(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Microglial activity during postnatal development is required for infantile amnesia in mice [1]

['Erika Stewart', 'School Of Biochemistry', 'Immunology', 'Trinity College Dublin', 'Dublin', 'Trinity College Institute For Neuroscience', 'Louisa G. Zielke', 'Antje R. De Boer', 'Gabrielle Guillaume Boulaire', 'Sarah D. Power']

Date: 2026-01 

Infantile amnesia, the inability to recall episodic memories formed during early childhood, is a hallmark of postnatal brain development. Yet the underlying mechanisms remain poorly understood. This work aimed to gain a better mechanistic understanding of infantile amnesia. Microglia, specialized macrophages of the central nervous system, are known to play an important role in synaptic refinement during postnatal development and have recently been implicated in memory-related functions. Using mouse models, we identified microglia as key regulators of memory accessibility in infancy. We profiled dynamic changes in microglial morphology across the postnatal window that paralleled the onset of infantile forgetting. We found that pharmacological inhibition of microglial activity during a specific postnatal window prevents infantile amnesia for a contextual fear memory, implicating microglia as active modulators of infant memory persistence. Using activity-dependent tagging of infant encoded engram cells, we demonstrated that microglial inhibition alters engram size and engram reactivation in the amygdala and results in changes in microglia–engram cell interactions. Furthermore, we characterized a relationship between microglial dysfunction and the lack of infantile amnesia in maternal immune activation offspring. Together, these findings reveal a novel role for microglia in regulating infant memory retrieval in mice and suggest that microglial dysfunction may contribute to altered memory trajectories in neurodevelopmental disorders.

Funding: This work was funded by the Lister Institute of Preventive Medicine ( https://lister-institute.org.uk/ , Fellowship to T.J.R.), European Research Council ( https://erc.europa.eu/ ; 715968 to T.J.R.), Irish Research Council ( https://research.ie/ ; GOIPG/2020/1294 to E.S.), Research Ireland ( https://www.researchireland.ie/ ; 15/YI/3187 and 24/FFP-A/13397 to T.J.R.), the Jacobs Foundation ( https://jacobsfoundation.org/ ; 2019-1356-2 to T.J.R.), the Canadian Institute for Advanced Research ( https://cifar.ca/) ; CF-0303 to T.J.R.), and the US National Institute of Health ( https://www.nih.gov/ ; 1R01NS121316 to T.J.R.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

In the present study, we examined the role of microglia in infantile amnesia. We first characterized changes in microglial morphology across the infantile window as proxy for changes in microglial activity. We used pharmacological and receptor-specific inhibition approaches and tested whether microglial inhibition alters memory persistence, engram dynamics, and microglia-engram interactions. Finally, we asked whether modulating microglial function in MIA offspring could reinstate infantile amnesia. Our findings revealed a previously unrecognized role for microglia in developmental memory regulation and suggest that microglial dysfunction may contribute to altered memory outcomes in neurodevelopmental disorders in mice.

Human epidemiological studies highlight a link between gestational inflammation and risk of neurodevelopmental disorders [ 14 , 15 ]. Animal models of maternal immune activation (MIA) show altered brain development and behavioral deficits reminiscent of autism-spectrum disorder (ASD) and schizophrenia [ 16 – 20 ]. We previously described a phenomenon in which MIA prevents infantile amnesia in male offspring [ 21 ]. Using this MIA model as a guide, we focused on potential mechanisms that may be mediating this change in infant memory. MIA is known to modulate the activity of microglia in the brains of offspring [ 22 – 24 ]. Thus, we hypothesized that microglia are one of the candidate mediators for the modulation of infant memory.

Memory is believed to be stored in specific cell ensembles known as engrams, that are active at the time of learning and reactivated during recall, to drive memory-specific behaviors [ 10 – 12 ]. Through the use of activity-dependent labeling we can investigate changes in engram cells across the life span [ 13 ]. However, how engram properties evolve during infancy, and what mechanisms render them inaccessible during infantile amnesia, has been scantly explored. By focusing on key developmental processes occurring during this period of early brain development, we aimed to identify potential mediators of infantile amnesia.

Infancy and childhood are developmental periods rich in novel experiences and stimuli as children actively explore the environment and assimilate new information to build internal representations of the world [ 1 ]. Yet, human infants rapidly forget early episodic and contextual memories, a phenomenon known as infantile amnesia [ 2 – 4 ]. This form of rapid forgetting is conserved across species, is well characterized in rodent models, and occurs during a period of robust brain development [ 5 ]. However, while the behavioral features of infantile amnesia have been well characterized, the underlying cellular and molecular mechanisms that drive it remain poorly understood [ 5 – 8 ]. Understanding how the brain regulates memory accessibility during early development is critical for uncovering fundamental principles of memory formation and for identifying mechanisms that may be disrupted in neurodevelopmental disorders [ 9 ]. However, the cellular substrates of infantile forgetting, and how they intersect with key developmental processes, are largely unknown.

