|Home | About | Journals | Submit | Contact Us | Français|
The transcription factor family of nuclear factor I (NFI) proteins is encoded by four closely related genes: Nfia, Nfib, Nfic, and Nfix. A potential role for NFI proteins in regulating developmental processes has been implicated by their specific expression pattern during embryonic development and by analysis of NFI-deficient mice. It was shown that loss of NFIA results in hydrocephalus and agenesis of the corpus callosum and that NFIB deficiency leads to neurological defects and to severe lung hypoplasia, whereas Nfic knockout mice exhibit specific tooth defects. Here we report the knockout analysis of the fourth and last member of this gene family, Nfix. Loss of NFIX is postnatally lethal and leads to hydrocephalus and to a partial agenesis of the corpus callosum. Furthermore, NFIX-deficient mice develop a deformation of the spine, which is due to a delay in ossification of vertebral bodies and a progressive degeneration of intervertebral disks. Impaired endochondral ossification and decreased mineralization were also observed in femoral sections of Nfix−/− mice. Consistent with the defects in bone ossification we could show that the expression level of tetranectin, a plasminogen-binding protein involved in mineralization, is specifically downregulated in bones of NFIX-deficient mice.
Nuclear factor one (NFI) proteins were identified in 1982 as site-specific DNA-binding proteins binding to regulatory elements of the chicken lysozyme gene (50). Later studies revealed that these proteins function as transcriptional activators and repressors of cellular and viral genes but are also involved as cellular host factors in viral replication (27). In vertebrates, the Nfi gene family consists of four closely related genes, named Nfia, Nfib, Nfic, and Nfix (37, 58). They encode proteins with a conserved N-terminal DNA-binding and dimerization domain (26) and a C-terminal transactivation/repression domain which exhibits a high variability due to extensive alternative splicing (29, 38, 48). NFI family members act as homo- and heterodimers (39) and bind with high affinity to the palindromic consensus sequence 5′-PyTGGCA-N3-TGCCAPu-3′ (6, 20, 51). Besides the original detection of binding sites for NFI proteins in regulatory regions of the lysozyme gene (6, 50), NFI-binding motifs were subsequently detected in promoters of genes expressed in almost every organ, including brain (4, 32, 36), lung (2), liver (33), intestine (69), muscle (61), and connective tissue and skeletal elements (1, 12, 52, 53, 64).
In order to gain insight into the individual and common functions of the murine Nfi genes, knockout analyses of each family member have been undertaken. Loss of NFIA is perinatally lethal and results in hydrocephalus, corpus callosum agenesis, and defects in midline glial structures (15, 59). NFIB-deficient mice die at birth of respiratory failure due to an arrest in late lung development and also show neuronal defects that are similar, but more severe, than in Nfia−/− mice (28, 63). Disruption of the murine Nfic gene leads to a lack of molar roots as well as thin and brittle mandibular incisors. Besides these tooth pathologies, no other defects are observed, and when fed soft chow, Nfic−/− mice have a normal life span (62). This, and the stringent spatiotemporal gene regulation of NFI family members during embryogenesis and in adult animals, shows that NFI proteins exert important functions during mouse development (11).
Here we report for the first time the inactivation of the Nfix gene in the mouse, by inserting a lacZ reporter gene into the reading frame of exon 2, thereby deleting the essential coding region of the DNA-binding domain. Nfix−/− mice are growth retarded and die between postnatal day 21 (P21) and P28. They exhibit an enlargement of the lateral and third brain ventricles and a partial agenesis of the corpus callosum. Additionally, NFIX-deficient mice show a kyphotic deformation of the spine and impaired endochondral ossification in the vertebrae and femur. The expression of tetranectin, a protein involved in bone metabolism, has been found to be downregulated in Nfix−/− mice. Taken together, our studies provide evidence that NFIX plays an essential role in the development of the brain and the skeleton.
A murine λ FIX II library (Stratagene) generated from Black Agouti 129 genomic DNA (57) was screened with an Nfix cDNA probe for clones containing exon 2. The isolated clones were characterized, and the localization of exon 2 was confirmed by restriction mapping, Southern blot analysis, and DNA sequencing. For the construction of the targeting vector, a 7.6-kb exon 2-containing ApaI fragment was used. Using site-directed mutagenesis, an EcoRI site was generated at the 5′ end of exon 2, thereby deleting the entire 3′ region and 32 bp of the following intron. This EcoRI site was then used to integrate a cassette containing the bacterial β-galactosidase gene (β-Gal) and a loxP-flanked neomycin phosphotransferase gene (neor), driven by the herpes simplex virus thymidine kinase (HSV-tk) promoter, into the reading frame of exon 2, resulting in a fusion of the first 4 nucleotides of exon 2 and β-Gal. The β-Gal-pA and pHSV-tk-loxP-neor-loxP cassette replaces 564 bp of Nfix, including all but 4 bp of exon 2. It is flanked by a 5.1-kb homology arm at the 5′ end and 2 kb at the 3′ end. In order to place an HSV-tk cassette as negative selection marker at each end of the vector, the construct was cloned into the BamHI site of the λKO vector (49).
