|Home | About | Journals | Submit | Contact Us | Français|
The Krüppel-like transcription factors (KLFs) are important regulators of cell proliferation and differentiation in several different organ systems. The mouse Klf7 gene is strongly active in postmitotic neuroblasts of the developing nervous system, and the corresponding protein stimulates transcription of the cyclin-dependent kinase inhibitor p21waf/cip gene. Here we report that loss of KLF7 activity in mice leads to neonatal lethality and a complex phenotype which is associated with deficits in neurite outgrowth and axonal misprojection at selected anatomical locations of the nervous system. Affected axon pathways include those of the olfactory and visual systems, the cerebral cortex, and the hippocampus. In situ hybridizations and immunoblots correlated loss of KLF7 activity in the olfactory epithelium with significant downregulation of the p21waf/cip and p27kip1 genes. Cotransfection experiments extended the last finding by documenting KLF7's ability to transactivate a reporter gene construct driven by the proximal promoter of p27kip1. Consistent with emerging evidence for a role of Cip/Kip proteins in cytoskeletal dynamics, we also documented p21waf/cip and p27kip1 accumulation in the cytoplasm of differentiating olfactory sensory neurons. KLF7 activity might therefore control neuronal morphogenesis in part by optimizing the levels of molecules that promote axon outgrowth.
Gene-targeted deletions in mice have demonstrated that members of the mammalian family of C2H2 zinc finger Krüppel-like factors (KLFs) play important roles in cell differentiation and embryonic development (5, 10, 26). These functions include control of erythroid cell proliferation and β-globin gene cluster activity (KLF1) (8, 45, 48); regulation of lung formation, blood vessel stabilization, and T-cell quiescence (KLF2) (30, 31, 61); terminal differentiation of dermal and intestinal epithelia (KLF4) (28, 53); involvement in cardiovascular remodeling (KLF5) (54); and modulation of uterine function (KLF9) (55). Additionally, KLF6 has been reported to be a tumor suppressor protein in prostate and colon cancers (44). KLF-like gene products have been also identified in lower vertebrate and invertebrate organisms, in which they appear to control cell differentiation during embryonic development (11, 24, 29, 46). Mammalian KLFs and the closely related group of Sp1-like proteins comprise 21 distinct molecules, which display highly homologous carboxy-terminal DNA-binding sequences and divergent amino-terminal domains that regulate gene transcription (5, 10, 26). KLF/Sp1-like proteins bind to similar “GT-box or CACCC element” sites on DNA and can function as activators, repressors, or both, depending on the promoter and cellular contexts (5, 10, 26).
KLF7 was originally identified during a PCR-based search of novel KLF transcripts and found to be broadly expressed at low levels in adult tissues, hence the early name of UKLF, for ubiquitous KLF (39). Subsequent gene expression studies of the developing mouse revealed that accumulation of Klf7 transcript is restricted to postmitotic neuroblasts of the developing central (CNS) and peripheral nervous systems (32, 35). Examples include the differentiating neuroblasts in the spinal cord, dorsal root ganglia (DRG), sympathetic ganglia, cerebral and cerebellar cortexes, hippocampus, olfactory system, and retina. Postnatal Klf7 expression was instead found to remain constitutively high only in the DRG, cerebellum, and olfactory system. Very recent studies demonstrated that KLF7 binds to and stimulates the activity of the proximal promoter of the cyclin-dependent kinase (cdk) inhibitor p21waf/cip gene (56). Based on these lines of correlative evidence, we proposed that KLF7 may be part of the genetic programs that regulate differentiation of progenitor cells, neuronal morphogenesis, and/or phenotype maintenance (32).
In order to elucidate the physiological function of KLF7 during mouse development, we have ablated its expression by gene targeting in embryonic stem (ES) cells. Here we report that loss of KLF7 activity leads to impaired axon projection in the olfactory and visual systems, cerebral cortex, and hippocampus, as well as reduced dendritic branching in the hippocampus. Consistent with previous findings, we found a significant downregulation of p21waf/cip gene expression in the olfactory sensory neurons (OSNs) of Klf7-null mice. We also observed a similar decrease in p27kip1 protein levels, which was associated with KLF7's ability to transactivate the p27kip1 promoter in cell transfection assays. Finally, we present correlative evidence suggesting that p21waf/cip and p27kip1 may contribute to the neuronal morphogenesis in the olfactory epithelium (OE).