Lastly, we evaluated the effect of MIA and early postnatal minocycline treatment on microglial activation ( Fig 4F and 4G ). In accordance with other reports, our histological analysis revealed that MIA results in decreased expression of CD68 in microglial cells suggesting a reduced phagocytic activity [ 23 ] ( Fig 4 ). Interestingly, early postnatal minocycline treatment of MIA offspring resulted in a significantly higher level of CD68 expression at P25 ( Fig 4G ). Together, these data support a model in which MIA induces alterations in microglial function, disrupting the developmental processes that normally give rise to infantile amnesia.

(A) Schematic diagram of experimental schedule. Pregnant dams received injection of Poly(I:C) at E12.5. Minocycline was administered from P0 to P14. Male offspring underwent CFC at P17 and were tested for memory recall 8 days later. (B) Quantification of freezing behavior of mice during recall test. (C–E) ASD phenotyping battery of behaviors. (C) Quantification of distance traveled in 3-chamber-test. (D) Time spent in the object or the mouse zone during test. (E) Quantification of % marbles buried. (F) Representative images of z-stacks and 3D reconstructions of microglial cells from MIA and MIA + Minocycline treated offspring, Scale bar 30 µm (upper panel), 10 µm (lower panel). Sections were stained with Iba1 (red) and CD68 (green). (G) Quantification of CD68 within Iba1+ cells. Data is presented as mean ± SEM. Points represent individual mice. For freezing behavior, N = 9–10 litters/group, n = 22/29 mice/group. For ASD phenotyping, n = 11–14 mice/group. For microglia histology, each point represents individual microglial cell. N = 4 mice/group, n = 8 microglia cells per group. Statistical comparison performed using (B) Student nested t test. (C, E, G) Student t test. (G) Two-way ANOVA with Bonferroni post hoc; n.s P > 0.05, *P < 0.05, **P < 0.01. Details of all statistical comparisons may be found in S1 Data . The data underlying this Figure can be found in S2 Data .

Our findings in Figs 2 and 3 describe a novel condition in which infantile amnesia is prevented through pharmacological inhibition of microglia. A common factor between these conditions and MIA, is their impact on microglia. Substantial evidence from both animal models of MIA and human samples from ASD or schizophrenia cases indicates a disruption to microglial function [ 24 , 53 – 56 ]. In line with other work, we found changes in embryonic microglial gene expression following MIA ( S6 Fig ). We observed changes in expression of some microglial-related genes in embryonic brains following MIA ( S6 Fig ). Inspired by work that targeted microglia to prevent or reverse behavioral deficits in MIA offspring, we then investigated the effect of microglial inhibition on infantile amnesia in our MIA offspring [ 22 , 23 , 57 ]. To test this, we inhibited microglial activity during the early postnatal period in MIA offspring. MIA offspring were treated with minocycline via drinking water from P0 to P14, a window selected to intervene prior to memory acquisition at P17 ( Fig 4A ). Control MIA offspring received regular drinking water. Our logic for targeting microglia in MIA offspring during this early postnatal window is based on previous work that has demonstrated premature shifts in the trajectory of microglia development and phenotype in MIA offspring [ 58 ]. Mice were trained on a CFC paradigm at P17 and tested 8 days later ( Fig 4A ). Minocycline-treated MIA offspring displayed significantly lower freezing levels compared to untreated MIA offspring ( Fig 4B ). Crucially, minocycline treatment had no effect on control offspring treated during the same developmental window of P0-P14 ( S6E and S6F Fig ). This result suggests that early postnatal microglial inhibition was sufficient to restore infantile amnesia in MIA offspring. Following a recall test at P25, mice were then either returned to their homecare to be evaluated in a battery of tests to assess for ASD phenotypic behaviors in adulthood or immediately perfused for histological analysis. Sociability was tested using the 3-chamber social interaction task and repetitive behavior was assessed using the marble burying assay ( Fig 4C – 4E ). Minocycline-treated MIA offspring demonstrated an increased preference for a social stimulus (a novel mouse) compared to a nonsocial object, whereas untreated control MIA offspring showed no preference ( Fig 4D ). We also observed a trend towards a decreased tendency for marble burying in minocycline-treated MIA offspring ( Fig 4E ).

