and misdirected proximal axon projections (Schuchardt et al, 1994; Nishino et al, 1999; Enomoto et al, 2001; Honma et al, 2002), suggesting that both RET and TRKC work in the sympathetic precursors to induce extension of proximal axons. As precursors diversify into defined neurotransmitter classes of sympathetic neurons RET is absolutely critical for defining the cholinergic phenotype as seen by the failure of Chat, VAChT, Sst and Vip expression in Ret mutant mice. These results are consistent with previous data which has established that an RET ligand, GDNF, induces ChAT in vitro (Brodski et al, 2002), and that loss of RET in mutant mice markedly reduces the number of ChAT⁺ and VAChT⁺ neurons at E16 (Burau et al, 2004). We found that the loss of cholinergic sympathetic neurons in Ret^CFP/CFP mice was paralleled by an increase in noradrenergic HMX1⁺ and TRKA⁺ neurons, which represents a true switch in fate, because cholinergic neurons that normally should have expressed Ret, acquired Hmx1 expression as determined in Ret^CFP/CFP;Hmx1^LcZ/+ mice. Hence, RET induces a cholinergic fate also by aborting expression of key factors necessary for noradrenergic fate. We identified HMX1 as one such factor, since HMX1⁺ neurons were markedly elevated in Ret^CFP/CFP;Hmx1^LcZ/+ mice. Consistently, TRKA⁺ neurons were increased proportionally to the increase in HMX1⁺ cells.

Tlx3 was found to be expressed in most or all sympathetic precursors but was rapidly downregulated in cells attaining Hmx1 expression and was later in development confined only to the cholinergic sympathetic lineage. In the sympathetic lineage, we observe that Tlx3 and Ret are always co-expressed, first in most or all precursors and later in the diversifying cholinergic lineage. The diversification of HMX1 and TLX3 along different sympathetic lineages can be explained by the suppressive activities of HMX1 on Tlx3, as observed by a continuous expression of Tlx3 in most neurons of Hmx1^fl/fl;Wnt1-Cre mice. RET can act as a positive regulator of Tlx3 expression in cultures of the chick sympathetic chain (Apostolova et al, 2007). Hence, a failure of RET repression in Hmx1^fl/fl;Wnt1-Cre mice might mediate the sustained Tlx3 expression. Thus, cholinergic neurons develop in a process involving sustained TRKC activity in sympathetic precursors, preventing Hmx1 induction and therefore persistent Tlx3/Ret expression that consolidates a cholinergic fate by RET-dependent induction and maintenance of cholinergic neuronal properties and a sustained repression of Hmx1.

In conclusion, our data support a model in which a hybrid noradrenergic/cholinergic sympathetic precursor diversifies in a highly controlled hierarchical segregation into distinct sympathetic types. This segregation of a common precursor into distinct neuronal types involves an intricate process based on repressive cross-regulatory interactions comprising both transcriptional activities and growth factor receptor activities, which defines the fate and function of the neurons.

Materials and methods

Generation of Hmx1 conditional null mice

Figure 3A and supporting information Supplementary Figure S5A and G show the targeting strategy used to generate mice carrying a conditional knockout allele of Hmx1 (Taconic-Artemis, Germany). The targeting vector with a 5.8-kb long arm and 3.0-kb short arm (Figure 3A; Supplementary Figure S5) was generated using BAC clones from the C57BL/6J RPCI-23 BAC library. The 5′ LoxP DNA sequence, the target for the Cre recombinase, was inserted in the first intron. The second LoxP site was placed 3′ to the second exon. The positive selection marker NeoR, which confers G418 (neomycin) resistance, was flanked by FRT sites that can be cleaved by the Flp-Deleter Flipase recombinase (Flpe); the negative selection marker Thymidine kinase (Tk) cassette, which confers Gancyclovir resistance, was placed 5′ to the long arm of homology. The linearized targeting vector was transfected into ES cells (Supplementary Figure S5A–E). Homologous recombinant clones were isolated using positive (neomycin) and negative (gancyclovir) selection. Hybridization of Southern blots was used on genomic DNA to detect homologous recombination and single integration at the 5′ side was performed using EcoRI, BclI and PflFI restriction enzymes and ila2 probe, which gave 13.6, 13.7 and 11.6 kb fragments, respectively, for the wild-type allele and a 7.8-, 11.9- and 8.4-kb fragment for the targeted allele (not shown). Southern hybridization on the 3′ side was performed by digestion with EcoRI, AflII, HindIII and 3e1 probe produced 13.6, 17.9 and 8.1 kb fragments, respectively, for the wild-type allele and a 5.4-, 5.2-, 9.2-kb fragment for the targeted allele (not shown). ES clones were injected into 3.5 days post coitum (dpc) blastocysts derived from super-ovulated BALB/c females mated with BALB/c males. Injected blastocysts were the transferred to each uterine horn of 2.5 dpc, pseudopregnant NMRI females. Chimeric mice were bred to strain BALB/c females in order to generate heterozygous mice carrying the targeted allele. These mice were bred with a Flp-Deleter mouse line, removing the neo cassette and creating mice carrying the Hmx1 floxed allele, referred to as Hmx1^fl. Hmx1^fl mice were then bred with Wnt1-Cre mice to remove the second exon (containing the homeobox domain) and generating a conditional knock-out mouse (Hmx1^fl/fl;Wnt1-Cre). Genotyping for mice carrying the Hmx1 floxed allele and Wnt1-Cre was performed by PCR (Supplementary Figure S5F). The following primers were used: 5′-CCTGGTGACATCCCTTGTACG-3′ and 5′-GGGTGACATTGGCACAACC-3′, to detect both the wild-type allele (218-bp band) and the floxed allele (375-bp band) of Hmx1 (Supplementary Figure S5G) and 5′-ACCAGGTTCGTTCACTCATGG-3′ (Forward) and 5′-AGGCTAAGTGC CTTCTCTACA-3′ (Reverse) to detect the Wnt1-Cre allele (200 bp). Successful elimination of Hmx1 in the SG was confirmed by in situ hybridization (Supplementary Figure S5H).