CJ7 embryonic stem (ES) cells (1 × 107) were electroporated with 25 μg of NotI-linearized targeting vector, plated on embryonic feeder cells on tissue culture dishes, and then selected in 0.35 mg/ml G418 (Gibco-BRL) and 2 μM ganciclovir (Sigma). After 5 to 7 days of selection, distinct colonies were picked and expanded in 96-well plates. Two plates were frozen, and one plate was used for DNA isolation. Southern blot analysis of EcoRV-digested genomic DNA revealed 5 out of 178 clones positive for homologous recombination, corresponding to a recombination efficiency of 2.8%.
To delete the neor selection marker, 1 × 107 targeted ES cells from two independent NfixlacZ-neor ES cell clones were transfected with 30 μg of supercoiled Cre-encoding plasmid (pPGK cre bpA) by electroporation. Cells were plated on 9-cm dishes for 24 h and then replated at 1 × 103 cells/9-cm plate. Four days after replating, 48 distinct colonies were picked and plated on one 96-well selection plate and one duplicate plate. As soon as colonies became established again, 0.4 mg/ml G418 was added to the medium on the selection plate. Dying G418-sensitive colonies became visible after 2 to 5 days, and corresponding clones could then be expanded from the duplicate plate (65). Deletion of the neor cassette was verified by Southern blot hybridization of NcoI-digested genomic DNA with a SalI-NcoI fragment as probe.
Mutant ES cells were injected into C57BL/6 blastocysts, and male chimeras were bred to C57BL/6 females to transmit the disrupted allele through the germ line.
Yolk sacs of embryonic day 10.5 (E10.5) to E18.5 or 0.5-cm tail tip biopsy samples of P21 to P28 mice were shaken overnight at 56°C in 0.5 ml of Laird's buffer (200 mM NaCl, 100 mM Tris-HCl pH 8.3, 5 mM EDTA, 0.2% sodium dodecyl sulfate) containing 0.15 mg/ml proteinase K. One μl of the lysate was analyzed by PCR in a total volume of 30 μl (1 U of Taq DNA polymerase S [Genaxxon], 0.1 mM deoxynucleoside triphosphate mix [Fermentas], 1× PCR buffer [Genaxxon, containing 100 mM Tris-HCl pH 9.0, 500 mM KCl, 15 mM MgCl2, 1.0% Triton X-100], Nfix-specific primer P1 at 5 pmol and P2 at 10 pmol, and lacZ-specific primer P3 at 15 pmol). The primer sequences were as follows: P1, 5′-GGAGCAAGTGGCAGGCGAAGCTGG-3′; P2, 5′-GGTTGATGAGAGCAACCTAC-3′; P3, 5′-CGCTTCTGGTGCCGGAAACC-3′. PCR conditions were 5 min at 94°C, 30 s at 94°C, 1 min at 60°C, and 1 min at 72°C for 40 cycles and 10 min at 72°C. PCR products were resolved on 1.5% agarose gels and visualized with ethidium bromide.
Embryonic and adult RNAs from different organs were isolated with TRIzol reagent (Invitrogen), and cDNA was generated from 1 μg RNA with the SuperScript first-strand synthesis system (Invitrogen). Nfix transcripts were analyzed by PCR using an exon 1-specific primer (5′-GTACTCCCCGTACTGCCTCACCC-3′) and an exon 4-specific primer (5′-CTGTCACATTCCAGACCCCA-3′). PCR products were resolved on 2% agarose gels and visualized with ethidium bromide. Primers for the analysis of type II pro-collagen α1 (col2a1), osteocalcin, and tetranectin transcripts were as follows: col2a1, 5′-GCGAGAGGGGACTGAAGGGACACC-3′ (forward) and 5′-CGGGGCTGCGGATGCTCTCAAT-3′ (reverse); osteocalcin, 5′-TCTGCTCACTCTGCTGAC-3′ (forward) and 5′-GGAGCTGCTGTGACATCC-3′ (reverse); tetranectin, 5′-GGATTTTGGGGCACCTACCTG-3′ (forward) and 5′-GCATTGCTTGTCGAACCAC-3′ (reverse); hypoxanthine phosphoribosyltransferase (HPRT), 5′-CACAGGACTAGAACACCTGC-3′ (forward) and 5′-GCTGGTGAAAAGGACCTCT-3′ (reverse).
Mice were anesthetized and investigated using a 3.0-T magnetic resonance imaging (MRI) machine (Trio; Siemens, Erlangen, Germany). The following sequences were used: T2-weighted high-resolution fast spin echo sequences with a 512 by 512 matrix and a field of view of 100 by 100 mm. The spatial resolution in plane was 0.19 mm. The slice thickness was 0.8 mm. The dimension of one voxel was therefore 0.19 by 0.19 by 0.8 mm. We performed 15 slices in sagittal and 15 slices in transversal orientations of the whole mouse body.