Our previous work characterized a developmental brain state in which infantile amnesia does not occur [ 21 ]. Gestational immune activation through the delivery of a viral-mimetic polynosinic-polycytidylic acid (Poly(I:C)) at embryonic day 12.5 (E12.5) results in male offspring that retain memories acquired in infancy, bypassing the typical forgetting observed during this period [ 21 ]. Here, we also tested the effect of another maternal environmental intervention on infantile amnesia, to determine whether this prevention of infantile amnesia was specific to a timed inflammatory insult at E12.5 or if lower-grade chronic inflammation was sufficient to produce the same effect on infant memory ( S5 Fig ). Maternal HFD had no effect on infantile amnesia, which suggests that there may be a sensitive window for the effects of maternal inflammation ( S5C Fig ).

Microglia are highly sensitive to changes in neuronal activity and two-photon imaging of microglia reveals targeted interactions between microglial processes and dendritic spines [ 52 ]. We hypothesized that microglial-engram interactions may regulate engram accessibility. To test this, we evaluated microglial interactions with engram cells in the AMG following minocycline treatment. Using 3D surface rendering and reconstruction of EYFP+ engram dendrites and Iba1+ microglial cells in the AMG, we quantified the number of direct contacts between microglia processes and engram cells ( Fig 3Z ). We found that Minocycline-treated mice exhibit a significant reduction in the number of microglia-engram contact points ( Fig 3B2 ), suggesting that decreased microglial engagement may facilitate engram stabilization.

Here, we found that minocycline treatment attenuated microglial activation, lead to the persistence of infant fear memory, and increased infant engram reactivation in the AMG. To further investigate the relationship between microglial plasticity and engram function, we quantified microglial morphology in the AMG following CFC and minocycline treatment ( Fig 3T ). Compared to controls, minocycline-treated mice exhibited increased microglial branching, with trends toward increased terminal points and filament length ( Fig 3U – 3W ). To assess how these morphological features related to memory performance and engram activity, we correlated branching complexity with freezing behavior and engram reactivation ( Fig 3X and 3Y ). Both measures were positively associated with microglial branching ( Fig 3Y ), linking reduced microglia activation to enhanced memory recall and engram reactivation.

We quantified engram reactivation in the hippocampal DG, the basolateral amygdala (BLA), the central amygdalar nucleus (CeA), and the retrosplenial cortex ( Figs 3C – 3G and S4 ). For each region, we quantified engram size (% EYFP+ cells), the number of active cells (% c-Fos+ cells), and engram reactivation (% double positive EYFP+ and c-Fos+ cells) at the time of recall in each region. In the DG and RSC, no differences were observed between groups in the number of EYFP+ cells, number of c-Fos+ cells, or overlap ( Fig 3H – 3J , 3Q - 3S ). However, in the BLA and CeA, minocycline-treated mice exhibited significantly increased engram reactivation compared to controls ( Figs 3M , 3P , and S4D ), consistent with our previous findings that artificial reactivation in the DG of an infant encoded engram post-amnesia also results in increased BLA activation ( Figs 3M and S4 ) [ 21 ]. Unexpectedly, we also found a difference in the number of engram cells in the BLA of minocycline-treated mice ( Fig 3K ). As a control experiment, we also evaluated engram reactivation following minocycline treatment in mice that received CFC (shock) and those that received only a contextual experience without shock (NS) ( S4 Fig ). We observed significant differences in levels of reactivation in both the DG and BLA of S and NS minocycline-treated mice ( S4 Fig ).

(A) Diagram of tamoxifen-driven inducible genetic strategy for engram cell tagging. (B) Schematic diagram of experimental schedule. Mice were administered 50 mg/kg minocycline from P16-P25 through drinking water. Mice were trained at P17 and underwent a recall test 8 days later and were scarified 45 min following recall. (C) Created in BioRender. Stewart, E. (2025) https://BioRender.com/gq535e6 . (D) Representative images of engram cell labeling (Green) and c-Fos+ cells (Red) and overlap in the DG, BLA, CeA, and RSC. Arrowhead indicates an example of overlapping cells. Large-scale bar, 150 µm. Small-scale bar, 75 µm. (H–J) Quantification of % Engram cells, % c-Fos, and engram overlap/reactivation in DG, (K–M) BLA, (N–P) CeA, and (Q–S) RSC. N = 4–6 litters/group, n = 4 slices per animal. (T) Representative images of Iba1+ microglial cells in the BLA and 3D reconstruction of individual microglial cells using IMARIS software. Scale bar 20 µm (left panel) 15 µm (middle and right panel). Created in BioRender. Stewart, E. (2025) https://BioRender.com/40c19e9 . (U–W) Quantification of measures of morphological plasticity in microglia between control and minocycline-treated mice. (X) Quantification of freezing behavior during P25 recall test in T2-Ai32 mice treated with either minocycline or vehicle. (Y) Scatter-plot of the relationship between microglial branching and engram reactivation or freezing behavior. (Z) Representative image of microglia-engram interaction. Scale bar, 20 µm (top panel) 5 µm (lower pane). (A2, B2) Quantification of % interaction between microglial and engram surfaces. Syringe symbol represents 4-OHT injection. Black lightning symbol represents foot-shocks. Data is presented as mean ± SEM. For engram quantification, each point represents individual litter. Counts per slice pooled per litter for statistical comparisons. For microglial quantification, n = 6–8 mice, 8 microglia/mouse. For interaction analysis, 6–10 mice/group, 2/3 sections per mouse. (U–W) Statistical comparison performed using (H–S) Nested t test, (U–W) Student t test, (X–Y) Pearsons correlation; n.s P > 0.05, *P < 0.05. Details of all statistical comparisons may be found in S1 Data . The data underlying this Figure can be found in S2 Data .

