Mutation in LBX1/Lbx1 precludes transcription factor cooperativity and causes congenital hypoventilation in humans and mice.

The respiratory rhythm is generated by the preBötzinger complex in the medulla oblongata, and is modulated by neurons in the retrotrapezoid nucleus (RTN), which are essential for accelerating respiration in response to high CO2 Here we identify a LBX1 frameshift (LBX1 FS ) mutation in patients with congenital central hypoventilation. The mutation alters the C-terminal but not the DNA-binding domain of LBX1 Mice with the analogous mutation recapitulate the breathing deficits found in humans. Furthermore, the mutation only interferes with a small subset of Lbx1 functions, and in particular with development of RTN neurons that coexpress Lbx1 and Phox2b. Genome-wide analyses in a cell culture model show that Lbx1FS and wild-type Lbx1 proteins are mostly bound to similar sites, but that Lbx1FS is unable to cooperate with Phox2b. Thus, our analyses on Lbx1FS (dys)function reveals an unusual pathomechanism; that is, a mutation that selectively interferes with the ability of Lbx1 to cooperate with Phox2b, and thus impairs the development of a small subpopulation of neurons essential for respiratory control.

Neurons represent the most diverse cell population in animals. How this diversity is specified and maintained is incompletely understood. Available evidence shows that multiple transcription factors cooperate to control common as well as neuron-specific gene expression programs (1, 2). The combinatorial binding of such factors to regulatory elements in chromatin is key for gene expression (3). Homeodomain transcription factors, among them Lbx1 and Phox2b, impose specific neuronal fates during development. In mice, Lbx1 specifies distinct neuronal subtypes in the spinal cord and hindbrain (4–7), and it is also essential for limb muscle development (8–10). Phox2b controls development of central and peripheral visceral neurons (11, 12). In the hindbrain, a single neuronal population (dB2 neurons) coexpresses Lbx1 and Phox2b and depends on both factors for proper development (6, 7, 13). A subpopulation of dB2 neurons forms the retrotrapezoid nucleus (RTN), a small group of cells in the ventral hindbrain that is central for the hypercapnic reflex; that is, the acceleration of breathing in response to increased partial pressure of CO2 levels (13–16).

Breathing is regulated unconsciously by the nervous system. The respiratory rhythm is generated by the preBötzinger complex located in the ventral hindbrain (17), and is modulated by RTN neurons and by other neuronal populations. Congenital central hypoventilation syndrome (CCHS, also known as Ondine’s curse; OMIM 209880) is a rare, life-threatening disorder characterized by slow and shallow breathing resulting from a deficiency in autonomic control of respiration. Patients with CCHS are hypercapnic; that is, they have abnormally high levels of CO2 in the blood and lack the hypercapnic reflex (18). Atypical heterozygous expansions of alanine repeats in PHOX2B are the most common genetic causes of CCHS (12, 19), but similar phenotypes can also be caused by genetic abnormalities in RET, EDN3, and MYO1H (20, 21). The introduction of a frequent PHOX2B (PHOX2B) mutation into the murine genome precludes development of RTN neurons, causes loss of the hypercapnic reflex, produces severe hypoventilation, and results in neonatal lethality (13). However, the selective elimination of RTN neurons accounts only for the loss of the hypercapnic reflex, but not for the severe hypoventilation and the neonatal lethality observed in Phox2b mutant mice (22).

In this study, we report on a consanguineous family with two CCHS-diagnosed children that tested negative for PHOX2B mutations. The children carried a homozygous frameshift mutation in LBX1 (LBX1) that alters the C terminus of the protein without affecting its homeodomain. Homozygous mice carrying the analogous mutation (Lbx1) displayed respiratory deficits that recapitulated the human phenotype. In Lbx1 mice, two Lbx1+/Phox2b+ neuronal subpopulations (in the RTN and in the dorsal hindbrain) were severely affected, but in contrast to Lbx1 null mutants, second-order somatosensory neurons and limb skeletal muscle formed correctly. Genomewide DNA binding analysis of Lbx1FS in a cell culture model showed that the mutant variant mostly binds to similar sites as the wild-type protein. However, in contrast to the wild-type protein, Lbx1FS is unable to correctly cooperate with Phox2b, and instead overrides its function. Thus, the Lbx1FS protein is selectively impaired in a transcriptional cooperativity with Phox2b during neuronal development, but functions correctly in other contexts.