The Hmx1^LcZ ES line was purchased from Velocigene, injected into 3.5 dpc blastocyst to generate chimeric mice that were bred for germline transmission and used as heterozygous mice for detection of Hmx1 expression (for further details, refer to Supplementary Figure S2A–C).

Other mice strains

Bax^−/−, TrkC^−/−, Wnt1-Cre and Gt(ROSA)26^{tm9(CAG-tdTomato)} mice were ordered from The Jackson Laboratory. Genotyping of the Gt(ROSA)26^{tm9(CAG-tdTomato)} mice was performed by PCR. The following primers: 5′-AAAGTCGCTCTGAGTTGTTAT-3′ (Forward wild-type) and 5′-GGAGCGGGAGAAATGGATATG-3′ (Reverse wild-type) and 5′-TGGCGTTACTATGGGAACAT-3′ (Reverse Tomato) were used to detect both the wild-type allele (500 bp band) and the tomato allele (407 bp band). Ret^CFP and Ret^ERT2 mice have been described previously (Uesaka et al, 2008; Luo et al, 2009).

Immunohistochemistry

Embryos were collected and fixed in 4% paraformaldehyde (PFA) in PBS (pH 7.4) at 4°C for 1–3 h. The lower thoracic sympathetic chain (vertebrae 9th to 13th) was dissected out together with the spinal cord and fixed in 4% PFA in PBS for 5 h. Samples were subsequently washed in PBS at 4°C for 1 h and cryoprotected by incubating at 4°C overnight in 20% sucrose followed by 30% sucrose in PBS. Tissues were then embedded in OCT and frozen at −20°C. Samples were sectioned at 14 μm and frozen at −20°C after drying at RT for 1 h. In order to have a complete representation of the 9th to 13th thoracic paravertebral SG, sections were serially collected on 10–12 slides. For every condition, each slide contained tissues from two embryos or pups, belonging to the same litter.

