Fibroblast growth factor‐2 (FGF‐2) has been implicated in various signaling processes which control embryonic growth and differentiation, adult physiology and pathology. To analyze the in vivo functions of this signaling molecule, the FGF‐2 gene was inactivated by homologous recombination in mouse embryonic stem cells. FGF‐2‐deficient mice are viable, but display cerebral cortex defects at birth. Bromodeoxyuridine pulse labeling of embryos showed that proliferation of neuronal progenitors is normal, whereas a fraction of them fail to colonize their target layers in the cerebral cortex. A corresponding reduction in parvalbumin‐positive neurons is observed in adult cortical layers. Neuronal defects are not limited to the cerebral cortex, as ectopic parvalbumin‐positive neurons are present in the hippocampal commissure and neuronal deficiencies are observed in the cervical spinal cord. Physiological studies showed that FGF‐2‐deficient adult mice are hypotensive. They respond normally to angiotensin II‐induced hypertension, whereas neural regulation of blood pressure by the baroreceptor reflex is impaired. The present genetic study establishes that FGF‐2 participates in controlling fates, migration and differentiation of neuronal cells, whereas it is not essential for their proliferation. The observed autonomic dysfunction in FGF‐2‐deficient adult mice uncovers more general roles in neural development and function.
Members of the fibroblast growth factor (FGF) family and their high affinity tyrosine kinase FGF receptors (FGFR) have been implicated in embryonic growth and patterning and, in particular, during central nervous system (CNS) development (reviewed by Eckenstein, 1994). Studies using Xenopus laevis embryos revealed roles for FGF signaling during establishment of initial posterior neural tube identities (reviewed by Sasai and De Robertis, 1997). Similarly, FGFR1‐deficient mouse embryos display neural tube defects, which most likely arise from incorrect specification of neuroectodermal cell fates (Yamaguchi et al., 1994). During initial patterning of the brain vesicle, FGF‐8 is expressed by the isthmus, the organizer which separates mid‐ from hindbrain (Crossley et al., 1996). Ectopic rostral FGF‐8 application or expression results in transformation of fore‐ into midbrain structures, indicating that FGF‐8 determines midbrain identity (Crossley et al., 1996; Lee et al., 1997). The spatial distribution of FGF‐3 in the developing hindbrain is consistent with roles in establishment of rhombomere boundaries (Wilkinson et al., 1988), but no changes in rhombomere identities are observed in FGF‐3‐deficient embryos (Mansour et al., 1993).
Roles for FGF‐2 during CNS morphogenesis have been postulated as high levels are detected from neurulation onwards (Dono and Zeller, 1994; Riese et al., 1995; reviewed by Eckenstein, 1994). Furthermore, overexpression of specific FGF‐2 isoforms in the neural tube of transgenic mouse embryos provided evidence for growth‐regulatory functions during spinal cord morphogenesis (Zuniga Mejia Borja et al., 1996). Such a proposed mitogenic role in neurogenesis was also supported by studies using cultured neural progenitors. These studies indicated that FGF‐2 levels regulate proliferation of specific neural cell types (reviewed by Vescovi et al., 1993; Temple and Qian, 1995). In contrast, other studies provided evidence that FGF‐2 participates in promoting differentiation and long‐term survival of neurons in culture. These in vitro studies suggested that FGF‐2 functions in a concentration‐dependent manner to regulate mainly proliferation, but also differentiation and survival of neural progenitors (Petroski et al., 1991; Ray et al., 1993; Ghosh and Greenberg, 1995; Vicario‐Abejon et al., 1995; Qian et al., 1997).
In addition to the CNS, FGF‐2 is expressed during other developmental processes such as limb, kidney and cardiovascular morphogenesis (e.g. Dono and Zeller, 1994). Of particular interest for the present study is expression by vascular smooth muscle cells, endothelial cells and post‐mitotic cardiomyocytes (Kardami and Fandrich, 1989; Lindner and Reidy, 1993; Cuevas et al., 1996). Both acute (Cuevas et al., 1991) and chronic (Lazarous et al., 1995) intravenous infusion of recombinant FGF‐2 in normotensive animals induces a significant reduction in blood pressure (hypotension), which is mediated partly by release of nitric oxide and activation of ATP‐sensitive K+ channels (Cuevas et al., 1991). In the spontaneously hypertensive rat, endogenous levels of FGF‐2 in endothelial cells are chronically reduced and blood pressure can be lowered by chronic infusion of FGF‐2 (Cuevas et al., 1996). These studies suggested that FGF‐2 functions in an antihypertensive vasodilator cascade. In contrast, other studies indicated that FGF‐2 acts as a potent mitogen for smooth muscle cells in vitro and regulates blood vessel growth in vivo (Lazarous et al., 1995; Davis et al., 1997). Furthermore, FGF‐2 is released from cardiomyocytes upon increased mechanical load (Clarke et al., 1995) and induces adaptive myocardial growth (Kaye et al., 1996). These studies suggested that FGF‐2 signaling functions in the cardiovascular adaptations leading to a chronic blood pressure increase (hypertension).
