Brain-derived neurotrophic factor from microglia regulates neuronal development in the medial prefrontal cortex and its associated social behavior

Abstract

Microglia and brain-derived neurotrophic factor (BDNF) are essential for the neuroplasticity that characterizes critical developmental periods. The experience-dependent development of social behaviors-associated with the medial prefrontal cortex (mPFC)-has a critical period during the juvenile period in mice. However, whether microglia and BDNF affect social development remains unclear. Herein, we aimed to elucidate the effects of microglia-derived BDNF on social behaviors and mPFC development. Mice that underwent social isolation during p21-p35 had increased Bdnf in the microglia accompanied by reduced adulthood sociability. Additionally, transgenic mice overexpressing microglial Bdnf-regulated using doxycycline at different time points-underwent behavioral, electrophysiological, and gene expression analyses. In these mice, long-term overexpression of microglial BDNF impaired sociability and excessive mPFC inhibitory neuronal circuit activity. However, administering doxycycline to normalize BDNF from p21 normalized sociability and electrophysiological function in the mPFC, whereas normalizing BDNF from later ages (p45-p50) did not normalize electrophysiological abnormalities in the mPFC, despite the improved sociability. To evaluate the possible role of BDNF in human sociability, we analyzed the relationship between adverse childhood experiences and BDNF expression in human macrophages, a possible proxy for microglia. Results show that adverse childhood experiences positively correlated with BDNF expression in M2 but not M1 macrophages. In summary, our study demonstrated the influence of microglial BDNF on the development of experience-dependent social behaviors in mice, emphasizing its specific impact on the maturation of mPFC function, particularly during the juvenile period. Furthermore, our results propose a translational implication by suggesting a potential link between BDNF secretion from macrophages and childhood experiences in humans.

Fig. 1. Juvenile social isolation reduces sociability and increases MG-Bdnf expression.

a Schema of the group-housed (GH) and juvenile social isolation (j-SI) mice time series. The experiments started at p56. b j-SI mice displayed reduced sociability in the three-chamber sociability test compared with GH mice. j-SI mice had a lower social interaction score than GH mice (t₍₂₆₎ = 3.948, p = 0.0005, unpaired two-tailed Student’s t test, GH: n = 12, j-SI: n = 16) and spent less time near novel mice (F_1,26 (interaction) = 17.16, p = 0.0003, two-way ANOVA followed by Bonferroni post hoc test, Bonferroni post hoc analysis of GH S vs. j-SI S, p < 0.0001 GH: n = 12, j-SI: n = 16). S, social; O, object. c (Left) Bdnf expression measured using real-time quantitative polymerase chain reaction (RT-qPCR) in microglia recovered from the cortex was higher in the j-SI mice than in the GH mice (t₍₂₂₎ = 4.876, p < 0.0001, unpaired two-tailed Student’s t test, GH: n = 12, j-SI: n = 12). (Right) Bdnf expression measured using RT-qPCR in microglia recovered from the medial prefrontal cortex (mPFC) was higher in j-SI than in GH mice (t₍₆₎ = 2.818, p = 0.0304, unpaired two-tailed Student’s t test, GH: n = 4 (from 16 mice), j-SI: n = 4 (from 16 mice)). The value was log-transformed with a base of 10 because of the right-skewed distribution. Each dot indicates the expression of Bdnf in microglia collected from the mPFC of four mice. d (Left) Schema of the GH and j-SI mice time series. The experiments started at P35. (right) Bdnf expression measured using RT-qPCR in microglia recovered from the cortex at P35 was higher in the j-SI mice than in the GH mice (t(14) = 4.353, p = 0.0007, unpaired two-tailed Student’s t test, GH: n = 8, j-SI: n = 8). *p  <  0.05, ***p  <  0.001, ****p  <  0.0001. Data are presented as the mean ± SEM. 2 M: two months of age, A.U.: arbitrary unit.

