Tumor necrosis factor alpha (TNFα) is a cytokine produced mainly by cells of the immune system. It is also expressed by brain neurons and glia. The physiological role of TNFα in the brain is not yet fully clear. Using TNFα-deficient mice, we have examined its role in hippocampal development and function. We report here that TNFα is involved in the regulation of morphological development in the hippocampus. TNFα-deficient mice exhibited an accelerated maturation of the dentate gyrus region and smaller dendritic trees in CA1 and CA3 regions in young mouse. In addition to its involvement in hippocampal morphogenesis, TNFα deficiency specifically improved performance of affected mice in behavioral tasks related to spatial memory. Moreover, lack of TNFα increased the expression of nerve growth factor (NGF), but not brain-derived neurotrophic factor (BDNF), following performance of the learning task. Our results suggest that TNFα actively influences hippocampal development and function. In adult mice, TNFα may interfere with memory consolidation, perhaps by regulating NGF levels.
The central nervous system was considered for many years as an immunologically privileged site, i.e. where cells of the immune system are absent. Recently, this concept was called into question when immune cells, as well as pro-inflammatory cytokines such as Tumor necrosis facto alpha (TNFα) and interleukin-1 (IL-1), were detected in the brain. In fact, it is now known that TNFα acts in the brain, through its receptor, TNFRI, a 55 kDa protein, to promote apoptosis and, through an isoform of the receptor, TNFRII (75 kDa) as a trophic factor (Neumann et al., 2002). Neurotrophins such as nerve growth factor (NGF), operating through their tyrosine kinase (Trk) receptors, block TNFα-induced apoptosis. According to the neurotrophic hypothesis, newly differentiating neurons compete for neurotrophins that are in limited supply and only those that succeed in establishing correct synaptic connections will acquire the trophic support to allow their survival (Levi-Montalcini, 1987). Thus, both death signals and survival signals could play a key role during neurogenesis. Recently, it was demonstrated that this death is induced partly by TNFα (Barker et al., 2001).
In the present study, we have examined the involvement of TNFα in the development and function of the central nervous system (CNS), concentrating on the hippocampal region. Using TNFα-knock-out (TNF-KO) mice, we have demonstrated, for the first time, that basal levels of TNFα in the brain regulates both NGF and brain-derived neurotrophic factor (BDNF) levels during hippocampal development. Modulation of these factors governs granular cell migration in the dentate gyrus and dendritic tree formation by pyramidal cells in regions CA1 and CA3. Furthermore, we have demonstrated that TNFα may specifically suppress hippocampal function in adult mice, through suppression of NGF expression.
Materials and Methods
Jackson Black C-57 mice and TNF-KO on JB C-57 background mice were used in the study. The TNF-KO mice (Pasparakis et al., 1996) were generously provided by Professor George Kollias, Hellenic Pasteur Institute, Athens, Greece. Absence of the TNFα gene was confirmed by ploymerase chain reaction (PCR) and by specific mTNFα enzyme-linked immunosorbent assay (ELISA) to serum of lipopolysaccharide (LPS)-injected mice (3 h after i.p. injection of 20 µg/ml of LPS) or to conditioned media of spleen cells activated with LPS (10 µg/ml) for 24 h. The mouse colony was maintained in a 12:12 light:dark schedule; food and water were provided ad libitum. All procedures were performed according to guidelines from the Israeli Council on Animal Care and approved by the Ben-Gurion University animal care and use committee.
Phenotype Aspects in the TNF-KO Mice
Mice were tested daily during the first month of life for general phenotypic and morphogenic aspects: body wt and length; brain length and width; hair growth; and date of eyelid opening and tooth eruption. A single animal per group was drawn from any single litter. Following phenotype examination, mice were anesthetized and thereafter killed. Brains were rapidly removed into ice-cold artificial cerebrospinal fluid (ACSF) and deeply frozen at –80°C for later analysis. Mice aged 1–7 days were anesthetized by hypothermia, while older mice were anesthetized by i.p. administration of ketamine and rompun.