Sacrificed mice were skinned, eviscerated, and fixed in 95% ethanol for 10 days. After incubation in acetone for 24 h, double staining with Alizarin red S and Alcian blue was performed (150 mg/liter Alcian blue [Sigma] in 1 volume of 70% ethanol; 50 mg/liter Alizarin red S [Sigma] in 1 volume of 95% ethanol, 1 volume of glacial acetic acid, and 17 volumes of 70% ethanol) at 37°C for 4 days. Specimens were washed for 2 h in tap water and then cleared in 1% KOH until the skeleton was clearly visible through the surrounding tissue. Following this incubation the skeleton was placed successively for 7 days into 20% glycerine-1% KOH, 50% glycerine-H2O, 80% glycerine-H2O, and 100% glycerine solutions.
Whole brains were fixed in 4% paraformaldehyde overnight, and spines were fixed in Bouin's fixative (Sigma) for 7 days. Both tissues were then dehydrated in ethanol and embedded in paraffin. Serial sections of 4 to 10 μm were made and stained with hematoxylin and eosin (Sigma) following standard procedures. For 5-bromo-4-chloro-3-indolyl-β-d-glucuronic acid (X-Gal) staining, brains and bones were snap-frozen in OCT Tissue Tek (Reichert-Jung GmbH) and cryosectioned using a Kryo Cut 2000 cryostat (Reichert-Jung). Fourteen- to 20-μm sections were mounted on SuperFrost slides and dried for 30 min. Sections were fixed in cold 2% paraformaldehyde-0.2% glutaraldehyde and stained overnight at 37°C in X-Gal staining solution (10 mM potassium ferricyanide, 10 mM potassium ferrocyanide, 2 mM MgCl2, 0.02% NP-40 in phosphate-buffered saline containing 1 mg X-Gal/ml of N,N-dimethylformamide). After X-Gal staining, slides were washed in phosphate-buffered saline, placed into tap water, and then transferred into a Nuclear Fast Red staining solution (Sigma) for 1 to 10 min. After staining, slides were dehydrated in ethanol and mounted for analysis.
Mice were anesthetized and investigated on a multislice computed tomography (CT) scanner (Somatom Sensation 16; Siemens, Erlangen, Germany). Surface-rendering and volume-rendering techniques were used for three-dimensional (3D) visualization of multislice CT data to evaluate changes in bone structure in NFIX-deficient mice. Data acquisition was performed using helical multislice CT (120 kV and 120 mA; 0.6-mm collimation).
Generation of an Nfix null allele was achieved by homologous recombination of a replacement-type targeting vector. The targeting construct was designed to disrupt Nfix gene expression by an in-frame insertion of a reporter gene (lacZ) and a loxP-flanked PGKneor-pA cassette into exon 2, thereby replacing all but the first 4 bp of exon 2 and 32 bp of the following intron (Fig. (Fig.1A).1A). Since exon 2 encodes the DNA-binding domain of the NFIX protein, deletion of this region results in a nonfunctional NFI fusion protein. The β-Gal activity of the fusion protein expressed under the endogenous Nfix promoter therefore reflects NFIX expression.
The targeting vector was introduced into CJ7 ES cells, and selected clones, which showed homologous recombination, were identified by Southern blot analysis (Fig. (Fig.1B).1B). Deletion of the selection marker gene was achieved by transient Cre-mediated recombination of the two loxP sites and verified by Southern blot analysis (Fig. (Fig.1C).1C). Two targeted NfixlacZ cell clones were injected into blastocysts to generate germ line chimeras. Male chimeras from one ES cell clone were used for matings with C57BL/6 females. Heterozygous offspring were crossed for homozygosity, and DNA from tail tip biopsy samples was taken to detect the mutated allele by PCR (data not shown) and by Southern blot analysis (Fig. (Fig.1D).1D). To determine whether the NfixlacZ allele was a null allele, RT-PCR was performed. Total mRNA transcripts from brain, liver, and lung from 3-week-old littermates of each genotype were reverse transcribed, and cDNA was analyzed with Nfix exon 1- and exon 4-specific primers (Fig. (Fig.1E).1E). A 622-bp fragment was amplified from wild-type and heterozygous cDNA corresponding to the full-length transcript containing exon 2 and exon 3. No such product could be amplified in tissues from homozygous animals. A smaller 147-bp PCR product was detected in homozygous and heterozygous samples and very weakly in wild-type samples. Sequencing of this smaller fragment showed that it resulted from a direct splice of exon 1 to exon 3 (data not shown). As the exon 1 to exon 3 junction is out of frame, the only possibility is that a 32-amino-acid missense polypeptide is synthesized that terminates in exon 4. It is remarkable that in tissues of Nfix−/− mice the exon 1 to exon 3 splice is more frequent, once the exon 2 to exon 3 splice is eliminated.