To study the essential in vivo functions of FGF‐2 during embryogenesis and in adult physiology, the FGF‐2 gene was inactivated in mice using homologous recombination in embryonic stem (ES) cells. FGF‐2‐deficient mice show neuronal defects in the cerebral cortex, hippocampal commissure and spinal cord. In particular, a fraction of neuronal precursors fail to colonize their normal target layers II and III during cerebral cortex development. Furthermore, adult FGF‐2‐deficient mice have a reduced blood pressure. Continued angiotensin II infusion induced hypertension of a similar magnitude as in wild‐type litter mates, demonstrating an intact responsiveness of their cardiovascular system. In contrast, infusion of isoproterenol into FGF‐2‐deficient mice uncovered an impairment of their baroreceptor reflex. Therefore, blood pressure regulation is affected by an autonomic dysfunction.
Mice homozygous for a loss‐of‐function FGF‐2 allele are viable and fertile
The first coding exon (e1, Figure 1A) of the murine FGF‐2 gene encodes three protein isoforms (Figure 1D), which are translated by alternative initiation. Therefore, the complete first FGF‐2 coding exon and parts of the 5′‐ and 3′‐flanking intronic sequences were replaced by the neomycin resistance gene (neo, Figure 1A) in the opposite transcriptional orientation using homologous recombination in ES cells. Chimeric mice were generated from correctly targeted ES cells and crossed to C57BL/6J mice for germline transmission. Subsequently, homozygous FGF‐2−/− mice were obtained in a mixed C57BL/6J×129/Sv genetic background. Genotyping of their offspring at birth and weaning age (Figure 1B and C) showed that FGF‐2−/− mice are born alive, mature normally to adulthood and are fertile (data not shown). As FGF‐2 is expressed from gastrulation onwards (Riese et al., 1995), it was essential to determine that a complete loss‐of‐function FGF‐2 allele had been generated. Normalized amounts of proteins from brains of newborn mice (Figure 1D) and embryos (data not shown) of all three genotypes were analyzed using specific FGF‐2 antibodies (Dono and Zeller, 1994). Western blot analysis showed that levels of all three FGF‐2 isoforms are reduced to about half in heterozygous mice (Figure 1D, compare lane 10 with 9), whereas no FGF‐2 proteins are detected in homozygous litter mates (Figure 1D, lanes 11 and 12, see also Figure 2B). Further analysis confirmed that no truncated proteins are expressed and that there is no embryonic rescue by maternal FGF‐2 protein (data not shown). In summary, expression of all three FGF‐2 isoforms is abolished and a complete loss‐of‐function allele has been generated.
Development of the cerebral cortex is impaired in FGF‐2‐deficient mice
FGF‐2 is expressed in the CNS from its earliest developmental stages onwards. In the forebrain, FGF‐2 proteins are first detected during embryonic day 9.5 (Nurcombe et al., 1993). Figure 2 shows the distribution of FGF‐2 proteins (Figure 2A–D) and transcripts of their high affinity FGF receptor isoforms [FGFR1‐IIIc (Figure 2E) and FGFR2‐IIIc (Figure 2F); reviewed by Miller and Rizzino (1994)] during embryonic days 14.5 and 16, respectively. At this developmental stage, the highest levels of FGF‐2 proteins are present in the cerebral hemispheres, whereas levels are lower in more caudal brain regions and in the spinal cord (Figure 2A). As expected from molecular analysis (Figure 1D), no FGF‐2 proteins are detected in FGF‐2‐deficient embryos (Figure 2B). In cerebral hemispheres, FGF‐2 protein levels are highest in the ventricular and sub‐ventricular zones of the lateral ventricles (Figure 2C and D), the region where neural progenitors are born and determined (Bayer and Altman, 1991). In contrast, cells within the cortical plate express much lower amounts of FGF‐2 (Figure 2D). Similarly, transcripts encoding FGFR1‐IIIc (Figure 2E) and FGFR2‐IIIc (Figure 2F) isoforms are co‐expressed in both ventricular and sub‐ventricular zones. FGFR2‐IIIc isoform transcripts persist in the cortical plate, whereas FGFR1‐IIIc expression is much lower (compare Figure 2F with E). No changes in FGFR transcripts were observed in brains of FGF‐2‐deficient embryos (data not shown).
To determine whether FGF‐2 is necessary for embryonic CNS morphogenesis, brains of wild‐type and FGF‐2‐deficient newborn litter mates were analyzed by comparative histology (Figure 3). Serial coronal sections of FGF‐2‐deficient mice showed no gross morphological abnormalities (compare Figure 3A with B and C). In newborn mice, differentiation of the cerebral cortex progresses from deep to superficial layers (VI–I, see Figure 3D; Bayer and Altman, 1991) and large pyramidal neurons are apparent (Figure 3D and G). No differences were observed between wild‐type (+/+) and heterozygous (+/−) mice (data not shown). However, FGF‐2‐deficient litter mates display defects in organization and differentiation of their cerebral cortex. In particular, the thickness of the cerebral cortex in homozygous (−/−) mice is reduced by ∼10% in comparison with their control litter mates at birth (+/+; +/−, Table IA and B; P <0.05). At the cellular level, all differentiating cortical layers appear compressed and less distinct (compare Figure 3D with E and F). Differentiated, large pyramidal neurons are apparent in newborn wild‐type mice (Figure 3D and G), whereas only a few are observed in FGF‐2‐deficient litter mates (Figure 3E and F and H and I). This phenotype reveals a requirement for FGF‐2 for embryonic cerebral cortex morphogenesis.