Results

A Homozygous LBX1 Frameshift Mutation Causes Recessive Congenital Central Hypoventilation Syndrome.

We identified two male siblings, offspring from a consanguineous marriage, who displayed hypoventilation during the neonatal period. The parents were unaffected and had a healthy daughter (Fig. 1). Two sisters from the father/mother side of the patients lost a child to cot death (). Both affected siblings studied here required continuous mechanical ventilation after birth because of respiratory insufficiency. They showed recurrent episodes of apnea and signs of central hypoventilation during sleep with no response to falling oxygen saturation or hypercapnia. The children were diagnosed with a severe pattern of classic CCHS. Sanger sequencing, microsatellite analysis, and multiplex ligation-dependent probe amplification of DNA from the children did not reveal any mutations in PHOX2B. Lbx1 ablation causes hypoventilation in newborn mice (6, 7). We therefore sequenced LBX1 in the affected individuals and identified a homozygous frameshift mutation in its exon 2 (LBX1 mutation; ). Sanger sequencing of the entire family confirmed that the LBX1 mutation segregated with the phenotype. The mutation was predicted to alter the LBX1 protein at the C terminus without affecting its homeodomain (). Furthermore, the LBX1 variant was absent in control cohorts such as Exome Aggregation Consortium and 1000 Genomes. Ablation of Lbx1 in mice results in a complex phenotype resulting from defects in the development of various hindbrain neuronal subtypes (dB1–dB4; see scheme in Fig. 1), as well as deficits in the formation of dorsal spinal cord neurons and limb skeletal muscle (4–10, 23). However, the children carrying the LBX1 mutation did not show any obvious change in limb musculature. We thus reasoned that the LBX1 mutation might selectively impair neurons that participate in the central control of respiration.

Fig. 1.

Genome-wide characterization of chromatin binding of a frameshift mutant LBX1 associated with CCHS in humans. (A) Pedigree of a family with two children diagnosed with CCHS (black). (B, Left) Transverse section of the developing brainstem stained with Olig3 (red) and Lbx1 (green) antibodies at E11.5. Olig3 is expressed in the dorsal brainstem, whereas Lbx1 is expressed in dB1–dB4 neurons. (B, Right) Scheme showing genes expressed in progenitor cells and neurons of the dorsal hindbrain. Lbx1 and Phox2b are coexpressed in dB2 neurons. (C, Left) Heat maps showing read tracks at sites occupied by Lbx1 and Lbx1FS ± 5 kb around the binding sites. (C, Right) Lbx1 (blue) and Lbx1FS (pink) mean read densities (MRDs) for distinct classes of binding sites. (D) ChIP-seq tracks illustrating Lbx1 and Lbx1FS occupancy on intergenic and intronic regions of Prxxl1. (E) ChIP-qPCR analysis using antibodies against flag-tag to validate Lbx1 and Lbx1FS occupancy on the highlighted regions displayed in D (n = 4 independent replicates). (F) H3K27ac (Left) and H3K27me3 (Right) ChIP-qPCR analysis performed on chromatin prepared from Lbx1 and Lbx1FS differentiated neurons (n = 4 independent replicates).

Similarities in the Genome-Wide Binding of Lbx1 and Lbx1FS.