For immunohistochemistry, fresh frozen sections were used and dried at RT for 1 h. For RET and ISL1 staining, particularly at postnatal stages, antigen retrieval was performed by immersing the sections in 80°C antigen retrieval solution (DAKO) for 20 min. Sections were then washed three times in PBS containing 0.1% Tween-20 (PBSt), incubated with primary antibodies diluted in PBSt at 4°C overnight and coverslipped with parafilm. Sections were then washed in PBSt and incubated with secondary antibodies diluted in PBSt at room temperature for 1 h, washed again in PBSt and mounted using glycerol. For detection of Ret^CFP, anti-GFP (FITC) antibody (Abcam) was diluted at 1:500 and incubated for 2 h following secondary antibody staining. The primary antibodies used were Goat anti-VAChT (Phoenix Europe GmbH, 1:400 dilution), Rabbit anti-VMAT2 (Phoenix Europe GmbH, 1:150), Sheep anti-TH (Novus Biologicals, 1:2000), Rabbit anti-TH (Pel-Freez, 1:1000), Goat anti-TRKC (R&D, 1:500), Goat anti-ChAT (Millipore, 1:100), Rabbit anti-DBH (Protos Immunoresearch, 1:100), Rabbit anti-PRPH (Chemicon International, 1:500), Goat anti-RET (R&D, 1:50), Sheep anti-NPY (Abcam 1:1000), Goat anti-TRKA (R&D, 1:500), Rabbit anti-TRKA (kind gift from Louis Reichardt), Rabbit anti-TRKA (Millipore, 1:1000), Goat anti-TRKC (R&D, 1:500), Goat anti-CGRP (Abcam, 1:1000), Rabbit anti-VIP (Immunostar, 1:200), Rat anti-SST (Millipore, 1:100), Chicken anti-βgal (Abcam, 1:1500), Rabbit anti-Phox2A/B (kind gift from Jean-François Brunet, 1:400), Rabbit and Guinea Pig anti-TLX3 (kind gift from Thomas Müller and Carmen Birchmeier, 1:10 000), Mouse anti-ISL1 (Developmental Studies Hybridoma Bank (DSHB), 1:100), ASMA-Cy3 (Sigma-Aldrich, 1:100), Rabbit anti-PGP9.5 (AbD Serotec, 1:1000). The secondary antibodies used were Alexa 555 or 647 conjugated donkey anti-mouse antibody, Alexa 488, 555 or 647 conjugated donkey anti-rabbit, Alexa 488, 555 or 647 conjugated donkey anti-goat, Alexa 488 or 555 conjugated donkey anti-sheep, Alexa 488 or 555 conjugated goat anti-rat, Alexa 549-donkey anti-chicken, Alexa 488 or 555 conjugated donkey anti-sheep. All secondary antibodies were purchased from Invitrogen and used at 1:1000 dilution.

In situ hybridization

Tissue was processed as for immunohistochemistry, and hybridizations were performed as previously described (Adameyko et al, 2009). pCRII-Hmx1 cDNA plasmid (kind gift from Thomas Lufkin, Genome Institute of Singapore) was used to synthesize digoxigenin antisense riboprobes according to supplier’s instructions (Roche, Mannheim, Germany). Visualization was carried out using alkaline phosphatase-conjugated anti-digoxigenin antibodies (Roche, 1:2000) followed by alkaline phosphatase staining developed with NBT/BCIP (Roche).

Tamoxifen injections

Tamoxifen (4-HT, Sigma) was dissolved in corn oil (Sigma). Tamoxifen solution was delivered via intra-peritoneal (i.p.) injection to pregnant females at E11.5 and 3-day-old pups (100 μg/g of 4-HT body weight). For all experiments, the day of the plug positive was considered as E0.5 and P0 the day when the mice were born.

Quantification and statistics

Hmx1 (recapitulated by βgal), TrkA, Ret, Vmat2, VAChT, Phox2a/b, TrkC, Sst, Prph2 and Npy expression was analysed in SG sections from Hmx1^LcZ/+ (E12.5, E13.5, E14.5 and E15.5) or RET^CFP/+;Hmx1^LcZ/+ (E18.5, P5 and P11) animals. In the latter case, Ret^CFP was detected using FITC-conjugated anti-GFP antibody. For quantification, at least six SG sections from at least two animals were counted for every developmental stage. For quantification of the RET⁺, TRKA⁺, TH⁺, PHOX2A/B⁺, SST⁺, VIP⁺, NPY⁺, ChAT⁺, VAChT⁺, DBH⁺, VMAT2⁺, CGRP⁺ and TLX3⁺ populations in control and Hmx1 mutants, at least six SGs from at least three animals for each genotype and stage were analysed. For Ret^−/− analysis, quantification of TRKC⁺, SST⁺, VIP⁺ and VAChT⁺ neurons was carried on control (RET^CFP/+) and mutant Ret (Ret^CFP/CFP) E15.5 embryos; quantification of Hmx1 and TrkA expression was carried out in control (Ret^CFP/+;Hmx1^LcZ/+) and mutant Ret (Ret^CFP/CFP;Hmx1^LcZ/+) E14.4 and E15.5 embryos. In both cases, at least eight SG sections from at least two embryos were counted. For quantification of (mRNA)Hmx1⁺, TRKA⁺, RET⁺, TH⁺ and ChAT⁺ neurons in control and TrkC^−/− and TrkC^−/−;Bax^−/− mutants at E15.5, E18.5 and P0 at least 15 randomly selected SG per animal per stage were counted. For all counts, the number of positive neurons was normalized to the total number of ISL1⁺ neurons in each ganglion. For quantification of the number of paravertebral sympathetic neurons in wild-type and Hmx1^fl/fl;Wnt1-Cre mice, 14–43 SG from three newborn (P0) pups were analysed. Data were analysed using GraphPad Prism 5 and expressed as mean±standard error of mean (s.e.m.). Unpaired Student’s t-test was performed to determine if there were significant differences between groups.