To investigate oro-buccal behavior, we counted the number of jaw openings elicited by an oral stimulation in Hoxa2^-/- mice and compared the phenotype to that of Krox20^-/- mice. In Krox20^-/- mice, a reduction by about 50% of the number of jaw openings indicated an alteration of the trigeminal pre-motor and/or motor control, normally originating in the r2-r3 region 17. In Hoxa2^-/- mutant mice, we found that the number of jaw openings during 30 s was decreased by 45% (P < 0.0001), ranging from 13.6 at postnatal day (P)0 to 16.6 at P0.5, significantly less than values of 24 at P0 and 32 at P0.5 measured in the wild-type (Table 1). Elicitation of jaw opening was normal in Hoxa2^+/- (Table 1) and Krox20 hypomorphic mutants (22 during 30 s (n = 7) versus 25 in the wild-type (n = 17)), indicating that full inactivation of either gene was required to affect behavior. The small number of evoked jaw openings at P0, coupled with the respiratory impairment described below, defined a highly penetrant (72%; 16 out of 22 homozygous mutants; Figure 1b, c) functional phenotype in Hoxa2^-/- mutant mice.

At P0.1, the tidal volume per body weight of Hoxa2^-/- homozygous mutants was twice as large as that of the wild-type or Hoxa2^+/- heterozygous mutant mice (Table 1). Samples of plethysmographic recordings (Figure 1a) show that this increase in respiratory amplitude (tidal volume (V_T)) was not compensated for by any decrease in respiratory frequency, which was similar in the three genotypes at P0.1 (Table 1). Consequently, Hoxa2^-/- animals exhibited a greater than normal average respired volume (minute ventilation (V_E) = f_R × V_T) at P0.1 (Table 1). Since there was no consistent difference in the duration of inspiration (Ti), these animals also showed a greater than normal inspiratory flow (V_T/Ti). This was very different from the observations in the Krox20 null mutant mice, which showed a V_T similar to their wild-type littermates (Figure 2b) during the first postnatal days.

In human infants, the first inspiratory efforts are the deepest breaths of the whole neonatal period 313233. The V_T is later reduced and increases afterwards progressively during several hours (see page 26 of 34). The same was observed in the wild-type (Figure 2a, b and inset) and Krox20^-/- (Figure 2b) mice, in which values of V_T were reduced to about 5 μl/g at P0.1 and increased to about 10 μl/g at P0.5, which is the normal value maintained during the postnatal week (Figure 2a, b). At P0.5, the absolute tidal volume values of all genotypes were very similar, ranging from 8–11 μl/g. The tidal volume of Hoxa2^-/- mutant mice was the same as at birth, indicating that the decrease in V_T observed in wild-type mice after the first breaths was impaired by the mutation (Figure 2a). Thus, the Hoxa2 invalidation perturbs a developmentally regulated mechanism involved in the control of the tidal volume during the first 12 postnatal hours.

Respiratory frequency (f_R) also changed during the first day but increased only by about 50% (Figure 3a). The values of respiratory frequency at birth were about the same in Hoxa2^-/- mutants as in wild-type animals and both showed an increase in frequency between P0.1 and P0.5 (Table 1, Figure 3a). The changes of V_T during the first hours were, therefore, mostly responsible for the modifications of the V_E and both were abolished in Hoxa2^-/- mutant mice. This contrasted with the respiratory phenotype associated with the Krox20 null mutation resulting in a respiratory frequency 60% less than in the wild-type (42 ± 24 breaths/minute at P0.5, n = 17; Figure 3b). Very long apneas (> 3 s) were frequently observed and an increase in inspiratory flow did not compensate for the decreased frequency, thus minute ventilation at P0.5 (0.48 ± 0.38 ml/g/minute, n = 17) was less than half the wild-type values (1.10 ± 0.79 ml/g/minute, n = 25).

