Early-life prefrontal cortex inhibition and early-life stress lead to long-lasting behavioral, transcriptional, and physiological impairments

Abstract

Early-life stress has been linked to multiple neurodevelopmental and neuropsychiatric deficits. Our previous studies have linked maternal presence/absence from the nest in developing rat pups to changes in prefrontal cortex activity. Furthermore, we have shown that these changes are modulated by serotonergic signaling. Here we test whether changes in prefrontal cortex activity during early-life affect the developing cortex leading to behavioral alterations in the adult. We show that inhibiting the prefrontal cortex of mouse pups leads to cognitive deficits in the adult comparable to those seen following maternal separation. Moreover, we show that activating the prefrontal cortex during maternal separation can prevent these behavioral deficits. To test how maternal separation affects the transcriptional profile of the prefrontal cortex we performed single-nucleus RNA sequencing. Maternal separation led to differential gene expression almost exclusively in inhibitory neurons. Among others, we found changes in GABAergic and serotonergic pathways in these interneurons. Interestingly, both maternal separation and early-life prefrontal cortex inhibition led to changes in physiological responses in prefrontal activity to GABAergic and serotonergic antagonists that were similar to the responses of more immature brains. Prefrontal activation during maternal separation prevented these changes. These data point to a crucial role of prefrontal cortex activity during early-life in behavioral expression in adulthood.

INTRODUCTION

Development is a period of heightened plasticity, during which neuronal activity shapes the connectivity of the emergent brain to adapt to the environment it will encounter. It is therefore also a period of heightened vulnerability to factors that can derail adaptive brain development. It is likely not coincidental that many psychiatric disorders are rooted in childhood. Childhood adversity (physical abuse, neglect, food insecurity, violence) is estimated to account for up to 45% of child-onset and 32% of adult-onset mental health disorders ^{1, 2}. Neglect is the most prevalent form of childhood maltreatment in the US [³. Early-life-stress has been associated with increased risk for major depression, bipolar disorder, schizophrenia, and post-traumatic stress disorder, all syndromes associated with cognitive impairments ⁴. Childhood trauma and neglect have also been associated with poor executive control including attention deficits and impulsivity ⁵. In rodents, maternal separation has been shown to affect recognition and location memory and this effect is associated with changes in prefrontal cortex (PFC) networks ⁶.

Experiences lead to enduring changes in physiology and behavior by modulating the organization and strength of synaptic connections and neural circuits. This synaptic plasticity is the basis for learning and memory in the adult but is also essential for the adaptive development of the postnatal brain. The postnatal period is one of high synaptic plasticity when the emerging brain can be shaped by the environment in adaptive or maladaptive ways ⁷. During this period, the main source of environmental input to a pup or infant is the mother or caregiver. Caregiver interactions during early life are thus critical for an individual’s lifelong mental health. However, the mechanisms by which maternal behavior regulates developing brain activity, and how these changes in activity lead to behavioral alterations in the adult offspring are not yet fully understood. Here we test the role of the PFC in this regulation.

The PFC is a critical structure subserving various cognitive functions ^{8, 9} including behavioral flexibility and working memory ^10–12. This brain region has a long developmental period ¹³, with high plasticity into adolescence, allowing it to adapt to environmental changes. However, this extended developmental period also makes it more susceptible to environmental insults that result in neural and behavioral deficits.

We recently showed that pups’ cortical activity is influenced by maternal interactions, the major environmental input at this stage. Using local field potential (LFP) recordings in behaving rat pups at postnatal day (P)10–P12, we and others observed that low-frequency cortical LFP power is greater when the mother is present in the nest than when she is absent ^{14, 15}. These changes were mimicked by exposure to the selective-serotonin-reuptake inhibitor (SSRI) fluoxetine and were blocked by ketanserin (an inhibitor of serotonin type 2 receptors) ¹⁴. These data support the role of serotonin in the regulation of brain activity in early life in response to maternal behavior. Here we tested the behavioral, and physiological consequences of changes in neural activity specifically in the PFC during a discrete period of development, and contrast those with the changes induced by maternal separation.

As reviewed in Short and Baram ¹⁶, our knowledge regarding how the environment impacts brain development has progressed significantly. From the 1950s to 1980s, the brain was believed to be resistant to stress as a safeguard against the trauma of birth ¹⁶. Then the concepts of epigenetics and plasticity were introduced and the effects of early-life-stress in the brain were documented ^{17, 18}. The recognition that early-life events could result in lasting changes in the brain, with the potential for transmission across generations, became widely accepted ¹⁹. Today, we can examine the maturation of brain circuits through synaptic pruning and strengthening and their relationship to environmental cues ^{16, 20}. Taking this a step further, here we manipulate the activity of these circuits during early development to understand their causal impact on neuronal activity and behavior in adulthood.

METHODS

Animals

Experiments were performed at the Nathan Kline Institute for Psychiatric Research (NKI). Experiments were conducted blind, in compliance with National Institutes of Health (NIH) guidelines, and under protocols approved by NKI. C57BL/6J mice were housed in groups (two to five mice per cage) and maintained on a 12 h light/dark cycle with access to food and water ad libitum. Approximately equal numbers of male and female mice were used in the experiments. Mice were tested as adults, between 3 and 5 months of age. The number of animals used in each experiment was based on previous experience with those tests. To identify the animals during maternal separation, viral injections and CNO injections we used toe clippings. Pups were randomly assigned to the treatment groups, by being randomly picked. To identify the animals after weaning we placed a numbered ear tag; these were used to blind the experimenter during testing.