The targeting vector was designed to replace most of exon 1 with the phosphoglycerate kinase (PGK)-neo cassette flanked by loxP sites (Fig. (Fig.1).1). Maintenance, transfection, and selection of mouse ES cells were performed as described previously (42). Two correctly targeted ES cell clones were selected to produce Klf7-null mice and yielded identical phenotypes. Electroporated ES cells and mutant mice were genotyped by Southern hybridization to probes upstream and downstream of the recombination site and by PCR amplification. Amplification primers included forward primer 5′-TTTCCTGGCAGTCATCTGCAC-3′ and reverse primer 5′-GGGTCTGTTTGTTTGTCAGTCTGTC-3′ to detect exon 1 of Klf7 and forward primer 5′-GCAGTCATCTGCACTGTACACG-3′ and reverse primer 5′-CGTTGTAAAACGACGGCCAGTG-3′ to detect the mutant allele without PGK-neo. About 200 ng of genomic DNA was PCR amplified for 35 cycles each under the following conditions: 95°C for 1 min, 63°C for 1 min, and 72°C for 1 min (first set of primers) and 95°C for 30 s, 60°C for 1 min, and 72°C for 30 s (second set of primers). Transgenic CMV::Cre mice were employed to excise the PGK-neo cassette in animals heterozygous for the targeted allele (42). The OMP::eGFP transgene included a 2-kb cassette made of the eGFP gene and the splice sites of the β-globin third intron, in addition to part of the 3′ untranslated region of the Drosophila melanogaster Klf7 transcript in order to monitor transgene expression in mice (11). This 2-kb cassette was placed downstream of the olfactory marker protein (OMP) promoter and translational start codon (a gift of F. Margolis, Baltimore, Md.) (9). The P2:lacZ transgenic mouse line was a gift of P. Mombaerts (New York, N.Y.) (41). Transgenic and targeted animals were generated by standard protocols at the Mount Sinai Mouse Genetics Shared Resource Facility (New York, N.Y.) (42).
In situ hybridizations were performed as described previously (32); probes included Klf7, Foxp1, Foxp2, Rorβ, Reelin (a gift from T. Curran, Memphis, Tenn.), and p21waf/cip (a gift from B. Vogelstein, Baltimore, Md.). For immunoblots, septa were isolated from four P1 mice and pooled together after removal of the airway epithelium. Antibodies against p21waf/cip and p27kip1 (1:1,000) were from Santa Cruz Biotechnology (Santa Cruz, CA), and those against TuJ (1:1,500) were from Chemicon (Temecula, CA); immunoblot assays were performed as described previously (32).
Functional assays employed a luciferase reporter gene construct driven by the 2.3-kb-long p27kip1 promoter sequence (a gift from G. Sonenshein, Boston, Mass.). The reporter gene construct was cotransfected into NIH 3T3 cells together with the Myc-tagged KLF7 expression vector, and luciferase activity was evaluated using a commercial kit (Promega, Madison, WI); transfection efficiency was normalized against the constitutively expressed Renilla luciferase gene (56). Functional assays were performed three times in duplicate, and the statistical significance of the resulting data was determined by the Mann-Whitney U test.
Chromatin immunoprecipitation (ChIP) was performed as described previously (56). 293T cells (~1 × 106) were transfected with 20 μg of Myc-tagged KLF7 expression vector or the control Myc-tagged plasmid, and ChIP was performed using a commercial kit (Upstate Biotechnology, Lake Placid, NY). Two consecutive PCRs were carried out, using 1 μl of the initial sample DNA (out of 20 μl total) and 1 μl of the first PCR mixture, respectively. Dimethyl sulfoxide was added at a final concentration of 10%. The cycling conditions were 25 cycles of 95°C for 1 min, 63°C for 1 min, and 72°C for 1 min, which was preceded by an initial denaturation step (95°C for 5 min) and a final extension step (72°C for 10 min). Amplified products were visualized by standard 2% agarose gel electrophoresis. The p27kip1 primers were forward primer 5′-AGGCCAGCCAGAGCAGGTTTGTTG-3′ and reverse primer 5′-TATGGCGGTGGAAGGGAGGCTGAC-3′.