Murine TNFα, NGF and BDNF ELISA
Immunoreactive murine TNFα was examined in serum and conditioned media of spleen cells activated with LPS; NGF and BDNF were examined in homogenates of hippocampi using a specific ELISA (R&D Systems, Minneapolis, MN). The sensitivity of the ELISA was 4 pg/ml for TNF and NGF and 47 pg/ml for BDNF.
Dendritic Tree Staining and Sampling
Four-day-old wild type (wt) and newborn TNFα-KO mice were cold-anesthetized, brains were rapidly removed to cold ACSF, hippocampi were isolated and transverse slices (300 µm thick) prepared with a tissue chopper. Following a 1 h incubation at 4°C, slices were placed in petri dishes coated with micro-crystals of the vital fluorescent dye DiIC18 (Molecular Probes), covered with a drop of modified Eagle medium (MEM) medium and incubated at 37°C, 5% CO2 for 10 min. After a brief rinse, the slices were incubated until appropriate staining was achieved, then fixed in 4% paraformaldehyde, washed in phosphate-buffered saline (PBS) and sampled. Neurons that could be separately visualized were sampled in a confocal laser scanning microscope (LSM510) at a magnification of 20×. Images were flattened and analyzed using NIHimage. Dendritic branch length and number of branch points per neuron were obtained for the apical dendrite.
Immunohistochemical and Cell Nuclei Staining
Sections for immunohistochemical staining were prepared from paraformaldehyde-fixed (4%), paraffin-embedded tissue. Transverse sections (4 µm thick) of medial hippocampus, were mounted on saline-coated slides, dried at 37°C for 48 h and stored at room temperature. Before the primary antibodies were applied, endogenous peroxidase was quenched by incubation in 3% H2O2 in 80% methanol for 15 min and non-specific staining blocked with a serum and albumin-containing solution. This solution was also used to dilute the primary antibodies and for washing the sections. Primary antisera (BDNF-IgG, R&D Systems, Minneapolis, MN; NGF-IgG, Alomone Labs, Jerusalem, Israel) were applied at dilutions of 1:100 and 1:50, respectively. The biotinylated antibodies and avidin–biotin conjugate were applied according to the suppliers’ directions. Development was performed with 0.06% 3′,3-diaminobenzidine + H2O2, with or without cobalt enhancement; Mayers haematoxylin was used to counterstain cell nuclei in corresponding sections. The sections were mounted with Eukitt. For immunofluorecence localization of BDNF, a secondary antibody conjugated to Cy-3 was used. For a negative control, normal serum replaced the primary antibodies. For cell density analysis, images of the dentate gyrus region of the hippocampus were sampled in an Olympus IX-70 microscope at 400× equipped with a SuperCam video camera (Applitec, Israel). Analysis was performed using NIHimage (Wayne Rasband, NINDS, NIH). The number of cell nuclei/(pixel)2 was determined in transverse sections through the approximate mid-rostral–caudal hippocampus of P1, P7, 14 mice. Cell density was categorized to: (i) more than one cell per 1000 pixel2, which represents the non-developed hilus and (ii) less than one cell per 1000 pixel2, which represents the normal development in the hilus of adult mice.
Reflex Development in Newborn Mice
Righting reflex was measured as the time, in seconds, required for a newborn laying on its back to right itself on four limbs (Crawley, 1999). Rotaroad (adjusted for neonates) tests the ability to hold a rotating bar at a rate of four cycles per minute. Latency to fall from the rotarod was recorded. Locomotion on inclination—the ability of newborn mice to climb on an inclined board—was examined at slopes of 45, 70 and 90°.
Behavior of Adult Mice
Open Field Task. Performance was tested in an open field arena: a 65 cm diameter gray wall, 30 cm high, over a period of 2 min. The following variables were observed: rearing (times), grooming (times), center/perimeter time and quantity of defecation delivered (Henderson, 1967).
Elevated Plus Maze. A ‘plus maze’, 15 cm high and with 40 cm long arms, was used to measure the time (s) spent in the open versus the closed arms (Lister, 1987a).
Hole Board. The number of holes into which the mice inserted their noses during a given 2 min interval was measured (Lister, 1987b).