Maternal separation (MS)

Breeding trios were set up and visibly pregnant dams were separated and singly housed before birth. Entire litters were randomized at postnatal day 1 (P1) and pups within each litter were divided into maternal separated and control groups. Each group was identified by toe clippings. MS with early weaning was performed as described in ²¹. Briefly, MS-assigned mice were separated from the dam and control littermates daily from P2-P17 and placed in individual dividers inside a temperature-controlled container (32°C). Separation was performed for 4 h/day from P2 to P5 and for 8h/day from P6 to P16. After separation, pups were returned to their home cages. On P17, the maternal separated groups were weaned. Control animals were weaned at P24. Manipulation of control animals was restricted to the handling necessary to check the toe clippings to separate MS-mice.

AAV injections into the PFC

Viral injections were performed at P1. Mice were anesthetized with isoflurane (1–2% in oxygen), and their heads were fixed in a stereotactic apparatus. Using a pulled borosilicate glass capillary (Borosilicate Glass, O.D.:1.0mm, I.D.:0.5m, 10 cm length, Sutter, B100–50-10), mice were injected bilaterally with 0.4 μl of AAV8-hSyn-hM4D(GI)-mCherry for inhibition (HM4D; Addgene #50475-AAV8), AAV8-hSyn-hM3D(Gq)-mCherry for excitation (HM3D; Addgene #50474-AAV8), or a control virus AAV8-hSyn-mCherry (mCherry; Addgene #114472-AAV8). The following coordinates were used (from the intersection of the inferior cerebral vein and the superior sagittal sinus): Anteroposterior: −1.3 mm, Lateral: +/−0.5mm, Dorsoventral: −1.2 mm. For PFC inhibition or excitation, mice were injected i.p. with Clozapine-N-oxide (CNO; RTI international, Batch ID: 13626–75) dissolved in 0.9% saline at 5 mg/kg/day from P2 to P17. Vehicle controls were injected with 0.9% saline (SAL).

Behavior

Object recognition:

In the novel object recognition test, subjects were first habituated for 5 min to an open-field (OF) arena (50 cm × 50 cm). Next, two identical objects were placed in the arena and the mice were allowed to explore them for 5 min. After a 5-min delay in the home cage, animals were placed back in the arena for 5 min with one of the objects replaced by a new one. We recorded the time spent exploring each object using ANY-maze software.

Delay non-match to sample (DNMS):

DNMS was performed on a T-maze (Plexiglass, long arm 63.5 cm x 10 cm, short left and right arms 55 cm x 10 cm) as described in ²². Pretraining: Mice were food restricted to maintain 80–85% of their free feeding weight. During the week of pretraining, mice were also given eight sucrose pellets (20mg dustless sucrose pellets, Bio-serv) in addition to their chow to habituate to the sucrose pellets. Apparatus habituation: Mice were placed in the maze for 20 min daily, for 5 days, and were allowed to freely explore the maze and collect 4 pellets from each arm. Training: Animals performed 4 trials per day (ITI 20 min) with each trial consisting of a forced choice and a free choice run. In the forced choice run, mice were placed in the start box at the end of the long arm of the T-maze and allowed to travel up the arm and enter a randomly selected open arm of the maze, with the other arm closed, to collect a sucrose pellet. After the forced run, mice were coached to return to the start box for a delay between the forced and the free choice trials. During training the delay was 4 s. After this delay, the door of the start box was opened, and mice were allowed to explore the maze and enter one of the arms. After making a choice, the other arm was closed. A correct choice was scored when the mouse chose the alternate arm to the forced run. Mice were trained until they reached criterion: 11 out of 12 correct choices for 3 consecutive days. On the days after reaching criteria, each mouse was tested daily (4 trials, ITI 20min) with longer delays. We tested 4 delays: 4s, 10s, 20s and 60 s, selected pseudorandomly.

Voltage sensitive dye (VSD)

At P13 or at >P70, mice were anesthetized with isoflurane, decapitated and 300 μm slices containing the PFC were sliced in High Sucrose Solution (Ecocyteshop, #LRE-S-LSG-1051). VSD staining and imaging was performed as described previously ²³. Brain slices were stained for 1 h by submersion in an aCSF solution (Ecocyteshop, #LRE-S-LSG-1000–2) containing the VSD (Di-4-ANEQ(F)PTEA, 2 μM, Potentiometric Probes) in an interface-type chamber. They were then transferred to an immersion-style recording chamber and continuously superfused with oxygenated aCSF. After recording a baseline, the slices were incubated with drugs designed to target specific receptors. We used the following drugs: 30 μM bicuculline (GABAAR antagonist ^{24, 25}; Sigma-Aldrich, #14340), 10 μM ketanserin (5HT2A/CR antagonist ^{25, 26}; Tocris, #83846–83-74), 10 μM serotonin (5HT; ²⁷; Sigma, #H9523), 30 nM WAY100635 (WAY; 5HT1A antagonist ²⁷, Sigma, #634908–75-1), 10 μM CNO (prototypical DREADD activator, ²⁸, RTI international, Batch ID: 13626–75). Drug perfusion was performed using a peristaltic pump to introduce the drug solution into the recording chamber while another tubing was used to drain the recording chamber. The solution flow rate was approximately 3 to 4 ml/min. To assess the effects of drug treatments, voltage images during drug application were normalized to the pre-drug baseline fluorescence. Imaging for bicuculline, serotonin (5HT), WAY100635 (WAY), and CNO was performed 5 min after the initiation of drug perfusion. Imaging for ketanserin was performed 30 min after the initiation of drug perfusion. Each recording utilized one hemisphere and only one drug was applied per slice. VSD imaging was conducted using a fluorescence microscope (Olympus BX51WI) with a Zyla camera (Andor, Zyla-4–2p-USB3) equipped with a 2x lens, focusing on layers 2/3 and 5 of the PFC. To acquire the fluorescence images, we utilized the Andor IQ3 live cell imaging software. Imaging was performed following the method outlined by ²⁹.