From a total of 35 heterozygous matings, Nfix homozygous mutant animals were obtained at a normal Mendelian ratio: 95 Nfix+/+ (31.6%), 130 Nfix+/− (43.3%), and 75 Nfix−/− (25%) mice. NFIX-deficient mice developed a dome-shaped head, were unable to fully open their eyes, and had an obvious deformation of the spine (Fig. (Fig.2A).2A). They showed an ataxic gait and, when lifted by their tails, they drew their limbs in toward their bodies (feet-clasping posture) (9), in contrast to the full extension of limbs observed with normal littermates, indicating neurological abnormalities (Fig. (Fig.2B).2B). With the exception of two animals that reached an age of 3 and 7 months, all other Nfix−/− animals died between P21 and P28 (Fig. (Fig.2C).2C). When born, Nfix−/− mice were indistinguishable from their wild-type and heterozygous littermates, but by P5, daily weighing showed that Nfix−/− mice were unable to gain weight as efficiently as their wild-type littermates. Between P16 and P19 their weight plateaued and then progressively decreased. Shortly before death the body weight amounted to merely 60 to 70% of the weight of their normal siblings (Fig. (Fig.2D).2D). Heterozygous mice showed a slight weight reduction but no obvious anatomical or behavioral defects (Fig. (Fig.2D2D).
To further characterize the defects in Nfix−/− mice, histological examinations of inner organs were performed, but apart from a generalized reduction in size, almost all organs appeared normal. However, in all affected animals the smaller digestive tract appeared pathological, showing thinning of the intestine walls and reduced blood supply and containing a yellowish/brownish fluid (Fig. (Fig.2E).2E). In histological preparations, myenteric neuronal innervation appeared normal, as well as the mucosa, submucosa, muscle layers, and serosa (data not shown). In addition to the intestinal anomalies, a general loss of muscle tissue was observed, but histological examination of muscle samples showed no evidence of a myopathy (data not shown). Taken together, death of Nfix−/− mice is most likely a consequence of the spectrum of neurological and intestinal problems which result in an inability to thrive.
For further analysis, magnetic resonance imaging was performed with 3-week- and 3-month-old NFIX-deficient animals. Hydrocephalus was revealed in 3-week-old Nfix−/− mice which was characterized by dilatation of the lateral brain ventricles and the third ventricle. The fourth brain ventricle showed a normal size (Fig. 3A and B). Analysis of 3-month-old mice showed that hydrocephalus progressed with increasing age (Fig. 3C and D).
Imaging of the spine revealed a progressive skeletal pathology. The mouse spine has two physiological curvatures in the ventral axis. There is a cervico-cervical lordosis and thoraco-thoracical kyphosis. Angles of both vertebral bowings were measured using the Cobb method (8) in 3-week-old and 3-month-old Nfix−/− mice and compared to those of wild-type control mice. The cervical angle in 3-week-old control mice was 74.9°, and the thoracic angle was 96.4°. In NFIX-deficient mice of the same age, corresponding angles were 52.8° and 81°, respectively (Fig. (Fig.3A).3A). At the age of 3 months, both angles in control mice became wider, measuring 88.2° for the cervical angle and 115° for the thoracic angle. In the 3-month-old Nfix−/− mouse the opposite effect was observed, as both angles became acuter and measured 41° and 52°, respectively. These changes caused severe cervico-cervical hyperlordosis and thoraco-thoracical hyperkyphosis. Because of the increased kinking of the spine, the truncal length of Nfix−/− mice appeared significantly reduced (Fig. (Fig.3C).3C). Scoliosis, which is characterized by an abnormal lateral curvature of the spine, was also observed.
To examine the presence of primary skeletal abnormalities, whole-mount Alizarin red and Alcian blue skeletal preparations of 3-week- and 3-month-old Nfix−/− and control mice were compared. No alterations in the number of vertebral bodies, no bone deformations, and no generalized skeletal dysplasias were observed (Fig. 3E and F). Other skeletal bones, like the femur and tibia, were 25 to 30% shorter compared to wild-type animals (data not shown), which contributed to the general reduction in body size of Nfix−/− mice. Alterations in cranial bones appeared to be caused by the progressive hydrocephalus and were characterized by macrocephaly, thinning of the cranial bones, and less-pronounced cranial sutures (Fig. 3E and G).
To further characterize the hydrocephalic phenotype, serial coronal sections of brains of adult Nfix−/−, Nfix+/−, and wild-type mice were performed. Severe dilatation of the lateral brain ventricles and the third ventricle was noted in Nfix−/− mice compared to sections of wild-type brains. The fourth ventricle of Nfix−/− mice was not altered. In Nfix+/− mice a slight enlargement of lateral and third ventricles was also observed (Fig. 4A to F). Analysis of sagittal sections and intraventricular dye injections indicated that communication between the lateral and fourth ventricle was not inhibited, and so stenosis of the aqueduct as a possible cause for the hydrocephalus can be excluded (data not shown). Another defect observed in Nfix−/− and Nfix+/− mice was a partial agenesis of the corpus callosum. In coronal sections of rostral brain regions the callosal body was present but rather thin compared to wild-type controls (Fig. 4B and F), whereas in more-caudal regions it was completely absent in NFIX-deficient (Fig. 4G and H) and heterozygous (data not shown) animals. To analyze the expression pattern of the Nfix gene in the brain, we stained coronal and sagittal sections of Nfix−/− brains with X-Gal to identify the NFIX::LacZ fusion protein. Expression was detected in the hippocampal dentate gyrus, in the ependymal cell layer of the ventricular system, and in the cortex (Fig. (Fig.4I).4I). Sagittal sections also revealed strong staining in the cerebellum (Fig. (Fig.4J4J).
