Early-Life Stress Specifically Impaired Cognition in Adolescent Male Mice
To investigate the effects of ES on cognition in adolescent mice, we first established the LBN model during PND 2–9 and then compared stressed and control mice on a series of stress-related physiological and behavioral measures (Fig. 1A-B). Immediately after stress (PND9), stressed mice had significantly lower body weight gain (Fig. S1A left) and the effect persisted into adolescence (Fig. S1A right). Mice exposed to ES also showed significant adrenal atrophy (Fig. S1B) and a tendency for thymus atrophy (Fig. S1C) in adolescence.
Cognitive performance was assessed in four tasks. In the temporary order memory task (TOM, Fig. 1C), unlike the control mice, stressed mice failed to distinguish the “remote” object from the “recent” object and had significantly lower discrimination index than control mice. In the Y-maze spontaneous alternation test (Fig. 1D), stressed mice showed lower spontaneous alternation rates and higher error rates of same arm return, indicative of spatial working memory deficits. In the novel object recognition test, stressed mice did not distinguish the “novel” object from the “familiar” object as control mice did (Fig. 1E), despite of no group differences in the discrimination index. Stressed mice also exhibited spatial object recognition deficits, as in this task they failed to discriminate the “displaced” object from the “stationary” object, and showed significantly lower discrimination index than control mice (Fig. 1F). The total probe time and distance traveled during the test phase of the recognition tasks or the total arm entries in the Y-maze were not affected by ES (Fig. S2A-D).
Besides cognitive behaviors, we also evaluated anxiety-like, social approach, and depression-like behaviors. In three tasks of anxiety-like behaviors, no significant differences were observed between stressed and control mice (Fig. 1G-I and S2E-G). Social approach was not affected by ES (Fig. 1J and S2H). ES did not significantly affect depression-like behaviors in the tail suspension test (Fig. 1K and S2I), sucrose preference test (Fig. 1M and S2K), or the latency to immobility in the forced swimming test (Fig. S2J), except that increased immobility was observed in the stressed mice in the forced swimming test (Fig. 1L).
Together, these behavioral and stress-related physiological results indicate that the adverse effects of ES emerge as early as in adolescent male mice, with the cognitive behaviors being particularly vulnerable.
To examine the involvement of mPFC PVI in the adverse effects of ES during adolescence, we quantified PVI density and activity in the following three analyses. First, immunohistochemistry revealed that ES significantly decreased the density of PVI in the mPFC, irrespective of subregions examined (Fig. 2A and S3A). The density of somatostatin-expressing interneurons (SST-INs) was not affected (Fig. 2B and S3B). Next, using immediate early gene c-fos immunostaining (which reflects neural activation[44]) and colocalization analysis, we found that stressed mice showed reduced density of activated PVIs in mPFC during the TOM test (Fig. 2C-D and S3C). Again, no stress effects were observed for SST-INs (Fig. 2E and S3D). Finally, to validate the effects of ES on mPFC PVI activity, we recorded the evoked and spontaneous action potential in mPFC PVIs in PV-cre::Ai14 mice through whole-cell voltage clamp (Fig. 2F). PVIs of stressed mice showed a significantly lower frequency of evoked action potentials in response to current injection compared with controls (Fig. 2G) and the interaction reflected that ES-induced suppression was mainly observed at currents greater than 200 pA. The spontaneous action potentials of PVIs were not altered by ES (Fig. 2H). Together, these results indicate that ES specifically reduced PVI density and activity (not SST-INs) in mPFC in adolescent male mice.
Prefrontal PVI Activity Mediates Early-Life Stress-Induced Cognitive Deficits in Adolescent Mice
Having shown that ES elicited both cognitive deficits and mPFC PVI activity reduction in adolescent mice, in this section we examined whether the reduced mPFC PVI activity causally mediates ES-induced cognitive deficits in adolescent mice using chemogenetic or optogenetic techniques. To start with, we carried out the following experiments to mimic ES-induced reduction PVI density and activity in mPFC and to evaluate the corresponding behavioral consequences.
First, a loss-of-function experiment was performed to selectively ablate mPFC PVIs using PV-Cre mice by injecting adeno-associated virus (AAV) expressing Cre-dependent Casp3 (a cell apoptosis effector molecule) into the mPFC at PND22 (Fig. S4A-B). This manipulation resulted in cognitive deficits, including lower discrimination indices in the TOM test (Fig. S4C) and in the novel object recognition test (Fig. S4E), and higher error rates of SAR in the Y-maze spontaneous alternation test (Fig. S4D). The spatial object recognition task was not affected by mPFC PVIs ablation (Fig. S4F). Anxiety-like behaviors in three tasks were also largely unaffected (Fig. S4G-I), except for reduced time spent in the light box in the light-dark box test (Fig. S4H).
