Disruption of the murine nuclear factor I-A gene (Nfia) results in perinatal lethality, hydrocephalus, and agenesis of the corpus callosum

Abstract

The phylogenetically conserved nuclear factor I (NFI) family of transcription/replication proteins is essential both for adenoviral DNA replication and for the transcription of many cellular genes. We showed previously that the four murine NFI genes (Nfia, Nfib, Nfic, and Nfix) are expressed in unique but overlapping patterns during mouse development and in adult tissues. Here we show that disruption of the Nfia gene causes perinatal lethality, with >95% of homozygous Nfia^−/− animals dying within 2 weeks after birth. Newborn Nfia^−/− animals lack a corpus callosum and show ventricular dilation indicating early hydrocephalus. Rare surviving homozygous Nfia^−/− mice lack a corpus callosum, show severe communicating hydrocephalus, a full-axial tremor indicative of neurological defects, male-sterility, low female fertility, but near normal life spans. These findings indicate that while the Nfia gene appears nonessential for cell viability and DNA replication in embryonic stem cells and fibroblasts, loss of Nfia function causes severe developmental defects. This finding of an NFI gene required for a developmental process suggests that the four NFI genes may have distinct roles in vertebrate development.

The nuclear factor I (NFI) family of site-specific DNA-binding proteins plays a dual role in adenoviral DNA replication and in the transcription of viral and cellular genes. Early studies showed that NFI proteins are essential for adenovirus DNA replication both in vitro (1–4) and in vivo (5, 6) and suggested that NFI may function in cellular DNA replication. More recently, expression of the four NFI genes has been shown to be developmentally regulated (7–9), and NFI proteins have been implicated in the regulation of transcription of genes expressed in many tissues, including liver (10, 11), mammary gland (12, 13), and brain (14–16). To assess the role of individual NFI family members in cell growth and development, we have begun a genetic analysis of NFI function in mice.

The NFI gene family has four members in vertebrates (NFI-A, NFI-B, NFI-C, and NFI-X) (17, 18), a single member in the nematode Caenorhabditis elegans (nfi-1) (19), and no homologues in any of the sequenced prokaryotic or simple eukaryotic genomes. A single NFI gene in the cephalochordate Amphioxus suggests that the four vertebrate genes were generated by duplications during vertebrate evolution (19). The vertebrate NFI genes share a conserved 220-aa N-terminal domain that is sufficient for dimerization, DNA-binding, and stimulation of adenovirus DNA replication in vitro (20–22). Products of the four NFI genes form homo- and heterodimers that bind to the canonical NFI binding site [TTGGC(N)₅GCCAA] with apparently identical affinities (23–26). Transcripts of all four NFI genes are alternatively spliced, yielding as many as nine different proteins from each gene (26, 27). This multitude of NFI isoforms has complicated the analysis of the role of specific NFI genes in transcription and development.

While all NFI proteins appear to bind to the same DNA sequence, differences in function among NFI gene products have been seen in a number of systems. NFI-C/CTF proteins (CTF, CAAT-box transcription factor) differ in their abilities to activate transcription in transfected Drosophila cells and yeast, depending on the sequence of their C-terminal domains (20, 27). In addition, gene-, cell-type-, and promoter-specific differences in transcriptional modulation have been demonstrated for various NFI isoforms (28–31). These different activities, together with the finding that the four NFI genes are expressed in distinct but overlapping patterns during mouse development (7), suggests that each NFI gene may play a unique role in mouse development. To address the role of the Nfia gene in development we have disrupted the gene by homologous recombination in embryonic stem (ES) cells. Mice with a disruption of the Nfia gene die shortly after birth and have neuroanatomical defects, indicating an essential role for NFI-A-regulated transcriptional pathways in the development of the brain.

Materials and Methods

Disruption of the Nfia Gene.