Tissue specimens were fixed in 4% paraformaldehyde; washed; decalcified in EDTA; and either infiltrated with sucrose, frozen in OCT, and sectioned on a cryostat or dehydrated, embedded in paraffin, and sectioned on a regular microtome. Secondary antibodies were conjugated to either Alexa dyes (Molecular Probes, Eugene, OR) or biotin and then probed using the ABC procedure (Vector Labs, Burlingame, CA). Staining was visualized on a Nikon microscope using bright-field or fluorescence optics or on a Zeiss confocal laser scanning microscope. Antibodies were against OMP (a gift from F. Margolis, Baltimore, Md.); TuJ (Sigma, St. Louis, MO); TAG-1 (DSHB, Iowa City, IA); p21waf/cip and p27kip1 (Santa Cruz Biotechnology, Santa Cruz, CA); and GAP43, green fluorescent protein (GFP), NCAM, L1, and phospho-histone 3 (Chemicon, Temecula, CA). Golgi-Cox staining was performed using a commercial kit (FD Neurotechnologies, Catonsville, MD).
Histomorphometric and cellular analyses of the OE were performed on three coronal sections taken from comparable levels of wild-type and mutant septa along the rostro-caudal axis. Sections were subjected to phospho-histone 3 immunochemistry or TUNEL (terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling) assay, and positive cells along the septum were counted using Metamorph software. To estimate cell density, nuclei were stained with DAPI (4′,6′-diamidino-2-phenylindole), imaged by Zeiss LSM510 laser scanning confocal microscopy, and counted within multiple adjacent fields. Overall OE thickness was determined by measuring the distance from the basal lamina to the apical limit of the OE at multiple points along the septum. The cortical thickness of P1 brains was determined by measuring the distance between ventricle and brain surface at its greatest extension using midsagittal sections. Cortical cell numbers in a 170-μm-wide column from the ventricle to the surface of the brain were measured using the same sections and coordinates. The number of phospho-histone 3-positive cells was determined in the ventricular part of this column. All measurements were performed with Metamorph software (Universal Imaging Corporation, Downingtown, PA). Unless otherwise indicated in the figure legends, four or more animals per genotype and time point were used for embryonic/perinatal analyses and three or more were used for adult phenotypes.
Primary OSNs were prepared and cultured as described previously (51). Cells were fixed for 10 min in 4% paraformaldehyde, and antibodies were blocked for 30 min in 2% normal goat serum-0.2% Triton X-100; primary and secondary antibodies were incubated for 12 h at 4°C and for 1 h at room temperature, respectively. Antibody dilutions in blocking solution were 1:800 for p21waf/cip and 1:400 for p27kip1. Neurons were identified based on their morphology and expression of neuronal tubulin (TuJ), using a specific antibody (1:1,000 dilution). Detection and quantification of fluorescent signals in the entire cell were performed using the Metamorph software package (Universal Imaging Corporation, Downingtown, PA). Anterograde DiI labeling of the retinal projection was performed as described previously (50). Briefly, a small DiI crystal (Molecular Probes, Eugene, OR) was placed on the optic nerve head and allowed to diffuse through the optic nerve and chiasm; heads were kept in phosphate buffer with sodium azide for ~3 weeks at 37°C. Specimens were dissected and photographed as whole mounts under a fluorescent microscope.
KLF7 activity in the developing mouse was abolished by replacing the first exon of the gene with the PGK-neo cassette flanked by loxP sites (ΔK7-neo allele); homozygous mutant mice without the PGK-neo cassette (ΔK7/ΔK7) were generated by crossing heterozygous ΔK7-neo animals with transgenic mice constitutively expressing Cre recombinase under the control of the cytomegalovirus promoter (Fig. (Fig.1a).1a). Segregation of the mutant Klf7 alleles was followed by Southern hybridization (ΔK7-neo allele) or PCR amplification (ΔK7 allele) (Fig. 1b and c). In situ hybridizations of E13.5 spinal cord and P1 brain to a Klf7 probe documented loss of gene expression in homozygous ΔK7-neo and ΔK7 mice, respectively (Fig. (Fig.1d).1d). This last result was independently confirmed by Northern blot hybridizations, as well as by the failure to identify shorter in-frame Klf7 transcripts in reverse transcription-PCR or 5′ rapid amplification of cDNA end products obtained from homozygous mutant tissues (data not shown). ΔK7/ΔK7 and ΔK7-neo/ΔK7-neo mice were born at the expected Mendelian frequency and displayed the same phenotype. Specifically, the vast majority of mutant homozygotes (98.5% of a total of 800 offspring) died within the first 3 days of life, showing little or no milk in the stomachs, hypopnea, cyanosis, and failed response to clamp stimulation of their tails (Fig. (Fig.1e).1e). Severely hypoplastic olfactory bulbs (OBs) were the only overt anatomical abnormality of Klf7-null mice (Fig. (Fig.1e).1e). By contrast, ΔK7/+ or ΔK7-neo/+ mice displayed no anatomical abnormalities and were fertile and viable. The analyses described below were performed on ΔK7-neo/ΔK7-neo mice in the mixed genetic background 129Sv;C57BL/6.
