Maternal brain TNF programs innate fear in the offspring

SUMMARY

Tumor necrosis factor alpha (TNF) is a cytokine that not only coordinates local and systemic immune responses [1, 2] but also regulates neuronal functions. Most prominently, glia-derived TNF has been shown to regulate homeostatic synaptic scaling [3–6], but TNF null mice exhibited no apparent cognitive or emotional abnormalities. Instead, we found a TNF-dependent intergenerational effect, as mothers with a deficit in TNF programmed their offspring to exhibit low innate fear. Crossfostering and conditional knockout experiments indicated that a TNF deficit in the maternal brain, rather than in the hematopoietic system, and during gestation was responsible for the offspring low-fear phenotype. The level of innate fear governs the balance between exploration/foraging and avoidance of predators, and is thus fundamentally important in adaptation, fitness and survival [7]. Since maternal exercise/activity is known to reduce both brain TNF [8] and offspring innate fear [9], while maternal stress was reported to increase brain-TNF [10] and offspring fear and anxiety [11, 12], maternal brain-TNF may report environmental conditions to promote offspring behavioral adaption to their anticipated postnatal environment.

eTOC Blurb

TNF regulates synaptic scaling, but Zupan et al. show that TNF-null mice exhibit no behavioral anomalies. Instead, a brain TNF deficit in dams results in low fear in the offspring. As maternal activity reduces brain TNF and downregulates offspring fear, brain TNF may report environmental conditions and help optimize offspring behavioral adaptation.

RESULTS

Parental TNF deficit programs low innate fear in the offspring.

The effect of genetic inactivation of Tnf on emotional behaviors, specifically on innate fear, learned fear and stress reactivity, was assessed by testing two groups of male WT offspring; one from WT and another from Tnf^+/− heterozygote (H) parents [2] (referred to as WT_offspring(WT_parents) and WT(H), respectively), and two groups of Tnf^−/− null (KO) male offspring; one from H and another from KO parents (i.e. KO(H) and KO(KO), respectively) (Figure 1A). First, innate fear was assessed in novel environments that inherently pose an approach-avoidance conflict, similar to that between foraging in open areas to feed vs. actively avoiding predators in nature [7]. While WT(H) and KO(H) littermates were not different from each other, KO(KO), relative to non-littermate WT(WT) mice exhibited more exploration, measured as an increase in the number of entries, time spent, and distance travelled in the center area of the open field (OF) and open arm of the elevated plus maze (EPM) (Figure 1B,C), with no change in total activity; interpreted as reduced innate anxiety of KO(KO) mice in novel environments. Principal Component Analysis (PCA) of all data (entry, distance, and time in OF and EPM) clearly showed the separation of KO(KO) offspring from the other groups (Figure 1D). Given the absence of behavioral differences between littermate WT(H) and KO(H) mice (i.e. offspring genotype differences), the reduced innate fear of KO(KO) relative to that of WT(WT) mice is not related to the offspring but rather to the parental KO genotype. As reduced innate fear in OF and EPM was seen only when parents were KO, i.e. KO(KO), but not when they were H, i.e. KO(H), a complete lack of parental TNF may be necessary to program the full phenotype in male offspring. No behavioral differences were found between littermate, i.e. WT(H) vs. KO(H), and non-littermate, i.e. KO(KO) vs. WT(WT), mice in cued fear conditioning and immobility time in forced swim test (FST) that assesses stress reactivity (Figure S1), suggesting that, in terms of programming stress response, the parental TNF effect is limited to innate fear.

Figure 1. Parental TNF deficit programs low innate fear in the offspring.

(A) Breeding strategy generating TNF WT and KO offspring derived from WT [WT_{offspring genotype}(WT_{parental genotype})], H [WT(H) and KO(H)], and KO [KO(KO)] parents. (B and C) Reduced parental TNF results in increased exploration of the center of OF and open arm of EPM. One Way ANOVAs of OF test; % center entry: F_3,53=5.89, P=0.0001; distance: F_3,53=15.78, P=0.0000; time: F_3,52=6.21, P=0.0011; Tukey HSD posthoc, *p<0.05, ^***p<0.001; n=15, 12, 13, 16 and of EPM test; % open arm entry: F_3,45=3.86, P=0.0154; distance: F_3,45=2.53, P=0.0688; time: F_3,45=3.58, P=0.0210; n=16, 9, 9, 15. (D) PCA plots of all OF and EPM data (entry, time and distance). Coordinates for the 95% confidence intervals are displayed around the barycenter of individual groups. See also Figure S1.