To model alterations of Lbx1FS function and its binding to DNA in a neuronal context, we looked for a suitable neuronal cell culture system. P19 murine embryonic teratocarcinoma stem cells differentiate into neurons that express Lbx1 and a HoxA gene code typical of the caudal hindbrain (rhombomeres 4–7) and anterior cervical spinal cord upon retinoic acid treatment () (24). In addition, they express Lmx1b, Pou4f1, and Prrxl1, the latter at low levels (); this combination is indicative of excitatory somatosensory neurons of the spinal cord and hindbrain. We used this model to analyze Lbx1 and Lbx1FS binding on a genome-wide scale. The endogenous Lbx1 locus was first mutated in these cells using CRISPR-Cas9, and the resulting Lbx1 mutant cells were transduced with retroviruses encoding flag-tagged Lbx1 or Lbx1 (referred to as Lbx1 and Lbx1FS cells). Cell clones that expressed comparable levels of Lbx1/Lbx1 were chosen for ChIP-seq analysis. In neurons differentiated from such cells (named Lbx1 and Lbx1FS neurons), we identified 7,537 binding sites for Lbx1, but considerably more (n = 12,343) sites for Lbx1FS. A large fraction (59%) of the Lbx1 sites was also bound by Lbx1FS (). To analyze how the 1.6-fold increase in Lbx1FS binding sites related to binding strength, another important variable for transcription factor function, we performed read enrichment analysis combined with k-means clustering for Lbx1- and Lbx1FS-bound sites. In general, the mean read density for Lbx1FS was lower than for Lbx1 (Fig. 1). Sites in which Lbx1FS bound more strongly than Lbx1 displayed, on average, low enrichment for both Lbx1FS and Lbx1 proteins (Fig. 1). Together, our data show that the Lbx1FS mutant protein can bind to most Lbx1 sites; however, the binding is weaker and less specific than that of the wild-type protein.

Inspection of Lbx1 and Lbx1FS ChIP-seq tracks revealed occupancy of both factors on intronic and intergenic regions associated with the somatosensory genes Prrxl1, Lmx1b, and Pou4f1 (Fig. 1 and ). The occupancy of Lbx1/Lbx1FS on such loci was confirmed by ChIP-qPCR (Fig. 1 and ). Similar gene expression levels for the three somatosensory genes were observed in neurons differentiated from Lbx1 and Lbx1FS cells, but they were not expressed in Lbx1 mutant neurons (). To test whether these intronic and intergenic regions correspond to enhancer elements, we performed ChIP-qPCR for H3K27ac and H3K27me3, two epigenetic marks associated with active or repressed enhancers, respectively (25). This showed strong enrichment for H3K27ac at the analyzed loci in neurons differentiated from Lbx1 and Lbx1FS cells, whereas H3K27me3 was not enriched (Fig. 1 and ). Hence, Lbx1/Lbx1FS binding sites on the Lmx1b, Prrxl1, and Pou4f1 loci correspond to active enhancers in Lbx1 and Lbx1FS neurons.

We next performed de novo motif analysis for Lbx1 and Lbx1FS binding sites. In both peak sets, various AT-rich motifs that aligned with previously identified Lbx1-binding sites were overrepresented (26). A closer inspection revealed subtle differences between Lbx1 and Lbx1FS sequence preferences (). The most significantly enriched motif in Lbx1FS peaks was a nonpalindromic 8-mer, possibly representing a monomer-binding site, which was identified in Lbx1 sites as the fifth most significant. In contrast, the most overrepresented motif in Lbx1 sites was a 12-mer palindrome, possibly a homodimeric site that was the third-most enriched motif for Lbx1FS. Finally, a 16-bp-long nonpalindromic motif was identified in Lbx1, but not Lbx1FS, sites (). This was composed of a partial Lbx1 site at the 3′ end, preceded with a distinct AT-rich 5′ sequence, which could represent the binding site of an Lbx1 cofactor. Interestingly, this AT-rich half-site was reported to be a preferred Phox2b binding site () (27). This raised the possibility that although the general DNA binding of Lbx1FS was only mildly compromised, its ability to interact with other factors is more severely impaired.

Hypoventilation and Lack of Hypercapnic Reflex in Homozygous Lbx1 Mice.

