Neonatal Lethality, Dwarfism, and Abnormal Brain Development in Dmbx1 Mutant Mice

Abstract

Dmbx1 encodes a paired-like homeodomain protein that is expressed in developing neural tissues during mouse embryogenesis. To elucidate the in vivo role of Dmbx1, we generated two Dmbx1 mutant alleles. Dmbx1⁻ lacks the homeobox and Dmbx1^z is an insertion of a lacZ reporter gene. Dmbx1^z appears to be a faithful reporter of Dmbx1 expression during embryogenesis and after birth. Dmbx1-lacZ expression was detected in the superior colliculus, cerebellar nuclei, and subpopulations of the medulla oblongata and spinal cord. Some Dmbx1 homozygous mutant mice died during the neonatal period, while others survived to adulthood; however, their growth was impaired. Both heterozygous and homozygous mutant offspring from Dmbx1 homozygous mutant females exhibited a low survival rate and poor growth. However, even wild-type pups fostered onto Dmbx1 homozygous mutant females grew poorly, suggesting a Dmbx1-dependent nursing defect. Dmbx1 mutant mice had an aberrant Dmbx1-lacZ expression pattern in the nervous system, indicating that they had abnormal brain development. These results demonstrate that Dmbx1 is required for postnatal survival, growth, and brain development.

The central nervous system (CNS) is derived from the neural plate, a tissue that arises from the anterior epiblast of the mouse gastrula (28). Complex morphogenetic processes transform the relatively flat neural plate into the neural tube. By the time the neural tube is formed, it has differentiated morphological segments along the anterior-posterior axis and is divided into several compartments called neuromeres. The neuromeres in the forebrain and hindbrain comprise six prosomeres and eight rhombomeres, respectively. This segmentation is induced and maintained by the expression of segment-specific genes, including homeobox genes that encode homeodomain-containing transcription factors.

The homeodomain is a conserved DNA-binding motif that consists of ∼60 amino acid residues. The homeoproteins are divided into two groups, clustered and dispersed, according to how they are distributed on the chromosomes. Mammals have four Hox gene clusters. Homeobox genes belonging to the dispersed class are further divided into several subclasses, primarily by the homeodomain sequences.

There are many homeobox genes that are known to be expressed in developing brains and/or mature nervous systems in mice (32). Several Hox genes are expressed in specific rhombomeres and are required for the identification of neuromeres. In the forebrain, several dispersed-class homeobox genes that belong to Dlx, Emx, Nkx, and paired-like subclasses are expressed in region-specific patterns. Based on the restricted expression pattern of regulatory genes and morphological observations, a prosomeric model of forebrain patterning was proposed (27). The embryonic midbrain is a compartment structure located between the forebrain and hindbrain and can be considered one of the neuromeres. Otx genes are expressed in the embryonic midbrain and mark the midbrain-hindbrain boundary.

To date, >30 paired-like homeoproteins have been identified in mice, and about two-thirds of them are involved in the induction or formation of neural tissues, as well as neuronal cell-type specification. For example, Otx1 is required for axonal projections of cortical neurons (36). Otx2 is essential for anterior neuroectoderm specification (1, 3, 17). Phox2a and Phox2b are essential for the development of adrenergic neurons (20, 25). Prop1 is required for the differentiation of pituitary lineage cells, and its functional deficiency causes Ames dwarfism in mice (29).

A novel paired-like homeobox gene, Dmbx1, expressed in neural tissues of the developing mouse embryo, was recently identified (24). Dmbx1 expression was initially detected at embryonic day 7.5 (E7.5) in the anterior head folds, and by E8.5, expression was detected in the prospective midbrain region. Between E9.5 and E10.5, the midbrain expression was maintained, and in addition, the expression was detected in the caudal diencephalon (prosomere 1) and nasal pits. At later stages of embryogenesis, subregions of the hindbrain and a subset of spinal cord neurons also expressed Dmbx1 (4). Dmbx1 orthologues have also been identified in other vertebrates, including chicken and zebrafish (8, 10). As in the mouse, chicken and zebra fish Dmbx1 expression was also detected in the midbrain and pretectum. A gene knockdown study using antisense oligonucleotides indicated that zebra fish Dmbx1 (mbx) is important for the development of the eyes and tectum (10).