Second, we inhibited mPFC PVI activity with a chemogenetic DREADDs manipulation, i.e., by bilateral injection of the AAV vector carrying the Cre-dependent hM4Di (Gi) into the mPFC in PV-Cre mice (Fig. 3A-B). Immunofluorescence staining combined with colocalization verified that the CNO group showed significantly lower percentage of virus-infected PVIs that co-express c-fos, compared with the Veh group (Fig. 3C), indicative of PVI activity inhibition. This manipulation again resulted in cognitive deficits in TOM and Y-maze spontaneous alternation tests. In the TOM task (Fig. 3D), mice in the CNO group showed reduced discrimination index than the vehicle group and they were unable to discriminate the “remote” object and the “recent” object. In the Y-maze spontaneous alternation test (Fig. 3E), CNO-treated mice showed higher error rates of SAR; no group differences were observed in SA or AAR. The novel object recognition and spatial object recognition tasks were not significantly affected by CNO treatment (Fig. S5A-B). CNO treatment did not affect anxiety-like behaviors in the open field test (Fig. S5C), but reduced the time spent in the light box in the light-dark box test (Fig. S5D) and the time spent in the open arms in the elevated plus maze (Fig. S5E), indicative of increased anxiety levels following PVI inhibition in mPFC, which are consistent with previous findings [16].
Optogenetic manipulation was then carried out to validate the above-mentioned chemogenetic results (Fig. 3F-I and S6A-B). As the previous two experiments support the consistent involvement of mPFC PVIs in the TOM and Y-maze spontaneous alternation tests, these two behavioral tests were carried out in the following experiments. Optogenetic inhibition of mPFC PVIs disrupted temporal order memory and spatial working memory (Fig. 3H-I). In the TOM task (Fig. 3H), eNpHR3.0-infected mice failed to discriminate the “remote” object and the “recent” object and exhibited lower discrimination index than control mice. In the Y-maze spontaneous alternation test (Fig. 3I), eNpHR3.0-infected mice displayed more errors of alternate arm return than EGFP-infected mice; no group differences were found for SA or SAR.
Finally, based on the three experiments above showing that PVI inhibition reproduced ES-induced impairments of PFC-dependent cognitive functions, we continued to investigate whether upregulating mPFC PVI activity could reverse the cognition-impairing effects of ES by bilateral injection of the AAV vector carrying the Cre-dependent hM3Dq (Gq) into the mPFC in PV-Cre mice (Fig. 3J-K). To test the efficiency and specificity of the DREADDs system, c-fos immunoreactivity was detected in mCherry-infected neurons (Fig. S7A) and PVIs (Fig. 3L) after a single CNO dose. The percentage of mCherry-infected neurons co-labeled with c-fos was significantly elevated by DREADDs (Fig. S7A). For the immunofluorescence staining for c-fos and PV co-labeling, two-way ANOVA revealed a significant stress × drug interaction (Fig. 3L). Further group comparisons revealed that c-fos expression in PVIs were decreased by ES, which was reversed by DREADDs. In terms of behavioral consequences, for the TOM task, two-way ANOVA revealed a significant stress × drug interaction (Fig. 3M). Selective activation of mPFC PVIs restored ES-induced impairment. No stress or CNO effects were observed in total probe time or total distances traveled in the acquisition phase (Fig. S7B). For the Y-maze spontaneous alternation task (Fig. 3N and Fig. S7C), two-way ANOVA showed significant stress × CNO interactions for SA and SAR. The negative stress effects induced by ES were attenuated by activation of mPFC PVIs. These results indicate that mPFC PVI upregulation is sufficient to alleviate ES-induced cognitive deficits in temporal order memory and spatial working memory.
By selectively downregulating and upregulating mPFC PVI activity, the four experiments above provide causal evidence that mPFC PVI activity mediates ES-induced cognitive deficits in adolescent mice.