To better understand the deficit in LBX1FS function, we introduced an analogous mutation into the mouse Lbx1 gene (). Heterozygous Lbx1 (Lbx1) mice were viable and fertile, and did not show an obvious phenotype. However, homozygous Lbx1 newborn mice displayed cyanosis and died (n = 18/18) within the first 2 h of life without displaying other apparent deficits in motor behavior. Plethysmographic recordings revealed pronounced respiratory deficits in Lbx1 mice; that is, shallow breathing with frequent and long apneas (Fig. 2 and ). In particular, Lbx1 mice displayed longer times between breathing cycles (Ttot), which led to reduced respiratory minute volumes (VE) and severe hypoventilation (Fig. 2 and ). Importantly, Lbx1 mice lacked the hypercapnic reflex and did not change ventilation (VE and Ttot) when exposed to high levels of CO2 in air (Fig. 2 and ). We concluded that the Lbx1 mutation in mice leads to a respiratory phenotype that resembles the one observed in the studied patients.

Fig. 2.

The Lbx1 mutation causes central hypoventilation and loss of CO2 sensitivity in mice. (A) Plethysmographic traces of control and Lbx1 mice in normal air (0.04% CO2; Top traces) and high CO2-containing air (hypercapnia, 8% CO2, Bottom traces). Numbers on the left of the traces indicate distinct individuals. (B) Quantification of VE of control (n = 9) and Lbx1 (n = 10) newborn mice in normal air and high CO2-containing air (unpaired nonparametric Mann–Whitney U test). (C, Left and Middle) Histological analysis of RTN neurons (arrowheads) using Lbx1 (green) and Phox2b (red) antibodies; these neurons are present in control mice but not in Lbx1 mice at birth. Antibodies against ChAT (blue) were used to distinguish RTN (Lbx1+/Phox2b+/ChAT−) neurons from facial (nVII) motor (Lbx1−/Phox2b+/ChAT+) neurons. Confocal tile scan modus was used to acquire photomicrographs and assembled using ZEN2012 software (10% overlap between tiles). (C, Right) Quantification of RTN neuron numbers in control and Lbx1 mice at E14.5 (n = 4 per condition; unpaired t test, t = 19,37; df = 6) and at birth (n = 4 per condition; unpaired t test, t = 26,08; df = 6). ***P < 0.0001.

The Lbx1 Mutation Interferes with RTN Formation.

We next assessed whether the lack of hypercapnic response in Lbx1 mice was a result of impaired RTN development. RTN neurons locate in the ventral hindbrain and coexpress Lbx1 and Phox2b, but not choline acetyl-transferase (ChAT) (Lbx1+/Phox2b+/ChAT−), and are thus distinguished from the neighboring facial motor neurons that coexpress Phox2b and ChAT, but not Lbx1 (Lbx1−/Phox2b+/ChAT+). In Lbx1 animals, Phox2b+/Lbx1+ cells were absent in the RTN region either at embryonic day (E) 14.5 or at birth (Fig. 2). However, several other Lbx1+ neuronal types were present and expressed Lbx1 at apparently normal levels. The absence of a functional RTN was confirmed by Ca2+ imaging (). Further analyses demonstrated that RTN precursors (i.e., Lbx1+/Phox2b+ dB2 neurons) were unchanged in Lbx1 mice at E11.5, but failed to initiate Atoh1 expression during their migration toward the ventral hindbrain at E12.5 (). preBötzinger complex neurons have no history of Phox2b or Lbx1 expression (28), and were present and functional in Lbx1 mice (). We conclude that in Lbx1 mice, dB2 neuronal precursors are correctly specified, but the subset destined to form the RTN fails to express Atoh1 and does not migrate into the position where the RTN normally resides.

The Lbx1 Mutation Does Not Preclude SpV and Limb Muscle Development.

Next we analyzed inhibitory and excitatory somatosensory neurons of the spinal trigeminal (SpV) nucleus. These neurons are absent in Lbx1 null mutant mice, where they instead assumed solitary tract nucleus and inferior olivary nucleus neuronal fates, respectively (7). Interestingly, the SpV was present in Lbx1 mice (Fig. 3). Furthermore, the solitary tract and inferior olivary nuclei appeared to have a normal size (Fig. 3). Finally, limb muscle development is severely affected in Lbx1 null mutant mice (8–10), but these muscle groups were present and appeared correctly formed in Lbx1 mice (Fig. 3). Together, our analyses demonstrate that the Lbx1 mutation selectively interferes with development of Lbx1+/Phox2b+ RTN neurons, but in other contexts, the mutant protein functions correctly, as in development of somatosensory SpV neurons and limb muscles.