The primary olfactory pathway consists of OSNs that project axons from the olfactory OE to the bulbs, where they synapse with projection neurons, which in turn relay odor information to higher brain centers (Fig. (Fig.2a)2a) (27). Development of the olfactory nerve in the mouse begins at ~E9.5 with the emergence of OSNs in the OE and progresses with neurite formation at ~E11.5 and contact of the forebrain by ~E12.5 (6). In situ hybridizations detected high levels of Klf7 transcripts in the OE as early as E11.5 (Fig. (Fig.2b).2b). Immunofluorescence analyses of coronal sections of OBs from newborn (P1) Klf7-null mice with markers for immature (GAP43) or mature (OMP) OSNs revealed lack of peripheral innervation. However, the number of OSNs within the olfactory epithelium of mutant mice did not appear to be substantially reduced at this developmental stage (Fig. (Fig.2c2c and data not shown) (2). By contrast, standard histology documented a seemingly normal nasal cavity in which the olfactory nerve had crossed the cribriform plate and entered the brain cavity (Fig. (Fig.2d).2d). Lack of peripheral innervation was further corroborated in nullizygous Klf7 mice harboring genetically marked OSNs. These mice were generated by intercrossing Klf7 mutant animals with transgenic animals expressing lacZ or enhanced GFP in a subset of or in all OE neurons under the control of the promoters of the P2 olfactory receptor or the OSN-specific OMP promoter, respectively (Fig. (Fig.2e2e).
Next, we followed nerve development in the olfactory cavity of mutant and wild-type mice in order to assess whether a transient contact might have been established between mutant axons and forebrain during embryogenesis. NCAM immunofluorescence, which marks axonal projections, showed that mutant neurites form at the right time (E11.5) but are fewer than in the wild-type sample; additionally, axon projections towards the forebrain are poor compared to wild-type controls (Fig. (Fig.3a).3a). Another axonal marker, GAP43, was employed to document that, unlike the wild-type counterpart, the mutant nerve fails to contact the E12.5 forebrain and later (E17.5) remains on the inner surface of the cribriform plate (Fig. (Fig.3a).3a). Together, this evidence favored the notion of a defect in axon growth. Parallel histomorphometric and immunohistochemical analyses failed to identify significant changes in OE thickness, cell density, proliferation, or apoptosis, thus excluding the possibility that loss of OSNs could account for the mutant phenotype (Fig. (Fig.3b).3b). We therefore concluded that the lack of OB innervation observed in newborn Klf7-null mice reflects mostly the involvement of the transcription factor in regulating OSN projection.
In the absence of other overt anatomical malformations, we examined axon projections in another sensory pathway, the neonatal visual system. Retinal ganglion cell (RGC) axons exit the retina at the optic nerve head and navigate through the optic nerve to reach the optic chiasm, at which point they diverge to form ipsilateral and contralateral projections (Fig. (Fig.4a).4a). In situ hybridization revealed that Klf7 is highly expressed in RGCs from the time that they first emerge (~E12) through E17.5 (Fig. (Fig.4b).4b). Initial histological analyses failed to reveal any obvious morphological differences between Klf7−/− and wild-type retinas. To visualize RGC axonal projections, we used the pan-neuronal marker GAP43. Immunofluorescence analyses revealed that mutant RGC axons project to the optic nerve head properly, but a small portion of them make aberrant projections and fail to exit the retina (Fig. (Fig.4c).4c). These findings were substantiated with another axonal marker, TAG-1, a protein related to the L1 family of cell adhesion molecules (20). These axon guidance errors are similar to those reported in netrin-1 mutants (12). Unlike GAP43 staining, TAG-1 staining revealed a surprising difference between Klf7−/− and wild-type animals. Whereas TAG-1 labeled the entire visual pathway from retina through optic tract in wild-type embryos (Fig. 4c and d), TAG-1 expression in Klf7−/− animals was strongly decreased at the chiasm and absent in the optic tract (Fig. (Fig.4d).4d). There is some evidence of moderately decreased TAG-1 levels in RGC axons after exit from the optic chiasm (38), but no studies have reported an absence of TAG-1 staining as seen in the Klf7 mutants. Interestingly, we also found that TAG-1 appears to be preferentially expressed in the ventral portion of the optic tract (Fig. (Fig.4d4d).