Neuronal cell migration is affected during cerebral cortex development in FGF‐2‐deficient mice
Early patterning and development of the embryonic brain appear normal in FGF‐2‐deficient embryos, as neither spatial nor temporal changes were detected in molecular markers during embryonic days 11–16 (BF1, Emx1, Otx1, Otx2, mtll; data not shown). After compartmentalization of the brain vesicle, the neurons of the cerebral cortex are generated by the pseudo‐stratified ventricular epithelium that lines the lateral ventricles in the telencephalon. Cortical neurogenesis is marked by the appearance of the first post‐mitotic neurons during embryonic day 11 and continues until embryonic day 17. Determined neurons leave the ventricular zone through the intermediate zone and migrate outwards along radial glial tracts. In the cortical plate, they colonize their respective target layers that form successively from deep to superficial (layer VI–II) as cortical development proceeds. After completing migration, cortical neurons undergo terminal differentiation which continues after birth (reviewed by Bayer and Altman, 1991; Hatten, 1993).
To determine whether proliferation of cortical progenitors is affected in FGF‐2‐deficient embryos, nuclei of proliferating cells were pulse‐labeled for 1 h with 5‐bromodeoxyuridine (BrdU). Quantitative analysis (Table IC) shows that proliferation of progenitor cells within the ventricular zone of FGF‐2‐deficient embryos (Figure 4B) is indistinguishable from that of wild‐type litter mates (Figure 4A) during embryonic days 14.5–15. Analysis of brains from embryonic days 13 (n = 2; data not shown) and 18 (n = 4; data not shown) confirmed that FGF‐2 is not essential for cell proliferation during cerebral cortex development.
Possible defects in cortical layer morphogenesis were studied using an in vivo cell labeling system, which allows one to examine the migratory history of individual cells (reviewed by Bayer and Altman, 1991). This system takes advantage of the fact that the prospective target layer of a neuron normally is linked directly to its time of birth in the embryonic ventricular zone. Therefore, pregnant females were injected with BrdU at precise developmental time points to label pre‐migratory neuronal cells. The fate of BrdU‐positive neurons was analyzed ∼1 month after birth on serial coronal sections of the rostral cerebral cortex (for details, see Material and methods). Pulse labeling at embryonic day 14.5–15 marks predominantly neuronal cells that colonize cortical layers III and II (Figure 4C), whereas cortical layers VI–IV are mostly devoid of labeled cells in heterozygous adult mice (Figure 4C and D). In contrast, about twice as many labeled neuronal cells are located in the deeper cortical layers and corpus callosum of FGF‐2‐deficient mice (Table ID and E; P <0.05; Figure 4E and F), indicative of their failure to reach target layers III and II. This defect is variable, as many more BrdU‐positive cells (25–36%) are detected in corpus callosum and deep layers of the more severely affected FGF‐2‐deficient mice (n/total = 4/12; Figure 4E and F), whereas a few others are indistinguishable from controls (for details, see scatter plot, Table ID). This could be either a consequence of incomplete penetrance of the embryonic FGF‐2 phenotype (see Discussion) or variable elimination of mislocalized cells by cellular apoptosis (see, for example, review by D'Mello, 1998). Concerning the latter, quantitative TUNEL analysis showed no significant differences in cell survival in either the embryonic (E16; n = 5), newborn (n = 4) or adult (n = 4) cerebral cortex (data not shown).
These defects should alter the distribution of specific neuronal populations. In particular, the distribution of the calcium‐binding protein parvalbumin, which marks a significant number of neurons in several cortical layers (Celio, 1990), was analyzed in adult brains (Figure 4G–I). An average reduction of ∼25% in parvalbumin‐positive neurons is seen in FGF‐2‐deficient mice in comparison with controls (compare Figure 4H with I; see Table IF). This reduction affects all layers (Figure 4I and data not shown) and is comparable with the proportion of neuronal cells which fail to reach their target layers as estimated by BrdU pulse labeling (Table IE). In contrast, no differences in the distribution of glial fibrillary acidic protein (GFAP), which marks astroglial cells, were detected (data not shown). Furthermore, no striking behavioral abnormalities such as seizures (Chae et al., 1997) or reeler‐like phenotypes (e.g. D'Arcangelo et al., 1995) were observed in FGF‐2‐deficient mice, although further tests are necessary to detect possible subtle defects.
Neuronal defects are not limited to the cerebral cortex
The hippocampal commissure, a brain region located beneath the cerebral cortex (Figure 5A and B), is devoid of parvalbumin‐positive neurons in wild‐type and heterozygous adult mice (Figure 5C; n = 8). However, FGF‐2‐deficient adult mice (n/total = 8/14) show varying degrees of ectopic parvalbumin‐positive neurons in this region (Figure 5D). In contrast, the distributions of calbindin‐positive neurons (Celio, 1990) and GFAP‐positive glial cells were not altered (data not shown). Analysis of serial coronal sections of the adult spinal cord also revealed neuronal abnormalities throughout the mantle layer of the cervical region (Figure 5E–H; n/total = 5/12). Fewer differentiated large neurons (compare Figure 5H with G; arrowheads in Figure 5H) were observed, whereas all other neural cell types appear normal (arrows in Figure 5H). These studies indicate that FGF‐2 also functions in subcortical and spinal cord neurons in agreement with its expression by the hippocampal formation (Figure 2C) and embryonic spinal cord (Figure 2A, see also Dono and Zeller, 1994).