Fig. 2. Overexpression of MG-BDNF leads to impaired sociability in adulthood.

a Diagram of the generation of MG-Bdnf-overexpressing mice. The tet-off system was activated in double transgenic mice (F1), which were obtained by crossing Iba1-tTA transgenic mice with Bdnf-tetO knock-in mice. Iba1-tTA(+)::Bdnf^(tetO/+) mice were the MG-Bdnf-overexpressing mice, and Bdnf^(tetO/+) mice were the control mice. b BDNF expression in microglia recovered from the cortex, as measured using ELISA. Iba1-tTA(+)::Bdnf^(tetO/+) mice had higher BDNF expression than Bdnf^(tetO/+) mice; this was normalized from day 5 using doxycycline (DOX) (The Kruskal-Wallis test, 9.086, p = 0.0118; n = 4 per group). c Schema of the behavioral experiments without doxycycline. The experiments were started at P61. d The Iba1-tTA(+)::Bdnf^(tetO/+) mice were less social in the three-chamber sociability test than the Bdnf^(tetO/+) mice. The Iba1-tTA(+)::Bdnf^(tetO/+) mice had a lower social interaction score than Bdnf^(tetO/+) mice (U = 47, p = 0.0226, Mann–Whitney U test, Bdnf^(tetO/+): n = 12, Iba1-tTA(+)::Bdnf^(tetO/+): n = 16) (left) and spent less time near novel mice (F_1,26(interaction) = 14.63, p = 0.0007, two-way ANOVA followed by Bonferroni post hoc test; Bonferroni post hoc analysis of Iba1-tTA(+)::Bdnf^(tetO/+) S vs. Bdnf^(tetO/+) S, p = 0.0001, Bdnf^(tetO/+): n = 12, Iba1-tTA(+)::Bdnf^(tetO/+): n = 16) (right). S, social; O, object. e Schematic of the Augmented Reality-based Long-term Animal Behavior Observing system (AR-LABO). Three novel C57BL/6 J mice were placed with a target mouse and observed for 1 h. f (Left) The cumulative number of approaches per 5-min bin for Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice. Iba1-tTA(+)::Bdnf^(tetO/+) mice accumulates significantly less cumulative approaches to other mice than Bdnf^(tetO/+) mice (two-way RM ANOVA, phenotype (control or Iba1-BDNF) × time (5 min bin) interaction F_(12,312)  =  3.926, p  < 0.0001; effect of phenotype F_(1,26)  =  6.670, p  =  0.0158; effect of time F_(12,312)  =  39.31, p  < 0.0001, Bdnf^(tetO/+): n = 15, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13). (Right) Iba1-tTA(+)::Bdnf^(tetO/+) mice approached other mice less than the Bdnf^(tetO/+) mice (t₍₂₆₎ = 2.717, p = 0.0116, unpaired two-tailed Student’s t test, Bdnf^(tetO/+): n = 15, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13). g The Iba1-tTA(+)::Bdnf^(tetO/+) mice were less approached by other mice than Bdnf^(tetO/+) mice (t₍₂₆₎ = 2.120, p = 0.0437, unpaired two-tailed Student’s t test, Bdnf^(tetO/+): n = 15, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13). h No differences in activity were observed during AR-LABO between the Iba1-tTA(+)::Bdnf^(tetO/+) and Bdnf^(tetO/+) mice (t₍₂₆₎ = 0.8059, p = 0.4276, unpaired two-tailed Student’s t test, Bdnf^(tetO/+): n = 15, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13). *p  <  0.05, ***p  <  0.001. Data are presented as the mean ± SEM. 2 M: two months of age, control: Bdnf^(tetO/+) mice, Iba1-Bdnf: Iba1-tTA(+)::Bdnf^(tetO/+) mice.