Anterograde DiI labeling of RGC fibers was performed to monitor projection errors at the chiasm. Unlike the very small ipsilateral projection in wild-type heads, mutant samples displayed an increase in ipsilaterally projecting fibers (Fig. (Fig.4e).4e). It is unknown if this new population of ipsilateral fibers arises from increased projections from the ventrotemporal retina (which normally project ipsilaterally) or ectopic projections from other regions of the retina. This point notwithstanding, we concluded that KLF7 plays an important role in RGC axon guidance at the optic nerve head and the optic chiasm and that its absence causes postchiasmatic TAG-1 expression to be downregulated.
Having demonstrated KLF7 involvement in neurite outgrowth in the olfactory and optic system, we searched for similar deficits in the Klf7−/− brain. Klf7 is highly expressed throughout the brain during embryogenesis and soon after birth, whereas high levels of Klf7 activity in the adult brain are mostly confined to the cerebellum, hippocampus, and OBs (32). Histological analyses of neonatal (P1) Klf7−/− brains revealed severe perturbations of the tracts that constitute the major forebrain connections (Fig. (Fig.5a).5a). Specifically, we found that the fibers of the corpus callosum do not cross the midline, the anterior commissure is missing or severely disrupted, and the fimbria is reduced in size (Fig. (Fig.5b5b and data not shown). TAG-1 immunofluorescence documented the failure of cortical efferent projections to enter the internal capsule (Fig. (Fig.5c)5c) (20). The comparable intensity of the TAG-1 signal in the mutant and wild-type brains suggested that the defect is not probably accounted for by a substantial reduction in the number of TAG-1-positive cortical neurons (Fig. (Fig.5c).5c). L1 is an adhesion molecule widely expressed in several neuronal populations, including those forming cortical afferent fibers (20). L1 immunofluorescence showed that cortical afferent fibers in Klf7-null brains are hyperfasciculated and that medial tracts are misrouted just before entering the cortical fiber layer (Fig. (Fig.5d).5d). GAP43 immunohistochemistry documented the same deficiencies in E16.5 mutant brains, thus suggesting that the cortical phenotype is probably not caused by excessive neuronal retraction and pruning after improper targeting (data not shown).
The cortical deficits described above could result from defective migration of progenitor cell migration, a process which is ultimately responsible for the formation of the multiple layers of the cerebral cortex (Fig. (Fig.5a)5a) (1). However, standard histology revealed a seemingly normal cortical layering in the mutant P1 brain (Fig. (Fig.6a).6a). This finding was extended at the molecular level by performing in situ hybridizations with probes specific for different layers of the cerebral cortex. They included Foxp2, which recognizes the neurons of the deeper layer 6; RORβ, which is specific for cells in layers IV and V; Foxp1, which is a marker for cells in layers III to V; and reelin, which is expressed by the Cajal-Retzius neurons in the outermost layer of the cortex (Fig. (Fig.6b)6b) (3, 18, 43). The results of these in situ hybridizations failed to detect significant differences between mutant and wild-type cortices (Fig. (Fig.6b).6b). However, a histomorphometric analysis estimated that the Klf7−/− cortex is ~30% thinner than the wild-type counterpart (Fig. (Fig.6c).6c). Although the phenotype could be caused in part by thinner white matter tracts, additional analyses revealed that there are ~15% fewer cells in the mutant than in the wild-type cortex (Fig. (Fig.6c).6c). Likewise, immunofluorescence using a generic neuronal marker, TuJ, showed a slightly less intense staining in the mutant than in wild-type cortex (Fig. (Fig.6d).6d). However, neither the TUNEL assay nor phospho-histone 3 immunohistochemistry correlated the morphological defect with statistically significant changes in cell death or proliferation (data not shown). Lacking additional evidence, we concluded that the dramatic loss of axon growth and consequently of target innervation in the Klf7 null brain may have compromised neuronal survival below the threshold of experimental detection.