Reduced blood pressure and impaired baroreflex function in FGF‐2‐deficient mice
Exogenous administration of FGF‐2 induces hypotensive effects in rats, rabbits and dogs (Cuevas et al., 1991; Lazarous et al., 1995), suggesting possible antihypertensive actions of endogenous FGF‐2 proteins (Cuevas et al., 1996). Therefore, the resting blood pressure of adult FGF‐2‐deficient mice was determined and compared with that of wild‐type (+/+) and heterozygous (+/−) litter mates to uncover possible FGF‐2 functions in blood pressure control. Initial statistical evaluation of resting blood pressure recordings showed that no differences are observed between wild‐type and heterozygous mice (data not shown). Therefore, these two genotypes (+/+; +/−) served as phenotypic controls. However, studies by Krege et al. (1995) revealed a gender difference in blood pressure levels using wild‐type mice of a similar genetic background to the present studies (mixed C57BL/6J×129/Sv). In agreement with these studies, male control mice display on average an ∼14 mmHg higher blood pressure than females (compare Figure 6B with A).
Analysis of the primary recordings of female FGF‐2‐deficient mice (Figure 6A) reveals a significant decrease in resting blood pressure in comparison with wild‐type and heterozygous female litter mates (Figure 6B). Similarly, many FGF‐2‐deficient males (Figure 6B) also display lower resting blood pressures than their wild‐type litter mates, but a substantial fraction (n/total = 4/12) have elevated blood pressures (see scatter plot, Figure 6B). This variation could be caused by incomplete penetrance of the phenotype or unknown modifiers of blood pressure in males. To assess the effects of the FGF‐2 deficiency on blood pressure independent of gender, the blood pressure differences (Δ blood pressure, Figure 6C and D) were calculated by subtracting the mean value of the respective control group (+/+; +/−) from the individual recordings. Analysis of the resulting histograms (including all mice from both sexes, Figure 6C and D) shows that the blood pressure differences of control mice are distributed normally around zero (Figure 6C). In contrast, the distribution is shifted by about −10 mmHg in FGF‐2‐deficient mice (Figure 6D). On average, the blood pressure of FGF‐2‐deficient adult mice is reduced by ∼10 mmHg in comparison with control litter mates (P <0.02).
This hypotensive phenotype might be the result of an insufficiency affecting smooth muscle and/or myocardial function (see Introduction). If this were the case, then the hypotensive phenotype should be aggravated further by chronic elevation of blood pressure. Therefore, hypertension was induced in FGF‐2‐deficient and control mice by continued infusion (6 days) of angiotensin II, which causes vasoconstriction and induces myocardial growth (Kim et al., 1995). As shown in Figure 7A, FGF‐2‐deficient mice respond to chronic angiotensin II infusion with an even more pronounced increase in blood pressure than controls (control, 19 ± 14 mmHg; −/−, 30 ± 13 mmHg; P = 0.05) to reach identical hypertensive levels. No differences in baseline plasma renin values, heart rates and body weights were observed between control and FGF‐2‐deficient mice (data not shown). These results demonstrate that FGF‐2‐deficient mice retain the ability to induce and maintain elevations of blood pressure in response to an exogenous vasoconstrictor such as angiotensin II.
Reduced blood pressure in combination with intact responsiveness to exogenously induced vasoconstriction is a common manifestation in autonomic dysfunction, which is characterized by an impaired baroreceptor reflex response to hypotensive stimuli (Bannister, 1988). To assess neuronal control of blood pressure in FGF‐2‐deficient mice, acute peripheral vasodilatation was induced by intravenous infusion of isoproterenol, a β1/β2‐adrenoreceptor agonist. The response to different doses of isoproterenol was determined in a pilot experiment using six male control mice (see Materials and methods). As expected, heart rates increased in a dose‐dependent fashion (data not shown), whereas blood pressure remained at its baseline value (Figure 7B) during administration of isoproterenol up to 0.05 ng/g body weight/min, indicating intact baroreceptor reflex regulation. Subsequently, the effects of this dose were studied in FGF‐2‐deficient male mice with a resting blood pressure <135 mmHg (Figure 7C and D). In contrast to the controls (Figure 7B), the blood pressure declined with a characteristic time course in FGF‐2‐deficient litter mates (Figure 7C). A significant drop in blood pressure was observed in all homozygous mutants but not in control mice (Figure 7D). Furthermore, while heart rates increased as expected in control male mice (Δ: +42 ± 35 beats/min), they failed to increase in FGF‐2‐deficient litter mates (Δ: −13 ± 34 beats/min; P <0.02). Taken together, these studies indicate that neural reflex control of blood pressure is impaired in FGF‐2‐deficient mice.