To determine whether the observed callosal defect was already evident during development, we examined serial brain sections of E17.5 embryos. In rostral Nfix−/− brain regions the corpus callosum was present but thinner than in wild-type animals (Fig. 5A and B), whereas more caudally NFIX-deficient mice exhibited complete agenesis of the callosal body (Fig. 5G and H). A slight reduction of callosal fibers was also observed in Nfix heterozygous mice compared to wild-type controls (Fig. 5C to F). In addition an enlargement of lateral brain ventricles was evident at this early developmental stage in brains of Nfix−/− mice (Fig. 5G and H).
We determined by RNase protection assays the expression of a series of neuronal marker genes, including myelin basic protein, neurofilament (NF-68), and glial fibrillary acidic protein (GFAP). The analysis revealed no alterations in the expression of these genes in brains of NFIX-deficient mice compared to control mice (data not shown).
Besides the observed brain malformations, NFIX-deficient mice also exhibited severe skeletal abnormalities. In order to investigate the observed deformation of the vertebral column in more detail, paraffin sections of spines of 14-day-old, 23-day-old, and 7-month-old NFIX-deficient and control mice were examined (Fig. (Fig.66 and Fig. Fig.7).7). Histological analysis revealed that in 14-day-old Nfix−/− mice larger amounts of cartilaginous material were present in the vertebral bodies in comparison to wild-type animals, indicating that the ossification process was delayed (data not shown). Sections of spines of 23-day-old Nfix−/− mice showed more pronounced morphological alterations. In regions of the most severe cervical and thoracic bowing, we observed that enlarged cartilaginous endplates associated with chondrocyte-like cells reached into the marrow cavity of the vertebral bodies, while control mice showed complete ossification of these structures (Fig. 6A and C). In addition, trabecular and cortical bone was thinner in bones of Nfix−/− mice than in wild-type animals (Fig. 6B and D). The intervertebral disks showed pathological alterations as well. They appeared irregular in shape, and the nucleus pulposus was smaller and sometimes fragmented (Fig. 6E and F). In some disks the lamellar orientation of the collagen fibers in the annulus fibrosus was also disturbed. X-Gal staining of cryosections of spines of Nfix−/− mice revealed Nfix gene expression in chondrocytes of the cartilaginous endplate and in trabecular bone (Fig. (Fig.6G).6G). To see if the impaired ossification was overcome during maturation, the spine of the one surviving 7-month-old mouse was analyzed and compared with a wild-type mouse (Fig. (Fig.7).7). Vertebral bodies were completely ossified, and no remaining chondrocyte-like cells were found in the vertebral marrow in the Nfix−/− mouse, though cartilaginous endplates were still abnormally enlarged (Fig. 7B and D). In addition, the anterior part of the intervertebral disks seemed compressed and the nucleus pulposus exhibited degenerative changes when compared to disks of the wild-type mouse (Fig. (Fig.77).
To analyze whether ossification was also impaired in other bones, femoral sections of 22-day-old Nfix−/− and wild-type mice were histologically analyzed (Fig. (Fig.8).8). Hematoxylin- and eosin-stained paraffin sections revealed an enlargement of the epiphyseal growth plate in Nfix−/− femurs compared to wild-type mice (Fig. 8A and B). This enlargement was due to an increase of the number of chondrocytes within the resting zone. In contrast, the zone of columnar proliferating chondrocytes was severely reduced and the zone of hypertrophic chondrocytes exhibited also a slight reduction (Fig. 8C and D). Trabecular bone mass was diminished, and an altered appearance of the bone marrow was evident compared to control mice (Fig. 8E and F). NFIX expression was detected in epiphyseal chondrogenic cells and also in the trabecular bone-forming cells, which corresponded to the observed defects (Fig. 8G and H). To analyze whether the impaired ossification and reduced mass of trabecular bone were associated with reduced bone density, radiographic analyses were performed. CT analysis revealed a mineralization deficiency in the hind limbs and jawbones of Nfix−/− mice compared to wild-type mice (Fig. (Fig.9).9). These findings indicate that NFIX plays an important role during endochondral ossification and mineralization.
In order to elucidate the molecular mechanism of delayed endochondral ossification and decreased mineralization in Nfix−/− mice, we analyzed specific skeletal marker genes. Semiquantitative RT-PCR analysis of RNA from bones of adult Nfix−/− and wild-type mice revealed no significant alterations in collagen 1α1, collagen 2α1, collagen 10α1, and osteocalcin gene expression (Fig. 10A and B and data not shown).