Thus, the respiratory phenotype of the Krox20 null mutation was dramatically distinct from that observed in Hoxa2 null mutants. Hoxa2 null mutation specifically affected the inspiratory control at birth, without modifying the respiratory frequency, whereas Krox20 null mutation had the exact opposite effect, deeply affecting the respiratory frequency without changing the respiratory amplitude.

Krox20 is required for the development of r3, where Hoxa2 is expressed under Krox20 control 4; nevertheless, the respiratory phenotype of the Krox20 null mutation was quite distinct from that observed in Hoxa2 null mutants (see above), suggesting that it was not induced by the loss of Hoxa2 expression in r3. However, as the Krox20 mutation results in the complete elimination of r3, it cannot be compared to the changes induced in the Hoxa2 null mutants, in which r3 is still present but displays patterning defects 1011. The abnormal inspiratory control in Hoxa2 mutants might, therefore, originate from an adaptive respiratory behavior following compensation for the loss of Hoxa2 in r3, while Krox20 is still functional.

To investigate this possibility we analyzed Krox20^Cre/flox hypomorphic mutants resulting in a reduced, though not absent, r3 territory. This hypomorphic mutant was obtained by combining two previously developed Krox20 alleles, a Cre knock-in and a floxed allele (see Materials and methods for detailed description). Compound heterozygous Krox20^Cre/flox mutants express Krox20 only transiently and, in the hindbrain, this results in a severe reduction of r3 (Figure 4). Analysis of the Krox20^Cre/flox mutants revealed that the tidal volume was normal (115% of controls) as well as the respiratory frequency (P0.5: 101 ± 6 breaths/minute (n = 7) compared to 114 ± 10 breaths/minute (n = 17) in wild-type littermates; Figure 4) and the duration of apneas during the first postnatal day. Therefore, partial impairment of Krox20 function, resulting in a severe reduction of r3, is compatible with a normal control of the respiratory rhythm at birth and does not reproduce the Hoxa2 null phenotype characterized by the absence of a transient decrease in tidal volume around birth.

Because development of rhythmic circuits in r3r4 has been ascribed to r3-related control of neurogenesis in r4 27, we analyzed the early pattern or neuronal differentiation in Krox20Cre/flox mutants (Figure 4c, d). We found that r3, although dramatically reduced in size, preserves its ability to delay neurogenesis and axonal invasion 35. We conclude that function of the anti-apneic control does not depend on the ability to maintain quantitatively the size of embryonic territories, but rather as previously suggested 27, on odd-numbered rhombomere properties required to qualitatively control neuronal circuit formation.

Taken together, these results establish that Hoxa2 does not mediate the Krox20 null respiratory phenotype in r3 since Hoxa2 null mutation does not reproduce Krox20 null mutation phenotypic traits. In addition, they suggest that the Hoxa2 mutant respiratory phenotype may be contributed by abnormalities in pontine areas more rostral than the Krox20 phenotype, since Hoxa2 has greater effects in its rostral domain of expression 1036.

At P0, the respiratory pattern was very irregular in Hoxa2^-/- mutant mice and the time spent in apneas (that is, respiratory pauses lasting more than 2 s) was about 10% of the total time of observation, similar to heterozygous and wild-type animals (Table 1). At P0.5, the time spent in apneas in wild-type animals normally decreased to 3.2 ± 2.0% of the recording time. In contrast, no decrease was observed in Hoxa2^-/- mutants (time spent in apneas at P0.5, 17.5 ± 5.5%) and at P0.5 these animals spent significantly more time in apneas than controls. However, apneas did not greatly influence the average volume inhaled by Hoxa2^-/- mutants. At P0.5, shortly before the death of the animals, the minute ventilation in homozygous mutants (1.36 ± 0.42 ml/g/minute) was not significantly different from that at birth (1.21 ± 0.14 ml/g/minute) or from that in the wild-type (1.71 ± 0.11 ml/g/minute) or heterozygous (1.43 ± 0.09 ml/g/minute) littermates.