Fig. 3.

Development of somatosensory neurons and limb muscle in Lbx1 mice. (A) Histological analysis of somatosensory neurons of the SpV, viscerosensory neurons of the nucleus of the solitary tract (NTS), and neurons of the inferior olive (IO) in control (Left), Lbx1 null (Lbx1; Middle), and Lbx1 (Right) mutant newborn mice. Pax2 (green) and Lmx1b (red) antibodies distinguish inhibitory and excitatory somatosensory neurons of the spinal trigeminal nucleus, respectively. NTS neurons express Lmx1b, and inferior olivary neurons express Foxp2 (blue). (B) Histological analysis of limb muscles in control, Lbx1, and Lbx1 newborn mice, using antibodies against laminin (Lam, red) and desmin (green). Confocal tile scan modus was used to acquire photomicrographs, and assembled using ZEN2012 software (10% overlap between tiles). Photomicrographs were mounted on a black frame to maintain figure panel proportions.

dB2 Neurons Are Responsible for the Breathing Deficits Observed in Lbx1 Mice.

To assess whether the breathing deficits observed in Lbx1 mice exclusively depend on dysfunction of dB2 derivatives, we conditionally restricted the Lbx1 mutation to the dB2 lineage by using Phox2b (Phox2b;Lbx1, named dB2-Lbx1 mice; see for a scheme of the strategy). In such animals, neurons with a history of Phox2b expression carried an Lbx1 genotype, but other cells (Lbx1) retained one copy of a fully functional Lbx1 allele (). In dB2-Lbx1 animals, RTN neurons were absent (Fig. 4). Plethysmographic recordings of dB2-Lbx1 animals showed a full recapitulation of the physiological phenotype observed in Lbx1 animals (i.e., severe hypoventilation, lack of the hypercapnic reflex, frequent apneas; ; summarized in Fig. 4), as well as lethality (n = 12/12) within the first 2 h of life. Thus, all respiratory deficits associated with the Lbx1 mutation are the result of a selective developmental deficit in the dB2 neuronal lineage.

Fig. 4.

Conditional mutagenesis restricts the Lbx1 mutation to specific neuronal subpopulations. (A) Analysis of Lbx1+ (red) Phox2b+ (green) RTN neurons (arrowheads) in control, dB2-Lbx1 and Egr2-Lbx1 newborn mice. DAPI (blue) was used as a counterstain, and the facial (nVII) motor nucleus is indicated. Confocal tile scan modus was used to acquire photomicrographs, and assembled using ZEN2012 software (10% overlap between tiles). (B, Top) Comparison of dB2 neuron numbers: RTN neurons [one-way ANOVA, F(3, 15) = 883.4]; dorsal Lbx1+/Phox2b+ neurons [one-way ANOVA, F(3, 15) = 370] and periV neurons [one-way ANOVA, F(3, 15) = 0.1562] in control (n = 6) Lbx1 (n = 4), Egr2-Lbx1 (n = 4), and dB2-Lbx1 (n = 4) mice. (B, Bottom) Comparison of ventilatory minute volumes in normal air and in hypercapnia in control (n = 20) Lbx1 (n = 11), Egr2-Lbx1 (n = 11), and dB2-Lbx1 (n = 12) mice (unpaired nonparametric Mann–Whitney U test).