Morris Water Maze. Mice were trained to swim to a hidden platform in the water maze (Morris, 1984). The maze consisted of a 70 cm diameter, 30 cm high, circular pool filled with milk and maintained at 25–26°C. The 10 cm diameter platform was 0.5–1 cm below the surface of the milk. The pool was placed in the center of a room with white walls and a colored geometric figure on each of the walls (triangle, square, circle, strip) that served as spatial cues. One day before the training, the mice were placed on the platform for 10 s, if they failed, they were replaced on the platform. Twelve male mice of each group (90 days old) were tested in a blind analysis and were randomly assigned for training to find the platform in a given quadrant. For three consecutive days, each mouse was placed in the pool six times, starting in a random order of direction — north, south, east or west — with 30 min intervals between trials. After locating the platform, the mice were dried and returned to their cages. If, after 60 s in the pool, the mice could not locate the platform, the trial was stopped and the escape latency recorded as 61 s. A probe test was conducted as follows. Following the last trial on the third day, the platform was removed from the pool. The mice were placed in the pool, starting in the location opposite the platform and allowed to swim for 60 s. The time spent in swimming in each quadrant was recorded. The amount of time spent swimming in the quadrant where the platform had been located was recorded as another index of spatial learning. All data were recored on videotape and analyzed blind. Mice were housed in individual cages on a 12 h light:dark cycle and were tested between 9:00 and 16:00, daily.
Visible Platform Training. Spatial cues were removed and the location of the platform made visible by use of clear water, marking of the submerged platform by blue paint and elevating it 1 cm above the surface of the water. The remaining details were similar to those described for the hidden platform test.
Effects of TNFα-deficiency on Mice Phenotype Development
The number of offspring per litter in the TNF-KO mice was similar to the wt mice (7 ± 2.74, n = 22 and 8.5 ± 2.9, n = 16, respectively; mean ± SD). TNF-KO mice were analyzed for gross developmental consequences through the first postnatal month. The mice so examined developed normal phenotypes with respect to body wt, body length and general brain size (depicted in Fig. 1). The days of eyelid opening, hair growth and teeth eruption were considered as the days in which 50% of newborn fit the criteria (Table 1). No significant differences were observed in these parameters.
In addition, development of locomotion reflexes, muscle strength and balance in the rotarod and inclining slope were similar in both groups. Righting reflex was fully developed in wt mice between postnatal days 2 and 9. However, in the TNF-KO newborns, it was fully developed by postnatal day 5 in all of the mice pups. Moreover, TNF-KO pups that succeed in righting themselves by postnatal days 4 and 5 were significantly faster than those in the wt group as depicted in Figure 1c (P < 0.01, t-test). In general, except for their superiority in the development of the righting reflex, TNF-KO mice exhibited phenotype and reflex development at the same rate as wt mice.
NGF and BDNF are differentially expressed in the hippocampus of TNFα-deficient and wt mice. Development of the hippocampus begins in the embryo and continues through the first month after birth (Bayer, 1980). One of the crucial molecules regulating neuronal survival and hippocampal morphogenesis is the neurotrophin, NGF (Levi-Montalcini, 1987). Hippocampal levels of NGF in rodents increase throughout the first month of life. In the present study, NGF protein was measured in hippocampal homogenates from both wt and TNF-KO mice at P1, P7, P14, P21 and P180 (depicted in Fig. 2b). During the first postnatal month, the NGF levels were significantly higher in the hippocampi of TNF-KO mice than in wt mice, as illustrated by immunohistochemical staining of the CA1 region of the hippocampus and neocortex (Ctx; Fig. 2a). NGF immunostaining was weaker in both CA1 and cortex of wt mice compared to TNF-KO mice (upper panels of Fig. 2a). Quantification of NGF in the homogenates of these tissues by ELISA confirmed the immunostaining results (Fig. 2b; P < 0.05, F = 3.8, one-way analysis of variance). This developmental difference was not observed in adult mice (Fig. 2b).
The levels of BDNF in the hippocampus were similar in both groups in the first postnatal week. However, in adult mice, levels of BDNF were lower, thought not reaching statistical significance. That could be seen in the immunofluorescent staining of hippocampal sections (Fig. 2c) and ELISA of homogenates (Fig. 2d) from P7 mice.