Prefrontal Pyramidal Neurons Are Involved in Early-Life Stress-Induced Cognitive Deficits through PVI
How does ES reduce mPFC PVI activity in adolescent male mice? The functional maturation of the PVIs during adolescence involves several mechanisms [21, 27], including increased glutamatergic inputs, increased density of perineuronal net (PNN) proteins, etc. PNNs as one component of the extracellular matrix preferentially surround PVIs and modulate their excitability [45]. We quantified the expression of PNNs in stressed and control mice at PND35 and did not observe significant group differences (Fig. S8A). As pyramidal neurons (PNs) form robust functional synapses onto PVIs, it is possible that ES may first reduce the PN activity, which in turn limits the excitatory inputs to PVIs. So, we examined the effects of ES on the excitatory inputs onto PVIs and the PN activity. The excitatory inputs onto PVIs were quantified by the expression levels of vesicular glutamate transporter-1 (VGluT1, responsible for loading glutamate into synaptic vesicles for future release [46] and involved in the regulation of excitatory neurotransmission [47] at different distances from the soma of PVIs (Fig. 4A-C). Two-way ANOVA revealed a main effect of stress for VGluT1 fluorescence intensity as well as a significant stress × distance interaction, which was driven by decreased VGluT1 expression in stressed group at the distances larger than 8.784 um. We also quantified the soma radius of PVIs, which was not significantly affected by ES and approximately 8.662 ± 1.211 um (Fig. S8B), which is consistent with VGluT1 result above to suggest that ES-induced excitatory input reduction occurred surrounding the soma of PVIs. We then measured the mPFC PN activity during TOM test using immunofluorescence staining combined with co-localization of c-fos and CamkIIa or neurogranin (two excitatory neurons markers [48, 49]). Compared with control mice, stressed mice showed reduced density of neurons showing the co-labeling of c-fos and CamkIIa (Fig. 4D) or neurogranin (Fig. 4E), indicative of ES-induced inhibition of PN activity. To examine the causal involvement of PNs in ES adverse effects in adolescent mice, we carried out the following chemogenetic experiments.
First, we tested whether inhibiting PN activity in adolescent mice could reproduce the cognitive deficits of ES (Fig. S9A-F). Specifically, we injected the AAV vector expressing Cre-dependent hM4Di (Gi) bilaterally into the mPFC of CamkIIa-Cre mice. Immunofluorescence staining for the co-labeling of mCherry+ and c-fos+ (Fig. S9B) showed that, compared with the Veh group, the CNO group showed decreased percentage of neurons co-expressing mCherry+ and c-fos+. In terms of cognitive performances, in the TOM task, CNO-treated mice failed to discriminate the “remote” object and the “recent” object, and exhibited lower discrimination index (Fig. S9C). In the Y-maze test, CNO-treated mice showed lower SA and higher error rates of alternate arm return compared with Veh-treated mice (Fig. S9E). No significant group differences were observed in the test phase of TOM task (Fig. S9D) or the total arm entries in the Y-maze test (Fig. S9F). That is, selective inhibition of mPFC PNs reproduces ES-induced deficits of temporal order memory and spatial working memory.
Second, we investigated whether activating PNs (Fig. 4F) could reverse ES-induced cognitive impairment. Immunofluorescence staining showed that the virus-infected (mCherry+) neurons co-labeled with excitatory neurons marker CamkIIa (Fig. 4G) and that the percentage of mCherry-infected neurons co-labeled with c-fos was significantly elevated by DREADDs (Fig. 4H-I), indicative of selective activation of PNs in vivo. For the cognitive tests, we observed a significant stress × CNO interaction (Fig. 4J) in the TOM task in that PN activation restored the discrimination index decreased in ES-treated mice. No stress or CNO effects were observed in the test phase (Fig. S10A). For the Y-maze spontaneous alternation task (Fig. 4K), two-way ANOVA showed significant stress × drug interactions on SA and SAR. The negative stress effects on these measures were attenuated by activation of prefrontal PNs. No differences were observed in total arm entries (Fig. S10B). That is, selective activation of mPFC PNs reverses ES-induced deficits of temporal order memory and spatial working memory.