RTN neurons arise from rhombomere 5 (29). We next restricted the Lbx1 mutation to rhombomeres 3 and 5, using Egr2 that only recombines cells in these rhombomeres (Egr2;Lbx1, named Egr2-Lbx1 mice; see for a scheme of the strategy). As expected, RTN neurons were absent in Egr2-Lbx1 animals (Fig. 4). Plethysmographic recordings of Egr2-Lbx1 mice showed that they were unable to respond to high CO2 levels in the air (). Nevertheless, Egr2-Lbx1 mice did not display apneas and survived the postnatal period (n = 11/11), with a mild hypoventilation that was observed in their early postnatal life (). The response of Egr2-Lbx1 mice to high levels of CO2 improved with maturation, but even adult mutants presented a blunted hypercapnic reflex (). This phenotype, largely similar to the one observed after conditional mutation of Phox2b in rhombomeres 3 and 5 (22), implies that several neuronal groups originating from dB2 precursors participate in the control of breathing.

We next used intersectional lineage tracing to specifically label dB2 derivatives with Tomato fluorescent protein, using Lbx1;Phox2b;Ai65 animals (see for a scheme of the strategy). Tomato+/Lbx1+/Phox2b+ cells were found, in addition to the RTN, around the trigeminal motor nucleus in rhombomere 1 and 2 (a population known as periV neurons), as well as in the dorsal part of rhombomeres 3–6 (). We compared development of these two dB2 derivatives (periV neurons and neurons in the dorsal part of the hindbrain) in strains displaying the most severe breathing phenotype (i.e., Lbx1, dB2-Lbx1) and the milder breathing deficit (Egr2-Lbx1). Lbx1+/Phox2b+ periV neurons were present in normal numbers in all analyzed strains (quantified in Fig. 4). However, the number of dorsally located Lbx1+/Phox2b+ neurons was severely reduced in Lbx1 and dB2-Lbx1 animals, but not obviously affected in Egr2-Lbx1 mice (Fig. 4 and ). Thus, the absence of the RTN combined with the reduction of the dorsal Lbx1+/Phox2b+ population correlates with the severe breathing phenotype observed in Lbx1 and dB2-Lbx1 mutants.

Ectopic Expression of Somatosensory Genes in Lbx1/Phox2b Expressing Neurons.

To assess whether the absent dB2 neurons in Lbx1 mice assumed an aberrant neuronal fate, we extended our intersectional genetic lineage tracing to Lbx1 (Lbx1;Phox2b;Ai65; see ) mutant mice. This demonstrated that ectopic Tomato+ cells appeared in the somatosensory SpV nucleus of Lbx1 mice, which were not observable in control animals (Fig. 5 and ). These ectopic Tomato+ cells coexpressed markers of excitatory somatosensory neurons such as Prrxl1 or Lmx1b (Fig. 5 and ). Thus, the Lbx1 mutation selectively affects the development of an Lbx1+/Phox2b+ dB2 subpopulation that adopts an aberrant somatosensory fate.

Fig. 5.

Mis-specification of dB2 neurons in Lbx1 mice. (A and B) Intersectional labeling of dB2 neurons in control (Lbx1;Phox2b;Ai65) and Lbx1 (Lbx1;Phox2b;Ai65) newborn mice (). (A) Histological analysis of dB2-Tomato+ (red) neurons located in the dorsal part of the SpV nucleus in control and Lbx1 newborn mice. DAPI (blue) was used to counterstain. Insets are magnifications (300×) of the boxed areas stained with the somatosensory specific Prxxl1 (green) and Tomato (red) antibodies. Confocal tile scan modus was used to acquire photomicrographs and was assembled using ZEN2012 software (10% overlap between tiles). (B) Quantification of Tomato+ and Tomato+/Prrxl1+ neurons in control and Lbx1 mice at birth (n = 4 per genotype; unpaired t test, t = 17.03; df = 6). (C) Hierarchical transcriptome clustering of Lbx1-, Phox2b-, Lbx1/Phox2b-, and Lbx1FS/Phox2b-expressing neurons (n = 3 independent replicates). Color code represents intersample distances (in reads per kilobase per million mapped reads). (D) Normalized read counts for Pou4f1, Lmx1b, and Prrxl1 transcripts in neurons expressing Lbx1, Phox2b, Lbx1/Phox2b, and Lbx1FS/Phox2b (n = 3 independent replicates). (E) ChIP-qPCR analysis performed on chromatin prepared from Lbx1/Phox2b and Lbx1FS/Phox2b expressing neurons using antibodies against flag-tag (for Lbx1 and Lbx1FS immunoprecipitation), HA-tag (for Phox2b immunoprecipitation), H3K27ac and, H3K27me3 (n = 4 independent replicates). Analyzed sites in the Prrxl1 locus are indicated schematically as blue triangles (Fig. 1). ***P < 0.0001.