Given that microglial inhibition preserved memory retention in infant mice past the onset of infantile amnesia, we next asked whether these behavioral effects were associated with changes in memory engram activity. We utilized a transgenic Cre-based engram tagging strategy to label infant engram cells and evaluate engram dynamics. We crossed a Targeted Recombination In Active Populations (TRAP2) mouse, in which the c-fos promoter drives expression of tamoxifen-inducible Cre recombinase (iCre), with an Ai32 reporter line expressing channelrhodopsin/enhance yellow fluorescent protein (Chr2-EYFP) [ 50 , 51 ]. This allowed for permanent tagging of infant engram cells with EYFP ( Fig 3A ). We hypothesized that microglial inhibition would lead to increased infant engram reactivation. To test this, we administered minocycline in drinking water while controls received regular drinking water. Mice underwent CFC at P17, receiving an injection of 4-OHT immediately following training to label active cells ( Fig 3B ). Mice were tested 8 days later and perfused 45 min following a recall test for tissue collection ( Fig 3B ).

To substantiate the above results, we used a more specific inhibitor of microglial-neuronal communication. Microglial activity and interaction with neurons are modulated by numerous factors, including soluble factors known as chemokines [ 48 ]. The chemokine CX3CL1 (fractalkine) is expressed by neurons and binds to its target receptor CX3CR1 that is expressed by microglia and is an important mediator of neuronal-glial communication that has been reported to play a crucial role in postnatal brain development [ 29 , 49 ]. We sought to investigate the potential role of CX3CL1-CX3CR1 signaling on infantile amnesia, using JMS-17-2, a potent and selective antagonist of CX3CR1 [ 44 ]. Infant mice were treated with JMS-17-2 or control through daily i.p injections from P16 to P25 ( Fig 2F ). Mice were trained on P17 in Context A and tested 8 days later at P25 ( Fig 2G ). Mice treated with JMS-17-2 displayed significantly higher levels of freezing 8 days post-training compared to control-treated mice ( Fig 2G ). To confirm that JMS-17-2 by itself or repeated daily i.p injections did not cause increased generalized anxiety or freezing, we compared mice that received CFC and JMS-17-2 (JMS-S) treatment with mice that received only a neutral contextual exposure with no foot-shock but also received JMS-17-2 treatment (JMS-NS) ( S1J Fig ). The lack of freezing in the JMS-NS group shown in S1L Fig indicates that freezing is due to retention of the fear memory. These results support our hypothesis that microglial activity throughout postnatal development may contribute to infantile amnesia potentially mediated by CX3CL1-CX3CR1 signaling.

To confirm that pharmacological treatment with minocycline was indeed influencing microglial activation, mice were euthanized and perfused following recall 8 days post-training and brain tissue was collected. Tissue was stained for Iba1 and CD68 ( Fig 2D ; top). We carried out 3D surface rendering of microglial cells and CD68 protein expression within the DG ( Fig 2D ; bottom). We found significantly less CD68 expression within Iba1+ cells in the hippocampus of subjects treated with minocycline compared to controls ( Fig 2E ).