Interestingly, administration of naloxone at P0.5 failed to improve survival of Hoxa2^-/- mutants (Figure 5), unlike Krox20^-/- and Hoxa1^-/- mutants 1718. In both wild-type and heterozygous Hoxa2 mutant animals, naloxone treatment had no significant effects on the breathing pattern. In the homozygous Hoxa2^-/- mutant mice (n = 7), naloxone injection at P0.5 slightly increased respiratory frequency from 135 breaths/minute to 177 breaths/minute (Figure 5b) and eliminated apneas (Figure 5a). Despite these stimulating effects upon ventilation, none of the treated Hoxa2^-/- mutants lived more than 18 hours (Figure 5c), the maximum lifetime observed in untreated Hoxa2^-/- animals. Altogether, these results demonstrate that apneas do not greatly influence the respiratory minute ventilation and are not responsible for lethality of the Hoxa2 mutation.

We analyzed the distribution of Enhanced Green Fluorescent Protein positive (EGFP+) cells in knock-in mice in which EGFP expression was selectively induced in r2-derived Hoxa2 expressing cells upon Cre-mediated recombination, by mating the Hoxa2^{EGFP(lox-neo-lox)} knock-in allele with a r2-specific Cre transgenic line 12 (Figure 6a, b, d, e, g; see also Materials and methods). This mating scheme resulted in full Hoxa2 targeted inactivation, and, concomitantly, allowed selective tracing of the Hoxa2-expressing EGFP⁺ cells only in the r2-derived progeny.

Using heterozygous Hoxa2^{EGFP(lox-neo-lox)/+};r2::Cre animals at E18.5, we showed that the estimated lateral parabrachial medial and Kölliker-Fuse nuclei (Figure 6b, e), as well as the more medial noradrenergic locus coeruleus, identified by PHOX2A or tyrosine hydroxylase (Figure 6g), were located outside, though adjacent to, the trigeminal r2-derived domain. Therefore, the putative inspiratory related domain is derived from r1, about two rhombomeres distant from the r3-r4 region, the most anterior site known to induce formation of a respiratory frequency controller 1726.

Analysis of EGFP expression in E18.5 Hoxa2^{EGFP(lox-neo-lox)/-};r2::Cre homozygous mutant mice (Figure 6c, f, h) indicated that the r2-derived peri-trigeminal nuclei were reduced while a ventral hyperplasia of the estimated Kölliker-Fuse (KF in Figure 6a–c) and parabrachial medial (PB in Figure 6d–f) nuclei was observed (arrowheads in Figures 5f and 6c). Other r1-derived structures, such as the locus coeruleus (LC in Figure 6c and Table 1) and the pediculo-pontine tegmental nuclei also expanded ventrally.

The anatomical structures modified in Krox20^-/-, Hoxa1^-/- and kreisler mutants 17182937 have also been investigated in Hoxa2^-/- neonates. Specifically, we found a normal morphology of the anterior fourth ventricle, a normal AP length of the dorsal Pons, and a normal position of the ambiguus and facial branchial motor nuclei (data not shown). The parvocellular reticular nucleus, a dorsal pontine structure originating in r3 and extending between the trigeminal motor nucleus and the facial nerve 171829 was also normal in Hoxa2^-/- mutants.

In conclusion, the brainstem defects likely responsible for the observed abnormalities of inspiratory control are restricted to the rostral pons of Hoxa2^-/- mutants and include the r1-r2 derived territory.