To investigate the in vivo function of mouse Dmbx1, we generated two mutant alleles of Dmbx1, including a homeobox deletion (Dmbx1⁻) and a lacZ insertion (Dmbx1^z). Here, we show that Dmbx1 is important for nervous system development, neonatal survival, postnatal growth, and nursing ability.

MATERIALS AND METHODS

Generation of Dmbx1 mutant mice.

A mouse 129/SvEv strain genomic library was screened to isolate Dmbx1 genomic DNA using a previously reported Dmbx1 homeobox probe (24). To generate the Dmbx1⁻ allele, we used 4.4-kb HindIII and 2.6-kb BglII genomic DNA fragments as the 5′ and 3′ homologous arms for a targeting vector, respectively. We replaced a ∼1.9-kb HindIII-BglII genomic sequence encoding the entire homeodomain and adjacent intron sequences with a floxed Neo cassette (Fig. 1A). To construct a Dmbx1^z targeting vector, we inserted a splice acceptor (SA)-internal ribosomal entry sequence (IRES)-lacZ-floxed Neo cassette into a StuI site located in intron 1 of a 5.8-kb XbaI genomic fragment (Fig. 1E). The lacZ sequence included a nuclear localization signal. Both targeting vectors have a herpes simplex virus thymidine kinase expression cassette located on the 3′ arms of homology for negative selection. Twenty-five micrograms of linearized targeting vector (Fig. 1A and E) were electroporated into 10⁷ protamine-Cre (PC3) ES cells (22). After 10 days of drug selection with 350 μg of G418/ml and 200 μM FIAU [1-(2′-deoxy-2′-fluoro-β-d-arabinofuranosyl)-5-iodouracil], ES cell colonies were picked and genotyped. Correctly targeted ES cell clones were injected into blastocysts from C57BL/6 (B6) mice and transferred into pseudopregnant females. The resulting chimeric male mice were mated with B6 females for germ line transmission of the Dmbx1 targeted alleles. Two independent ES cell clones were confirmed to be germ line transmitted for both Dmbx1⁻ and Dmbx1^z alleles. All mice were examined on a B6-129 mixed genetic background. A 5′ probe used for Southern analysis was PCR amplified using the primers 5′-TTC CAG ACC CTG GTG TTA G-3′ and 5′-GAT TAA ACA TAC ATG CAA TGT G-3′. A 0.2-kb HindIII-BglII fragment was used as a 3′ probe. The PCR primers used for genotyping were as follows: a, 5′-AAA CAA CGT CGG AGT CGC-3′; b, 5′-ATC TTC TGG GGC AGA CTC-3′; c, 5′-CCT TTC CCA TGT GCT GTG-3′; d, 5′-GTT CCA GGA CAG CTT GGG-3′; e, 5′-AAG TAC TGG AGC GTC TGG GGA TT-3′; and f, 5′-CTG ACC AGG AGG ACA ACC GAA G-3′. The PCR conditions were 94°C for 2 min 30 s, followed by 35 cycles of 94°C for 30 s, 55°C for 1 min, and 72°C for 1 min, and a final extension of 72°C for 10 min.

FIG. 1.

Generation of Dmbx1 targeted mice. (A and E) Strategies to generate Dmbx1⁻ and Dmbx1^z alleles are shown. Dmbx1 consists of four coding exons (boxes). To generate the Dmbx1⁻ allele, the second and third coding exons encoding the homeodomain (solid boxes) was replaced by a floxed Neo cassette as described in Materials and Methods. For the generation of the Dmbx1^z allele, an SA-IRES-lacZ-floxed Neo cassette was inserted into an intron between the first and second coding exons. Solid triangles, loxP sequences; TK, thymdine kinase gene for negative selection; Neo, neomycin resistance gene for positive selection; LacZ, lacZ gene containing a nuclear localization signal; Cre, Cre recombinase. Essential diagnostic restriction enzyme sites are indicated by vertical bars. B, BglII; I, BglI. Positions of probes used for Southern analysis and PCR primers (see Materials and Methods) are indicated by horizontal bars and arrowheads, respectively. (B and F) Southern analyses of ES cell DNA. +, wild-type allele; −, Dmbx1⁻ allele; z, Dmbx1^z allele. (C and G) Genotyping of Dmbx1 targeted mice. (D and H) PCR genotype of Dmbx1 targeted mice.