FGF‐2 is required for normal CNS development
FGF‐2‐deficient mice display neurological defects consistent with essential functions during development of the CNS. In particular, lack of FGF‐2 signaling affects migration of a fraction of neuronal precursors during cerebral cortex development (see below). Lack of FGF‐2 also induces ectopic parvalbumin‐positive neurons in the hippocampal commissure and results in neuronal deficiencies affecting the cervical spinal cord. Previous studies suggested major roles for FGF‐2 in glial cell proliferation and survival (see, for example, Petroski et al., 1991; Qian et al., 1997). The present genetic analysis failed to reveal defects in astroglial cells, suggesting that the cellular defects primarily affect neuronal cell types. Furthermore, an autonomic dysfunction impairing blood pressure regulation points to general roles in neural development and function. Previous studies using cultured CNS progenitors have suggested a major role for FGF‐2 signaling in cell proliferation (Temple and Qian, 1995). However, the present study establishes that neuronal proliferation is normal during cerebral cortex morphogenesis in FGF‐2‐deficient embryos. Therefore, FGF‐2 either does not function as a mitogenic signal for neural progenitors or its deficiency is functionally compensated by other FGFs in vivo. Indeed, several other FGFs are expressed by the ventricular zone of the developing cerebral cortex, including FGF‐1, the closest relative of FGF‐2 (Wilcox and Unnerstall, 1991). It is possible that FGF‐1 compensates and/or acts as a mitogenic signal for neural progenitors. Functional compensation may also account for the variable and/or rather restricted phenotypes observed in FGF‐2‐deficient mice. Indeed, it has been proposed that such compensation could account for both incomplete penetrance and restricted phenotypes observed in mouse embryos deficient for either FGF receptors (see, for example, FGFR1: Yamaguchi et al., 1994) or specific ligands (see, for example, FGF‐3: Mansour et al., 1993; FGF‐5: Hebert et al., 1994). However, the identification of shared functions has to await further genetic study and analysis of compound mutant embryos. Alternatively, phenotypic variation could be caused by non‐uniform segregation of unknown modifiers in a mixed genetic background.
Genetic evidence that FGF‐2 participates in regulating fate, migration and differentiation of neuronal progenitors
The neuronal precursors of the cerebral cortex are generated in the ventricular zone, and glial fibers provide a generic guide for radial migration to their target layers (reviewed by Hatten, 1993). One of the earliest determinative events of cortical layer formation is the colonization of layer I by Cajal–Retzius (CR) neurons, which promote differentiation of radial glial cells prior to the onset of neuronal migration (Soriano et al., 1997). CR neurons produce reelin, an extracellular matrix protein, whose disruption by the murine reeler mutation causes inversion of all cortical layers (D'Arcangelo et al., 1995). Several additional components of a genetic cascade controlling establishment of the cortical plate and migration of neuronal precursors have been identified. Inactivation of mouse disabled‐1, a component of the Src signal transduction cascade, Cdk5 and p35, which are all expressed by neuronal cells, also resulted in reeler‐like phenotypes (Ohshima et al., 1996; Chae et al., 1997; Howell et al., 1997). Therefore, these genes could be part of the same or parallel signaling pathways, which regulate cell migration and organization of the cerebral cortex.
These mutations, however, do not seem to alter neuronal cell fates, which are specified during the final cell divisions in the ventricular zone (reviewed by Bayer and Altman, 1991; see also Howell et al., 1997). The migratory path of a neuronal progenitor is defined by cell–cell interactions and exposure to unknown signals during its cell fate determination within the ventricular zone (reviewed by Hatten, 1993). The present genetic study indicates that FGF‐2 is one of these signals as its expression is highest in the ventricular zone and a fraction of neuronal progenitors fail to reach their target layers in FGF‐2‐deficient mice. Therefore, FGF‐2 most likely participates in defining both the cell fate (see also Ghosh and Greenberg, 1995; Vicario‐Abejon et al., 1995) and migratory path of neuronal progenitors. As Src family kinases can transduce FGF‐2 signals (Zhan et al., 1994; Yayon et al., 1997), it is possible that FGF‐2 is part of a common signaling cascade involving both rather general (reeler‐like, see above) and specific cues (such as FGF‐2) to regulate neuronal cell migration directly during cortical layer formation.
Our genetic study also reveals that ectopic parvalbumin‐positive neurons are present in hippocampal commissure of FGF‐2‐deficient mice. Therefore, FGF‐2 normally should function to suppress differentiation of parvalbuminpositive neurons in the hippocampal commissure. Indeed, previous studies indicated that FGF‐2 levels might control the fate and differentiation of cultured cortical stem cells (Ghosh and Greenberg, 1995; Qian et al., 1997). Low levels of FGF‐2 promote neuronal differentiation, possibly in combination with other factors such as neurotrophins (Temple and Qian, 1995), while high levels induce glial cell types (Qian et al., 1997). Therefore, our genetic analysis shows that FGF‐2 functions in the control of neuronal cell fates and/or differentiation, whereas it is not essential for proliferation.
An evolutionarily conserved function of FGF signaling in regulating embryonic cell migration
Genetic analysis in the mouse demonstrated a role for FGFR1 signaling in the morphogenetic movements of mesodermal cells through the primitive streak (Yamaguchi et al., 1994). Furthermore, expression of a dominant‐negative FGFR1 transgene in oligodendrocyte precursors disrupted their migration in newborn brains (Osterhout et al., 1997). Our analysis of FGF‐2‐deficient embryos demonstrates that FGF signaling controls aspects of cell migration during patterning of the vertebrate neocortex. In Drosophila, FGF signaling also controls morphogenetic movements such as those occurring during gastrulation, neurogenesis and tracheal development (Klämbt et al., 1992; Gisselbrecht et al., 1996). Interestingly, the FGF ligand branchless (Sutherland et al., 1996) seems to act as a chemoattractant to guide migration of tracheal cells expressing the cognate FGF receptor breathless (Klämbt et al., 1992). In contrast, during vertebrate neocortex layer formation, neurons are exposed to FGF‐2 signals prior to or during onset of migration to their target layers as no protein is detected in the cortical plate. Therefore, it is unlikely that FGF‐2 functions as a chemoattractant (see also Armstrong et al., 1990). These distinct modes of action probably reflect mechanistic differences in the underlying morphogenetic movements. Indeed, neurons moving along radial glial fibers use different cell adhesion and cytoplasmic streaming mechanisms than, for example, migrating neural crest cells (discussed by Fishell and Hatten, 1991). Therefore, FGF‐2 most likely regulates migration indirectly by modulating the adhesive properties of neuronal cells as evidenced by previous studies of neuroepithelial cells in culture (Kinoshita et al., 1993). In summary, genetic analysis of FGF ligands and receptors in different species points to general roles for FGF signaling in cell migration during embryogenesis.