Fig. 3. MG-BDNF overexpression affects the inhibitory synaptic inputs to mPFC pyramidal cells and the excitability of such cells.

a Electrophysiological analysis of mPFC layer V pyramidal cells in adulthood without doxycycline administration. The experiments were started at P62. b Representative traces recorded from mPFC pyramidal cells at 200-pA injection. c Iba1-tTA(+)::Bdnf^(tetO/+) mice had lower spike frequency (U = 93.50, p < 0.0001, Mann–Whitney U test) (left) and amplitude (U = 131, p = 0.0026, Mann–Whitney U test) (middle) than Bdnf^(tetO/+) mice at 200-pA injection. No significant differences existed in spike threshold (t₍₄₆₎ = 1.997, p = 0.0518, unpaired two-tailed Student’s t test) (right). (n  =  18 cells from three biologically independent Bdnf^(tetO/+) mice, n  =  30 cells from four biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice). d Representative traces of spontaneous excitatory postsynaptic currents (sEPSCs). e (Left) The Iba1-tTA(+)::Bdnf^(tetO/+) mice had lower sEPSC frequency than Bdnf^(tetO/+) mice (U = 130, p = 0.0436, Mann–Whitney U test). (Right) No significant difference existed in sEPSC amplitude (U = 185, p = 0.5761, Mann–Whitney U test). f Representative traces of spontaneous inhibitory postsynaptic currents (IPSCs). g (Left) Iba1-tTA(+)::Bdnf^(tetO/+) mice showed increased sIPSC frequency compared with Bdnf^(tetO/+) mice (U = 85, p = 0.0010, Mann–Whitney U test). (Right) There was no significant difference in sIPSC amplitude (t₍₃₉₎ = 1.440, p = 0.1579, unpaired two-tailed Student’s t test). e, g n  =  18 cells from three biologically independent Bdnf^(tetO/+) mice, n  =  23 cells from four biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice. h Representative traces of miniature EPSCs (mEPSCs). i (Left) Iba1-tTA(+)::Bdnf^(tetO/+) mice had lower mEPSC frequency than Bdnf^(tetO/+) mice. (t₍₃₅₎ = 3.949, p = 0.0004, unpaired two-tailed Student’s t test). (Right) There was no significant difference in mEPSC amplitude (U = 158, p = 0.7074, Mann–Whitney U test). j Representative traces of miniature IPSCs (mIPSCs). (k) The Iba1-tTA(+)::Bdnf^(tetO/+) mice had increased mIPSC frequency (U = 53, p = 0.0002, Mann–Whitney U test) (left) and amplitude (U = 101, p = 0.0335, Mann–Whitney U test) (right) than the Bdnf^(tetI/+) mice. i, k n  =  18 cells from three biologically independent Bdnf^(tetO/+) mice, n  = 19 cells from three biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. Data are presented as the mean ± SEM. 2 M: two months of age, control: Bdnf^(tetO/+) mice, Iba1-Bdnf: Iba1-tTA(+)::Bdnf^(tetO/+) mice.

Fig. 4. MG-BDNF overexpression affects the complement system.

a RNA-seq analysis of the mPFC in adult Bdnf^(tetO/+) (n = 4) and Iba1-tTA(+)::Bdnf^(tetO/+) (n = 4) mice without doxycycline. The experiments were started at P64. b Principal component analysis revealing gene expression differences between the Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice. c Volcano plot of differentially expressed genes (DEGs). The thresholds are log₂ fold change >1.5 and p < 0.05. d Heatmap and hierarchical clustering of DEGs between Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice. The thresholds are log₂ fold change >1.5 and p < 0.05. Gene expression levels are indicated in the heatmap by the Z-scores in the legend. e The Gene Ontology analysis of down-regulated DEGs in the Iba1-tTA(+)::Bdnf^(tetO/+) mice compared with Bdnf^(tetO/+) mice suggested the involvement of the Wnt signaling pathway (Wnt-activated receptor pathway, p = 0.0006; Wnt-protein binding, p = 0.0013) and complement component C3a receptor activity (p = 0.0018) in molecular function. f Differences in expression levels of selected complement genes in Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice in RNA-Seq analysis. Gene expression levels are presented in the heatmap by the Z-scores in the legend. 2 M: two months of age, control: Bdnf^(tetO/+) mice, Iba1-Bdnf: Iba1-tTA(+)::Bdnf^(tetO/+) mice.