(A) Schematic diagram experimental schedule. Mice were administered minocycline (50 mg/kg) through drinking water from P16-P25. (B) Schematic of behavioral schedule. Male and female mice were trained at P17 and underwent recall test 1 and 8 days post-training. (C) Quantification of freezing behavior of minocycline or control mice 1 and 8 days post-training. (D) Representative image of 3D rendering of CD68 expression within microglia. Scale bar 30 µm (upper panel) 10 µm (lower panel). (E) Quantification of CD68 volume within Iba1+ cells. (F) Schematic diagram experimental schedule. Mice were administered JMS-17-2 (10 mg/kg) through i.p injection from P16-P25. (G) Quantification of freezing behavior of control of JMS-treated mice 8 days post-training. Black lightening symbol represents foot-shocks. Males are represented as circle symbol and females as triangle symbol on graphs. Data is presented as mean ± SEM. (C) 21 mice per group. (E) Each point represents an individual mouse. n = 5–6 mice/group. 4–5 microglia were analyzed from 2 ROI images per mouse. Individual cells pooled per mouse for statistical comparisons. (G) n = 8/9 mice per group. Statistical comparison performed using (C) RM Two-way Anova with Bonferroni (E, G) Students t test *P < 0.05, **P < 0.01. Details of all statistical comparisons may be found in S1 Data . The data underlying this Figure can be found in S2 Data .

Given the changes we observed in microglial morphology and activation state across the infantile window ( Fig 1 ), in addition to their well-described developmental roles and their implication in memory modulation, we hypothesized that microglial activity drives infantile amnesia. To test this, we treated mice postnatally with minocycline hydrochloride, a second-generation tetracycline antibiotic widely used to inhibit microglial activity [ 41 – 44 ]. Minocycline was administered either via drinking water from P16 to P25 ( Fig 2A ) or through daily intraperitoneal (i.p) injections ( S1A Fig ). Control groups received regular drinking water or saline injections, respectively. Mice were trained at P17 on a CFC paradigm in context A and tested for memory recall 1 day later (P18) in the same context to confirm memory formation and 8 days later (P25) to confirm infantile amnesia ( Figs 2B and S1B ). Both control and minocycline-treated mice displayed equivalent levels of freezing 1-day post-training demonstrating intact fear memory ( Figs 2C and S1C ). However, 8 days post-training, the minocycline group displayed significantly higher levels of freezing compared to control-treated group ( Figs 2C and S1C ). The high level of freezing exhibited by the minocycline-treated group at P25 indicates preserved fear memory despite the typical onset of infantile amnesia, suggesting that microglial inhibition prevents infantile amnesia. To verify memory specificity, we tested mice in novel context B. Minocycline-treated mice did not exhibit higher levels of freezing compared to controls indicating context-specific memory recall rather than a generalized freezing phenotype resulting from minocycline treatment ( S1C Fig ). As an additional control, we demonstrated that minocycline-treated mice, that did not receive a foot shock during training do not freeze in Context A, again confirming that this freezing behavior is as a result of fear memory recall ( S1F Fig ). We did not observe any effect of sex on the effect of minocycline treatment on infantile amnesia ( S1I Fig ). Both male and female mice treated with minocycline showed significantly higher freezing levels 8 days post-training compared to controls ( S1I Fig ).

We stained brain sections for a well-established marker of microglia, ionized calcium-binding adapter molecule 1 (Iba1) ( Fig 1F and 1G ; top). We performed surface rendering and 3D reconstruction of Iba1+ labeled cells in the DG and AMG and we quantified numerous morphometric parameters of microglia at these different time points ( Fig 1F , and 1G ; bottom). In the DG, we observed a significant difference in microglial branching and filament length between 3 (P20) and 8 days (P25) ( Fig 1H – 1J ). In the AMG, we observed significant differences in branching, the number of terminal points and filament length between 3 (P20) and 5 days (P22) ( Fig 1K – 1M ). In adults, measures of microglial morphology in DG remained consistent ( Fig 1N – 1P ). Interestingly, in the adult AMG, we observed significant differences in the number of branch points and terminal points between the 3 and 8-day recall group suggesting some changes in microglia following adult recall at 8 days post-training ( Fig 1Q – 1S ). The observed changes in microglial morphology follows a pattern of change that mirrors memory retention in both infants and adults. We also note that the differences in quantity of morphological changes between adults and infants likely reflect the relative levels of cell maturity [ 34 , 40 ]. To further explore changes in microglial state across the infantile window, we quantified the expression of CD68, a marker of phagolysosomes often used as a proxy of microglial activity and phagocytosis, at 3, 5, and 8 d post-CFC ( Fig 1T ). We observed changes in microglial CD68 expression within microglial cells that also mirrored infant memory retention ( Fig 1U and 1V ). These findings highlight microglia as a prime candidate for further mechanistic investigation in the context of infantile amnesia and memory.