An autonomic dysfunction affects blood pressure regulation in FGF‐2‐deficient mice
FGF‐2‐deficient adult mice display a lower blood pressure than their wild‐type or heterozygous litter mates. Recently, Zhou et al. (1998) also reported a similar hypotensive phenotype independently derived in FGF‐2‐deficient mice. These genetic observations are rather surprising as previous studies demonstrated that exogenous FGF‐2 induces hypotension in normotensive and hypertensive animals (Cuevas et al., 1991; Lazarous et al., 1995). Furthermore, reduction of endogenous FGF‐2 levels in endothelial cells was correlated to elevated blood pressure in spontaneously hypertensive rats, a genetic model for studying hypertension (Cuevas et al., 1996). Accordingly, one might expect increased blood pressure levels rather than a hypotensive phenotype in FGF‐2‐deficient mice. A possible explanation for this apparent discrepancy may be that signaling by endogenous FGF‐2 does not affect blood pressure directly under normal physiological conditions. In the cardiovascular system, the major stimulus for FGF‐2 release appears to be transiently increased cell membrane permeability (Clarke et al., 1995; Kaye et al., 1996). Indeed, recent observations indicate that FGF‐2 release from vascular smooth muscle cells requires mechanical strains well above the physiological range of normal arteries (Cheng et al., 1997). Thus, only under pathophysiological conditions associated with elevated mechanical strain, such as hypertension, may the endogenous release of FGF‐2 be sufficiently high to lower blood pressure. Such an interpretation would also explain the exaggerated increase in blood pressure in FGF‐2‐deficient mice during chronic angiotensin II infusion.
In fact, several of our observations indicate that the reduced blood pressure in FGF‐2‐deficient mice is a consequence of an autonomic dysfunction, which characteristically is associated with hypotension (Bannister, 1988). First, FGF‐2‐deficient mice show an exaggerated increase in blood pressure in response to an exogenous vasoconstrictor (see above). This observation argues strongly against a chronic smooth muscle and/or myocardial insufficiency being the primary cause of the reduced blood pressure. Second, an acute vasodilatory stimulus (isoproterenol), which does not trigger a hypotensive response in control mice, induces a drop in blood pressure in FGF‐2‐deficient mice. This response, together with an impaired increase in heart rates, indicates that vasodilatation is not compensated by increased sympathetic nerve activity, a phenomenon observed in most human patients with autonomic failure (Mathias, 1988). Finally, FGF‐2 is expressed by the intermediolateral autonomic nuclei (Stapf et al., 1997) in the spinal cord. Even though preliminary analysis failed to reveal striking reductions in the numbers of pre‐ganglionic sympathetic neurons, neuronal lesions were observed in various other CNS regions in FGF‐2‐deficient mice. Such neuronal deficiencies have been reported in human patients with autonomic failure and are clinically defined as ‘multiple system atrophy’ (Bannister, 1988).
Further studies will be necessary to define at which cellular and/or molecular level the neural control of blood pressure is impaired in FGF‐2‐deficient mice. Despite the fact that most clinical cases of autonomic failure in humans are secondary forms with a different pathogenesis (Bannister, 1988), FGF‐2‐deficient mice appear to be the first defined genetic model of autonomic dysfunction. These mice will enable embryological, cellular and physiological analyses of normal and pathological development of the autonomic innervation and neural regulation of blood pressure.
Materials and methods
Targeting vector and generation of FGF‐2‐deficient mice
Genomic clones were isolated by screening a 129/Sv λ phage library (provided by R.Klein) with a cDNA probe encoding the ATG‐initiated FGF‐2 isoform (Hebert et al., 1990). A 15 kb genomic clone spanning the first FGF‐2 coding exon was used to generate the targeting construct in a pPNT‐based vector (Figure 1A, Tybulewicz et al., 1991; provided by R.Klein). An 8 kb fragment of intronic sequences upstream of the first coding exon (long arm) was inserted into the SalI restriction site located between the PGK‐neo and PGK‐tk marker genes. Subsequently, a 1.5 kb fragment of intronic sequences located downstream of the exon (short arm) was inserted into an XhoI–NotI restriction site vector. Therefore, the resulting targeting construct induces replacement of a 2.3 kb fragment containing the first FGF‐2 coding exon by the PGK‐neo gene in the opposite transcriptional orientation (Figure 1A). Transfection and selection of correctly targeted R1‐ES cells (Nagy et al., 1993) were performed as previously described (Wurst and Joyner, 1993). PCR‐positive ES cell clones were obtained and accurate targeting was confirmed by Southern blotting. Five to 10 μg of genomic DNA was digested with PstI and probed with a genomic fragment located immediately downstream of the 3′ end of the short arm (Figure 1C), a neo probe and a 5′ probe. Three correctly targeted clones were injected into C57BL/6J blastocysts as previously described (Wurst and Joyner, 1993) and chimeras from one clone resulted in germline transmission. Mice were genotyped either by PCR using specific oligos or by Southern blotting. Mice of mixed C57BL/6J×129/Sv backgrounds were generated as follows: chimeras (composed of ES cells of 129/Sv background) were crossed to C57BL/6J females and their heterozygous (+/−) F1 progeny crossed to generate all three genotypes. These mice and subsequent offspring were used for all studies. Litter mates of the same genotype always contributed randomly to the data points obtained by analyzing several litters. The statistical significance of all phenotypic differences between control and FGF‐2‐deficient mice was validated using the unpaired Student's t‐test.