Furthermore, we analyzed the expression of the plasminogen-binding extracellular matrix protein tetranectin, because it had been postulated earlier that tetranectin plays a crucial role in mineralization during osteogenesis (68). Furthermore, tetranectin deficiency in the mouse leads to kyphosis due to an asymmetric development of growth plates and intervertebral disks (30). We detected a significant decrease in the level of tetranectin mRNA transcripts in lung, skeletal muscle, and femurs of adult Nfix−/− animals compared to wild-type tissues (Fig. 10C). These data indicate that NFIX is involved in the regulation of the activity of the tetranectin gene.
To investigate the function of the transcription factor NFIX, we generated NFIX-deficient mice by replacing the main part of exon 2 with an in-frame insertion of the lacZ reporter gene (Fig. (Fig.1).1). Nfix−/− mice die between postnatal days 21 and 28, exhibiting brain malformations (Fig. (Fig.44 and and5)5) and defects in endochondral ossification (Fig. (Fig.66 to to8).8). Hydrocephalus is a severe brain malformation which is known to be lethal in many cases; thus, early death of NFIX-deficient mice could be a direct result of the hydrocephalus. By the time of death, Nfix−/− mice are undernourished and exhibit some pathological changes of the digestive tract (Fig. (Fig.2).2). As it is known that hydrocephalus is often associated with reduced appetite, it might be possible that the smaller digestive tract is a consequence of impaired food uptake. Although we observed that NFIX-deficient mice showed a visible effort to suckle and eat, calorie uptake was obviously not sufficient to survive. In addition, the kyphotic deformation of the spine might lead to difficulties in swallowing, which would also contribute to malnutrition. Tooth defects as seen in Nfic−/− mice (62) could not be found.
Alternatively, malnutrition could also be a direct cause of gastrointestinal defects. It has been shown that NFI proteins regulate muscle- and extracellular matrix-specific genes, such as the α-sarcoglycan gene in myoblasts (18), the human elastin gene (17), the mouse smooth muscle myosin heavy chain gene (73), and the OCI-5/glypican 3 gene in intestinal epithelial cells (43). Thus, it is possible that defects in intestinal connective tissue or enteric muscle function lead to the pathological appearance of the intestine in NFIX-deficient mice. The high expression levels of the Nfix gene in mucosa, submucosa, muscle layers, and serosa (K. Driller, S. Krück, A. Gründer, and A. E. Sippel, unpublished data) further contribute to this suggestion.
It is also known that NFI proteins activate the human intestinal type IIb Na+-Pi cotransporter gene in the small intestinal epithelium (69) and bind to the enhancer region of the adenosine deaminase gene expressed in the intestinal epithelium at the suckling-weaning transition (47). NFI-binding sites have also been identified in the promoter region of the human sodium-dependent multivitamin transporter (19) and in the promoter region of the Na+/H+ exchanger gene NHE3 in intestinal epithelial cells (34). Thus, it is also possible that NFIX-deficient mice die as a result of impaired nutrient absorption due to biochemical defects in the function of intestinal cells which then would not be manifested as a gross histopathological defect. Further studies will be necessary to determine whether the malnutrition observed in Nfix−/− mice is influenced by an altered expression of any of these genes or if other factors are involved. In summary, we are presently unable to clarify whether early death of Nfix−/− mice is due more to hydrocephalus, more to the result from gastrointestinal defects, or to a combination of both.
Implications that NFI proteins are involved in neuronal development arose from the identification of NFI-binding sites in promoter regions of a large number of neuronal genes (32, 36, 42, 67) and also from the analysis of NFIA- and NFIB-deficient mice, which both exhibited neuronal disorders (15, 63). We showed that loss of NFIX also leads to brain malformations, including hydrocephalus and partial agenesis of the corpus callosum (Fig. (Fig.44 and and55).
It is known that hydrocephalus can be caused by several alternative mechanisms, like overproduction of cerebrospinal fluid (CSF) (44), defects in reabsorption of CSF in the subarachnoid space (14), or impaired flow of CSF due to dysfunction of ependymal cilia or aqueductal stenosis (31). As we found no obstructions of the ventricular spaces, we presumed that CSF flow is not disturbed in Nfix−/− mice and therefore not the cause of hydrocephalus formation. Cerebrospinal fluid is mainly produced by the choroid plexus, and overproduction of CSF has been reported in mice deficient for the transcription factor E2F-5 (44) and in mice carrying a frameshift mutation in the Hydin gene (16). Both genes are specifically expressed in ependymal cells, and mutant mice develop severe communicating hydrocephalus. Defects in reabsorption of CSF due to impaired function of ependymal cells or disturbed extracellular matrix can also lead to the development of hydrocephalus, as seen in transgenic mice overexpressing the cytokine transforming growth factor β1 (13). Misexpression of other factors such as matrix metalloproteinases and their specific inhibitors (tissue inhibitors of matrix metalloproteinases) have also been implicated in promoting hydrocephalus formation by altering the extracellular matrix environment (70). Since the Nfix gene is highly expressed in ependymal cells (Fig. (Fig.4)4) and in the choroid plexus (K. Driller, S. Krück, A. Gründer, and A. E. Sippel, unpublished data), it might be possible that NFIX is directly or indirectly involved in the regulation of CSF production and/or reabsorption.