Slice electrophysiology

P2 mice were decapitated, and coronal cortical slices (300 μm) from the prefrontal cortex area were obtained using an OTS-5000 slicer. Slicing was performed in ice cold high sucrose brain slice cutting solution (aCSF; 254 mM sucrose, 10 mM D-glucose, 24 mM NaHCO₃, 2 mM CaCl₂, 2 mM MgSO₄, 3 mM KCl, 1.25 mM NaH₂PO₄, pH 7.4 ²⁷). Subsequently, the slices were recovered in oxygenated aCSF (128 mM NaCl, 10 mM D-glucose, 26 mM NaHCO₃, 2 mM CaCl₂, 2 mM MgSO₄, 3 mM KCl, 1.25 mM NaH₂PO₄, pH 7.4 ²⁷) at 31–33 °C for at least 2 hours. Patch electrodes (3–7MΩ) were pulled from borosilicate glass. The recording setup involved using aCSF oxygenated with a mixture of 95%O₂ /5%CO₂ and perfused at a rate of approximately 3 to 4ml/min. The internal solution for recording electrodes contained 5 mM KCl, 2 mM MgCl₂, 4 mM K₂-ATP, 0.4 mM Na₃-GTP, 10 mM Na₂-phosphocreatine, 10 mM HEPES buffer (pH 7.3). HM3D (pAAV-hSyn-hM3D(Gq)-mCherry, Addgene, 50474-AAV8) or control virus (pAAV-hSyn-mCherry, Addgene, 114472-AAV8) positive cells, in layers 2–3, were visualized with a fixed-staged microscope (Olympus BX51WI) and targeted based on the expression of mCherry. Whole-cell recording was conducted in current-clamp mode using a Multiclamp 700A amplifier (Molecular Devices). Data acquisition, set at 20kHz and low pass filtered at 3kHz, was performed through pClamp11 and Digidata 1440A. The current-clamp recording involved injecting current into cells to maintain them close to a threshold of −65mV. To measure CNO (10 μM, ²⁸, 2016, RTI international, Batch ID: 13626–75) effects on spiking, a 300 pA 500ms depolarizing current injection was administered before and during drug applications ²⁷. Input-output frequencies were determined by counting the number of action potentials elicited by each depolarizing current step in a protocol that entailed 50pA 500ms depolarizing current injections. Using a current sweeps protocol, where cells were held at the same current, frequencies were determined by counting the number of action potentials per second and compared before and during the application of CNO.

In vivo recordings

At P1 mice were injected with AAV8-hSyn-hM3D-mCherry or AAV8-hSyn-hM4D-mCherry into the PFC (0.4 μl/side). At P7 animals were anesthetized with urethane (0.25g/kg) and their heads were fixed in a stereotactic apparatus. Tungsten microelectrodes were placed in the PFC. Local field potentials (LFP’s) were recorded, filtered (0.3–300Hz) and digitized (100Hz) for acquisition and analysis in Spike2 (CED, Inc). After recording a 15 min baseline, mice were administered saline, 1, 5 or 10 mg/kg CNO and LFPs were recorded for 1h. LFP power was normalized in relation to the baseline, as previously reported ³⁰.

Immunohistochemistry and viral expression

To assess viral mCherry expression, tissue was fixed in 4% paraformaldehyde (PFA) overnight and incubated in 20% sucrose for at least 20h. Tissue was embedded in OCT (Tissue-tek, 4583), cryosectioned at 50 μm, washed with PBS three times and incubated in blocking solution (3% triton, Sigma, 9002–93-1 and 3% BSA, Sigma, 10735078001) at room temperature for at least 1 hour. Sections were then incubated overnight with PBS containing 3% triton X-100 (Sigma, 9002–93-1), 3% BSA (Sigma, 10735078001), and anti-mCherry (1:500, Millipore, AB356483). Sections were then mounted, and cover slipped using DAPI mounting medium (SouthernBiotech, 0100–20). Sections were stained with the Alexa dye-conjugated secondary anti-rabbit antibody (1:1000; Life Technologies, A10042) for 2 h. mCherry expression was analyzed for viral spread using an Olympus BX43 microscope. The spread of viral expression was delineated for each animal by gross anatomical comparison against images of the Allen Mouse Brain Atlas ³¹. All delineated images were superimposed and a heatmap of viral expression was plotted using MATLAB. A higher number of animals expressing virus in a determinate location is represented in warm colors, while a lower number of animals expressing virus in a certain location is represented by cooler colors.

Multiplexed single-nucleus RNA-sequencing (snRNAseq)

C57Bl/6J mice were subjected to the maternal separation protocol described above. The PFC of 4 control and 4 MS male mice was dissected at P13; another set of 4 control and 4 MS mice were allowed to grow until P70 at which point we dissected the PFC. Tissue was dissected using a tissue puncher and flash frozen in dry ice. Samples were homogenized, and the nuclei were purified using an iodixanol cushion. Each sample was stained with HTO-tagged antibodies ³². Pools of eight hashed samples were processed simultaneously. We performed library preparations using the 10X Chromium platform and sequenced them on the Illumina NovaSeq. Each run consisted of two P13 controls, two adult controls, two P13 MS and two adult MS samples. One adult mouse from the MS group was removed from analyses because it was a female misidentified as male.

After sequencing, 10x Genomics Cell Ranger (6.0.0) was used to perform alignment, cell calling, and gene quantification on the RNA library ³³. For the hashing (HTO) library, the antibody sequence list was pre-processed using the “kite” pipeline (https://github.com/pachterlab/kite) to allow pseudo-alignment to both the exact sequences and those with one base difference. A reference index was created using Kallisto (v0.46.0), and then Kallisto and bustools (v0.39.3) were used to obtain a cell x tag UMI count matrix ³⁴.