Genomic clones of the Nfia gene were isolated from a 129/Sv library by using the Nfia gene-specific primer described previously (7), and a combination of restriction mapping, PCR, DNA sequencing, and Southern analysis with exon 1a- and exon 2-specific probes was used to map the locations of exons 1a and 2. The targeting construct pNFI-A5′3′ was prepared by cloning a 7.2-kb EcoRI fragment containing exon 1a of Nfia and a 2.4-kb EcoRV–SpeI fragment containing the 3′ half of exon 2 of Nfia and adjacent intron sequences separated by a 1.8-kb PGK-neo expression cassette from pNT (32) (Fig. 1A). A PGK-TK negative selection cassette from pNT was cloned upstream of the Nfia gene segments (Fig. 1A). E14–1 cells (33) were electroporated with the targeting vector and neomycin- and ganciclovir-resistant clones were isolated (34). Four of 91 clones had a correctly targeted disruption of Nfia as shown by Southern blot and PCR analyses with probes and primers outside of the targeting vector. Nfia⁻ ES cells were injected into C57BL/6 blastocysts and chimeric mice were bred to Black Swiss mice to transmit the disrupted allele. The Nfia⁻ allele was transmitted from chimeras from two different ES cell clones, and both lines exhibited identical perinatal lethal phenotypes in homozygous Nfia^−/− animals.

Figure 1.

Disruption of the murine Nfia gene and demonstration of “exon-skipping.” (A) The structure of a portion of the wild-type Nfia gene is shown with exons 1a and 2 as black boxes, EcoRI sites marked E, and the 7.0-kb HindIII fragment shown with a bracket. The targeting vector was constructed as described in Materials and Methods, and recombination within the 5′ and 3′ regions with the wild-type gene would yield the disrupted Nfia gene missing most of exon 2. (B) Clones of ES cells electroporated with the targeting vector were lysed and DNA was isolated, digested with HindIII, transferred to membranes, and probed with the probe shown as P in A. This probe lies outside of the targeting vector and detects a 7.0-kb band from the wild-type Nfia allele and a 5.0-kb band from the disrupted allele. Lanes 1–5, 7, and 8 show a wild-type pattern, while lane 6 shows both the wild-type and disrupted alleles. (C) Genomic DNA from the tails of +/+, +/−, and −/− animals was digested with HindIII and analyzed as in B. (D) The nested primers shown as arrows below the disrupted gene in A were used to amplify genomic DNA of the progeny of a mouse heterozygous for the disrupted Nfia⁻ allele. Lanes 2, 3, 6, and 8 are positive for the 2.3-kb PCR fragment, while lanes 1, 4, 5, and 7 are negative. Size markers of HindIII-digested λ DNA are shown in the leftmost lane, marked λ. The rightmost set of nested primers from A lies outside of the targeting construct. (E) RNA from the brains of wild-type (lanes 3 and 4), heterozygous (+/−, lanes 5 and 6), and homozygous Nfia⁻ mice (−/−, lanes 7–10) was isolated, reverse transcribed, and PCR amplified by using primers in exon 3 and either exon 1a or 1b of Nfia as denoted above the lanes. The products were analyzed on a 2% agarose gel along with 123-bp ladder markers (lane M). The arrows E1b-E2-E3 and E1a-E2-E3 show the positions of correctly spliced NFI-A mRNAs, while the arrows E1b-E3 and E1a-E3 show the positions of the products lacking exon 2. Lanes 1 and 2 contained no reverse transcribed product.

Analysis of Nfia-Deficient Mice.