(A) Schematic diagram of experimental schedule. Infant mice were trained on P17 on a CFC task and underwent a recall test either 3, 5, or 8 days later before being sacrificed for tissue collection. Adult mice underwent the same CFC and underwent a recall test either 3, 5, or 8 days later before being sacrificed for tissue collection. (B) Quantification of initial shock responses during CFC. (C, D) Quantification of freezing behavior during recall tests. ( E) Iba1+ cells in DG and AMG. Created in BioRender. Stewart, E. (2025) https://BioRender.com/hf3ttgz . (F, G) Representative images of Iba1+ stained images and 3D filament reconstructions of microglia. Scale bar, 20 µm. (H–M) Quantification of microglial morphology in infant mice. (N–S) Quantification of morphological measures in adult mice. (T) Representative image IBA1 and C68 staining and 3D rendering of CD68 expression within microglia. Scale bar 30 µm (upper panel) and 20 µm (lower panel). (U, V) Quantification of CD68 volume within Iba1+ cells in infants DG and AMG. Black lightning symbol represents foot-shocks. Data is presented as mean ± SEM. Each data point represents individual mice. n = 4–11 mice/group. 2 region of interest (ROI) images were analyzed per mouse from either the DG molecular layer or BLA. 3–5 microglia were analyzed per 2 ROI image. Individual cells were pooled per mouse for statistical comparisons. One-way ANOVA with Tukeys post hoc. n.s P > 0.05, *P < 0.05, **P < 0.01. Details of all statistical comparisons may be found in S1 Data . The data underlying this Figure can be found in S2 Data .

To assess memory retention in mice we employed a classical Pavlovian contextual fear conditioning (CFC) paradigm ( Fig 1A ). Infant mice were trained on postnatal day 17 (P17) and tested for recall in the same context either 3, 5, or 8 days later on P20, P22, and P25, respectively ( Fig 1A ). Adult mice were trained between 8 and 9 weeks of age (P56) and similarly underwent a recall test either 3, 5, or 8 days later ( Fig 1A ). We first quantified the baseline and shock responses of both infant and adult mice during training ( Fig 1B ). Infant and adult mice showed equivalent baseline levels of freezing, but infants demonstrated a higher response to the first shock with overall comparable levels between ages ( Fig 1B ). We then quantified freezing behavior during recall as a readout of fear memory ( Fig 1C and 1D ). Consistent with previous findings, infant mice exhibited a graded decrease in the levels of freezing across this developmental window, with high levels of freezing observed at both P20, reduced levels at P22 before the onset of amnesia at P25 ( Fig 1C ) [ 8 ]. In contrast, adult mice displayed stable memory retention demonstrated by high levels of freezing at all time points ( Fig 1D ). To investigate morphological changes, subjects were sacrificed, perfused, and brains were collected for histological analysis ( Fig 1E ). Morphometric profiling of microglia has been classically used as a proxy of microglial activation state in addition to maturation [ 34 , 35 ]. Here, we characterized microglial morphology across the infantile amnesia window following CFC to identify changes in microglia that may coincide with the onset of infantile amnesia. We focused specifically on two brain regions of interest; the hippocampal dentate gyrus (DG) which plays a critical role in supporting the encoding and retrieval of contextual information [ 36 ] and the amygdala (AMG); an integrative hub that plays a key role in processing emotional information [ 37 , 38 ]. Both areas are recruited during contextual fear memory formation [ 39 ].

Discussion

The findings presented in this study demonstrate clear evidence for a role for microglia in infantile amnesia in mice. We characterized microglia morphology throughout the developmental window of interest and found that microglia exhibit changes in morphology and CD68 expression across this window of interest suggesting changes in their activity and/or function. Through pharmacological inhibition of microglial activity, we have shown that microglial activity is necessary for infantile amnesia of a contextual fear memory to occur, and microglia may modulate engram activity through the microglia-neuronal axis via CX3CL1-CX3CR1 signaling. By tagging infant engram cells during learning, we further illustrate that microglial inhibition results in differences in engram size and reactivation specifically in the AMG. Our data aligns with existing evidence that microglia contribute to decreased fear expression in adult mice [43] but build on these findings by demonstrating a specific role for microglia in the rapid forgetting experienced throughout infancy, and delineating their impact on infant engram dynamics. Additionally, our data indicates that microglial dysfunction may underpin developmental and behavioral changes in MIA offspring and presents original evidence that links microglial activity with our previous published finding on changes in infant memory in MIA offspring [21].