Seurat (v3.2.1) was used for hashing demultiplexing (assigning sample identities) using the HTODemux module ³⁵, including all cells called by CellRanger in “filtered_feature_bc_matrix” with at least 10 total HTO nUMI counts. Each of the two pools was run separately through HTODemux. To be extra sure that none of the cells from the female mouse were included, cells in the second pool with a raw UMI count of 3 or more for the XIST transcript were also excluded from further analysis. Next, doublets and “negatives” (cells that could not confidently be assigned to any sample identity) were also excluded.

At this point, the cells for the two pools were combined. Seurat ³⁶ was also used for normalization (NormalizeData), variable gene selection (FindVariableFeatures), z-scoring (ScaleData), and dimensional reduction (RunPCA). The top 30 PCs were used as input to clustering (FindNeighbors and FindClusters), clustering at resolution=0.3. These top 30 PCs were also used as input to further dimensional reduction for visualization (UMAP).

Clusters were manually annotated based on cluster-level expression (mean expression and percentage nonzero) of canonical markers as calculated using the Seurat DotPlot module. These canonical markers included those for: astrocytes (Acsbg1, Aqp4, Mlc1), endothelial cells (Cldn5, Flt1, Pglyrp1, Rgs5), microglia (C1q1a/b/c, Ctss), oligodendrocytes (Aspa, Mog), oligodendrocyte precursors (OPCs) (Cacgn4, Matn4, Pdgfra, Neu4), excitatory neurons (Slc17a6, Slc17a7), and inhibitory neurons (Gad1, Gad2). Then, all cells in a cluster were assigned to the corresponding cell type identity for that cluster. Two clusters did not show sufficient expression of any of the sets of canonical markers and so were left as unannotated (Supplementary Fig. 1).

Differential expression was performed using Seurat’s FindMarkers module, with logfc.threshold=0.10. Analysis of differentially expressed genes (DEGs) was performed using Ingenuity Pathway Analysis with a cutoff of FDR 0.05 (IPA, Qiagen) ³⁷.

Statistics

Parametric statistical tests were used in the behavioral and slice physiology experiments with alpha=0.05, two-tailed. Student’s t-test was used for comparisons between two groups. ANOVA was used for comparisons of more than two groups, followed by a corrected multiple comparisons test (Sidak, Tukey or Dunnett’s, as recommended by the software) in Prism (GraphPad).

RESULTS

Modulation of PFC activity during early life using DREADDs

To test the time-point post-surgery at which DREADD expression was sufficient to cause changes in activity in response to clozapine-n-oxide (CNO), we injected an inhibitory DREADD using an adeno-associated virus (AAV) (AAV8-hSyn-hM4D-mCherry; HM4D) into the PFC of P1 mouse pups (Fig. 1a). We dissected the brains at P2, P4, P7 and P14. We prepared PFC slices for VSD ²³ and tested for changes in activity in response to 10 μM CNO. We observed a decrease in activity, in response to CNO, in slices expressing AAV8-hSyn-hM4D-mCherry starting at P2. As expected, as time from surgery progressed, the inhibitory effects of HM4D+CNO increased. CNO alone did not induce a change in activity in control slices from HM4D-negative animals (PN age x Treatment: F_{(3, 49)}= 3.479, P=0.0227; P2_{HM4DvsControl}: Sidak multiple comparisons adjusted P=0.0193; Fig. 1b–c).

Figure 1.

DREADDs expression and response to CNO in early-life. a-c) Time course of AAV-hSyn-hM4D response to CNO. a) Mice were injected with AAV8-hSyn-hM4D-mCherry or a control virus (AAV8-hSyn-mCherry) into the PFC at P1. b) PFC slices were collected at several points after infusion (P2, P4, P7 and P14). The response to CNO was measured by VSD imaging and normalized to pre-CNO baseline. c) Illustration of the anatomical reference of Layer2/3, marked in blue, for analysis for voltage dye imaging. Gray images: baseline fluorescence. Colored images: Variation in fluorescence in response to CNO. d) mCherry expression in the PFC was observed at P2, 24 hours after surgery. DAPI (blue) and mCherry (red) are shown. Scale bar 50 μm in 20x figure. e-h) Electrophysiological response of HM3D-positive neurons, compared to mCherry controls, to CNO bath application at P2 (24h post viral infusion). e) HM3D (n=6) and mCherry (n=6) cells in response to a 300 pA depolarizing steps, representative current-clamp traces (right) from PFC positive cells for HM3D before and after CNO application. After CNO application, HM3D neurons fire more action potentials indicating increased intrinsic excitability compared to before CNO application and to mCherry-positive cells. f) The resting membrane potential Vm (mV) of mCherry and HM3D-positive cells at baseline and after exposure to CNO showed no statistical difference. F) Spike threshold (mV) of mCherry and HM3D-positive cells at baseline and after exposure to CNO during the recording indicates that HM3D neurons were more excitable when exposed to CNO. h) The majority of HM3D neurons (n=11) responded to CNO with increased spiking, while mCherry neurons (n=7) showed no changes in spiking when exposed to CNO. Pre-CNO activity was analyzed for 20 seconds, while post-CNO was analyzed for 60 seconds. *p<0.05, **p<0.01, ***p<0.001. Data presented as mean ± SEM.