In addition to defects in axon projections, we also identified abnormalities in dendritic organization in Klf7-null brains, notably in the adult hippocampus (Fig. (Fig.7)7) (27). Golgi-Cox staining in fact showed that the complexity of both apical and basal dendritic trees of CA1 pyramidal cells is reduced (Fig. 7c and d) (27). There were also abnormal arbors of cortical neurons, granule cells of the dentate gyrus, and the major fiber tracts (data not shown). In contrast to the cerebral cortex and hippocampus, no defects were noted in another location of high Klf7 expression in the brain, the cerebellum (data not shown). Together, these observations therefore documented the role of KLF7 in guiding neurite outgrowth at selected anatomical locations of the CNS.
We have recently shown that KLF7 stimulates p21cip/waf transcription by directly binding to the proximal promoter sequence (56). Others had previously reported accumulation of p21cip/waf in the differentiating neurons of the postnatal OE (34). Based on these lines of evidence, we compared the expression levels of p21cip/waf in mutant and wild-type OE from embryos and newborn mice. In situ hybridizations and immunoblot assays estimated that p21cip/waf gene activity in the mutant OE is significantly less than normal (Fig. 8a and b). Furthermore, immunoblots of protein extracts from wild-type and mutant OE of P1 mice revealed a similar decrease in p27kip1 protein levels and no detectable changes in p57kip2 (Fig. (Fig.8b).8b). This last result raised the possibility that p27kip1 may be another gene targeted by KLF7. To test this hypothesis, a KLF7 expression plasmid was cotransfected with a luciferase reporter gene construct harboring the proximal promoter of the p27kip1 gene (40). These in vitro assays showed that KLF7 can stimulate transcription from the p27kip1 promoter nearly 10-fold (Fig. (Fig.8c).8c). Additionally, we performed ChIP experiments in 293T cells transfected with the KLF7 expression plasmid and found that the transcription factor binds to the proximal promoter sequence of the endogenous p27kip1 gene (Fig. (Fig.8d).8d). Together with the in vivo findings, transfection and ChIP experiments established that two members of the Cip/Kip family of cdk inhibitors are directly regulated by KLF7.
Recent studies have demonstrated that the Cip/Kip proteins have additional functions outside the nucleus that are unrelated to the cell cycle, such as regulating neurite remodeling (14). We therefore examined the intracellular distribution in wild-type OSN cultures of the two Cip/Kip proteins whose expression is affected by loss of Klf7. Unlike immature neurons or nonneuronal cells, differentiating OSNs displayed significant amounts of p21cip/waf and p27kip1 in the cytosolic compartment, including axonal and dendritic extensions (Fig. (Fig.8e).8e). The quantity, but not the intracellular localization, of p21cip/waf was significantly reduced in Klf7−/− OSNs (Fig. (Fig.8f).8f). We interpreted these results to suggest that KLF7 may act on neuronal morphogenesis in part by reducing the relative levels of molecules that may promote neurite outgrowth.
The results of the present study imply a function for KLF7 in neuronal morphogenesis and thus in the establishment of connectivity at several distinct anatomical sites in the CNS and peripheral nervous system. A role for the transcription factor in neurogenesis was previously suggested by gene expression studies that documented high Klf7 activity in selected neuronal subtypes of the developing embryo and adult organism (32, 35). Here we have shown that absence of KLF7 affects neurite outgrowth. This observation is consistent with the prevalent expression of the transcription factor in postmitotic and not in proliferating neurons (32). The most evident anatomical consequence of loss of KLF7 activity was seen in the olfactory system, in the form of severely hypoplastic bulbs lacking peripheral innervation. In vivo evidence indicates that part of KLF7 action on OSN differentiation is to establish optimal levels of p21cip/waf and p27kip1 synthesis. We speculate from correlative data in the literature and our own in vitro analyses that KLF7 stimulation of p21cip/waf and p27kip1 transcription may ultimately impact on cytoskeletal dynamics that promote neurite outgrowth. Irrespective of the underlying mechanism, our genetic evidence adds the nervous system to the list of organs and tissues whose development is regulated by members of the KLF family.