β-Galactosidase staining and histological analysis.

Embryos and brains were stained for β-galactosidase activity as described previously (23). Seven-micrometer-thick paraffin-embedded tissue sections were cut and stained with hematoxylin and eosin Y or cresyl violet. Antibodies used for immunostaining were anti-growth hormone (anti-GH) and anti-thyroid-stimulating hormone (anti-TSH), obtained from the National Hormone and Peptide Program, and anti-neurofilament, 2H3, obtained from the Developmental Studies Hybridoma Bank, Iowa City, Iowa. Immunological reactions were visualized using a Vector ABC kit or by peroxidase-diaminobenzidine reaction.

Section in situ hybridization and Northern analysis.

Twelve-micrometer-thick paraffin sections were prepared and hybridized with a digoxigenin (Dig) 11-UTP-labeled anti-sense or sense (control) Dmbx1 homeobox probe (24). After being washed in SSC buffer (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), the sections were incubated with alkaline phosphatase-conjugated anti-Dig-11-UTP antibodies (Roche) and visualized using nitroblue tetrazolium and BCIP (5-bromo-4-chloro-3-indolylphosphate) as substrates. Northern analysis was performed using a filter purchased from ORIGENE hybridized with a Dmbx1 homeobox probe (24).

RESULTS

Generation of Dmbx1 mutant mice.

We generated two targeted Dmbx1 mutant alleles, Dmbx1⁻ and Dmbx1^z, by homologous recombination in the PC3 mouse ES cell line, which is homozygous for a Prm1-Cre transgene (Fig. 1B and F). Dmbx1⁻ lacks the second and third exons, which encode the homeodomain. Dmbx1^z is an insertional mutation of a SA-IRES-lacZ cassette into intron 1. Correctly targeted ES cell clones were injected into blastocysts to generate chimeric mice that transmitted the targeted mutations to their offspring. When PC3 ES cells differentiate into spermatids in male chimeras, Cre recombinase will be expressed, causing the floxed Neo cassettes in the Dmbx1 targeted loci to be removed prior to the fertilization of oocytes (Fig. 1). Mice heterozygous for the Dmbx1⁻ and Dmbx1^z alleles appeared normal and were fertile.

Dmbx1-lacZ expression in the CNS.

We examined lacZ expression in Dmbx1^z heterozygous mutant mice (Dmbx1^z/+). At E8.5, lacZ expression was detected in the prospective midbrain region, and later, at E9.5 and E10.5, lacZ expression was detected in the caudal diencephalon, midbrain, and nasal pits (Fig. 2A to C). At E12.5 and E14.5, in addition to the above-mentioned regions, subregions of the hindbrain and spinal cord expressed lacZ (Fig. 2D to F). These expression sites are similar to the previously reported Dmbx1 mRNA expression pattern detected by whole-mount in situ hybridization, indicating that the Dmbx1^z allele recapitulates the endogenous Dmbx1 expression pattern and may be a useful marker to identify potential Dmbx1 expression sites.

FIG. 2.

(A to Q) X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) staining of Dmbx1^z heterozygous mutant mice. The solid and open arrowheads indicate the nasal pit and spinal cord, respectively. The solid and open arrows indicate the inferior colliculus and cerebellum, respectively. (R and S) In situ hybridization of Dmbx1. (T) Northern analysis of adult tissues. (A) E8.5. (B) E9.5. (C) E10.5. (D) E12.5. (E) E14.5. (F) Cross-section of E12.5 spinal cord. (G) Dorsal view of P7 brain. (H) Ventral view of P7 brain. (I) Dorsal view of P21 brain. (J to M) X-Gal and eosin Y costained paraffin sections. The blue cells are Dmbx1-lacZ-positive cells. (N to Q) X-Gal- and cresyl violet-costained paraffin sections. (R and S) Superior colliculus of postnatal brain counterstained with eosin Y.