We next modeled the (dys)function of Lbx1FS in Phox2b+ neurons, using our cell culture model. For this, Lbx1 mutant P19 cells were transduced with retroviruses encoding a HA-tagged version of Phox2b (hereafter Phox2b cells) alone or in combination with flag-tagged Lbx1 or Lbx1FS (Lbx1/Phox2b and Lbx1FS/Phox2b cells). We then sequenced the transcriptomes of neurons differentiated from these cells. Hierarchical expression clustering showed that Phox2b, Lbx1/Phox2b, and Lbx1FS/Phox2b neurons were clearly distinct from Lbx1 neurons and clustered separately (Fig. 5). Nevertheless, Lbx1 and Lbx1FS/Phox2b neurons were more closely related to each other than to Lbx1/Phox2b or Phox2b neurons (Fig. 5). Interestingly, the Prrxl1, Lmx1b, and Pou4f1 somatosensory genes were among the most significant and differentially expressed genes in Lbx1FS/Phox2b+ neurons compared with Phox2b or Lbx1/Phox2b+ neurons (Fig. 5). Thus, Phox2b represses these somatosensory genes alone or even when Lbx1 is present, but this does not occur when Lbx1FS and Phox2b are coexpressed, a change reminiscent of the one observed in vivo where Lbx1FS+/Phox2b+ (dB2) neurons assumed an aberrant somatosensory fate.

Next we analyzed chromatin modifications of the previously characterized enhancers of Prrxl1, Lmx1b, and Pou4f1 somatosensory genes. In Lbx1/Phox2b neurons, ChIP-qPCR showed a modest enrichment of Lbx1 and Phox2b at the analyzed loci (Fig. 5 and ). Moreover, the chromatin mark H3K27me3 was enriched in those sites, demonstrating that the enhancers are repressed. However, when the chromatin of Lbx1FS/Phox2b neurons was used for ChIP-qPCR experiments, Lbx1, Phox2b, and H3K27ac were significantly enriched at the Prrxl1, Lmx1b, and Pou4f1 enhancers (Fig. 5 and ). Thus, enhancer sequences of the Prrxl1, Lmx1b, and Pou4f1 genes are activated when Lbx1FS and Phox2b are recruited to these sites, but repressed when Lbx1 and Phox2b are recruited.

Discussion

Respiratory disorders in humans range from irregular and unstable respiration to the complete loss of breathing control. The most common causes of congenital hypoventilation are dominant mutations in PHOX2B that affect the formation of the RTN. Here we show that a homozygous frameshift mutation in LBX1 causes severe congenital hypoventilation that resembles classical CCHS. We used cell culture and mouse models to investigate the (dys)function caused by the frameshift mutation, which alters the C-terminal sequence of the protein but spares its homeodomain. In most developmental contexts, the mutant protein exerts its role correctly; that is, the mutation only interferes with small subsets of Lbx1 functions. Our analysis has thus revealed a very unusual pathomechanism of a transcription factor mutation that results in a severe respiratory disorder.

Lbx1FS Protein Correctly Functions in Most Developmental Contexts.

Our cell culture modeling of Lbx1FS binding showed that Lbx1FS and Lbx1 largely bind to similar sites genome-wide, which is in agreement with conserved functionality of the Lbx1FS protein in most developmental contexts. Motif analyses revealed subtle differences between the binding preferences of Lbx1 and Lbx1FS. In particular, a specific motif was present in Lbx1, but not in Lbx1FS binding sites, which consists of a 16-bp-long nonpalindromic sequence that is composed of an Lbx1-monodimer site combined with a half-site of another factor. Interestingly, the sequence that represents the second half-site corresponds to the preferred binding motif previously identified for Phox2b (27). This observation suggested a failure of Lbx1FS to cooperate productively with Phox2b.