The olfactory nerve defect in Klf7−/− mice is apparent as early as the stage in which OSNs first project towards the forebrain in an otherwise anatomically normal olfactory cavity (6). This last observation together with Klf7-restricted expression in the OE excludes the possibility that loss of signals emanating from the surrounding mesenchyme may contribute to the mutant OE phenotype. We have also presented evidence indicating that the mutation has no significant negative effects on OE cell survival or proliferation and consequently on the thickness and cell density of the mutant tissue. Although absence of afferent innervation has been also observed in the olfactory system of Pax6, Gli3, Emx2, and Dlx5 mutant mice, the phenotype is significantly more severe in these animals than in Klf7-null mice (25, 36, 59, 63). This and preliminary evidence showing normal expression of the aforementioned genes in the Klf7-null OE and vice versa suggest that KLF7 operates on a separate pathway from transcription factors Dlx5 and Gli3/Emx2 during olfactory neurogenesis. OB hypoplasia in Klf7−/− mice, on the other hand, is in line with the current model of bulb development, which postulates that onset of neurogenesis and bulb evagination are innervation-independent processes (25, 36, 59, 63). Work in progress is examining the olfactory system of those rare Klf7-null mice that survive into adulthood in order to assess whether a potential contribution of peripheral innervation to bulb maturation and morphogenesis may exist.
By analogy to the olfactory system, restricted Klf7 expression in the forming RGCs and the overall normal anatomy of the mutant eye argue for a cell-autonomous defect causing impaired axonal growth in the visual system. A number of transcription factors have been identified that modulate the formation of the optic nerve and optic tract, but only some appear to affect RGC morphogenesis in a cell-autonomous manner (17, 22). For example, Brn-3.2 displays the same expression pattern in the eye as Klf7 does, and mice lacking this transcription factor are characterized by aberrant RGC projections towards the optic disk (17). Erkman et al. (17) have shown that Brn-3.2 controls axon pathfinding in part by regulating the expression of the actin-binding protein mabLIM. Whether Brn-3.2 and/or mabLIM expression is affected in the Klf7−/− retina remains to be determined. The absence of expression of TAG-1 in the optic tract in the Klf7-null embryos is reminiscent of TAG-1 in the spinal cord, where the TAG-1 gene is downregulated in commissural axons after decussation (15). It is possible that guidance errors or disorganization of axons at the chiasm, as demonstrated with the increase in ipsilateral projections, results in this decreased TAG-1 expression. Although preferential expression of TAG-1 in the ventral optic nerve is a novel finding, two TAG-1-related cell adhesion molecules, PSA-NCAM and L1, have been recently reported to be more strongly expressed in the dorsal than the ventral portion of the optic tract (7).
Similar to the olfactory and optic systems, brain abnormalities in Klf7-null mice appear to involve distinct pathways compared to similar phenotypes of mice lacking other transcription factors, such as Coup-tf1 or Tbr-1 (23, 64). Coup-tf1-null mice display a lack of thalamocortical input to the cortex and aberrant differentiation and apoptosis of subplate neurons (64). Since subplate neurons are crucial for proper thalamocortical projections, the thalamocortical defect of Coup-tf1-null mice is likely to represent a secondary effect of the mutation (64). By contrast, the fact that the subplate in Klf7−/− brains is unaffected gives credence to the notion that the abnormal thalamocortical trajectory in these mutant mice is a primary rather than a secondary defect. Thalamocortical and subplate abnormalities, as well as defects in the corpus callosum and corticothalamic projections, characterize Tbr1−/− mice (23). However, these animals also display cortical cell migration defects leading to cortical inversion, which may conceivably contribute to subplate and projection abnormalities (23). We have tested Klf7−/− mice for potential cortical migration defects and found none. On the other hand, we were unable to explain the precise origin of the cellular deficit in the Klf7-null cortex. Finally, the finding of significant neurite deficits in the hippocampus raises the possibility that maintenance of KLF7 activity throughout adulthood may be crucial for neuronal wiring and implicitly for learning processes localized to this area of the brain (27).