Consistent with the original observations of the Hoxa2^-/- defects being restricted to the anteriormost domain of Hoxa2 expression 1036, the function of respiratory rhythm controllers located in the caudal pons and medulla was normal in Hoxa2^-/- animals. The phenotypic defects of the Hoxa2^-/- mice that were not observed in Krox20 mutants are likely to be related to the reorganization of neural structures derived from the alar plate at the r1-r2 level of the brainstem 10. The distribution of cells expressing EGFP in the pons of Hoxa2^{EGFP(lox-neo-lox)/-)};r2::Cre homozygous mutants pons suggests that the territories derived from r2 are reduced along the AP axis or have lost r2-specific characteristics. Lack of an r1/r2 boundary 101112 explains that territories derived from r1 and adjacent to r2 are also modified, with, for example, an ectopic projection of the ascending branch of the sensory trigeminal nerve to the cerebellum 13. The abnormal location of locus coeruleus neurons is also consistent with an altered positioning of neurons within the r1-derived territory 11. Thus, anatomical observations indicate that a large part of the rostral pontine territory deriving from r1-r2 is reorganized in Hoxa2^-/- mutants.

The r1-derived territory adjacent to r2 includes the estimated location of medial parabrachial and Kölliker-Fuse nuclei where a pontine respiratory group regulates adult mammalian inspirations 38394041 (for review see 2842). Therefore, Hoxa2 might be required for normal development of the pontine respiratory group. Our data further suggest that the pons contributes in shaping inspirations during the first postnatal hours in intact mice in vivo.

In addition, catecholaminergic 43 and cholinergic 44 control of breathing are probably affected. Interestingly, a large V_T with normal respiratory frequency in vivo has been observed after the inactivation of the gene encoding acetylcholinesterase, a procedure found to greatly reduce the muscarinic and nicotinic control of the respiratory generator in vitro 44. It is possible, therefore, that abnormalities of the pediculo-pontine tegmental nucleus (PPTg), the major source of ponto-bulbar cholinergic neurons, contribute to the control of inspirations in Hoxa2 mutants.

Selective inspiratory control by the rostral pons has been postulated to explain postnatal respiratory deficits following the exposure to sub-teratogenic doses of retinoic acid at the onset of hindbrain segmentation (E7.5) 29. We presently show that development of inspiratory control requires rhombomere-related expression of the most rostral Hox gene in the neural tube. These results further suggest that the genetic network controlling hindbrain segmentation is involved in the formation of neuronal circuits controlling breathing, thereby giving the modular organization of the ponto-bulbar respiratory network 26.

As shown recently, Hoxa2 is required in r2-r3 for trigeminal nerve pathfinding at early stages and axonal arborization of afferents at later stages 13. Here we show that the anomaly of facial circuits in Hoxa2^-/- mutant embryos is associated with an impairment of the reflex-induced rhythmic oro-buccal behavior at birth. Because the Hoxa2^-/- oro-buccal behavior resembles that induced by the elimination of r3 in Krox20^-/- 17, trigeminal behavioral deficits may involve both r2- and r3-derived processes, including the Hoxa2-dependent arborization of sensory axons in the rostral principal trigeminal nucleus 13. In addition, Hoxa2^-/- and Krox20^-/- deficits may involve rhythmic pre-motor reticular neurons in the vicinity of the trigeminal motor nucleus controlling rhythmic opening of the jaw during alimentary behaviors 14151618http://www.jneurosci.org/cgi/external_ref?access_num=000077737200006&link_type=ISI. This deficit in suction might participate in the early death of Hoxa2 null mutant mice, impeding the correct feeding behavior together with extensive transformation of the first branchial arch-derived facial skeleton 236.