Parental programming of innate fear is associated with the prenatal period.

To specify if the parental TNF effect is associated with the prenatal or postnatal period, WT(WT) and KO(KO) offspring were crossfostered at birth to either WT or KO foster mothers (Figure 2A). WT(WT) mice raised by WT or KO mothers (i.e. WT_offspring(WT_prenatal/WT_postnatal) and WT(WT/KO), respectively) were indistinguishable in their innate fear responses in EPM and OF (Figure 2B,C), indicating that the postnatal KO environment is not a significant factor in programming reduced innate fear. Instead, a clear difference emerged when the control WT(WT/WT) group was compared to either the KO(KO/WT) or KO(KO/KO) groups, which both differ from the control by their KO prenatal environment. PCA also showed the separation of the two prenatal KO groups from the control and postnatal groups (Figure 2D). This suggests that parental programming of innate fear is associated with the prenatal period.

Figure 2. Prenatal parental environment is responsible for programming innate fear.

(A) Crossfostering strategy generating subjects exposed to TNF-deficient prenatal KO_offspring(KO_prenatal/WT_postnatal), postnatal [WT(WT/KO], pre/postnatal [KO(KO/KO] and control [WT(WT/WT] developmental environment. (B and C) Parental TNF deficit during the prenatal period results in increased exploration of the center of OF and open arm of EPM. One Way ANOVAs of OF test; % center entry: F_3,52=8.33, P=0.0001; distance: F_3,52=3.92, P=0.0135; time: F_3,52=23.40, P=0.0000; Tukey HSD posthoc, *p<0.05, ^**p<0.01^***p<0.001; n=13, 15, 16, 12 and of EPM test; % open arm entry: F_3,86=3.09, P=0.0313; distance: F_3,86=2.32, P=0.2363; time: F_3,86=.99, P=0.3998; n=22, 21, 25, 22. (D) PCA of all OF and EPM data.

TNF deficit in maternal brain programs reduced innate anxiety.

TNF is produced in the periphery by immune cells that, by acting on the germline and/or embryo, could program fear responses in the offspring. TNF is also produced in the brain (primarily by glia but also neurons [3]) and is believed to regulate synaptic scaling [3–5]. Indeed, we detected both TNF protein and mRNA in various brain regions during pregnancy (Figure S2). To determine the source of TNF in the prenatal programming of innate fear levels, we used a conditional KO approach (Figure 3A). This strategy also allowed us to test if programming is maternal, paternal, or both. Using a loxP-flanked Tnf allele in combination with a nestin-cre or polyIC-induced Mx1-cre transgene, we generated females with either brain or hematopoietic cell specific Tnf deletion respectively [13]. Nestin is expressed in neural stem cells producing both neurons and glia during development. Nestin-cre:Tnf^loxP/loxP mice had a 60–85% reduction of the WT Tnf allele in the cortex, hippocampus, isolated CA1 and dentate gyrus neurons, and glia rich corpus callosum but no change in spleen (61%, 64%, 72%, 88%, 82%, and 20%, respectively) [13]. Consistent with these data, Tnf mRNA was reduced by ~80% in the cortex and hippocampus but not in spleen (80%, 82%, and 20%, respectively). Male Tnf^loxP/loxP offspring (WT TNF expression) of nestin-cre:Tnf^loxP/loxP mothers (TNF deficient) and Tnf^loxP/loxP fathers (WT TNF expression), i.e. cre^-_offspring(cre⁺_mother), as compared to control cre^-(cre^-) offspring, exhibited increased center field and open arm entries, time, and distance traveled in OF and EPM (Figure 3B,C). This maternal brain KO effect on reducing innate fear was independent of the offspring genotype. Indeed, PCA clearly separated the two maternal cre⁺ offspring, i.e. cre^-(cre⁺) and cre⁺(cre⁺) from the control group (Figure 3D). In contrast, hematopoietic deletion of Tnf via polyIC-induced Mx1-cre:Tnf^loxP/loxP (that resulted in a ~70% reduction of the WT Tnf allele and mRNA in the spleen but not in brain; 75%, 90%, and 24% of WT Tnf allele and 94%, 89%, and 31% of WT Tnf mRNA expression, respectively [13]), caused no apparent change in innate fear in the offspring (Figure 3E-G). These data demonstrate that TNF in the maternal brain programs offspring innate fear. The brain specific maternal deficit in TNF caused no behavioral change in offspring cued fear conditioning and FST immobility time (Figure S3), indicating again that it does not affect learned fear and stress-induced escape behavior and that programming is specific for innate fear.