To test whether viral expression was visible at P2 we injected virus at P1 and stained for mCherry at P2. Viral expression was already prominent at P2 (Fig. 1d). To assess changes in activity caused by DREADD expression in response to CNO, we infused an excitatory DREADD (AAV8-hSyn-hM3D-mCherry; HM3D), or a control virus (AAV8-hSyn-mCherry; mCherry) into the PFC of P1 mouse pups. We then recorded the brains at P2 and tested for changes in activity in response to 10 μM CNO. We conducted whole-cell patch-clamp electrophysiology on cortical slices, targeting fluorescent mCherry and HM3D-mCherry cells in layer 2/3 of the PFC (Fig. 1e–h). HM3D neurons fired at a greater frequency in response to depolarizing current injections with CNO compared to baseline (ANOVA_{Virus x CNO}: F_(18,140)=1.962, P=0.0156; Fig. 1e). Passive membrane characteristics, such as the resting membrane potential (RMP), between layer 2/3 HM3D and mCherry neurons did not differ (Fig. 1f). The threshold shift in HM3D-positive cells upon CNO application was significantly higher compared to HM3D-positive cells before application of CNO (ANOVA_{Virus x CNO}: F_(1,20)=4.794, P=0.0406; HM3D_{BaselinevsCNO:} Sidak multiple comparisons adjusted P<0.0001; Fig. 1g). Next, we tested CNO-induced depolarization in layer 2/3. CNO elicited depolarization increased the firing rate in HM3D-postive cells (ANOVA_{Virus x CNO}: F_(1,16)=11.62, p=0.0036; HM3D_{BaselinevsCNO:} Sidak multiple comparisons adjusted P<0.0001; Fig. 1h).

To choose the lowest dose at which CNO had an effect on mouse pups, we injected AAV8-hSyn-hM3D-mCherry (HM3D) or AAV8-hSyn-hM4D-mCherry (mCherry) into the PFC of P1 mouse pups. At P7 we recorded in vivo anesthetized LFPs in the PFC of these pups in response to saline and 1, 5 or 10 mg/kg CNO. As previously described in adult rats ³⁰, we found a sustained increase in LFP power in the 5 mg/kg group expressing HM3D. No change was observed in the mCherry control groups injected with CNO nor in the saline groups (Supplementary Fig. 2). Here we limited our recordings to 1h, although previous studies have shown that, despite rapid decreases in plasma CNO levels ³⁸, the behavioral and electrophysiological effects of CNO in DREADD-expressing animals are evident for close to 6 h ³⁹.

Maternal separation and transient inhibition of PFC activity during early life affect adult cognition

Early-life stress has been associated with cognitive deficits. To assess cognition during adulthood in our model, we employed the object recognition test (OR) and the working memory delayed-non-match-to-sample task (DNMS). Mouse pups were subjected to the MS protocol described in Methods. MS and their littermate controls were tested in the OR and in the DNMS starting at P70 (Fig. 2a). There were no significant differences between sexes (Supplementary Fig. 3). MS mice showed reduced locomotion in the open field (t₍₃₈₎=2.623, P=0.0125, Fig. 2b). No differences were found in time exploring objects (Fig. 2c). We found that performance on OR was impaired in MS mice when compared to their littermate controls (Fig. 2d; t₍₃₈₎=11.37, P<0.0001). MS animals reached criteria in the DNMS more quickly (Fig. 2e; t₍₃₈₎=2.193, P=0.0345) but their performance deteriorated as delays increased (Fig. 2f; ANOVA_{DelayxTreatment}: F_{(3, 114)}=21.59, P<0.0001). To test the effect of early-life PFC inhibition on cognitive task performance, we injected an adeno-associated virus expressing the inhibitory designer receptor HM4D into the PFC of P1 mice. CNO (5 mg/kg) or control vehicle (saline, SAL) was injected daily in HM4D and control mCherry mice from days P2 to P17; animals were tested starting at P70 (Fig. 3). There were no significant differences between sexes (Supplementary Fig. 4). Comparable to what we observed in MS animals, performance in the OR and DNMS tasks was impaired in HM4D/CNO animals compared to controls (Fig. 3e,g); OR_preference: ANOVA: F_{(3, 65)}=118.6, P<0.0001; DNMS: ANOVA_{DelayxTreatment}: F_{(9, 189)}=4.046, P<0.0001).

Figure 2.

Maternal separation leads to cognitive deficits. a) Experimental design; created with https://biorender.com. b) Maternally separated animals displayed no statistical difference from littermate controls on distance travelled during habituation. c) Maternally separated animals displayed no statistical difference in exploration of two equal objects in the sample phase. d) Maternal separated animals performed significantly worse than standard reared controls in the object recognition task. e) Standard reared control mice took longer to reach the delayed non-match to sample test (DNMS) criteria than maternally separated males. f) Maternal separated animals performed significantly worse than standard reared controls in the DNMS. In MS animals’ performance declined as the delay increased. NOR: nCTR=9M+13F; nMS=8M+9F. *p<0.05, **p<0.01, ***p<0.001. Data presented as mean ± SEM.

Figure 3.

Early-life PFC inhibition leads to cognitive deficits. a) Experimental design; created with https://biorender.com. b) Right: Density plot showing the viral spread in all animals. Red colored areas indicate a higher number of animals with expression in that area. Left: Representative image of viral spread. DAPI (blue) and mCherry (red). c) Distance travelled during habituation was similar in all groups. d) Time exploring the objects was similar in all groups during the sample phase. e) HM4D/CNO mice performed significantly worse than standard reared control animals in the object recognition test. f) Days to criteria in the DNMS was similar in all groups tested. g) HM4D/CNO animals performed significantly worse than controls in the DNMS. In HM4D/CNO animals’ performance declined as the delay increased. nmCherry/SAL = 8M+8F; nHM4D/SAL = 7M+7F; nmCherry/CNO = 12M+10F; nHM4D/CNO = 7M+7F. *p<0.05, **p<0.01, ***p<0.001. Data presented as mean ± SEM.

PFC excitation during maternal separation rescues cognitive deficits induced by maternal separation

To determine whether increasing activity of the PFC during MS could prevent the development of cognitive deficits, we used HM3D (Fig. 4). At P1, mice were injected with AAV8-hSyn-hM3D(Gq)-mCherry, or the control virus AAV8-hSyn-mCherry. Half of the mice in these groups were subjected to MS and the other half experienced standard rearing (CTR). All animals received CNO from P2 to P17. Interestingly, we observed that the performance of mice expressing HM3D and exposed to maternal separation (MS/HM3D) was identical to that of control animals (CTR/mCherry), both on OR and DNMS (Fig. 4e,g; OR: ANOVA: F_{(3, 92)}=215.6, P<0.0001; DNMS: ANOVA_{DelayxTreatment}: F_{(9, 261)}=1.948, P=0.0458). There were no significant differences between sexes (Supplementary Fig. 5).