Microglia are the central nervous system representatives of the immune system that mediate immune responses, promote injury repair, and fight invading pathogens [59]. Beyond their well-described immunological roles, microglial actively interact with neurons to modulate neuronal circuits with resulting impacts on cognition and behavior [60–62]. A role for microglial activity has been described in the regulation of numerous forms of behavior, for example, the extinction of anxiety-like behaviors or juvenile social and play behaviors [44,63]. Specifically, the synaptic modulatory functions of microglia position them as ideal plasticity managers [32,33]. Indirect evidence for a microglial role in memory primarily stems from their association with the pathogenesis of aging-related cognitive decline, traumatic brain injury (TBI) and Alzheimer’s disease (AD), in which memory deficits are a defining feature [64–68]. However, recent studies have provided evidence for a direct role for microglia in a time-dependent decline in adult fear memory expression and have detailed specific microglia-engram interactions in the adult mouse brain [43,69]

Here, we provide the first evidence for a role of microglia in infantile amnesia. We show specific developmental changes in microglia activity following fear memory learning (Fig 1). We demonstrate that inhibiting microglial activity not only prevented infantile amnesia behaviorally (Fig 2) but also resulted in changes in engram activity (Fig 3). We observed enhanced reactivation of AMG engram cells in minocycline-treated mice (Fig 3K). These data indicate that changes in microglial activity influences engram formation and reactivation. Interestingly, we also observed a significant difference in engram size in the BLA of mice treated with minocycline compared to controls. Engram size has been shown to vary with learning, with larger engrams supporting enhanced memory precision [70]. Relatedly, we previously reported a similar increase in DG engram size related to memory retrieval in MIA offspring, suggesting a common underlying mechanism [21].

These findings of larger engram underpinning memory retention appear somewhat contradictory to work by [71] who found that engram sparsity, namely in the CA1, is associated with the emergence of precise, adult-like memory. However, this may reflect that the memory retention we observe in minocycline-treated mice may not be due to an acceleration of memory maturation, but instead due to a prevention of the normal developmental trajectory that would otherwise lead to infantile amnesia. Supporting this interpretation, we observed a reduced number of PNNs in the DG and CA1 of minocycline-treated mice compared to controls. In contrast, [71] demonstrated a developmental increase in PNNs underpins the emergence of engram sparsity and memory precision (S2 Fig). Based on these findings, we propose that by inhibiting microglial activity during this developmental window, we prevent the normal trajectory of development that would otherwise lead to infantile amnesia.

Our previous work found that MIA in mice prevents infantile amnesia in male offspring. MIA is characterized by neuroinflammatory changes in the embryonic brain that can alter the trajectory of brain development and can also lead to persistent inflammatory changes in offspring postnatally and beyond into adulthood [20,72]. There is mounting evidence from rodent work that microglial function is perturbed by MIA and other prenatal environmental stressors, and this dysfunction may underpin many of the brain and developmental and behavioral abnormalities reported in MIA offspring [23,24,54,73]. This motivated us to investigate the potential relationship between MIA, microglia, and infantile amnesia given that we have observed changes in infant memory following both MIA or microglial inhibition. We found changes in some microglial-related genes in the embryonic brain following maternal Poly(I:C) injection, most notably a decrease in the microglial innate immune receptor TREM2 that governs microglial phagocytosis (S6 Fig). Interestingly, microglial TREM2 is reduced in human autistic patients and mice lacking TREM2 demonstrate altered brain connectivity and ASD-like behaviors [74]. Several rodent studies have targeted microglia in MIA offspring either through pharmacological inhibition with minocycline or complete depletion and have found that this leads to a prevention or reversal of some ASD behavioral deficits [22,57,75]. Our goal was to target microglia during an earlier postnatal window from P0-P14 (Fig 4A). Matcovitch-Natan and colleagues revealed that discrete developmental stages are underpinned by specific microglia transcriptomic profiles that likely reflect their respective developmental niche and MIA prematurely shifted microglia to a more “adult-like” state [58]. We hypothesized that inhibiting microglia during early postnatal development may realign them to their normal development stage permitting the emergence of normal behaviors such as infantile amnesia. The inhibition followed by removal may also result in a “rebound” effect of microglial activity, in a similar way to what has been suggested for depletion and repopulation that corrects microglial function [57]. In our investigation, we found that this early postnatal minocycline treatment prevented the persistence of infant fear memory in MIA offspring, likely through the modulation of microglia (Fig 4). This view is supported by our histological results which demonstrated that MIA resulted in a blunted microglia phenotype that was reversed by minocycline treatment (Fig 4G). Overall, our data is coherent with well existing evidence highlighting a crucial role for microglia dysfunction in MIA-induced behavioral deficits, including our novel memory phenotype of persistent infant engram expression. This result also strengthens our claim that microglia play an important role in infantile amnesia and additionally demonstrates important overlap between the developmental and cognitive changes in MIA offspring and microglia function. Yet, there is a need for future research to accurately define the phenotype and activity of microglia in MIA offspring during early postnatal development and how this contributes to changes in infant memory.