Immunoblot analysis of FGF‐2 proteins
Immunoblot analysis was performed as previously described (Dono and Zeller, 1994) with the exception that ECL detection was used (Amersham). FGF‐2 proteins were detected using affinity‐purified polyclonal FGF‐2 antibodies (Dono and Zeller, 1994).
Whole mount antibody and section RNA in situ analysis
Embryos were collected and genotyped using DNA prepared from extra‐embryonic membranes. Brains were dissected in ice‐cold phosphate‐buffered saline (PBS), cut along the medial plane into left and right hemispheres and processed for whole‐mount antibody staining as previously described (Riese et al., 1995). Alternatively, heads were isolated and paraffin embedded for subsequent RNA in situ hybridization analysis on sections. Coronal sections (6 μm) were processed and hybridized using digoxigenin‐labeled riboprobes as described (Neubuser et al., 1997). Riboprobes for FGFR1 (450 bp PCR fragment) and FGFR2 (360 bp) are complementary to the isoform‐specific sequences encoding the immunoglobulin domain IIIc (plasmids provided by G.Holländer; unpublished).
Newborn mice were decapitated and their heads fixed for ∼12 h in 4% phosphate‐buffered paraformaldehyde (4% PFA) at 4°C. Adult mice were perfused with 4% PFA, and dissected heads and spinal cords were fixed overnight at 4°C. Subsequently, brains and spinal cords were dissected from skulls and bones, fixed for an additional 12–24 h and then either embedded in paraffin wax (newborns) or processed for cryo‐sectioning (adults). For cryo‐sectioning, the fixative was replaced with 30% sucrose in PBS and tissues were cryo‐protected in tissue freezing medium (Tissue‐Tek) prior to sectioning. Serial (7 μm, paraffin wax) or coronal (25 μm, frozen) sections were mounted on gelatin‐coated slides and dried at 42°C or at room temperature. Sections were stained subsequently with 0.1% cresyl violet (Nissl‐stain).
Measurement of cerebral cortex thickness at birth
The thickness of the cerebral cortex was determined using Nissl‐stained coronal brain sections (7 μm, paraffin). Measurements of the frontal cortex were done using sections at identical antero‐posterior positions. Duplicate measurements of both hemispheres were made ∼80 μm apart. Variations among the measurements from one brain were <5% in all cases.
BrdU labeling of embryos for cell proliferation studies
At defined developmental time points, pregnant females were injected intraperitoneally with BrdU solution (7.5 μg/g body weight, Sigma) and sacrificed 1 h later. BrdU‐labeled nuclei were detected as described below. Proliferating cells were quantitated using age‐matched wild‐type and homozygous litter mate embryos. Four equally spaced coronal sections were analyzed per brain. BrdU‐positive cells in the dorso‐lateral aspect of the ventricular zone in both brain hemispheres were counted using an ocular micrometer to permit quantitation of equal squares.
Cell migration studies
In a series of pilot experiments, embryos were BrdU pulse‐labeled between embryonic days 12 and 16 to study cell migration during cortical layer formation. These initial studies established that progenitors labeled between embryonic days 14.5 and 15 colonize predominantly layers III and II, while deeper cortical layers VI–IV are mostly devoid of labeled cells. Accordingly, cells which fail to migrate to their target layers and remain in deeper layers can be identified readily. In contrast, during earlier developmental stages (embryonic days 14 and earlier), cells colonize predominantly cortical layers VI–IV, which renders detection of only a fraction of mis‐localized neuronal cells impossible (data not shown). Therefore, pregnant females between embryonic days 14.5 and 15 were injected with BrdU. Offspring were analyzed 4–5 weeks postnatally to ensure that cell migration and cortical layer formation were completed (Bayer and Altman, 1991). Heterozygous females were crossed with homozygous males to enable comparison of a large number of heterozygous offspring with homozygous litter mates. Only litters in which all brains of both genotypes showed predominant labeling of cortical layers II and III (assessed by comparison with Nissl‐stained parallel sections) were included in the analysis shown in Table ID and E. Anesthetized mice were perfused and processed for cryo‐sectioning as described before. Serial coronal sections (25 μm) were collected individually in PBS. Sections were incubated free‐floating in solution and endogenous peroxidases were blocked by treatment with 2% H2O2 in H2O for 15 min. Subsequently, sections were pre‐treated in 2 M HCl solution (1 h at 37°C), trypsin digested (5 mg/ml for 6 min at room temperature) and permeabilized with Tween‐20 (0.5% in PBS for 20 min at room temperature). Following washing in PBS and post‐fixation in 4% PFA, sections were blocked in 5% goat serum, 3% bovine serum albumin (BSA) and 0.3% Tween‐20 in PBS (BS) for 1 h at room temperature. Incubation with antibodies was done overnight at 4°C (1 U/ml in BS). Immune complexes were visualized using 1 mg/ml diaminobenzidine tetrahydrochloride (Sigma) and 0.03% H2O2 (Aldrich) in 0.1 M Tris pH 7.2. Afterwards, sections were mounted on gelatin‐coated glass slides, dried overnight at room temperature and mounted in Eukitt medium (Riedel‐de‐Haën).