We also examined Dmbx1-lacZ expression in the postnatal CNS. At postnatal day 7 (P7), lacZ expression was detected in the tectum of the midbrain and subregions of the medulla oblongata, cerebellum, and spinal cord (Fig. 2G and H). In the midbrain at P21, the superior colliculus was positively stained, while staining of the inferior colliculus was absent (Fig. 2I and data not shown). Histological analyses revealed that the Dmbx1-lacZ-expressing cells were primarily located in the superficial layer of the superior colliculus (Fig. 2J). Dmbx1-lacZ expression was also detected in deep nuclei of the cerebellum, medulla oblongata, and spinal cord (Fig. 2K to M). Cresyl violet staining demonstrated that these Dmbx1-lacZ-expressing cells had relatively large nuclei, suggesting that the cells were neurons (Fig. 2N to Q). Dmbx1 mRNA expression was detected in the superior colliculus at P19 by section in situ hybridization (Fig. 2R and S). Northern analysis showed that Dmbx1 is expressed in the adult brain (Fig. 2T). These results indicate that Dmbx1-lacZ expression is consistent with the endogenous Dmbx1 expression pattern in the postnatal CNS.

Neonatal lethality and growth defects of Dmbx1 mutant mice.

To examine the in vivo role of Dmbx1, we generated Dmbx1 homozygous mutant mice by intercrossing Dmbx1 heterozygotes. Pups obtained from Dmbx1⁻ heterozygous (Dmbx1^+/−) mutant intercrosses were initially genotyped at P9. Only 13.0% of the progeny were Dmbx1⁻ homozygous (Dmbx1^−/−) mutants, which was significantly lower than the expected Mendelian ratio (Table 1). However, at E18.5, the ratio of wild-type, Dmbx1^+/−, and Dmbx1^−/− mice was nearly the expected Mendelian ratio. These results suggest that some Dmbx1^−/− mice died postnatally before P9. Similar results were obtained when Dmbx1^+/z mice were intercrossed (Table 1). This indicates that the insertion of the SA-IRES-lacZ cassette disrupts Dmbx1 function and that the Dmbx1^z allele behaves phenotypically like the Dmbx1⁻ allele.

TABLE 1.

Summary of Dmbx1 heterozygous intercrosses

Mutant	Stage	No. (%)					Total	Avg litter size
		+/+	+/−	−/−	+/z	−/z
Dmbx1+/−	E 18.5	13 (27.1)	22 (45.8)	13 (27.1)			48	6.0
	P9a	47 (32.2)	80 (54.8)	19 (13.0)			146	7.7
Dmbx1z/+	P9a	47 (38.5)			63 (51.6)	12 (9.8)	122	7.2

^aP < 0.01 (chi-square test).

Those Dmbx1 homozygous mutant mice that did survive were consistently smaller than their heterozygous and wild-type littermates. Therefore, we measured body weights at P7, P14, and P21 (Fig. 3A). Dmbx1^−/− mice were significantly lighter in weight than wild-type or Dmbx1^+/− mice and were proportionally smaller. However, body weights at E18.5 and P0 were similar among wild-type, Dmbx1^+/−, and Dmbx1^−/− mice, indicating that the observed growth defect is a postnatal event (Fig. 3B). The growth impairment of Dmbx1^−/− mice continued throughout adulthood, although growth rates were comparable to those of wild-type or Dmbx1^+/− mice after weaning (data not shown). The growth defect was also observed in Dmbx1-lacZ homozygous mutant (Dmbx1^z/z) mice, confirming the functional similarity between the Dmbx1⁻ and Dmbx1^z alleles (Fig. 3C).

FIG. 3.

(A) Body weights of wild-type (solid bars; n = 9 for P7, 8 for P14, and 10 for P21), Dmbx1^+/− (shaded bars; n = 8 for P7, 14 for P14, and 8 for P21), and Dmbx1^−/− (open bars; n = 8 for P7, 12 for P14, and 12 for P21) mice. Standard deviations are indicated by the error bars. Statistical analyses were scored by t test. (B) Body weights of wild-type (solid bars; n = 8 for E18.5 and 9 for P0), Dmbx1^+/− (shaded bars; n = 14 for E18.5 and 8 for P0), and Dmbx1^−/− (open bars; n = 10 for E18.5 and 9 for P1) mice. (C) Body weights of wild-type (solid bars; n = 5 for P7, 4 for P14, and 5 for P21), Dmbx1^z/+ (shaded bars; n = 7 for P7, 5 for P14, and 8 for P21), and Dmbx1^z/z (open bars; n = 7 for P7, 4 for P14, and 7 for P21) mice. Statistical analyses were scored by t test. (D) Appearance of neonates (P0). +/+, wild type; −/−, Dmbx1^−/− littermate. The arrows indicate the positions of the stomachs. (E) Immunohistochemistry of pituitary glands at P21. +/−, Dmbx1^+/− mice; −/−, Dmbx1^−/− mice. α-GH, anti-GH; α-TSH, anti-TSH.