Figure 4.

Early-life PFC excitation blocks cognitive deficits induced by maternal separation. a) Experimental design; created with https://biorender.com. b) Right: Density plot showing the viral spread in all animals. Red colored areas indicate a higher number of animals with expression in that area. Left: Representative image of viral spread. DAPI (blue) and mCherry (red). c) CTR/HM3D and MS/HM3D mice, presented significantly higher distance traveled in the habituation phase. d) Time exploring the objects was similar in all groups during the sample phase. e) Maternal separated animals (MS/mCherry) performed significantly worse than standard reared animals in the object recognition test. Mice in which we excited the PFC during MS performed at the same level as standard reared animals. f) Days to criteria in the DNMS was similar in all groups tested. g) Maternal separated animals (MS/mCherry) performed significantly worse than standard reared animals in the DNMS. In MS/mCherry animals’ performance declined as the delay increased. nCTR/mCherry=8M+16F;nMS/mCherry=13M+12F; nCTR/HM3D=9M+19F; nMS/HM3D=11M+8F. *p<0.05, **p<0.01, ***p<0.001. Data presented as mean ± SEM.

Maternal separation leads to differences in interneuron gene expression

To understand how early-life stress affects gene expression in the PFC, we performed multiplexed single-nucleus RNA sequencing (snRNAseq; Fig. 5a). Fig. 5b–c presents a landscape of this data including UMAP visualization. We can observe that, as expected, the clustering difference between ages (P13 vs. P70) was much more accentuated than the differences between treatments (control reared vs. MS) within the same age groups. Furthermore, we found that most of the differentially expressed genes in MS-adults vs. Control-adults were in glutamatergic (excitatory) and GABAergic (inhibitory) neurons (Fig. 5d-right). As expected, we found a larger number of differentially expressed genes between age groups across all cell types (Fig. 5d-left).

Figure 5.

Transcriptome-based cell classification and differential comparison by age and rearing in the mouse PFC. a) Multiplex single-nucleus RNAseq procedure; created with https://biorender.com. b-c) UMAP showing the clustering of cells by broad cell type (b) and age and rearing status (c). d) Number of significantly up or downregulated genes for each cell type in different comparison groups (adjusted p value<0.05). nMS/P70 = 3; nControl/P70 = 4; nMS/P13 = 4; nControl/P13 = 4.

Ingenuity Pathway Analysis was performed on DEGs from P13 standard-reared inhibitory neurons from the PFC as well as from DEGs from MS P70 inhibitory neurons, both compared to adult P70 controls (Fig. 6). Pathway analysis showed several key neuronal pathways were modulated by normal aging that also showed dysregulation in the MS animals compared to controls in inhibitory neurons. Significant pathways are indicated by log-fold change (LFC), with z-score assessment indicating down- (shown in green shading) and up- (shown in pink shading) regulation within these key neuronal pathways (Fig. 6a). White represents those pathways that were significantly modulated by condition but lacked sufficient gene-specific information for z-score assessment. We observed changes in synaptic plasticity and excitability related pathways as well as in several neuromodulatory pathways, including the cannabinoid and dopaminergic signaling pathways. Importantly, we observed changes in glutamatergic, GABAergic and serotonergic signaling pathways. IPA assessment indicates that glutamatergic signaling was differentially modulated in adult MS versus control-reared inhibitory neurons (Fig. 6b; Blue text = downregulated genes, red text = upregulated genes), with downregulation of Grm7 and Grid1& 2, with concomitant upregulation of Gria4, while in immature inhibitory neurons vs. adult neurons, alterations of Grid2 match MS/P70 with additional alterations in specific AMPA, NMDA and KA subunits (Fig. 6b). Similarly, we found changes in adult-control versus adult-MS gene expression in GABAergic signaling, specifically in expression of the GABA_B subunit and its downstream regulators (Fig. 6c). These changes were distinct from the differences observed in this pathway in P13-controls versus P70-controls, which showed differential expression levels of GABA_A and GABA_B subunits and downstream regulators (Fig. 6c). Finally, in examining the serotonergic signaling pathway, the 5HT2_C and 5HT4 receptors were significantly downregulated in GABAergic cells (receptors represented with blue text) of MS-adults compared with Control-adults (Fig. 6d). There were no significant changes in the serotonergic signaling pathway in inhibitory neurons in P13-controls versus P70-controls. These data support that MS can lead to long lasting change of excitatory/inhibitory balance in the PFC by altering the gene expression profile of inhibitory cells.

Figure 6.

Maternal separation affected pathways related to development, glutamatergic/GABAergic and serotonergic function, in interneurons. a) Pathway analysis of DGE in interneurons in development (P13 vs P70; left) and in maternal separated versus standard-reared animals (right). IPA z-score indicates down- (green shading) and up- (pink shading) regulation of pathways b-d) Schematic of gene expression changes in interneurons. Blue text on image represents downregulated genes, red text on image represents upregulated genes. b) Glutamatergic receptor signaling changes in interneurons in adult MS versus control (top) and control-P13 versus control-adults (bottom). c) GABAergic receptor signaling changes in interneurons in adult MS versus control (top) and control-P13 versus control-adults (bottom). d) Serotonergic pathway changes in interneurons in adult MS versus control. Created with https://biorender.com. nMS/P70 = 3; nControl/P70 = 4; nMS/P13 = 4; nControl/P13 = 4.