Involvement of TNFα in the Regulation of Hilar Maturation (Morphogenesis)
Since NGF and BDNF are involved in various aspects of neurogenesis and in our present study we found that these factors are differently expressed in TNF-KO and wt, we therefore asked whether loss of TNFα also affects hippocampal morphogenesis. The dentate gyrus of the hippocampus is an area that undergoes extensive development postnatally (Bayer, 1980). During the first postnatal month and even later, cells in the hilus continue to divide, differentiate and migrate toward the stratum granulosum layer (Gould et al., 1998). Hence, at birth, the hilus is densely populated with cell somata, which, over the following days, with the end of mitosis and continued cell migration, reduces, indicating maturation of the area. In the present study, we used cell density in the hilar region as an indicator of the state of dentate gyrus development and maturation. Figure 3a–c depicts the dentate gyrus at the different stages of postnatal development in wt mice. The right-hand panels (Fig. 3d,e) depict representative images of this area in the TNF-KO mice of the same ages. The percentage of mice with mature hippocampal hilus (see Materials and Methods) was significantly higher in the TNF-KO mice compared to wt during the first week (P1 and P7, P = 0.0004 and 0.001, respectively). However, by P14, both groups had achieved equivalent developmental stages (Fig. 3f). This would appear to suggest that at birth, the dentate gyrus and hilus of the TNF-KO mice are already at advanced maturational stages.
Decreased Arborization of the Apical Dendrites in CA1 and CA3 of TNF-KO Mice
At birth, pyramidal neurons in the CA1 region of the hippocampus vary in size and dendritic tree arborization. Some cells present only short apical dendrites, while others present a more developed dendritic tree (Tyzio et al., 1999). The number of branch points in the apical dendrites of pyramidal neurons and the length of dendritic branch segments (as shown in Fig. 4bii) were employed as measures of development. As depicted in Figure 4c, most of the cells in the CA1 region of P4 mice of both groups developed, on average, <10 branch points per neuron. On the other hand, in the wt group, 37% of the neurons developed >10 branching points per neuron compared with 5% in the TNF-KO group (Fig. 4c, n = 27, 49, P < 0.004, χ2 test). A similar trend was observed in CA3. The number of neurons with a small number of branch points in CA3 was significantly higher in the TNF-KO mice compared to wt mice (n = 52, 47, respectively, P < 0.0001, χ2 test). In the wt mice, 60% of neurons developed dendritic trees with >10 branch points. However, only 15% of neurons from the KO mice exhibited the same number of branch points. Thus, the pyramidal neurons of both CA1 and CA3 from TNF-KO mice were less arborized than those from wt mice.
The average lengths of segments of dendritic tree branches are shown in Figure 4e,f as percentages of total segments of branches. No significant difference was observed in this parameter in CA1 and CA3 regions of both study groups. Thus, although no significant differences were observed between wt and TNF-KO in gross phenotype and development of some motor reflexes during the first month of life, the TNF-KO animals did exhibit specific differences in morphological development.
Effect of TNFα-deficit on Behavior in Adult Mice
We have examined whether the developmental differences observed in hippocampi of wt and TNF-KO mice could lead to differences in their behavior as adults. In the open field arena (Fig. 5a), TNF-KO and wt mice were found to be similar in their rearing and grooming behaviors. Nevertheless, the ratio of perimeter versus center time was higher, 0.83 ± 0.17 (mean ± SD), for the TNF-KO compared to 0.63 ± 0.2 for the wt mice (n = 12, P = 0.03, t-test; Fig. 5a) and they produced a larger numbers of fecal pellets (2.25 ± 1.4 and 1.16 ± 1.3, respectively, n = 12, P = 0.03, t-test). These results may indicate that the TNF-KO mice are more distressed than the wt mice when exposed to an open field environment. This possibility was assessed in the elevated plus maze (Fig. 5b). The time that mice from both groups spent in the dark arms did not differ (Fig. 5c). In order to evaluate the basal activity levels in addition to the exploratory behavior of the mice; activity was examined in the holes board (Fig. 5c). Again, no statistical difference was noted between the groups (24 ± 8.9 and 30.66 ± 9.4 holes/2 min, for the wt and TNF-KO mice, respectively). Since developmental differences were observed in the hippocampi of young mice, another task, based on the neural circuitry of the hippocampus, i.e. spatial learning in the Morris water maze, was examined. Twelve mice from each group (each one from a different litter) were examined in the maze (Fig. 5d,e). The performances of the TNF-KO mice, on the first day, were better than performances of the wt mice. No errors in platform finding occurred during the first trial, compared to 5/12 errors for the wt mice (Fig. 5d). Performance of the task by the KO group was also characterized by shorter latency in finding the platform (26.4 ± 6.6 s, compared to 44.1 ± 6.7 s for the wt, n = 12, P < 0.05, t-test, Fig. 5e). On days 2 and 3, these differences were masked, due apparently to the learning process in the wt group. Though not demonstrated by latency, the TNF-KO mice better remembered the platform location, as indicated by the results of the probe test conducted at the end of the experiment. In these experiments, the TNF-KO mice spent a significantly longer time (32.2 ± 5.6 s) in the correct quarter compared to the wt mice (20.4 ± 5.7 s) (P < 0.01, t-test). No differences were observed between the two groups in the performance of the cued version of this task.