We observed little or no milk in the stomachs of Dmbx1 homozygous mutant mice after birth, suggesting that impaired milk intake caused the neonatal lethality and poor growth (Fig. 3D). Because it is well known that pituitary deficiency can result in growth deficiency, we examined the expression of GH and TSH by immunohistochemistry (Fig. 3E). GH and TSH immunostaining was detected in the pituitary glands of Dmbx1^−/− mice.

Aberrant expression of Dmbx1-lacZ in Dmbx1 mutant mice.

We next examined Dmbx1-lacZ expression in Dmbx1 mutant mice. To control for the number of lacZ alleles used, we compared lacZ expression levels in Dmbx1^z/+ and Dmbx1^z/− mice. At P7, the Dmbx1-lacZ-expressing region in the inferior colliculus was expanded in Dmbx1^z/− in comparison to Dmbx1^z/+ mice (Fig. 4A and B). On the other hand, the number of Dmbx1-lacZ-expressing cells in the ventral region of the medulla oblongata was reduced in Dmbx1^z/− mice (Fig. 4C and D). These observations were also found at P1 and P14, suggesting that the aberrant expression of Dmbx1-lacZ was maintained throughout the neonatal period (Fig. 4E to H and data not shown). Histological analysis of serial sections of P21 brains revealed that a substantial number of lacZ-positive cells were observed in the inferior colliculus in Dmbx1^z/− mice, while no LacZ-positive cells were detected in Dmbx1^z/+ mice (Fig. 4I and J). In the cerebellum, the number of lacZ-expressing deep nuclei was reduced in Dmbx1^z/− mice compared to Dmbx1^z/+ mice, and the ratio of lacZ-positive cells per section was 1:0.24 (P < 0.001; t test) (Fig. 4K and L). In the ventral region of the medulla oblongata, the lacZ-expressing cells were also reduced in number, and the ratio was 1:0.45 (P < 0.001; t test) (Fig. 4M and N). Since the aberrant expression of Dmbx1-lacZ was noticed at P1, we also examined Dmbx1-lacZ expression during embryonic stages. At E9.5 and E11.5, the regions that express lacZ were comparable between Dmbx1^z/+ and Dmbx1^z/− embryos; however, the intensity of the staining was stronger in Dmbx1^z/− embryos than in Dmbx1^z/+ embryos (Fig. 4O to R). In spite of this difference in Dmbx1-lacZ expression, overall brain structure was maintained in Dmbx1^z/− embryos, and no obvious defects were observed at E9.5 or E10.5 using a Phox2 probe as an early marker for catecholaminergic neurons or antineurofilament staining, respectively (Fig. 4S and T and data not shown).

FIG. 4.

Comparison of Dmbx1-lacZ expression (A to R) and anti-neurofilament staining (S and T). (A and B) Dorsal views of P7 brains. (C and D) Ventral views of P7 brains. (E and F) Dorsal views of P14 brains. (G and H) Ventral views of P14 brains. (I to N) X-Gal- and cresyl violet-costained paraffin sections of P21 brains. (I and J) Inferior colliculus. (K and L) Cerebellum. (M and N) Ventral regions of medulla oblongata. (O and P) E9.5. (Q and R) E11.5. (S and T) E10.5. The arrows in panels B and F indicate the Dmbx1-lacZ-positive cells in the inferior colliculus. (A, C, E, G, I, K, M, O, Q, and S) Dmbx1^z/+. (B, D, F, H, J, L, N, P, R, and T) Dmbx1^z/−.

Nursing defect of Dmbx1 mutant mice.