Mice heterozygous for the Nfia⁻ allele (Nfia^+/−) were bred to generate homozygous Nfia^−/− animals. Embryos were harvested by cesarean section at different times of gestation and RNA was prepared from samples frozen in liquid nitrogen (Trizol, Life Technologies). Reverse transcription (RT)-PCRs (Superscript II, Life Technologies) were analyzed by using primers specific for exon 1a (NFI-AE1, TGTATTCTCCGCTCTGTCTCAC) or exon 1b (NFI-AE1b, CTCCCGTGAGTTAGCATCAGATG) and exon 3 (NFI-AE3C, TGGCTGGGACTTTCAGATTGAC). RT-PCRs were analyzed by electrophoresis on 0.8% and 2% agarose gels. Some embryos and fetuses were fixed in 4% paraformaldehyde, Histochoice (AMRESCO), or snap frozen in OCT and cryosectioned or embedded in paraffin and sectioned. Immunostaining was performed as described previously (35) by using monoclonal antibodies to glial fibrillary acidic protein (GFAP) (G-A-5; Sigma) and myelin phospholipid protein (PLP) (M2; ref. 35).

Results

Disruption of the Nfia Gene.

The Nfia gene is on mouse chromosome 4 flanked by the Jun and Cyp4a10 genes (19). The Nfia gene was disrupted by homologous recombination of the flanking arms of the pNFI-A5′3′ vector with the wild-type allele, deleting the 3′ splice acceptor site and 219 bp of exon 2 (Fig. 1A). Disruption was confirmed by Southern blotting (Fig. 1 B and C) and PCR analyses with probes outside of the targeting vector (Fig. 1D). Because exon 2 of the Nfia gene encodes essentially the entire N-terminal NFI DNA-binding domain, any proteins produced from transcripts lacking this exon would lack DNA-binding or dimerization activity. Also, because the splice donor and acceptor sites of exon 2 of Nfia are spliced in different reading frames (19), transcripts lacking exon 2 would generate a missense polypeptide containing alternate first exons fused to an out-of-frame 37-aa polypeptide stopping at a termination codon in exon 3. To confirm that the disrupted Nfia allele generates a transcript lacking exon 2, RT-PCRs were performed on whole embryos homozygous for the Nfia disruption by using primers in exons 1a, 1b, and 3 (Fig. 1E). Exon 1a is the first exon of most NFI-A cDNAs, whereas exon 1b is the first exon in a subset of NFI-A cDNAs from brain and olfactory epithelium (ref. 14; data not shown). In brain RNA from homozygous wild-type embryos (Nfia^+/+), RT-PCR products consistent with the expected NFI-A transcripts were seen (Fig. 1E, lanes 3 and 4, upper arrows). In Nfia^−/− mice, only PCR products of a size consistent with transcripts lacking exon 2 were seen (Fig. 1E, lanes 7–10, lower arrows), whereas in heterozygotes both transcripts were detected (Fig. 1E, lanes 5 and 6). Both restriction enzyme digestion and direct sequencing of the PCR products confirmed the predicted splicing of exons 1, 2, and 3 at the wild-type allele and direct splicing of alternate exons 1 and 3 of the mutant allele (not shown). Thus the disruption of exon 2 generates a functionally null allele of Nfia.

Homozygous Nfia⁻ ES Cells and Fibroblasts Are Viable.

Since NFI proteins are required for adenoviral DNA replication and have been implicated in the replication of other DNA viruses and of cellular DNA, we determined whether ES cells lacking NFI-A were viable. ES cells containing a single targeted Nfia⁻ allele were cultured in the presence of high concentrations of G418 to select for cells containing only the targeted allele (36). Several clones were obtained which by both Southern blotting and PCR analysis contained only the targeted Nfia⁻ allele and no wild-type allele (data not shown). These apparently homozygous Nfia⁻ ES cell lines grew at a rate comparable to that of the heterozygous ES cells (data not shown). Fibroblasts from Nfia −/−, +/−, and +/+ embryos (see below) have also been prepared and show no obvious growth differences in culture (data not shown). These data indicate that the Nfia gene products are not required for DNA replication in the cell types examined.

Phenotype of Nfia⁻ Mice.