Maternal separation and early-life PFC inhibition led to changes in GABAergic and serotonergic responses which were rescued by early-life PFC activation

To determine whether early-life MS and PFC inhibition leads to long-lasting changes in mPFC circuit function, we used slice VSD to measure PFC responses to GABAergic and serotonergic receptor inhibition. To test the effects of MS on PFC inhibition/excitation balance and on the function of serotonergic signaling, we prepared a cohort of maternal separated animals and littermate controls as described in Methods (Fig. 7a). We observed that while bicuculline increased excitation in the PFC of controls, it had a negligible effect on MS animals (Fig. 7b; PN_treatment: F_{(1, 13)}= 35.72, P<0.0001), suggesting that GABA signaling was affected in these mice. As expected, serotonin in adult controls led to an overall inhibition ⁴⁰. However, in MS animals, serotonin led to net excitation (Fig. 7b; PN_treatment: F_{(1, 13)}= 25.51, P=0.0002). Furthermore, we found that ketanserin, a 5-HT2_A/C receptor antagonist frequently used in slice electrophysiology ^{26, 41}, increased excitation in control animals. However, in MS animals, ketanserin did not lead to net excitation (Fig. 7b; PN_treatment: F _{(1, 13)} = 60.59, P<0.0001). Finally, we found a similar overall response to WAY100635, a 5HT1A antagonist (Fig. 7b; PN_treatment: F_{(1, 13)}= 15.92, P=0.0015). Interestingly, we observed that the network responses from MS animals to the different drugs tested were similar to the effects seen in young animals (Fig. 7c–d; Bicuculline: PN_treatment: F_{(1, 19)}=7.022, P=0.0158; 5HT: PN_treatment: F_{(1, 19)}=8.596, P=0.0086; Ketanserin: PN_treatment: F_{(1, 15)}=5.926, P=0.0279; WAY: PN_treatment: F_{(1, 28)}= 106.6, P<0.0001) ^{40, 42}.

Figure 7.

PFC responses to GABAergic and serotonergic modulation. a, c) Experimental designs; created with https://biorender.com. b) Bicuculline application increased excitation in slices from control mice but not from MS mice. 5HT application excited MS and inhibited slices from control animals. Ketanserin and WAY application excited slices from control animals. d) Bicuculline application increased excitation in slices from adult animals. 5HT.application excited P13 and inhibited slices from adult animals. Ketanserin and Way application excited slices from adult animals. nCTR=10–9, nMS=5; nPAdult=10, nP13=11. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Data presented as mean ± SEM.

CNO/HM4D animals, in which we inhibited the PFC during early life, exhibited a pattern of drug response similar to MS and immature animals (Fig. 8a–b; Bicuculline: PN_treatment: F_{(3, 104)}=22.99, P<0.0001; 5HT: PN_treatment: F_{(3, 98)}=29.39, P<0.0001; Ketanserin: PN_treatment: F_{(3, 52)}=13.79, P<0.0001; WAY: PN_treatment: F_{(3, 50)}=24.58, P<0.0001). Importantly, PFC excitation during MS rescued these physiological phenotypes. Drug responses from CTR/mCherry did not differ significantly from the responses of MS/HM3D (Fig. 8c–d; Bicuculline: PN_treatment: F_{(3, 32)}=13.48, P<0.0001; 5HT: PN_treatment: F_{(3, 32)}=39.35, P<0.0001; Ketanserin: PN_treatment: F_{(3, 32)}=11.87, P<0.0001; WAY: PN_treatment: F_{(3, 32)}=14.20, P<0.0001).

Figure 8.

PFC responses to GABAergic and serotonergic modulation. a, c) Experimental designs; created with https://biorender.com. b) Bicuculline application increased excitation in slices from control animals but not from CNO/HM4D animals. 5HT application excited slices from CNO/HM4D animals. ketanserin and Way application excited slices from control animals but not from CNO/HM4D animals. d) Bicuculline application increased excitation in slices from all animals except MS/mCherry. 5HT application excited slices from MS/mCherry animals. Ketanserin and Way application excited slices from all animals except MS/mCherry. nSAL/mCherry=13, nSAL/HM4D=14, nCNO/mCherry=16, nCNO/HM4D=13; CTR/mCherry=8, MS/mCherry=8, CTR/HM3D=11, MS/HM3D=9. ***p<0.001. Data presented as mean ± SEM.

DISCUSSION

Complex behaviors involve the correct activation of pathways at the molecular, cellular and circuit levels. These pathways mature with time and inputs from the environment via activity-dependent mechanisms ^{16, 43}. These critical maturations take place during sensitive developmental periods ⁴⁴. We previously found that maternal presence/absence from the nest regulates PFC activity ¹⁴. Here, we hypothesized that inhibition of PFC activity during early-life would lead to deficits that mimic those observed in MS. Interestingly, we found that: (1) MS and early-life inhibition of PFC both led to cognitive deficits; (2) activation of the PFC using DREADDs during MS prevented those deficits; (3) MS led to changes in gene expression in interneurons, affecting GABAergic and serotonergic pathways; (4) MS and early-life PFC-inhibited animals had neural responses resembling those of immature animals with reduced responses to the GABAa antagonist bicuculline, excitation by serotonin, and reduced excitation caused by both the 5HT2R antagonist ketanserin and 5HT1AR antagonist WAY; and (5) activation of the PFC using DREADDs during MS prevented the MS-induced changes in physiological responses to GABAergic and serotonergic agents.