NGF and BDNF are Differently Regulated by TNFα in Brains of Adult Mice Following Learning
Neurotrophins, in addition to supporting survival and growth of neurons, are also involved in learning and memory consolidation (Woolf et al., 2001). Therefore, we have examined NGF and BDNF levels in adult mouse brains in both groups following learning in the Morris water maze. Twenty-four hours following the probe test, NGF levels in mouse brains were determined by ELISA. Left and right hippocampi were examined separately and the frontal cortex was analyzed as a negative control area, since it is not primarily involved in the spatial learning processes. NGF levels in TNF-KO hippocampi were two to three fold higher than those in the wt mice (n = 5, P < 0.01, t-test, for NGF in whole hippocampus of wt as compared to TNF-KO). This change is apparently region specific, since no difference in NGF levels was observed in the frontal cortex (Fig. 6). In contrast, 24 h after learning, BDNF levels did not differ in the two groups, either in hippocampus or frontal cortex (Fig. 6b). This might suggest that expression of NGF protein is specifically regulated in the hippocampus by TNFα during memory consolidation.
TNFα is Involved in Hippocampal Development
TNFα deprivation during embryonic and early postnatal development did not result in gross phenotypic alterations. However, it did affect protein levels of neurotrophic factors and resulted in morphological variation in the hippocampus. The major roles of neurotrophic factors in neurogenesis are well established. They are involved in cell differentiation, migration and morphogenesis (for reviews, see Huang and Reichardt, 2001; Sofroniew et al., 2001). Since NGF was shown to induce cell differentiation, it is possible that elevated levels of NGF in the TNFα-deprived mice may result in an accelerated maturation of the dentate gyrus region. The appearance of a ‘mature’ dentate gyrus in the P1 TNF-KO newborn may reflect acceleration of embryonic development. Similarly, exogenous BDNF has been shown, in various brain areas, to enhance axonal growth and branching (McAlister et al., 1995) and regulation of growth and maturation of the dendritic tree in developing cerebral cortex (McAlister et al., 1995, 1997) and also to influence synaptogenesis in the hippocampus (Martinez et al., 1998). Our finding that TNFα deprivation reduced hippocampal BDNF levels and the number of branching points of the apical dendrite of CA1 and CA3 pyramidal cells is consistent with these reports. Although, TNFα was previously shown to be involved in the regulation of NGF and BDNF (Aloe et al., 1999), we report here the first in vivo evidence linking TNFα to hippocampal morphogenesis and, more specifically, to dentate gyrus maturation and dendritic tree arborization in CA1 and CA3. A recent paper reported that TNFα has a direct effect on dendritic branching in cultured neurons (Neumann et al., 2002), probably through its TNFRII receptor, which is known to transduce the trophic effect of TNFα (Yang et al., 2002). However, it is unclear if the basal levels of TNFα in these regions are sufficient to activate this receptor. Moreover, since we observed significant changes in neurotrophin levels, we suggest that TNFα could have a direct influence on the morphological changes we observed, which might be effected in a synergistic manner with the neurotrophins. Reduction in the branching of the pyramidal cell dendrites of TNF-KO mice contradicts the suppressive effect of TNFα on dendritic branching reported by Neumann et al. (2002). This could be explained by the difference in the culture (i.e. in vitro) versus the intact brain (i.e. in vivo) environment. TNFα induction of AMPA receptor surface expression was recently reported (Beattie et al., 2002). However, its biological relevance to our findings is not yet clear. All the above developmental changes may alter the performance in hippocampus-dependent tasks, such as those observed in water maze test.