Heterozygous Nfia^+/− mice were bred to generate animals homozygous for the Nfia disruption. In the random-bred Swiss background in which the mice were generated, ≈95% of the Nfia^−/− animals died within 2 weeks of birth, with most animals dying on the day of birth (Table 1). There was also a loss of heterozygous Nfia^+/− animals, suggesting potential haploinsufficiency at the Nfia locus. Consistent with haploinsufficiency, far fewer than 50% of the animals generated from breeding wild-type males with Nfia^+/− females contained the Nfia⁻ allele (Table 2). Surprisingly, no loss of Nfia^+/− animals was seen when Nfia^+/− males were bred to wild-type females, indicating a maternal effect in loss of Nfia^+/− animals. When postnatally the Nfia⁻ heterozygotes are lost and the nature of the maternal effect require further investigation. Loss of >90% of Nfia^−/− mice has also been seen in N5 back-crosses into the inbred C57BL/6 strain. Of the rare Nfia^−/− survivors, all were runted for the first 2 weeks of life (Fig. 2A, −/− vs. +/+) and had cranial defects suggesting brain abnormalities. The most obvious phenotype was a “dome head” frequently seen in mice with hydrocephalus as compared with the roughly wedge-shaped heads of wild-type mice (Fig. 2B, −/− vs. +/+). These obvious cranial defects diminished with time, and adult animals appeared relatively normal but possessed a tremor indicating neurological defects. Adult Nfia^−/− males appeared sterile, whereas the females had very low fecundity but were able to give birth and to suckle young when bred to wild-type males (not shown). It is unclear whether the sterility in males was behavioral or physiological, since histological examination of the testes revealed no obvious abnormalities.

Table 1.

Loss of Nfia⁻ homozygous and heterozygous mice from wild-type (WT)/Nfia⁻ x WT/Nfia⁻ crosses

Progeny	Number of progeny with genotype
	WT/WT	WT/Nfia−	Nfia−/Nfia−
Male	17	17	8
Female	18	23	5
Total	35	40	13*
(Expected)	(35)	(70)	(35)

Heterozygous WT/Nfia⁻ mice were bred and the number of progeny of each genotype was scored. These numbers exclude any lost mice for which DNA was not obtained (<5%). The expected values are calculated assuming that the WT/WT homozygotes all survive and should represent 25% of the progeny. The apparent loss of heterozygotes (40/70) is not seen in male WT/Nfia⁻ × female WT/WT crosses and may represent a maternal effect phenotype. 

^*Only 4 Nfia⁻/Nfia⁻ mice survived more than 6 weeks after birth (4/102 total pups from nine litters). Most Nfia⁻/Nfia⁻ pups die within 1 day of birth with a few surviving up to 2 weeks and <5% surviving 6 weeks. 

Table 2.

Loss of heterozygous Nfia⁻ progeny from heterozygous Nfia⁻ mothers

Progeny	Number of progeny with genotype
	Heterozygous Nfia− male × WT female		Heterozygous Nfia− female × WT male
	+/+	−/+	+/+	−/+
Male	60	46	33	26
Female	41	40	35	13
Total	101	86	68	39*
(Expected)	(101)	(101)	(68)	(68)

Heterozygous Nfia⁻ males were crossed with wild-type (WT) females and heterozygous Nfia⁻ females were crossed with wild-type males. The genotype and sex of the progeny of the different crosses were recorded. All animals were in the outbred Black Swiss background. The reduction in the total number of heterozygous Nfia⁻ animals compared to expected in the latter cross is significant (∗, P < 0.01 by χ²) and may be caused by preferential loss of heterozygous female progeny. 

Figure 2.

Hydrocephalus in homozygous Nfia⁻ mice. (A) A homozygous Nfia⁻ pup (−/−) and a wild-type littermate (+/+) are shown. Note the smaller size and foreshortened head of the −/− animal. (B) Enlargement of A showing the characteristic “dome head” of the Nfia⁻ (−/−) animal compared with the more wedge-shaped head of wild-type animal. (C and D) Cresyl violet stained coronal sections through the brains of 6-month-old Nfia^−/− (−/−) and wild-type (+/+) animals. Note the dilation of the ventricles (arrows in C). (Bar in D = 2 mm.)