PFC development and its role in cognition

The PFC is central for cognitive functions like learning, memory, inhibitory control and decision making ⁴⁵. A substantial literature in humans has shown that early-life-stress is a risk factor for behavioral and neurobiological deficits ^{5, 46, 47}, and early-life-stress has been linked to changes in adult PFC function ^48–51. Significant impairments in short- and long-term memory, lower IQ, and impaired academic performance are consistently observed in individuals exposed to childhood adversity ^52–55, including childhood neglect and institutional rearing ^56–59. Deficits in attention and inhibitory control are also consistently observed in post-institutionalized children ^{56, 60–62} and individuals with a history of childhood maltreatment and early-life stress ^{54, 61, 63–65}. The PFC plays a major role in higher-order behaviors such as inhibitory control, attention, and working memory ⁴⁵. Due to its late development ¹³, the PFC is vulnerable to early-life adversity. PFC alterations are found in subjects who experienced varied forms of early-life trauma ^48–51, including inhibitory control deficits and differences in PFC activation in adolescents with a history of maltreatment ^{65, 66}.

Role of activity during development in behavioral function in adulthood

Having previously found that maternal presence/absence regulates PFC activity ¹⁴, we tested here whether changes in PFC activity during early-life are causal to cognitive deficits in adulthood. We observed that inhibition of the PFC during the first weeks of life using the DREADD HM4D led to cognitive deficits in adulthood. Interestingly, we also observed that cognitive deficits induced by maternal separation could be reversed by early-life PFC activation using HM3D. These data support a causal relationship between PFC activity in early-life and later behavioral function. Maternal separation entails many changes in pups beyond changes in PFC activity, including food insecurity and lack of sensory stimulation ⁶⁷. Notably, using HM4D we could isolate one factor, PFC activity during early-life, and test its role in adulthood. Interestingly, a recent publication has linked adolescent thalamic inhibition to decreased thalamic-prefrontal function and PFC-dependent cognitive deficits ⁶⁸. Two other papers have also highlighted the impact of adolescent activity of interneurons in adult working memory performance ⁶⁹ and of early-life PFC activity in adult emotional behavior ⁷⁰. Contrasting with our findings, Bitzenhofer et al. ⁷¹ found that excitation of the PFC of developing mice led to cognitive deficits. In control animals, we did not find a deficit in performance on our tests after early-life excitation of the PFC. Bitzenhofer et al.’s protocol differs from ours in a few key aspects. First, their MS period was shorter (P7-P11 vs. P2-P17 in ours). Second, they applied transcranial light stimulation that involved differential handling of animals. Last, and perhaps most important, in Bitzenhofer et al. stimulation was restricted to L2/3 pyramidal neurons, while in our experiments we targeted both excitatory cells and interneurons. Our data in control mice may not have led to changes possibly due to the simultaneous activation of all neuron types in this region. Importantly, inhibition of the PFC caused cognitive deficits in our model. Together, these data strongly support the importance of early-life neural activity, and an optimal developmental E/I balance, in the establishment of behavioral phenotypes later in life.

Early life stress and changes in PFC activity

Evidence is accumulating that stress is associated with an imbalance of excitation-inhibition in the PFC ⁷². However, the changes in excitation-inhibition balance in the PFC in adult mice in response to early-life-stressors are not fully understood. We chose to measure the long-term effects of early-life stress on the PFC using VSD imaging because it allows us to measure global activity changes. Consistent with the predominantly inhibitory role of serotonin in the adult PFC ^{73, 74}, here we observed that serotonin in adult controls led to an overall inhibition. However, in MS animals as well in the HM4D group, serotonin led to net excitation in adulthood. These data are consistent with earlier work ²⁶ which found enhanced excitatory inward currents in a subset of neurons evident only in maternal separated animals. Furthermore, we found that ketanserin, a 5-HT2_A/C receptor antagonist frequently used in slice electrophysiology ^{26, 41}, increased excitation in control animals. However, in MS animals, ketanserin did not lead to net excitation. Additionally, we observed that blocking 5HT1AR increased overall excitation in adult controls, possibly by blocking 5HT1A receptors that are highly expressed in PFC pyramidal cells ⁷⁵. However, this increase was absent in MS, HM4D/CNO and P13 mice. Interestingly, we observed that the network responses from MS animals and the HM4D group to the different drugs tested were similar to the effects seen in young animals in which GABA and serotonin are both excitatory ^{40, 42}.

Recently, the GABA-switch from excitation to inhibition has been found to be accelerated by early-life-stress ⁷⁶. In rodents, this GABA-switch occurs around P15. However, in mice subjected to early-life-stress, this switch occurs earlier, during P6-P9 ⁷⁶. If GABA receptors become inhibitory earlier due to MS, then we can expect that early inhibition of the PFC could produce effects similar to those of MS. With the disruption of the E/I balance, long-lasting effects could be due to compensatory changes, e.g., downregulation in receptors. Follow-up studies, using more sensitive methods to probe cell-specific receptor changes in the PFC, will be important to elucidate these mechanisms.

Early life stress and changes in circuit maturation

A growing literature has focused on whether early-life stress alters brain maturation ^77–79, with some studies showing that early-life stress leads to “more mature” connectivity between the amygdala and the PFC ⁸⁰. Here we found a pattern of more “immature” physiological responses in the PFC. However, although we found similar physiological responses in MS and HM4D-treated adults as in young animals, our data does not suggest that these physiological responses are due to a similar mechanism. We observed an alteration in the transcriptome specifically in interneurons. When we examined specific DEGs in interneurons in immature compared to adult controls, we found genes showing differential regulation during maturation, which were not replicated in the dysregulated DEGs seen in MS adults compared to adult controls. We observed this in both excitatory and inhibitory synapses onto inhibitory neurons. Finally, serotonergic synapses were only significantly dysregulated in the MS animals and were not significantly changed during maturation in inhibitory neurons.

Together, these data point to early life as a window during which neural activity is critical for brain development. Specifically, we causally link PFC neural activity in early life with PFC network responses and cognitive performance in the adult.

Data availability:

snRNA-seq data have been deposited in the NCBI Gene Expression Omnibus (GEO) database and are publicly accessible through GEO accession number GSE254342.