Fig. 5. Normalizing MG-BDNF during the juvenile period does not impair sociability or abnormal inhibitory inputs in the mPFC.

a–l The Iba1-tTA(+)::Bdnf^(tetO/+) mice were administered doxycycline from p21 at weaning to normalize MG-BDNF. Bdnf^(tetO/+) mice were also administered doxycycline from p21 as the control. Behavioral and electrophysiological experiments were started at p61. b Normalizing MG-BDNF from p21 did not reduce sociability in adult Iba1-tTA(+)::Bdnf^(tetO/+) mice in the three-chamber social test. Both groups had no differences in the social interaction score (U = 57, p = 0.2701, Mann–Whitney U test, Bdnf^(tetO/+): n = 12, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13) (left) or social investigation time (F_1,23(interaction) = 2.626, p = 0.1187, two-way ANOVA, Bdnf^(tetO/+): n = 12, Iba1-tTA(+)::Bdnf^(tetO/+): n = 13) (right). S, social; O, object. c Representative traces recorded from mPFC pyramidal cells at 200-pA injection. d No differences existed between Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice treated with doxycycline from p21 in the spike frequency (U = 110, p = 0.5192, Mann–Whitney U test) (left), spike amplitude (U = 120, p = 0.7944, Mann–Whitney U test) (middle), or threshold (t₍₃₀₎ = 2.026, p = 0.0517, unpaired two-tailed Student’s t test) (right) at 200-pA injection (n  =  17 cells from three biologically independent Bdnf^(tetO/+) mice, n  =  15 cells from four biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice). e Representative traces of spontaneous EPSCs. f No differences were observed between Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice treated with doxycycline from p21 in sEPSC frequency (U = 85, p = 0.2671, Mann–Whitney U test) (left) or amplitude (U = 89, p = 0.3453, Mann–Whitney U test) (right). g Representative traces of spontaneous IPSCs. (h) Normalizing MG-BDNF from p21 did not increase sIPSC frequency (U = 107, p = 0.8381, Mann–Whitney U test) (left) or amplitude (U = 69, p = 0.0742, Mann–Whitney U test) (right) in Iba1-tTA(+)::Bdnf^(tetO/+) mice. f, h n  =  15 cells from five biologically independent Bdnf^(tetO/+) mice, n  =  15 cells from three biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice. i Representative traces of miniature EPSCs. j No differences existed between Bdnf^(tetO/+) and Iba1-tTA(+)::Bdnf^(tetO/+) mice treated with doxycycline from p21 in mEPSC frequency (t₍₂₈₎ = 1.988, p = 0.0566, unpaired two-tailed Student’s t test) (left) or amplitude (t₍₂₈₎ = 1.729, p = 0.0948, unpaired two-tailed Student’s t test) (right). k Representative traces of miniature IPSCs. l Normalizing MG-BDNF from p21 did not increase mIPSC frequency (t₍₂₈₎ = 0.01783, p = 0.9859, unpaired two-tailed Student’s t test) (left) or amplitude (U = 76, p = 0.1370, Mann–Whitney U test) (right) in Iba1-tTA(+)::Bdnf^(tetO/+) mice. j, l n  =  15 cells from five biologically independent Bdnf^(tetO/+) mice, n  =  15 cells from three biologically independent Iba1-tTA(+)::Bdnf^(tetO/+) mice. Data are presented as the mean ± SEM. 2 M: two months of age, control: Bdnf^(tetO/+) mice, Iba1-Bdnf: Iba1-tTA(+)::Bdnf^(tetO/+) mice.

Fig. 6. Correlation between human macrophage BDNF expression and adverse childhood experiences.

(Left) No correlation existed between the total Child Abuse and Trauma Scale (CATS) score and BDNF expression in M1 macrophages (rs=0.0694, p = 0.6746, Spearman’s rank correlation, N = 39). (right) A significant correlation existed between the total CATS score and BDNF expression in M2 macrophages (rs=0.4088, p = 0.0098, Spearman’s rank correlation, N = 39). rs; Spearman’s rank correlation coefficient. The false discovery rate was controlled using the Benjamini–Hochberg method to adjust for multiple comparisons in the CATS total score and Sub-item scores; values with q = 0.033 and p < 0.033 were considered significant. **p  <  0.01.