As a second severe brain defect, NFIX-deficient mice showed a partial agenesis of the corpus callosum (Fig. (Fig.44 and and5).5). The corpus callosum is the largest fiber tract in the brain that connects the two brain hemispheres and extends from rostral to caudal brain regions. Callosal axons of the rostral part of the developing corpus callosum cross the midline at E15.5 by following so-called pioneering axons. The caudal region develops approximately 1 day later at E16.5, and callosal axons require the presence of the hippocampal commissure to cross the midline (56). In Nfix−/− mice the rostral part of the callosal body is present but rather thin, while more caudally it is completely absent, indicating that NFIX is involved in the development of both callosal regions. Given the expression of the Nfix gene in the cortex (Fig. (Fig.4),4), one could speculate that loss of NFIX results in impaired or defective outgrowth of callosal axons from cortical regions, resulting in a thin and partly absent callosal body. If axonal outgrowth is not affected, it might be that callosal axons are unable to cross the midline efficiently (rostral) or not able at all (caudal) due to defects in the pioneering axons or axons of the hippocampal commissure, respectively. In NFIA- and NFIB-deficient mice it was shown that the total absence of the corpus callosum resulted from abnormal development of midline glial populations, which was also reflected by reduced expression of the glial marker GFAP (15, 59, 63). In Nfix−/− mice histological analysis of midline glial cells did not indicate any obvious cellular defects and, also, GFAP gene expression was not altered (data not shown). Although it was postulated that NFI proteins bind to the GFAP promoter and NFIX regulates GFAP expression in astrocytes (10, 25), our data indicate that the absence of NFIX in glial cells might be compensated by other NFI family members and has no effect on GFAP expression in Nfix−/− mice. Taken together, we show that NFIX is a regulatory transcription factor crucial for correct brain formation.
Currently there exists only limited knowledge about the molecular regulatory mechanisms of the complex process of bone formation. Our data indicate that NFIX is a new factor involved in this process. NFIX-deficient mice are growth retarded and exhibit a deformation of the spine (Fig. (Fig.22 and and3).3). Histological analysis of spines and femurs of Nfix−/− mice revealed a delay in endochondral ossification characterized by an enlarged zone of resting chondrocytes and a reduced zone of proliferating and hypertrophic chondrocytes (Fig. (Fig.66 to to8).8). Trabecular bone formation and calcification are significantly reduced (Fig. (Fig.99).
NFI-binding sites are often found in close proximity to binding sites of other transcription factors, such as the activator protein 1 complex (AP1) and members of the specificity protein (Sp) family. Delayed endochondral ossification was also observed in mice with a conditional knockout of the AP1 transcription factor component Fra2 (35). The ossification delay described for Fra2−/− mice was due to impaired hypertrophic chondrocyte differentiation in vertebrae and femurs and resulted in the development of kyphosis, similar to the phenotype seen in Nfix−/− mice. Fra2 is a member of the Fos family of basic leucine zipper domain proteins, which all appear to be involved in bone development (66). They function as dimers with proteins of the Jun family by forming the transcription factor AP1. The predominant AP1 dimer in vertebral development seems to be the dimer formed by Fra2 and c-Jun, because the conditional knockout of c-Jun alters intervertebral disk development, which also leads to a kyphotic deformation of the spine (5). In addition, members of the Sp family appear to be involved in bone development as well. Sp3-deficient mice exhibit defects in skeletal ossification (24). Sp1 and Sp3 are involved in the regulation of skeletal genes like collagen type X in hypertrophic chondrocytes (46) and RANKL in osteoblasts and bone marrow stromal cells (45), and Sp7 plays a role in osteoblast differentiation and bone formation (71). This raises the question whether NFI, AP-1, and Sp family members interact in transcriptional regulation of skeletal genes during bone development. In this respect it is noteworthy to mention that we had earlier observed a local connection of four Jun and four NFI genes on three different human chromosomes (1, 9, and 19) and two different mouse chromosomes (4 and 8) (54). On the other hand, an interplay among Sp1, Sp3, and NFI was already demonstrated for several basal and tissue-specific genes, such as the P2X1 gene in megakaryocytes (72), the secretin gene in neuronal cells (42), the poly(ADP-ribose) polymerase gene (40), and the α1 (I) and (III) procollagen gene in fibroblasts (1, 12). Rafty and colleagues (55) delivered the first evidence for the physical interaction of Sp1 and NFIX during regulation of the rat PDGF-A gene in vascular smooth muscle cells. That AP1-, Sp1-, and NFI-binding sites are also found in regulatory regions of skeletal genes was shown for the mouse tetranectin promoter (60). The analysis of tetranectin mRNA levels revealed altered gene expression in NFIX-deficient mice, with reduced tetranectin transcripts in lung, muscle, and bone (Fig. (Fig.10).10). Tetranectin is a plasminogen-binding protein with a role in tissue remodelling and a potential function in mineralization during osteogenesis (68). Tetranectin-deficient mice exhibit kyphosis due to intervertebral disk defects (30), which indicates that tetranectin might be required for stabilizing the spine. It is assumed that the loss of tetranectin leads to the softening of the extracellular bone matrix and the intervertebral disk, which then results in spinal deformation. In accordance with that assumption, reduced tetranectin expression was also reported in mice deficient for the G-protein-coupled receptor GPR103, and these mice also developed kyphosis and reduction of trabecular bone (3). Taken together, the observed kyphotic phenotype in NFIX-deficient mice is consistent with reduced expression of tetranectin and suggests that tetranectin is a new downstream target of NFIX.