Additional evidence for TNFα involvement in the development of neuronal circuits was demonstrated in the righting reflex. Accelerated reflex development was observed in the TNFα-deficient mice, possibly due to earlier maturation of cerebellar circuits. We suggest that TNFα had no effect on the musculature development, since similar performance to the wt was observed in the rotaroad task and inclining slope locomotion, both tests involving muscle strength.
Involvement of TNFα in Learning and Memory
Our data indicate specific involvement of TNFα in spatial learning and memory as measured in the Morris water maze and possibly on other behavioral aspects. These finding are in agreement with a recent report by Aloe et al. (1999), who demonstrated, in two lines of mice that over-expressed human recombinant TNFα, a significant impairment in spatial learning, as demonstrated by the Morris water maze. In addition, our data are also in agreement with several other studies that have examined the suppressive effect of TNFα on synaptic plasticity in the hippocampus. In those studies, application of exogenous TNFα to hippocampal slices suppressed LTP (Tancredi et al., 1992; Cunningham et al., 1996), a cellular manifestation of learning and memory. Besides the specific improvement in performance of the spatial learning task, TNFα-deficient mice in the present study showed a region-specific increase in NGF levels following learning. It was previously reported that NGF levels correlate with an animal’s capacity for spatial learning (Henrikkson et al., 1992) and that NGF infusion facilitates spatial memory performance (Janis et al., 1995). Involvement of NGF in memory consolidation was also reported by Woolf et al. (2001). However, in that study, elevated levels of NGF in rats were obtained only a week following the contextual learning paradigm. Comparing NGF levels in the adult mice before (Fig. 2b) and after learning (Fig. 6a) indicates that in the wt mice, there was a 70% suppression of NGF production after learning. In contrast, in the TNF-KO mice, NGF levels were 26% higher following learning as compared to levels in adult mice (Fig. 2b). Thus, it is possible that under normal conditions (wt mice), basal levels of TNFα partially impede plasticity of hippocampal circuitry and that removal of this suppressing factor (TNF-KO mice) enables exposure of a larger portion of it capacity.
Besides the specific increase in NGF levels in the hippocampus of TNF-KO mice, no differences were observed in BDNF levels in either hippocampus or frontal cortex. An increase in BDNF levels in rat hippocampus following a spatial learning task was previously reported following intensive training in an eight-arm radial maze (Mizuno et al., 2000). It is possible to suggest that the relatively short training period required for the Morris water maze was not sufficient to induce an increase in BDNF protein level. No differences were observed in BDNF levels in either group, suggesting that TNFα regulation during neuronal activity could be specifically associated with NGF levels.
The present data suggest that basal levels of TNFα may be specifically involved in the regulation of hippocampus development and function. The interaction between TNFα and NGF may indicate that the levels of NGF in young mice may regulate higher sensitivity to NGF in adults, for example, by up-regulating expression of Trk A.
Our results indicate that TNFα specifically and significantly affects the hippocampus. They further support the possible involvement of TNFα in the regulation of learning and memory under physiological and pathological conditions. Since TNFα has been implicated in the development of some neurodegenerative diseases related to memory, such as Alzheimer’s, it is possible that regulation of TNFα may be considered in development of new therapeutic strategies.
Supplementary material can be found at: http://www.cercor.oupjournals.org
We thank Dr G. Kolias from the Pauster Institute, Athens Greece for kindly providing the TNF-KO mice and Mr T. Green for technical assistance in the analysis of DiI staining. We also thank Dr Zeev Silverman, from the Morphology Department in Ben-Gurion University, for critical reading of the manuscript. The study was supported by The National Institute for Psychobiology in Israel.
Postnatal day in which 50% of newborns fit criteria, n = 10–30.