Loss of KLF7 activity is predicted to affect the expression of multiple genes and intracellular pathways either directly or indirectly. Significant downregulation (but not loss) of p21cip/waf expression in the mutant OE is an important finding of our study, which demonstrates a functional relationship between this multifunctional protein and OSN differentiation. The finding is in line with the well-established contribution of p21cip/waf to terminal differentiation of several cell lineages through negative or positive mechanisms and in connection with or independent of cell cycle regulation (16). Legrier et al. (34) have interpreted gene expression data to indicate that individual cdk inhibitors fulfill distinct roles in OE neurogenesis, including maintenance of a quiescent phenotype (p21cip/waf) and control of cell cycle withdrawal (p27kip1). These last data, however, seem to contrast the apparent lack of OSN defects in p21waf/cip- and p27kip1-deficient mice, a problem which probably reflects the different focuses of these earlier analyses (13, 19).
We can only speculate from correlative evidence in the literature and our study about the mechanistic relationship between downregulation of p21cip/waf and p27kip1 and impaired neurite outgrowth in Klf7-null OSNs. A large body of work has established the multiple functions of the actin cytoskeleton in axon initiation, growth, guidance, and branching, as well as the role of the Rho family of small GTPases in regulating neuronal morphogenesis (21, 37). Evidence has been presented that signaling of many axon guidance receptors converges onto Rho, Rac, and/or Cdc42, which in turn activate downstream targets, like Rho-kinase (ROCK), which remodel the cytoskeleton (37). Recent studies of a variety of cell systems, including differentiating neurons, have shown that Cip/Kip proteins regulate actin dynamics through inhibition of the Rho-ROCK-LIMK pathway (4, 14, 57, 62). One of them, in particular, has documented the ability of cytoplasmic p21cip/waf to promote axonal regeneration and functional recovery in a rat model of spinal injury (58). In the present study, we have shown that p21cip/waf displays a cytoplasmic localization in differentiating OSNs; additionally, we have made an indirect connection between loss of KLF7 activity and impaired neurite growth by associating p21cip/waf downregulation in the mutant OE and cultured OSNs. Accordingly we propose that KLF7 may promote neuronal morphogenesis in part by increasing the overall (and cytosolic) levels of p21cip/waf and p27kip1. A similar mechanism may also operate in the accessory olfactory system of the Klf7-null mice in which there are fewer and poorly projecting mature sensory neurons and substantial p21cip/waf downregulation (our unpublished data). Furthermore, this functional relationship may extend to other regions of the neonatal and adult brain where Klf7 and Cip/Kip genes are coexpressed, such as the OBs, cerebral cortex, and hippocampus (32, 47, 52, 60). Irrespective of the underlying mechanism in OE neurogenesis and its relevance to other regions of the nervous system, our findings are the first to demonstrate a role for p21cip/waf and p27kip1 in the differentiation of olfactory neurons.
As we have previously shown for p21cip/waf (56), cotransfection and ChIP experiments have established that the p27kip1 gene is a transcriptional target of KLF7. Several KLFs have been shown to stimulate transcription from the p21cip/waf promoter, thus raising the possibility of functional redundancy in cell types where they are coexpressed (5, 10, 26). KLF6 and KLF7 are a case in point. They constitute a structurally distinct group within the KLF family, which shares a common progenitor in invertebrates (11); they have identical transactivation domains, which stimulate the p21cip/waf promoter in cotransfection assays (32, 44, 56); and the corresponding genes are coexpressed in a few locations of the nervous system, including cortical neurons, DRG, and neural tube (32, 33). Although we noted Klf6 upregulation in Klf7-null neurons, early embryonic lethality of Klf6-null mice has hampered the analysis of the genetic interaction between these two loci (our unpublished data). Finally, it is interesting that KLF7 function in neurogenesis is consistent with a recent report of a chromosomal deletion that includes the KLF7 gene in a human patient with neurodevelopmental abnormalities (49). That Klf7 loss impairs neurite outgrowth without disturbing overall tissue architecture and morphology of the nervous system makes the Klf7−/− mice an informative model in which to dissect the histological complexity of mammalian axogenesis.
We thank F. Margolis, P. Mombaerts, G. Sonenshein, T. Curran, and B. Vogelstein for providing critical reagents; T. Sakurai for invaluable comments on the manuscript; S. Lee-Arteaga and C. Else for excellent technical assistance; and K. Johnson for preparing the manuscript.
This work was supported by NIH grants NS33199 (L. F. Parada) and AR38648, the New York State Spinal Cord Injury Research Program, the St. Giles Foundation, and the James D. Farley family.