Lbx1 and Phox2b are known to functionally repress each other: When Lbx1 is mutated, supernumerary Phox2b viscerosensory neurons arise (7). Vice versa, mutation of Phox2b results in the appearance of supernumerary somatosensory Lbx1 neurons (30). Remarkably, development of the dB2 lineage depends on both Lbx1 and Phox2b and relies on the repression of the somatosensory genes (7, 13). Lbx1-dependent differentiation of somatosensory neurons can be modeled in vitro and occurs in the presence of Lbx1 and Lbx1FS. Interestingly, coexpression of Phox2b represses somatosensory genes in Lbx1+ but not Lbx1FS+ neurons. In the presence of Phox2b, the altered C-terminal sequence of Lbx1FS might impede the correct recruitment of coregulatory factors, thus accounting for the fact that Lbx1FS is unable to correctly cooperate with Phox2b.

LBX1/Lbx1 in CCHS.

Here we demonstrate that the hypomorphic Lbx1 mutation selectively interferes with the development of specific dB2 neuronal populations. Physiologically, Lbx1 mice display a plethora of respiratory deficits: slow and irregular breathing, lack of hypercapnic reflex, and frequent and prolonged apneas. Together, these deficits appear to result in neonatal lethality. We observed that the conditional restriction of the Lbx1 mutation to the dB2 lineage (dB2-Lbx1 mice) fully recapitulates the physiological phenotypes observed in Lbx1 mice. In contrast, the conditional restriction of the Lbx1 mutation to rhombomeres 3 and 5 (Egr2-Lbx1FS mice) impaired RTN neuron development, abolished the hypercapnic reflex, and caused mild hypoventilation, but not abnormal apneas or neonatal mortality. Interestingly, similar or even identical phenotypes are observed when the Phox2b mutation is restricted to rhombomeres 3 and 5 (22). Thus, the Lbx1 mutation causes respiratory deficits that are in part, but not completely, a result of the loss of RTN neurons.

Last, we report in this study that dB2 precursors produce, in addition to the RTN and periV cells, an additional not previously described group of Lbx1+/Phox2b+ neurons that locate dorsally in rhombomeres 3–6. Because of the complexity of the developmental deficits displayed by Lbx1 null mutant mice, the contribution of individual cell populations to respiratory deficits had previously not been assessable. We used here intersectional genetic strategies to show that the combined deficits in development of RTN and the dorsal Lbx1+/Phox2+ population correlated with severe hypoventilation and neonatal lethality. Further studies will be needed to define the connectivity and the exact function of this dorsal neuronal population.

Materials and Methods

Research Involving Humans and Mice.

Venous blood and genomic DNA samples from humans were obtained by standard procedures. Written informed consent was obtained from all individuals. Experimental procedures and animal handling were conducted according to institutional protocols and guidance approved by the Max Delbrueck Center (Berlin), CNRS (Gif sur Yvette), Max Planck Institute for Genetics (Berlin), and the Ethic Committee of the Charité Universitätsmedizin (Berlin). Details on mouse strains are provided in .

Histology.

Development of dB2 neuronal derivatives was assessed on 20-μm transverse hindbrain sections from control and mutant mice. Details on antibodies and in situ probes used in this study are provided in .

Cell Cultures.

P19 embryonic teratocarcinoma cells were obtained from ATTC (CRL-1825) and differentiated into neurons using 1 μM retinoic acid (Sigma), as described (31). Details on CRISPR-CAS9 mutation of Lbx1 in P19 cells, retroviral infection, ChIP, and deep sequencing experiments are provided in .

Physiology.

Unrestrained plethysmographic recordings of individual mouse pups were carried out as described (32). Further details on plethysmographic recordings and Ca2+ imaging studies can be found in .