Based on the above-mentioned results that ES reduced the functional activity of both PNs and PVIs and the excitatory inputs onto PVIs and that both types of neurons are causally involved in ES-induced cognitive impairment in adolescent mice, we then hypothesized that the cognition-improving effects of PN activation may be mediated by PVI. To test this hypothesis, we carried out an experiment to activate PNs and inhibit PVIs in mPFC and examined whether PVI inhibition could block the reversal effect of PN activation on ES-induced cognitive deficits. The double chemogenetic manipulation was achieved by bilateral injection of a mixture of two viruses (AAV-CamkIIa-Gq; Cre-dependent AAV-DIO-Gi) into the mPFC in adolescent PV-Cre mice (Fig. 4L). Immunofluorescence staining for mCherry, PV, and c-fos was performed to examine the c-fos-positive neurons that are mCherry+ (non-PV+) and PV+ for validation. For mCherry+ (non-PV+) & c-fos+ neurons (Fig. 4M), we observed both the main effect of CNO and the stress × CNO interaction. Post hoc tests with Bonferroni correction showed that activation of PNs was observed in both control and stressed mice. For PV+ & c-fos+ neurons (Fig. 4N), we also observed both the main effect of CNO and the stress × CNO interaction. That is, PVI activity was inhibited in control mice after CNO administration and was reduced in ES-exposed vehicle mice; CNO did not further downregulate the PVI activity in stressed mice. For cognitive effects, in the TOM task (Fig. 4O), two-way ANOVA revealed a tendency for main effect of CNO and stress × CNO interaction. That is, selective inhibition of mPFC PVIs blocked the reversal effect of activation of PNs on the TOM impairment induced by ES. Besides, the double chemogenetic manipulation significantly reduced the discrimination index in control mice, which resembled the previous results of PVI inhibition. No stress or CNO effects were observed in the total probe time and distance traveled (Fig. S10C). For the Y-maze spontaneous alternation task (Fig. 4P), two-way ANOVA showed a significant main effect of stress on SA, indicating the absence of the reversal effects of PN activation after PVI inhibition. Main effects of CNO were also observed for AAR and SAR in that CNO significantly increased AAR and decreased SAR. No differences were observed in total arm entries (Fig. S10D). Together, these results indicate that inhibition of the PVIs in mPFC could block the reversal effect of the PN activation on ES-induced deficits in temporal order memory and spatial working memory, supporting the indispensable role of PVIs in the cognition-improving effects of PNs.
CRHR1 Downregulation Reverses Early-Life Stress-Induced Cognitive Deficits in Adolescent Mice by Restoring PVI Activity in mPFC
Our previous studies have highlighted the critical role of the CRH-CRHR1 system in ES-induced behavioral and neural abnormalities [28, 40–42, 50], including corticolimbic neurons [28], in postnatal and adult mice. Here we examined whether the CRH-CRHR1 system is involved in ES-induced negative effects in adolescent mice. As CRHR1 has been found to be mainly expressed in PNs in cortex [51], we first validated the localization of Crhr1 mRNA in the mPFC using single molecule fluorescence in situ hybridization (RNAscope ISH). Colocalization of Crhr1 mRNA with Slc17a7 (the mRNA of VGluT1, an excitatory neuron marker) and Slc32a1 (the mRNA of GABA vesicular transporter, an inhibitory neuron marker) in the mPFC (Fig. 5A) demonstrated that Crhr1+ neurons mainly co-localized with Slc17a7+ neurons (2313 out of 2800 Crhr1+ neurons, 82.61% and out of 2489 Slc17a7+ neurons, 92.93%). Only a small fraction of Crhr1+ neurons co-localized with Slc32a1+ neurons (43 out of 2800 Crhr1+ neurons, 1.54% and out of 520 Slc32a1+ neurons, 8.27%). This observation confirmed that Crhr1 mRNA is primarily expressed in pyramidal, not inhibitory, neurons, in the mPFC.
We then investigated whether blocking the mPFC CRH-CRHR1 system could reverse ES-induced deficits on PVIs and temporal order memory. We constructed an AAV vector carrying sgRNA targeting Crhr1 to achieve CRISPR-Cas9-mediated deletion of CRHR1 (Fig. 5B). Five gRNAs were screened using in vitro cellular assays and the sgRNA3 sequence showing the most effective transfection was chosen for packaging (Fig. S11A). HA tag staining confirmed that the virus was mainly infected in the mPFC (Fig. S11B). For the TOM task, significant reversal effects were observed: CRHR1 deletion restored temporary order memory impaired by ES (Fig. 5C). No significant effects of ES or CRHR1 deletion were observed in the test phase (Fig. S11C). For the PVI activity (Fig. 5D-E), two-way ANOVA revealed a significant stress × virus interaction in the mPFC: CRHR1 deletion in the mPFC reversed ES-induced reduction of PVI activity. Furthermore, discrimination index in the TOM task significantly correlated with PVI activity in the mPFC across all the animals (Fig. 5F). Together, these results support our hypothesis that early-life stress may first upregulate CRH-CRHR1 signaling in prefrontal PNs, which reduces PN activity, inhibits excitatory inputs to PVIs, and in turn downregulates PVI activity and leads to cognitive impairments.