Genotyping of embryos at different times of gestation indicated that few or no Nfia^−/− or Nfia^+/− animals were lost prior to birth (data not shown). To determine whether major morphological defects present in embryos might contribute to the perinatal death, embryos were fixed, sectioned, and stained, and some embryos were used for skeletal and cartilage preparations. No obvious morphological defects were detected in such preparations, indicating similar skeletal and cartilage formation in the Nfia^−/− and wild-type animals (data not shown). However, several clear brain defects were seen upon serial sectioning of the brains of embryonic day 18.5 (E18.5), newborn, and adult Nfia^−/− animals. The rare adult Nfia^−/− animals examined (8 total) had severe hydrocephalus, as noted by dilation of the lateral and third ventricles (Fig. 2 C vs. D, arrows denote expanded ventricles). Both intraventricular dye injections and serial sectioning of adult brains indicated that the hydrocephalus was communicating, with no obstructions detected in either the foramens of Munro or the aqueduct (data not shown). The cause of the hydrocephalus is not yet known.

The second major defect seen in adult Nfia^−/− animals was the absence of a corpus callosum. The corpus callosum is a major fiber tract formed during embryogenesis that connects the two hemispheres of the brain. This surprising finding of agenesis of the corpus callosum was noted during examination of serial sections of adult brains for the investigation of the hydrocephalic phenotype. Since the corpus callosum develops between ≈16 and 18 days post coitus (d.p.c.) in the mouse (37, 38), we examined serial sections of the brains of E18 C57BL/6 fetuses. By 18 d.p.c., the corpus callosum was well developed in wild-type and Nfia^+/− animals but was absent from Nfia^−/− animals (Fig. 3 A vs. B, expanded box in 3 C vs. D, −/− vs. +/+, labeled cc). Slight expansion of the lateral ventricles was frequently detected in coronal sections of brains of Nfia^−/− animals suggesting that hydrocephalus begins to develop by E18, near the time of the development of the corpus callosum. While the corpus callosum was absent from the Nfia^−/− mice, the anterior and hippocampal commissures appeared normal in these animals (data not shown).

Figure 3.

Absence of corpus callosum in Nfia^−/− mice. Cresyl violet-stained coronal sections thorough the brains of C57BL/6 fetuses 18 days post coitus. (A and B) Wide-field pictures of the brains of −/− and +/+ littermates with the region surrounding the corpus callosum boxed. (C and D) Expansion of A and B showing the corpus callosum in the +/+ animal (labeled cc in D) and the absence of the corpus callosum in the −/− animal. Serial sections throughout this region showed a complete absence of callosal development in −/− animals. (Bars at the bottom right of B and D equal 2 and 1 mm, respectively.)

Since binding sites for NFI proteins have been implicated in the expression of GFAP and PLP in the brain (39, 40), we examined the expression of these proteins in adult Nfia^−/− mice by immunocytochemistry. Wild-type mice showed normal patterns of expression of both PLP (Fig. 4B) and GFAP (Fig. 4D) with strong PLP expression in the fiber tracts of the corpus callosum (Fig. 4B, horizontal band above hippocampus) and fimbria (Fig. 4B, tracts below to the left and right of the hippocampus). GFAP expression was seen in astrocytes throughout the cortex and hippocampus (Fig. 4D). Nfia^−/− mice showed relatively normal expression of PLP (Fig. 4A; note, however the expanded ventricles and absence of the corpus callosum vs. 4B) but had reduced levels of GFAP expression in both the cortex and hippocampus (Fig. 4 C vs. D; higher magnification of dentate gyrus in 4 E vs. F). Immunostaining with another astrocyte marker, S100, indicated similar numbers of total astrocytes in the Nfia^−/− and Nfia^+/+ animals (data not shown), suggesting a down-regulation of GFAP expression in astrocytes in specific regions of the brain (see Discussion).