Figure 3. TNF deficit in maternal brain programs reduced innate anxiety.

(A) Brain and hematopoietic deletion of Tnf by conditional KO strategy; cre-(cre-)=TNF WT offspring, cre(cre+)=TNF WT offspring derived from TNF deficient dams, cre+(cre+)=TNF deficient offspring. (B and C) Maternal nestin-cre genotype t-test (2 tailed) for OF test; % center entry: t(21)=2.41, p=0.025; distance: t(21)=2.54, p=0.019; time: t(21)=2.24, p=0.036; n=7 (cre^-), 16 (cre⁺) and for EPM; % open entry: t(43)=1.69, p=0.098; distance: t(37.59)=4.25, p=0.0001; time: t(37.95)=4.47, p=0.0001; n=13 (cre^-), 32 (cre⁺). (D) PCA of all nestin-cre OF and EPM data. (E-F) Mx-1 cre OF and EPM; no significant differences, n=12, 14, 19.and 15, 15, 15, respectively. (G) PCA of all mx-cre OF and EPM data. See also Figures S2 and S3.

DISCUSSION

Given the fundamental neuronal functions of TNF, particularly in the establishment, function, and modification of synaptic connections in neuronal and slice cultures [3–5], and its expression in the hypothalamus, hippocampus, cortex, and striatum, it is surprising that deletion of TNF in mice has no or only modest impact on emotional (see Results) and cognitive behaviors [13]. Similar to our findings with littermate WT and KO mice (Figure 1; WT(H) vs. KO(H)), targeted inactivation of TNF was reported to have no effect on innate fear in EPM and memory [14] (but caused increased rearing and grooming), and genetic inactivation of the TNF receptor genes Tnfr1 and Tnfr2 caused no change in innate fear either [15]. Also, we previously reported no effect of Tnf KO in spatial memory [13]. We note that two papers reporting spatial memory and visual cortex neuroplasticity changes in TNF KO mice used non-littermate WT and KO mice; thus could not differentiate between the effects of the parental and offspring KO genotypes [16, 17]. Overall, these data indicate that the in vitro identified effect of TNF on synaptic scaling is not apparent at the behavioral level, perhaps because of compensation in vivo or because of the insensitivity of the behavioral tests used. In contrast, the programming effect of brain TNF was robust in the next generation. Although the mechanism connecting the maternal brain TNF and offspring brain/behavior is not known, TNF in mothers may be involved in the top-down regulation of hormonal or immune pathways, such as the hypothalamic regulation of the HPA axis or the cortex-brainstem-autonomic regulation of the immune system [18]. In turn, hormones and cytokines can reach the fetus and influence offspring brain development [12, 19].

Our data can also be interpreted in an ecological context such that physically active dams in a complex environment (like wild relatives) may program, via lower brain TNF, less avoidance and higher exploration to their offspring, because of the adaptive value of reduced innate fear in a competitive environment with limited food availability and extensive foraging. In contrast, dams living in an environment with continuous and predictable food supply and no need for extensive foraging and activity, become relatively sedentary and may program, via higher brain TNF, more avoidance and less exploration of open areas to their offspring. As these behaviors also have adaptive advantages in the given environment (i.e. less exposure to predators in the wild), maternally programmed innate fear may be fundamentally important in adaptation, fitness and survival [7]. This interpretation is consistent with the concept that maternal effects, via fetal programming, facilitate the adaptation of the offspring to their future anticipated environment [20].

Alternatively, because activity is part of life in the wild, low innate fear may be the evolutionarily optimal behavioral state. Hence, the increased fear in the offspring of sedentary mothers, and in more extreme circumstances of stressed mothers, could be the result of compromised or maladaptive programming [19, 20]. Indeed, chronic stress shifts maternal brain proinflammatory cytokines, including TNF, towards borderline high or even pathological levels [10] (more so than sedentary conditions), which may explain the high anxiety/fear of the offspring of mothers experiencing sustained stress during pregnancy [11, 12](Figure 4).