Quantitation of cell migration experiments and parvalbumin‐positive neurons
In 4‐ to 5‐week‐old mice, large round nuclei (indicative of neurons) were counted in the target layers III and II and the area between lateral ventricle and cortex layer III using an ocular micrometer to enable quantitation of equal squares. Both left and right cortical hemispheres were analyzed using four sections spaced at roughly equal intervals in the anterior part of either heterozygous or FGF‐2‐deficient mice. For each cerebral cortex, the cells which fail to colonize their target layers (i.e. positive cells remaining in the corpus callosum and layers VI–IV) were determined as a percentage of the positive cells in target layers III and II (Table IE). Weakly labeled small or irregular nuclei were not counted as they most likely reflect cells of endothelial and/or glial origin (Bayer and Altman, 1991).
Parvalbumin‐positive neurons were detected using commercial primary antibodies (Sigma) on 25 or 30 μm floating cryosections. Total numbers of parvalbumin‐positive neurons were determined by analyzing left and right hemispheres of four cerebral cortex sections at identical positions. All sections were processed identically to enable direct comparison.
Mean arterial blood pressure and heart rate were measured in awake, resting adult mice of 3–6 months of age. On the day of recording, mice were anesthetized with a single dose of intraperitoneal fentanyl (0.03 μg/g body weight), fluanisone (10 μg/g body weight) and midazolam (10 μg/g body weight). For blood pressure measurements, a polyethylene catheter (outer diameter: 400 μm) was inserted into the left carotid artery. If drugs were infused during the experimental protocol, a second polyethylene catheter of similar size was placed into the left jugular vein. All catheters were fixed to the vessel wall and led to the neck of the animal. Mice usually were awake <60 min after induction of anesthesia and were allowed to recover from surgical procedures for 5 h prior to recording baseline blood pressure. Each recording session lasted at least 10 min, and only episodes of physical rest were analyzed. During recordings, mice were placed in Plexiglas tubes to partially restrict their movements.
To induce chronic hypertension in female mice, angiotensin II (2 ng/g body weight/min) was infused continuously over 6 days using osmotic mini pumps (alzet, model 1007D). The osmotic pumps were implanted through a small neck incision during general anesthesia (see above). In a separate group of mice, a sham operation was performed. The dose of angiotensin II was adjusted to induce an increase in mean arterial blood pressure of ∼20 mmHg in normal mice as determined in a series of pilot experiments. To minimize effects of confounding biological variables (initial body weight, age, genetic factors), female litter mates of the same genotype were assigned to both treatment groups in a paired fashion. On the day of recording, mice were given a carotid artery catheter during general anesthesia (see above), and blood pressure was measured as described above. Statistical comparisons between the different experimental groups were made using a two‐way repeated measures analysis of variance (ANOVA) followed by the Bonferroni test for repeated comparisons.
Baroreflex control of blood pressure was studied in male mice using infusions of the β1/β2‐adrenoceptor agonist isoproterenol. After baseline blood pressure was determined for at least 10 min, isoproterenol (0.05 ng/g body weight/min) was infused at a constant rate (0.25 μl/g body weight/ min) intravenously for 3 min. Blood pressure responses were averaged over the last 60 s of the infusion period. Then the infusion was stopped, and blood pressure was followed for another 3 min to assess the reversibility of the effects. To determine the critical isoproterenol dose, a dose–response curve was established as follows: six wild‐type mice were infused with increasing doses of isoproterenol (0.05, 0.1 and 0.2 ng/g body weight/ min). This procedure was repeated twice and yielded the following results: 0.05 ng/g/min, −0.3 ± 1.7; 0.1 ng/g/min, −5.7 ± 7.1 mmHg; 0.2 ng/g/min, −22.7 ± 3.4 mmHg. Statistical comparisons were made between baseline and infusion periods using ANOVA (dose–response curve) or the paired Student's t‐test. All animal experiments were approved by the local authorities and performed in accordance with the EMBL guidelines on animal experimentation.
The authors thank F.Casagranda and R.Klein for providing genomic libraries and ES cells for homologous recombination, G.Martin for providing the mouse FGF‐2 cDNA, and A.Plück‐Becklas, K.Brennan and M.Lemaistre of the EMBL TCF facility for generating mouse chimeras. We are indebted to K.Vintersten, J.Helppi, A.Reid and the EMBL animal technicians for excellent mouse care and help with animal procedures. We are grateful to G.Holländer and P.Monaghan for providing in situ probes, A.Neubüser for advice on non‐radioactive section in situ hybridization, and J.Peters for performing the renin analysis. We are indebted to A.Haramis, A.Hohn, R.Klein, W.Kriz, F.Maina, L.Minichiello, D.Orioli, A.Postigo and H.Seller for helpful discussions during the analysis of the CNS phenotype. G.Davidson, F.Guillemot, A.Just and A.Zuniga are thanked for very helpful discussions and suggestions on the manuscript. This study was partially supported by an EU grant (CI1*CT93‐0017) to R.Z.
- Copyright © 1998 European Molecular Biology Organization