The most rostral rhombomere that has been shown to be required for the development of normal respiratory rate and to prevent apneas is r3. Elimination of Krox20 expression in r3 reduces quiet breathing frequency by 50% and multiplies by 10 the time spent in apnea, eventually leading to opioid-sensitive lethality 1726. Because, in contrast, hypomorphic Krox20 neonates breathe at a normal rate without apneas, Krox20 expression must be entirely eliminated to alter respiration. Our results, therefore, suggest that the transient expression of Krox20 and the subsequent reduction of r3 size observed in hypomorphic Krox20 mutants do not prevent the development of the para-facial anti-apneic system. This result extends to mice previous observations in chick embryos using loss- and gain-of function strategies 27. Krox20 expression in r3 was found necessary for the non-cell autonomous induction of a neuronal rhythm controller from r4 27. In contrast, it was found to be unlikely that this induction requires the generation of a specific population of r3 neurons because KROX20 inhibits (rather than stimulates) neurogenesis and neuronal differentiation 35. Furthermore, ectopic Krox20-expression in a limited territory (for example, obtained by unilateral electroporation in a rhombomere) was sufficient to induce a fully functional rhythm generator, although the population of neurons specified by Krox20 was certainly smaller than normal r3 in these preparations 27. We presently show that, in mice as in chicks, induction of the anti-apneic parafacial function by Krox20 is a robust process that persists despite quantitative alterations of pontine cell populations. Our working hypothesis is, therefore, that Krox20 may act by initiating a cell non-autonomous control of specific neuronal fates rather than by generating a respiratory-related neuronal population in a cell autonomous manner. In addition, rhythm generators may compensate for minor abnormalities during fetal development to restore normal function at birth 2944. To further explore the parafacial development in mouse embryos, experiments are in progress in our laboratories to identify cell lineages that could be targets of the induction initiated by Krox20 expression.

High levels of Hoxa2 are expressed in r3 and are required at late stages (E13) in trigeminal principal sensory neurons to induce arborization of whisker-related maxillary primary afferents 13. In contrast, we show that Hoxa2 is not necessary in r3 for the development of the Krox20-dependent anti-apneic respiratory frequency controller; because in Hoxa2^-/- neonates respiratory frequency was normal, apneas were not life-threatening and treatment with naloxone had no effect on survival. In r3, Hoxa2 function downstream of Krox20 may be partially redundant with that of its paralogue Hoxb2 11, also a direct target of Krox20 3. Functional redundancy of Hox2 paralogs can be expected at early developmental stages (end-segmental stages, about E9.5), during which Krox20 expression initiates formation of rhythm generators. Indeed, synergistic genetic interaction of Hoxa2 and Hoxb2 has been shown for the early patterning of r3 11. Redundancy might be less at later stages (about E13), when Hoxa2 exerts cell type-specific functions 13. Previous observations 36 have shown that the anti-apneic activity abolished in Krox20^-/- and Hoxa1^-/- mutants is preserved in kreisler mutants lacking r5. Altogether, the available data support the location of the anti-apneic activity to be within the para-facial r3r4-derived territories and support Krox20, but not Hoxa2, as a major player in this process.

A total of 38 wild-type, 58 Hoxa2^+/- and 29 Hoxa2^-/- littermates resulting from crosses between heterozygous animals 101112 were used in the present study. These numbers are in accordance with Mendelian repartition of genotypes (χ² test, P = 0.469). The Hoxa2^{EGFP(lox-neo-lox)} mouse knock-in allele allows selective activation of EGFP expression from the Hoxa2 locus only upon Cre recombinase mediated recombination, as described in 12. The r2::Cre transgenic line allows selective expression of Cre in r2 and its derivatives, as described in 45. The original Hoxa2 null mutation is described in 36. The phenotypes of Hoxa2^-/- and Hoxa2^{EGFP(lox-neo-lox)}/^{EGFP(lox-neo-lox)} homozygous mutants are indistinguishable. The analysis of Hoxa2^{EGFP(lox-neo-lox)}/⁺ and Hoxa2^{EGFP(lox-neo-lox)}/^- specimens allowed, therefore, the comparison of heterozygous and homozygous Hoxa2 mutants carrying only one dose of EGFP in both genotypes. DNA was extracted from the tail of the neonate mouse and the genotype was subsequently determined by a PCR assay using specific sets of oligonucleotide primers, as described in 12. All homozygous Hoxa2^-/- animals did not feed, lost 6.6% of their birth weight in their first 18 hours of life and died within 12–20 hours after birth, whereas their heterozygous or wild-type littermates fed, gained about 15% of their birth weight during the same period, survived (Table 1; see also 3645) and were, therefore, studied during the first week after birth.