We next examined the phenotypes of mature Dmbx1 homozygous mutant mice. Both male and female Dmbx1 homozygous mutant mice were fertile. However, while breeding Dmbx1 homozygous mutant mice, we noticed that the number of surviving offspring from Dmbx1 homozygous mutant females was lower than that from Dmbx1 heterozygous mutant females (Table 2). When we crossed Dmbx1^−/− males to Dmbx1^+/− females, most of the Dmbx1^+/− pups survived to P21. However, ∼50% of the Dmbx^−/− mice died during the neonatal period, confirming the neonatal lethality of Dmbx1 mutant mice. When we crossed Dmbx1^+/− males to Dmbx1^−/− females, even the percentage of Dmbx1^+/− pups was reduced, and the Dmbx1^−/− pups rarely survived to P21. Interestingly, Dmbx1^+/− pups raised by Dmbx1^−/− females grew poorly and were similar in weight to Dmbx1^−/− pups nursed by Dmbx1^+/− females (Fig. 5A). The growth defect was not due to the genotypes of the pups, because growth impairment was observed even when wild-type pups were cross-fostered to Dmbx1^−/− females (Fig. 5B). These observations indicate that Dmbx1^−/− females have a nursing defect. Maternal behaviors, such as nest building and pup retrieval, seemed normal in Dmbx1^−/− females, and they did not abandon their pups (data not shown). We also examined the mammary glands of postpartum Dmbx1^−/− females (Fig. 5C). Milk production was clearly evident, and well-differentiated duct formation was detected in Dmbx1^−/− females.

TABLE 2.

Survival of pups fostered by Dmbx1^−/− mice

Genotype of parents (female × male)	No. of litters	No. of pups of genotype:
		+/−			−/−
		P0	P21	Ratioa (%)	P0	P21	Ratioa
+/− × −/−	8	31	30	97	35	19	54
−/− × +/−	7	27	16	59	13	0	0

^aSurvival ratio (P21/P0).

FIG. 5.

(A) Body weights of Dmbx1^+/− pups nursed by Dmbx1^+/− females (solid bars; n = 14 for P7, 14 for P14, and 22 for P21), Dmbx1^−/− pups nursed by Dmbx1^+/− females (shaded bars; n = 11 for P7, 10 for P14, and 15 for P21), and Dmbx1^−/− pups nursed by Dmbx1^−/− females (open bars; n = 9 for P7, 6 for P14, and 6 for P21). Standard deviations are indicated by the error bars. Statistical analyses were scored by t test. (B) Body weights of wild-type pups cross-fostered by Dmbx1^+/− females (solid bars; n = 5) and wild-type pups cross-fostered by Dmbx1^−/− females (open bars; n = 4). Statistical analyses were scored by t test. (C) Histological examinations of mammary glands. Paraffin sections were stained by hematoxylin and eosin Y. +/+, wild-type; +/−, Dmbx1^+/−; −/−, Dmbx1^−/− females.

DISCUSSION

To investigate the in vivo role of Dmbx1, we generated Dmbx1 mutant mice. The Dmbx1^z allele recapitulated the previously reported Dmbx1 mRNA expression patterns during embryonic stages and identified several sites in the postnatal CNS as potential Dmbx1-expressing regions. After birth, Dmbx1-lacZ expression was detected in subpopulations of the superior colliculus, inferior colliculus, medulla oblongata, cerebellum, and spinal cord. These observations suggest that Dmbx1 is expressed postnatally in the CNS and marks a unique population of neuronal cells. The Dmbx1-lacZ expression in the postnatal brain was confirmed by section in situ hybridization and Northern analysis of Dmbx1 mRNA. The numbers of Dmbx1-lacZ-expressing cells in the inferior colliculus were gradually reduced during the neonatal period, indicating that the expression of Dmbx1 is temporally and spatially regulated or that Dmbx1-expressing neurons are progressively lost after birth.