Given the emerging role of microglia in natural forgetting and the possibility that infantile amnesia represents an accelerated form of this process, it is plausible that microglial modulation of engram cells may act as a general mechanism of modulating engram expression. Under homeostatic conditions, baseline levels of microglial activity may govern natural forgetting, but during periods of elevated microglial activity, such as early postnatal development, may drive infantile amnesia. Furthermore, it is reasonable to suggest that disproportionate microglial activation could also result in maladaptive or pathological forgetting similar to what we observe in AD or TBI, while underactive microglia may give rise to excessive memory recall leading to disorders such as post-traumatic stress disorder or the blunted microglia phenotype in MIA may explain the lack of infantile amnesia in those mice [76,77].

One possible cellular mechanism through which microglia drive infantile amnesia or forgetting in general is direct synaptic modulation through pruning or synaptic remodeling. This idea is supported by reports of complement-driven synaptic elimination underpinning forgetting in adult mice [43]. Microglia engage in high levels of synaptic pruning throughout early postnatal development that is crucial for establishing accurate connectivity, which makes this a plausible mechanism through which microglia mediate infantile amnesia [29,30,78]. Synaptic maturation and elimination also peak around the third postnatal week in mice, which coincides with the boundary of infantile amnesia [31]. We observed a decrease in CD68 expression following minocycline treatment which indicates a decreased level of phagocytic activity in microglia. In further support of synaptic pruning as a mechanism, a disruption in synaptic pruning and increased synaptic density has been reported in MIA offspring [22,23,79]. We previously reported an increase in spine density on DG engram cells of MIA offspring that could possibly be related to a reduction in synaptic pruning by microglia which as a result leads to the persistence of infant memories [21]. In addition to direct synaptic elimination, microglia can interact with and remodel ECM structures such as PNNs, that may indirectly influence synaptic plasticity and neuronal function and subsequently effect memory [80,81]. We found a significant difference in PNNs in DG and CA1 hippocampal subregions between minocycline and control-treated mice at P25 (S2 Fig). PNNs significantly increase and stabilize between P21 and P28 in the hippocampus [47,71]. Importantly, disrupted microglial homeostasis can lead to decrease PNNs, as has been observed in mice with Csfr1 haploinsufficiency [82]. Moreover, a decreased number of PNNs has been observed in MIA offspring [83–85]. A disruption of engram connectivity and memory accessibility may also occur due to alterations in the level of neurogenesis, and notably microglia are known regulators of neurogenesis [6,86–88]. Indeed, the neurogenic hypothesis of infantile amnesia suggests that high levels of postnatal neurogenesis lead to accelerated forgetting and is supported by work demonstrating that inhibiting neurogenesis prevents forgetting in infancy [6,89]. Similarly, an acceleration of neurogenesis in mice increases infant spatial amnesia [90]. This bi-directional effect of neurogenesis makes it a potential synergistic mechanism through which microglia modulate memory. This possibility resonates with findings of disrupted neurogenesis in offspring following prenatal immune challenge [91–93].

Finally, another possibility is that rather than impeding the target engram, microglia facilitate interference of competing engram ensembles. Interference describes a common type of forgetting where competing information impairs encoding or consolidation or retrieval of memory traces [94,95]. The presence of competing engram ensembles has been shown to interfere with original engram activation and memory recall, resulting in forgetting [95–97]. Microglia are tightly attuned to neuronal activity and can directly feedback onto neurons to control activity [33,98]. It is conceivable that microglia also modulate the activity of competing engram ensembles to promote their activation resulting in suppression of the original engram.

A limitation of the present study is the use of minocycline hydrochloride as a primary method of microglial manipulation. However, this was the most effective pharmacological agent as our attempts with Csfr1 inhibitors PLX5622 and BLZ945 at the required developmental stages either did not successfully deplete microglia or required diluents such DMSO which can have negative impacts on brain development and plasticity. We recognize that minocycline does not have specificity for microglia, and we cannot rule out the possible antimicrobial effects on the gut microbiome. However, we demonstrate that maternal and postnatal HFD, which is known to alter gut microbiome composition does not affect infantile amnesia (S5 Fig) [99,100]. Also, inhibition of microglial CX3CR1 results reproduced the same effect on infantile amnesia, giving us confidence in our findings. Additionally, tools for microglial manipulation remain limited, in particular for developmental studies, however, more recent developments in genetic manipulation may permit deeper investigation into the role of microglia in memory and forgetting in the future [101]. We acknowledge the importance of including female mice for sex-balanced experiences. In this case, we did not see any sex differences in our original experiments and chose to continue with males for our MIA questions as the prevention of infantile amnesia in MIA offspring is a male-specific phenotype. Finally, in this investigation, we focused specifically on the role of microglia in infantile amnesia, but we do not discount the possibility of an involvement of other nonneuronal cells such as astrocytes which have also recently been implicated in memory and engram modulation [102].

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3003538

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/