Figure 4.

Reduced GFAP expression in Nfia^−/− mice. Coronal cryostat sections of the brains of 3-month-old Black Swiss Nfia^−/− (−/−) and wild-type mice (+/+) were fixed and stained with antibodies against PLP and GFAP. (A and B) PLP expression in Nfia^−/− (A, −/−) and wild-type (B, +/+) mouse brains. Note expanded ventricles indicating hydrocephalus and absence of the corpus callosum (tract lying directly above the hippocampus spanning the two hemispheres) in A vs. B. Expression of PLP appears normal in both animals in the tracts of the fimbria (dark sac-like regions on lower left and right of the hippocampus). (C and D) GFAP expression in Nfia^−/− (C, −/−) and wild type (D, +/+) mouse brains. Note the severe reduction in GFAP staining in the cortex and hippocampus (as assessed by reduced levels of punctate cellular brown precipitate), the expanded ventricles, and the absence of corpus callosum in C vs. D. Loss of GFAP staining is regional with reduction of staining in the cortex and hypothalamus but retention of staining with some altered morphology in the fimbria (sacs below the hippocampus) and thalamic nuclei (not shown). (E and F) Higher magnification of GFAP expression in the left dentate gyrus regions of C and D. Note distorted dentate gyrus appearance (probably secondary to hydrocephalus) and reduced GFAP expression in the dentate gyrus in E vs. F. (Bars = 1 mm.)

Discussion

These studies demonstrate that Nfia plays an important role in normal postnatal survival in the mouse and in the development of the corpus callosum in the developing brain. The ≈95% penetrance of lethality in the random-bred Swiss background suggests that while Nfia is important for normal postnatal survival, a small fraction of animals can survive to adulthood without Nfia, but they suffer severe neurological consequences. Given the previously described expression of Nfia at day 9 of embryonic development (7), it was somewhat surprising that Nfia-deficient mice survived embryonic development only to die perinatally (Table 1). The lack of major morphological defects in Nfia^−/− embryos (excluding the brain) and the characteristic deformities in surviving Nfia^−/− mice indicate that Nfia is not essential for development of most tissues and body structures (Figs. 2 and 3 and data not shown).

Even though the four NFI genes are expressed in broadly overlapping patterns during embryogenesis and adult life (7), there are apparently unique functions of the Nfia gene that cannot be compensated for by the other three NFI genes. It is of interest that the levels of expression of the other three NFI genes (Nfib, Nfic, and Nfix), as assessed by using quantitative real-time PCR (ABI Prism 7700), are about the same (<1.4-fold difference, data not shown) in brain RNA from adult wild-type and Nfia^−/− mice. Thus we have no evidence for compensation for loss of Nfia by increased levels of the other three gene products in the brain. It is possible that increased levels of the other three NFI genes in other tissues may account for the lack of phenotype seen in these organ systems. Since products of the four NFI genes bind to the same DNA sequence but differ in their promoter-specific activation and repression of transcription, the proposed defect in gene expression caused by loss of Nfia gene products may be due either to loss of the Nfia gene product function per se, or conversely to increased binding of other NFI family members at promoters normally occupied by NFI-A proteins. Because NFI proteins bind DNA as either homo- or heterodimers (41), it is also possible that heterodimers of NFI-A proteins with other NFI proteins, which would be missing in the Nfia^−/− mouse, may be important in postnatal development. Since NFI proteins differ in their abilities to both activate and repress transcription from reporter genes (30, 31), it will be important to determine to what extent the Nfia^−/− ES cells or fibroblasts exhibit changes in NFI-dependent gene expression. Such studies, together with analysis of global gene expression changes in tissues of the Nfia^−/− mice, may allow us to determine the changes in NFI-A-responsive transcriptional pathways that lead to the morphological phenotypes seen.