Although Dmbx1 is expressed in the forebrain, midbrain, and hindbrain during embryogenesis, Dmbx1 homozygous mutant mice were born without any overt alterations in brain structure. This is in contrast to other homeobox genes that are expressed in embryonic brains, such as Otx2, En1, and Gbx2. Otx2 homozygous mutant mice lack rostal brain structures, and both midbrain and rostal hindbrain tissue is missing in En1 homozygous mutant mice (1, 3, 17, 18, 30, 37). Gbx2 is required for the development of the midbrain-hindbrain junction (34). A previous report suggested that Otx2 is necessary for the expression of Dmbx1 in the midbrain (16). However, the phenotypes of Otx2 mutant mice are not due only to the loss of Dmbx1, because Dmbx1 mutant mice have overtly normal brains and Dmbx1 is dispensable for survival during embryogenesis.

Although Dmbx1 homozygous mutant mice manifested no obvious malformations of the brain, they appeared to have abnormalities in the development of cell lineages in the CNS, because Dmbx1^z/− mice exhibited aberrant Dmbx1-lacZ expression patterns. Dmbx1-lacZ-expressing cells in deep nuclei of the cerebellum and the ventral region of the medulla oblongata were reduced in number, while in the inferior colliculus, an increased number of Dmbx1-lacZ-expressing cells were detected at birth and through the neonatal period. These observations suggest that Dmbx1 is required for the proper generation or function of specific cell populations in the CNS.

We compared the Dmbx1-lacZ expression patterns during embryogenesis and noticed that the intensity of the lacZ staining was stronger in Dmbx1^z/− than Dmbx1^z/+ embryos at E9.5 and E11.5. However, the regions that express Dmbx1-lacZ were comparable in Dmbx1^z/− and Dmbx1^z/+ embryos, with a sharp boundary between the midbrain and the midbrain-hindbrain junction at E9.5. These observations indicate that Dmbx1 has an autoregulatory mechanism to down-regulate its expression; however, the autoregulation is dispensable for the boundary formation of Dmbx1-expressing compartments. Previously it was reported that Dmbx1 expression was suppressed by Gbx2 in the midbrain-hindbrain junction (16). Thus, our results suggest that the repression of Dmbx1 expression by GBX2 occurs independently of DMBX1.

Several Dmbx1 homozygous mutant mice were found dead within 24 h after birth without milk in their stomachs. Once they suckled, they survived for at least several days, although some of them died before being weaned, gradually losing weight. These observations support the possibility that there is a Dmbx1-dependent impairment of milk intake that causes mortality during the neonatal period. It is likely that the milk intake deficit also influences growth, because all of the Dmbx1 homozygous mutant mice manifested a growth defect. The growth retardation was noticed within 3 days after birth, and this is in contrast to dwarfisms caused by pituitary deficiencies, in which growth defects are usually detected 1 week after birth. Hypogonadal and hypothyroid mice lacking the common α subunit of TSH, luteinizing hormone, and follicle-stimulating hormone demonstrate apparently reduced growth between 2 and 3 weeks of age (11). Otx1-null mice, which manifest low levels of GH, follicle-stimulating hormone, and luteinizing hormone, exhibit dwarfism around P7 (2). Laron-type dwarfisms, which are caused by GH insensitivity, also exhibit significant weight differences after the first week of life (26, 38). We have not observed any skeletal abnormalities in Dmbx1 mutant mice, such as tail kinks or bends, rounded heads, or misaligned teeth, which are characteristics of Fgfr3 mutant dwarf mice (6, 33). These findings, in conjunction with the observation that the production of GH and TSH is maintained in the pituitary glands of Dmbx1 homozygous mutant mice, suggest that the growth defects were caused by insufficient milk intake and were not due to pituitary dysfunction or skeletal dysplasia. The abnormal development of the CNS seems to be related to the deficit of milk intake in Dmbx1 mutant mice.

The body weights of adult Dmbx1 homozygous mutant mice never catch up with those of wild-type or Dmbx1 heterozygous littermates, and in most cases, the homozygous mutant mice are 20 to 50% lighter in weight than the others. Nevertheless, the responses to starvation stress of adult Dmbx1^−/− mice were virtually normal. They lost and gained weight as much as Dmbx1^+/− mice after a 24-h starvation and refeeding, respectively (data not shown). The recovery of body weight was due to hyperphagia, and Dmbx1^−/− mice ate more during the first day of refeeding than the second day, like Dmbx1^+/− mice (data not shown). These observations indicate that adult Dmbx1 homozygous mutant mice manifest normal eating behaviors. The reduced body weight of adult Dmbx1 homozygous mutant mice can be attributed to insufficient milk intake prior to weaning, because an increased growth rate was observed after weaning in Dmbx1 homozygous mutant mice, like wild-type and Dmbx1 heterozygous mutant mice.