Figure 4. Schematic representation of the direct relationship between maternal brain TNF during gestation and the innate fear response of the adult offspring.

Low maternal brain TNF, as a result of conditional KO of Tnf, programs low, while high TNF level in sedentary mothers programs high innate fear in the offspring. Since maternal exercise reduces while stress increases brain TNF levels, environment-induced changes in maternal brain TNF could be responsible for the low and high anxiety of the offspring of exercising and stressed mothers, respectively.

The positive correlation between maternal brain TNF and offspring fear/anxiety, revealed in our model, has translational implications as exercise (in motorized treadmill) elicits a ~70% reduction in brain Tnf mRNA [8] (similar to the deficit created genetically in our experiments), and because maternal environmental enrichment and associated physical activity reduce innate fear in the adult offspring [9] (Figure 4). This suggests that TNF may mediate, at least partly, the moderating effect of maternal physical activity on offspring innate fear.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Miklos Toth (mtoth@med.cornell.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

All procedures were performed in accordance with the guidelines of the Weill Cornell Medical College Institutional Animal Care and Use Committee. Animals were group housed at 2–5/cage, except breeding cages which contained one male, one female and one litter of pre-weaning pups. Mice were weaned at 3 weeks of age and maintained under standard environmental conditions with food and water ad libitum and a 12hr light cycle with lights on at 600h. TNF^–/– N1F2 hybrid (B6;129S-Tnf^tm1Gkl/J) mice and their appropriate WT controls (B6129SF2/J) were obtained from The Jackson Laboratory, Bar Harbor, ME. They were backcrossed onto the B6/Tac background 6 times. Littermate and non-littermate Tnf WT and KO mice were obtained by crossing Tnf WT(♀) x WT(♂) or KO(♀) x KO(♂) and H(♀) x H(♂), respectively. Nestin-cre mice [21] (B6.Cg-Tg(Nescre)1kln/J) and mx1-cre mice [22] (B6.Cg-Tg(Mx1-cre)1Cgn/J) were purchased from The Jackson Laboratory. TNF^flox/flox mice on the B6J background were obtained from Dr. Sergei Nedospasov (Engelhardt Institute of Molecular Biology, Russian Academy of Sciences)[23]. Mx1-cre recombinase expression was induced in females by the intraperitoneal injection of polyIC (250 μg Sigma-Aldrich) at the age of 6 weeks as described previously [22]. Three weeks later, females were bred to generate conditional TNF mutants. Crossfostering was performed within 24hrs of birth by transfer into the nest of an early post-partum adoptive dam.

METHOD DETAILS

Behavioral testing

All testing was performed between 8am and 5pm on mice 8–14 weeks of age. Exploration of the center of an open field (39W×54L cm²), a measure of innate fear, was assessed as previously described [24]. Subjects’ 10 min exploration of the arena was tracked and analyzed for center entry and distance using EthoVision (Noldus Information Technology, Wageningen, The Netherlands). Innate fear/anxiety was additionally assessed two days later using the elevated plus maze (EPM) as previously described [24]. Briefly, animals were placed into the middle of a four-armed maze (5Wx30.5L cm²) facing an open arm and allowed to explore for 10 minutes. Number of arm entries and distance covered was tracked and analyzed by EthoVision. One week after the completion of the EPM test, stress-induced escape behavior was assessed using the Porsolt swim test [25], with mice placed into a cylinder filled with room temperature water and allowed to swim for 6 minutes. Immobility duration was scored from video by a blinded observer. An additional week later, animals were tested for cued fear conditioning according to the protocol of Pattwell et al. [26]. Mice were fear conditioned in two chambers (Context A or B) contained within a sound-attenuated chamber and cleaned between each animal with peppermint- or lemon-scented (0.1%) ethanol (70%)(Coulbourn Instruments, Allentown, PA). On day 1, following a 2-min acclimation period to Context A mice were presented with three trials (30-sec ITI) consisting of a 30-sec tone (5 kHz, 70dB) that co-terminated with a 1-sec, 0.7mA foot shock. After the final tone-shock pairing, mice remained in the conditioning chamber for 1 min before being returned to their home cages. On day 2, mice were returned to conditioning chamber (Context A) where freezing behavior was scored during the last 3.5 min of the total 5.5 min spent in the chamber. On day 3, mice were placed into a novel chamber (green cylinder, Context B) and following a 2-min acclimation period, were presented with 3 X 30-sec tones (5 kHz, 70-dB, 30sec ITI). Freezing behavior was scored during each of the 30-sec tone presentations.