One indication of the cause of the postnatal lethality is the phenotype of the rare Nfia^−/− survivors. A tremor seen in all of the Nfia^−/− survivors indicates neurological defects. The “dome-head” phenotype of young Nfia^−/− animals (Fig. 2B) is lost as the animals age, and it is unclear whether it plays an important role in later phenotypes, including sterility and tremor. Also, agenesis of the corpus callosum does not in general cause postnatal lethality; indeed, several inbred stains of mice show hypoplasia or agenesis of the corpus callosum (37, 42–44). However, the combination of callosal agenesis and hydrocephalus bears some resemblance to the CRASH syndrome in humans (corpus callosum hypoplasia, mental retardation, adducted thumbs, spastic paraplegia, and hydrocephalus), which results in a number of developmental defects in the brain. Since mutations in the L1 gene have been found in many patients with CRASH syndrome (45–47) and mutations in the mouse L1 gene cause similar neuroanatomical defects (48, 49), it would be of interest to determine whether expression of the L1 gene is affected in the Nfia⁻ mice. However, hydrocephalus and/or agenesis of the corpus callosum is seen in several targeted and spontaneous mouse mutants [e.g., E2F-5 (50), Otx2 (51), Sek4 (52), βAPP (44), and CREB (53)], which suggests that a complex multigenic developmental pathway is disrupted in the Nfia^−/− mice. The observed runting that the rare Nfia^−/− survivors undergo during the first several weeks of postnatal life may also contribute either directly or indirectly to the phenotypes seen in the adult Nfia^−/− animals. Thus it will be important to establish the molecular pathways that lead to each of these diverse phenotypes (perinatal lethality, hydrocephalus, agenesis of the corpus callosum, reduced GFAP expression, tremor, male sterility) and to determine whether early runting contributes to later defects.

Agenesis of the corpus callosum can occur through a number of mechanisms (54), including defects in (i) generation or survival of the cells whose axons form the callosum, (ii) axonal outgrowth, (iii) axonal pathfinding, (iv) axonal crossing of the midline, and others. Some clues as to which process(es) may be affected by loss of Nfia come from the embryonic expression pattern of Nfia, the brain-specific promoters that contain NFI binding sites, and the final neuroanatomy of the callosal defect. Nfia is expressed from embryonic day 9 in the developing nervous system, and strong expression continues in the cerebral cortex and other regions up to birth (7). Since the neurons that eventually extend processes across the midline to generate the corpus callosum originate in regions of the neocortex, it is possible that loss of Nfia gene products affects the growth, differentiation, or viability of these neurons. However, the precise number and location of the cell bodies of neurons that pioneer the corpus callosum is still under investigation, and markers specific for this subset of neurons are not yet available (37, 38, 55). Furthermore, in some acallosal strains of mice it has been suggested that the inability to cross the midline may result from deficiencies in the substrates found by axonal growth cones at the midline (56). Thus the aberrant expression of genes in nonneuronal cells that are important for growth cone guidance could generate this acallosal phenotype in the Nfia^−/− mice. Given the previously reported high level of expression of Nfia at embryonic days 15–18 throughout the regions of the cerebral cortex known to contribute to the corpus callosum (7), the acallosal phenotype of the Nfia⁻ mice suggests an important role for NFI-A proteins in either developing neurons or glial cells in this region of the brain. However, it will be necessary to assess cell generation and death rates, as well as the expression levels of the large number of molecules known to influence axon guidance and axon outgrowth to determine the specific (and likely multiple) transcriptional pathways affected by the loss of Nfia gene products in the brain.