Both male and female Dmbx1 homozygous mutant mice were fertile. However, pups from Dmbx1 homozygous mutant females showed a high mortality ratio compared to pups from Dmbx1 heterozygous mutant females. Furthermore, the surviving pups manifested the growth impairment even if they were Dmbx1 heterozygotes. These phenotypes did not depend on the genotypes of the pups or the sizes of litters and were caused by a maternal defect, because similar results were obtained when wild-type pups were cross-fostered by Dmbx1 homozygous mutant females. It was previously reported that deficits of maternal behaviors can cause lethality or growth defects in offspring. fosB mutant mice are deficient in retrieval responses and crouching behaviors, and most pups die within 2 days after birth (5). Two imprinted genes, Mest and Peg3, are related to maternal behaviors, and heterozygous mutant mice are defective in nursing and their offspring are growth impaired and show a low survival rate (14, 15). Mbd2 homozygous mutant mice show poor retrieval responses, and their pups are growth retarded (9). Dbh-deficient mice, which lack the enzyme responsible for the synthesis of (nor)epinephrine, manifest impaired maternal behaviors, and their pups die within several days after birth (31). In Dmbx1 homozygous mutant mice, the expression of TH, the enzyme essential for the synthesis of dopamine and (nor)epinephrine, was detected by immunohistochemistry, and Dmbx1-lacZ-expressing cells are noncatecholaminergic neurons (data not shown). We have also observed normal maternal behaviors, such as nest building and pup retrieval, in Dmbx1 homozygous mutant females. Hence, the growth impairment and lethality in pups nursed by Dmbx1 homozygous mutant females are not likely caused by deficits of maternal behavior.

Although oxytocin-deficient mice exhibit normal maternal behaviors, their pups die within 24 h of delivery with no milk in their stomachs (21). Offspring of Id2 homozygous mutant mothers die within 2 days after birth due to a lactation defect (19). Examination of the mammary glands of postpartum Dmbx1 homozygous mutant females revealed well-differentiated duct formation and milk accumulation. Therefore, an impaired milk ejection reflex may be responsible for the neonatal lethality and growth defects in pups nursed by Dmbx1 homozygous mutant females.

It is interesting that Foxb1 (a winged helix gene)-deficient mice share some phenotypes with Dmbx1 mutant mice (7, 13, 35). Foxb1 mutant mice are postnatally growth retarded, which is noticeable within 2 to 3 days after birth, and remain proportionally smaller in adulthood, although the growth curve is parallel to that of wild-type or heterozygous animals. Thirty to 50% of Foxb1 mutant mice die before being weaned, and the peak of lethality is observed in the first 2 days after birth. They have normal circulating GH and TSH, and the growth retardation is not due to a pituitary deficiency. Surviving Foxb1 mutant mice are fertile, but Foxb1 mutant females have a defect in the milk ejection reflex. Thus, we examined Foxb1 expression in the neural tissues of Dmbx1 mutant mice during embryogenesis and after birth (data not shown). However, we found no differences in Foxb1 expression, suggesting that Foxb1 does not mediate the Dmbx1 mutant phenotype. It was proposed that morphological defects in the inferior colliculi were responsible for the lactation defect of Foxb1 mutant mice (12). This proposal suggests a similar cause for the Dmbx1 maternal phenotypes, since an aberrant expression of Dmbx1-lacZ was observed in the inferior colliculus in Dmbx1 mutant mice. Determining the neuronal networks mediated by Dmbx1-expressing neurons will be useful for understanding the phenotypes of mutant mice.

In summary, we have demonstrated that Dmbx1 is expressed in the neuronal tissues during both the embryonic and postnatal periods and is required for survival, growth, normal brain development, and maternal activity in mice. Since the sequence of Dmbx1 is highly conserved between mouse and human (24), the impaired function of human DMBX1 might be related to neurological disorders, such as impaired food intake and milk ejection reflex, that are relevant to the phenotype of Dmbx1 mutant mice.