Quantification of DNA levels.

Quantification of genomic DNA levels was performed using a qPCR approach as described previously [13]. Briefly, using primers for the WT and floxed TNF alleles, 5′-TGAGTCTGTCTTAACTAACC-3’ and 5′-CCCTTCATTCTCAAGGCACA-3’ [23], levels of Tnf allele expression were normalized to levels expressed in Cre- controls.

Tissue harvest for brain TNF expression

Sexually naïve C57Bl6/J females were housed with males and checked daily for vaginal plugs (around 7 am). Plugs were noted daily, males were removed after 4 days and pregnancy was confirmed post-mortem. Control females were housed without males. Bilateral frontal cortex, hippocampus, hypothalamus, and striatum were harvested at E7 and E15 from pregnant (vaginal plug and post-mortem confirmation) and non-mated control female mice (n=6/group). Samples from one hemisphere were placed into RNAlater (ThermoFisher Scientific) for subsequent RNA analysis and stored at 4°C. Samples from the other hemisphere were placed into 200uL lysis buffer (ice-cold PBS and complete protease inhibitor cocktail) and immediately homogenized at 4°C. Samples were spun down at 4°C for 17 min at 13,000 rpm and the supernatant was stored at −80°C.

qPCR.

Total RNA was extracted using standard Trizol (ThermoFisher Scientific) protocol, then reverse transcribed into cDNA using Verso cDNA Synthesis Kit (ThermoFisher Scientific). Real-time PCR was performed in triplicate using iTaq Universal SYBRgreen Supermix (BioRad) and gene expression analyzed using the ΔΔcT method, normalizing expression of the average TNF cT from two primer sets to the geometric mean of GAPDH and L13alpha. Cycling conditions: denaturation for 30sec at 94°C, annealing for 30sec at 59°C, extension for 45sec at 72°C with 35 cycles for all except negative control samples (45 cycles).

ELISA

Protein samples were thawed on ice and diluted 1:50 (for BCA protein analysis, Micro BCA Protein Analysis Kit, ThermoFisher Scientific) or 1:8 (for ELISA, Mouse TNF-α High Sensitivity ELISA, ThermoFisher Scientific) in ice-cold ddH20. Total protein and TNF-α quantification were performed in duplicates following manufacturers’ protocols. TNF-α levels were calculated as pg/mg total protein, then normalized to non-pregnant control group.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data are shown as means ± SE with n values listed in figure legends. At least 3 litters/group were used for all behavioral assays and variability within and between litters was found to be similar across test groups (Levene’s test for homogeneity of variance). One or two way ANOVA or t-test was used in the analyses to compare groups. LSD or Tukey HSD posthoc analyses were used to assess statistical significance. Differences between groups were considered to be significant when P<0.05. Principal Component Analysis (PCA) plots were done using the PCA function in FactoMineR package in R. The input included the Frequency, Duration, and Distance endpoints for both Open Field (OF) and Elevated Plus Maze (EPM). Missing values (i.e. animals that were involved in one test, but not the other) were substituted with the average of all animals in a group. Coordinates for the 95% confidence intervals were also computed around the barycenter of individuals using coord.ellipse function in FactoMineR.

KEY RESOURCES TABLE

Tumor necrosis factor α	Forward 1	5’-GCCTCTTCTCATTCCTGCTTG-3’
	Reverse 1 (114bp)	5’-CTGATGAGAGGGAGGCCATT-3’
	Forward 2	5’- CCACCACGCTCTTCTGTCTA −3’
	Reverse 2 (102bp)	5’-AGGGTCTGGGCCATAGAACT −3’
L13alpha	Forward	5’-CACTCTGGAGGAGAAACGGAAGG-3’
	Reverse (181bp)	5’-GCAGGCATGAGGCAAACAGTC −3’
GAPDH	Forward	5’-ACCACAGTCCATGCCATCAC-3’
	Reverse (450bp)	5’- TCCACCACCCTGTTGCTGTA −3’

Highlights.

• Dams with a deficit in TNF program their offspring to exhibit low innate fear

• Programming is mediated not by hematopoietic but by brain TNF during gestation

• Brain TNF is known to be down- and upregulated by physical activity and stress

• Brain TNF may report environmental conditions to program offspring fear