It is supposed that the four mammalian NFI genes arose through gene duplication and diversification from a putative NFI ancestral gene. In accordance with that, only one NFI gene (nfi-I) was found in the nematode Caenorhabditis elegans (22). The high degree of homology in gene sequence and structure is consistent with a general functional role as DNA-binding and regulatory proteins. Their specific and overlapping expression patterns indicate common, but also unique, functions for each family member. Comparison of the different knockout phenotypes of each Nfi gene in mouse and C. elegans helps to elucidate their role during development. These animal models revealed that NFI family members seem to regulate processes that occur relatively late in development, since none of the strains of knockout mice shows embryonic lethality (15, 28, 62, 63). This is surprising, given the early start of expression at E9 (Nfia), E9.5 (Nfib), E11.5 (Nfic), and E11.5 (Nfix) (11). Even loss of the putative NFI ancestral gene nfi-I in C. elegans is only lethal in adult life, though expression occurs throughout early embryogenesis (41). This indicates that during early development, NFI proteins seem to regulate downstream targets of minor importance for embryonic survival or that, in mice, other members of the Nfi family can compensate for the loss of one family member.
Furthermore, it is notable that three NFI-deficient strains, Nfia, Nfib, and Nfix null mice, exhibit similar brain defects (agenesis of the corpus callosum and hydrocephalus), which implies a common but not redundant function in brain development. This is consistent with the fact that the first organ where these three genes are expressed is the cortex in the developing forebrain. In NFIC-deficient mice no neuronal defects have been reported, although the Nfic gene is also expressed in the adult mouse brain (K. Driller, unpublished observation). Additionally, it has been shown that NFIC binds to the promoter of the secretin gene in neuronal cells (42) and is involved in the regulation of the GFAP gene in astrocytes (25), which raises the question whether the evolutionary distance of the Nfic gene in relation to the other family members is more progressed. In nfi-I-deficient worms, neurological defects have also been discussed to be the cause of the behavioral defects that were identified, indicating that regulation of neurological genes seems to be a conserved function of NFI family members throughout evolution.
It is striking that, besides the observed brain malformations, knockout mice for each Nfi gene show also very distinct defects in other organs, which is in accordance with the tissue-specific expression pattern of the four genes in different organs. Nevertheless, it is possible that related downstream targets or signaling pathways underlie these diverse defects. For example, proper lung and tooth development both depend on specifically regulated epithelial-mesenchymal interactions. This mechanism seems to be affected in Nfib−/− and Nfic−/− mice, since they show failure of lung maturation and impaired molar root formation, respectively, indicating that NFIB and NFIC contribute to epithelial-mesenchymal cross talk in the developing lung and in tooth buds. So far we have no indications that NFIX function also fits into this regulatory pattern. To gain further insight into the general function and possible redundancy of the Nfi gene family members, it is now possible to generate Nfi gene double- and triple-knockout mice. Further investigation is also needed to determine whether misexpression or mutations in NFI genes lead to human disorders. Brain defects seen in NFIA-, NFIB-, and NFIX-deficient mice resemble human malformation syndromes often associated with severe mental retardations (23). Skeletal defects such as the observed deformation of the spine in Nfix−/− mice appear similar to defects seen in Scheuermann's disease (7), and the multiple defects caused by NFIX deficiency resemble a new syndrome characterized by osteochondrodysplasia, associated with hydrocephalus and callosal agenesis (21). As knowledge of genetic defects in patients exhibiting these various disorders is very limited, Nfix might be a very interesting candidate gene to look for.
We are grateful to the Knockout Core Facility of the Albert-Ludwigs-Universität Freiburg (C. Peters) for blastocyst injections and the generation of founder chimeras. Special thanks go to Gudrun Krüger for excellent technical assistance and to J. Mason for reading the manuscript.
This work was supported by grants from the Deutsche Forschungsgemeinschaft to A.E.S. (SFB 592/B5).
Published ahead of print on 12 March 2007.