We also analyzed 7 hypomorphic Krox20 mutants at P0.5 and P3-4 and we reinvestigated the breathing behavior of 17 Krox20^-/- animals 17 during the first hours following birth. The hypomorphic Krox20 mutant was obtained by combining two previously developed Krox20 alleles: the Krox20^Cre allele consists of an insertion of the gene for the Cre recombinase into the Krox20 locus, resulting in Krox20 inactivation and expression of the Cre gene with a pattern that faithfully recapitulates the normal Krox20 pattern 46; in the Krox20^flox allele, the second Krox20 exon is flanked by loxP sites – this allele behaves like the wild-type until excision of the floxed exon by the Cre recombinase results in inactivation of Krox20 3046. The compound heterozygous animals, Krox20^Cre/flox, express Krox20 only transiently, due to subsequent elimination of the second exon. In the hindbrain, this combination behaves as a hypomorphic mutation, resulting in a severe reduction in, but not elimination of, r3, and a slight reduction of r5 (Figure 4).

Respiratory activity was measured using a modified barometric method, previously employed in neonates, called whole-body plethysmography 37. The plethysmograph chamber (20 ml) equipped with a temperature sensor was connected to a reference chamber of the same volume. The pressure difference between the two chambers was measured with a differential pressure transducer connected to a sine wave carrier demodulator. The spirogram was stored on a computer using a Labmaster interface at a sampling frequency of 1 kHz. Calibrations were performed at the end of each recording session by injecting 2.5–5 μl of air in the chamber with a Hamilton syringe.

Measurements started 0.5–2 hours after birth and were repeated every 4–6 hours. Neonates were removed individually from the litter and placed in the plethysmograph chamber, which was kept hermetically closed and maintained at 31°C during the recording session (165 s). In each sample, periods of quiet breathing were identified by the absence of limb or body movements. Periods of limb, body and head movements were measured to determine neonatal activity during recordings. During quiet breathing, a computer-assisted method was used to measure the duration of inspiration and expiration from which respiratory frequency (f_R) is derived and the tidal volume (V_T, μl/g) from which minute ventilation (V_E = f × V_T/1,000, ml/g/minute) is derived. Total duration of apneas per recording session (given in percentages) provided an estimation of the apneic breathing. Apneas were included in the calculation of the minute ventilation, giving a mean value of this physiological parameter. In the case of naloxone treatment, separate calculations of respiratory frequency with and without inclusion of apneas were performed. Naloxone (1 mg/kg in saline) was administered subcutaneously using a Hamilton syringe, and the next recording carried out one hour after injection.

Before plethysmographic recordings, the number of jaw openings within 30 s was counted by introducing a catheter into the mouth of the newborns, therefore testing oro-buccal reflexes 17. In order to obtain comparable data, counting was done by the same person for the whole study. Both plethysmographic recordings and jaw opening counting were done blindly, without knowing the genotype. Values are given as mean ± standard error of the mean. Comparisons between two sets of data were performed by paired or unpaired Student's T-tests.

E18.5 dpc (days post coitum) Hoxa2^{EGFP(lox-neo-lox)/+};r2-Cre and Hoxa2^{EGFP(lox-neo-lox)/EGFP(lox-neo-lox)};r2-Cre foetuses were obtained by caesarean section. Brains were dissected out, fixed in 4% paraformaldehyde, 1 × phosphate buffer saline (PBS) at 4°C overnight, cryoprotected, and sagittally sectioned. Adjacent sections were processed for immunohistochemistry using anti-EGFP (1:2000; Molecular Probes (Carlsbad, California, USA)) or anti-Phox2a (1:1,000; a gift from Jean-François Brunet, ENS, Paris, France) polyclonal antibodies. Sections were then incubated for 2 hours with biotinylated secondary antibodies (1:200, pH 7.4; Vector

(Burlingame, California, USA)