While the promoter regions of several nervous system-specific genes are known to contain NFI-binding sites [e.g., MBP (14), NF(M) (57), and 5HT3R (15)], many of these genes are expressed in tissues only at times later than callosal formation (15–18 days post coitus) and thus are unlikely to play a role in the process. Furthermore, the apparent absence of Probst bundles in the Nfia^−/− mice suggests that a relatively early step in the formation of the corpus callosum is affected (54). Probst bundles are aggregations of fibers that result from a failure of axons to cross the midline after their initial extension to the medial hemispheric walls. No Probst bundles were seen in serial sections of E18 brains, even in very rostral sections where bundles are seen at this stage in some other acallosal mice (52, 58). In adult brains the distortion of the ventricular walls due to extensive hydrocephalus made identification of Probst bundles problematic, although bulging of the medial ventricular wall was noted in some but not all older animals (data not shown). Such bulges may be compensatory changes caused by the hydrocephalus or indicate a relatively late onset of Probst bundle formation. Thus, while the absence of embryonic Probst bundles suggests a defect in a step prior to this accumulation of aberrant axons, future studies must focus on the expression of genes known to be involved in neuronal proliferation, survival, outgrowth, and pathfinding, either by genetic analysis of mutant mice defective in these processes or by in vitro analysis of these processes in callosal neurons from the Nfia-deficient mice.

The down-regulation of GFAP in the Nfia⁻ mice is intriguing because in some congenital hydrocephalic disorders mild to strong astrogliosis has been noted (59), with GFAP being found in the cerebrospinal fluid of some hydrocephalic patients (60, 61). Also, in the congenital hydrocephalic HTX rat, reactive astrocytes are found in the more severely hydrocephalic rats relative to normal or mildly affected animals (62). Thus the absence of reactive gliosis in the Nfia^−/− mice may reflect either an adaptation by astrocytes to the hydrocephalic neuropathology or possibly a direct role for Nfia gene products in GFAP expression. Indeed, binding sites for NFI proteins have been identified in the GFAP promoter, and mutation of one of these sites reduces GFAP gene expression (39, 63). It would therefore be of interest to determine whether the reactive astrogliosis seen in some mouse models of hydrocephalus and multiple sclerosis is abrogated in Nfia^−/− mice (64).

The apparent loss of heterozygous Nfia^+/− progeny only when the Nfia⁻ allele is transmitted from the female parent suggests an unusual maternal effect phenotype. Such specificity of loss of Nfia^+/− progeny could arise by a number of mechanisms, including (i) a defect in the Nfia⁻ female parent that influences the survival only of Nfia^−/− pups (true haploinsufficiency in both female parent and progeny); (ii) the Nfia alleles may be differentially expressed from the maternal and paternal genomes (imprinting); (iii) loss of maternally inherited Nfia gene products causes partially penetrant postnatal lethality (maternal effect phenotype); or (iv) a combination of the above. We have initiated long-term studies of fostering pups of Nfia⁻ females and assessment of Nfia levels in Nfia^+/− animals containing either maternal or paternal Nfia⁻ alleles to determine the cause of the perinatal lethal phenotype. Indeed, since all heterozygous Nfia⁻ animals examined have normal corpus callosums, a detailed analysis of the perinatal lethality in the heterozygotes may yield important insights into the proximate cause of death in the absence of potential physiological compensation for callosal agenesis and hydrocephalus.

The rare survival of Nfia^−/− mice may provide insights into genes or conditions that can modify the lethal phenotype in Nfia⁻ mice. Survival may be because of either a stochastic process in which a small fraction of homogeneous animals survive, or genetic heterogeneity in the random-bred Black Swiss strain used in these studies. We are currently breeding the Nfia⁻ allele into various inbred strains of mice to determine whether the degree of lethality is affected by genetic background. If a strain is found where lethality is significantly decreased, it should be possible to map the genes that affect lethality in Nfia⁻ mice. Such genes would be good candidates for cellular factors that interact with Nfia or other NFI gene products and affect the transcription or expression of NFI-dependent genes.

Abbreviations

NFI

nuclear factor I

ES

embryonic stem

RT

reverse transcription

GFAP

glial fibrillary acidic protein

PLP

myelin phospholipid protein

En

embryonic day n