Motivated by the recently established association between the CRH-CRHR1 system and cognition in the literature [32], we then carried out two pharmacological intervention experiments using the CRHR1 antagonist, antalarmin. In the first experiment, intraperitoneal injection of antalarmin was performed daily during stress procedure (PND2-8, Fig. 5G). The TOM task and the activity of PVIs were then assessed in adolescent mice. For TOM, two-way ANOVA revealed a significant stress × drug interaction, as antalarmin restored the ES-induced reduction of discrimination index (Fig. 5H). No significant group differences were observed in the test stage (Fig. S11D). For PVI activity in the mPFC, immunofluorescence staining for c-fos and PV showed significant main effects of stress and antalarmin, without stress × drug interaction (Fig. 5I-J). That is, the percentage of PVIs co-labeled with c-fos was significantly reduced by ES and upregulated by antalarmin. Similar ES and antalarmin effects were observed for the density of neurogranin+ neurons co-labeled with c-fos (Fig. S11E-F). Furthermore, discrimination index in the TOM task significantly correlated with PVI (Fig. 5K) and PN (Fig. S11G) activity in the mPFC. In the second experiment, to examine the acute pharmacological effects of antalarmin, the drug was given 30 minutes prior to the TOM test in adolescent stressed and control mice (Fig. 5L). Similar stress × drug interactions were observed for TOM and mPFC PVIs. Acute antalarmin injection blocked ES-induced reduction of discrimination index (Fig. 5M and S11H) and of mPFC PVIs (Fig. 5N-O). Significant correlation between the discrimination index and PVI activity in the mPFC was also observed (Fig. 5P).
To further test the hypothesis that the cognition-improving effects of antalarmin may be mediated by PVI activity in the mPFC, we carried out an experiment to examine whether PVI inhibition could block the reversal effects of acute antalarmin treatment on ES-induced cognitive deficits (Fig. 5Q-R). For PVI activity in the mPFC in stressed mice receiving antalarmin treatment, immunofluorescence staining for c-fos and PV showed that the percentage of PVIs co-labeled with c-fos (Fig. 5O) was significantly reduced by DREADDs (Fig. 5S). In the TOM task, compared with the stressed mice that received antalarmin treatment and exhibited intact temporal order memory, the CNO mice could not discriminate the “remote” object from the “recent” object and had significantly lower percentage time exploring the “remote” object (Fig. 5T), indicating that the reversal effect of antalarmin on ES-induced TOM impairment (Fig. 5M) was blocked by mPFC PVI inhibition. Together, these pharmacological experiments indicate that CRHR1 blockade could successfully reverse ES-induced temporal order memory deficits by restoring mPFC PVI activity, supporting the therapeutic potentials of antalarmin on ES-related cognitive impairments.
Environmental Enrichment Alleviates Early-life Stress-induced Cognitive Deficits through Activation of Prefrontal PNs and PVIs in Adolescent Mice
Besides pharmacological intervention, here we also tested whether EE, a commonly used non-pharmacological intervention for animals exposed to ES, could reverse ES-induced negative effects in adolescent mice. Stressed and control mice were exposed to three-week enriched or standard housing environments after weaning (PND21-42) and were then tested in the TOM task (Fig. 6A-B). As shown in Fig. 6C, despite lack of significant main effects and interactions in the two-way ANOVA, only the stressed mice kept in a standard environment failed to distinguish the “remote” object from the “recent” object, while mice in the other three groups exhibited intact recognition memory, which indicates that EE partially reversed ES-induced temporal order memory deficits.
We then investigated the neural correlates underlying EE cognition-improving effects by measuring the activity of PNs and PVIs in mPFC. Regarding the density of neurogranin+ cells that are c-fos+ (indicative of the PN activity), we found a significant stress × environment interaction in the mPFC (Fig. 6D-E). Specifically, ES significantly reduced neurogranin+ neuron activity, which was reversed by environment enrichment. Similar with the results of PNs, for the density of PVIs that are c-fos+ (indicative of the PVI activity), we also observed a significant stress × environment interaction (6G-H). ES significantly reduced PVI activity and EE significantly upregulated the activity of PVIs in stressed mice. Importantly, the discrimination indices in the temporal order memory test were significantly correlated with both the density of activated PNs (Fig. 6F) and PVIs (Fig. 6I) in the mPFC.
These data support the beneficial effects of EE in alleviating ES-induced temporal order memory deficits and reduced activity of PNs and PVIs.