Chronic Glucocorticoids Consumption Triggers and Worsens Experimental Alzheimer’s Disease-Like Pathology by Detrimental Immune Modulations

Animals

Adult male Sprague-Dawley rats (Janvier Lab., Le Genest-Saint-Isle, France) weighing 260–280 g (8 weeks) at the beginning of the experiments were housed 1 week before experiments in a standard animal facility of the University of Montpellier (CECEMA, registration No. D34-172-23) (12H/12H light/dark cycle with lights on at 07H00; 21 ± 1°C, food and water ad libitum). All experiments, including sacrifices, were performed in conscious rats between 09H00 and 12H00, during the diurnal trough of the HPA axis circadian rhythm. Male has been preferred to female because the latter may present a confounding factor of the estrous cycle, a parameter that needs to be systematically controlled. Although unlikely, a possibility still exists that mixing males and females in each experimental group may introduce a variable difficult to account for. Furthermore, doubling the number of groups and rats goes beyond the 3R rule. We do recognize the prospective importance of using both genders, and results will need to be verified in female.

Amyloid-β Peptide

In patients, soluble Aβ oligomers contain mainly the sequences Aβ_1–40 and Aβ_1–42 [2]. Nevertheless, they also contain peptides with shorter sequences such as Aβ_25–35 or Aβ_25–35/40 [33^-35], with no difference between human and rodent [36]. These shorter peptides can be produced in AD patients by enzymatic cleavage of Aβ_1–40 [33^, 35]. They contain extracellular and transmembrane residues that have been reported to be a biologically active region of Aβ [37^-39] and to contain the highly hydrophobic region forming stable aggregates [38]. Interest in this undecapeptide, which itself shows a β-sheet structure [17^, 38], has grown over the last 15 years, mainly because it induces neurite atrophy, neuronal cell death, synaptic loss, as well as synaptic plasticity and memory deficits in a similar way to Aβ_1–40 and Aβ_1–42 [40], but with better solubility and efficiency [41]. Aβ_25–35 and scrambled Aβ_25–35 peptides (Eurogentec, Angers, France) were dissolved in sterile water (1 µg/µL) and stored at −20°C. Since soluble Aβ oligomers correlate better with the progression of the disease [42], Aβ_25–35 and scrambled peptides were pre-aggregated by an in vitro incubation at 37°C (4 days) to obtain a solution mainly composed (more than 95%) of a mixture of soluble oligomer species (oAβ_25–35), as previously characterized [18].

Experimental Procedures

First, to validate the chronic CORT consumption model, animals received drinking water containing 1, 5, or 10 mg/kg CORT (Sigma-Aldrich, Lyon, France) dissolved in 0.45% β-cyclodextrin (Sigma-Aldrich, Lyon, France) [27] for 4 weeks (online suppl. Fig. 1a; for all online suppl. material, see http://www.karger.com/doi/10.1159/000521559). A control cohort received the vehicle alone (0.45% β-cyclodextrin) in drinking water. All animals were weighed once per week. The general behavior or anxiety state of the different groups of rats was tested at the end of the treatment (after 4 weeks) in an open field or in an elevated plus maze (EPM) test, respectively. Animals were sacrificed by decapitation. Pituitary and adrenals were rapidly harvested to be weighed and blood samples were collected for CORT assay.

Then, a new cohort of rats received drinking water containing the effective dose of CORT (10 mg/kg) or the vehicle (β-cyclodextrin) for 5 weeks. At week 4, animals were divided into 3 groups. The first groups had no surgery (control groups), the second received an icv injection of incubated scrambled peptide (10 µg/rat; scrambled groups), and the third received an icv injection of oAβ_25–35 (10 µg/rat; oAβ groups) (Fig. 1a). The animals were anesthetized with an intraperitoneal (ip) injection of 1 mL of a mixture of ketamine and xylazine (80 and 10 mg/kg b.w., respectively). As previously described [17^, 18], oAβ_25–35 was injected directly into the lateral ventricles using a David-Kopf stereotaxic apparatus (Phymep, Paris, France) (coordinates: AP −1 mm, L ±1.5 mm, DV −3.5 mm) [43]. All animals were weighed before, 4 weeks after the treatments, and 1 week after the icv injection of oAβ (week 5) (Fig. 1a). Short-term memory was tested in a T-maze 1 week after the icv injection of oAβ (Fig. 1a). Then, animals were sacrificed by decapitation. Blood samples and hippocampi were rapidly collected during the diurnal part of the HPA axis rhythm (between 09H00 and 12H00) for CORT assay and WB analysis. Naive rats received no treatment but were manipulated in the same manner as treated rats.

Open multimedia modal

General Behavior

The open-field test was used to characterize general behavior of rats, as previously reported [44]. The apparatus consists of a squared open-field (1 m² × 0.6 m high) made of white PVC with an infrared light emitting floor. An infrared sensitive CCD camera was placed above the field and connected to a video tracking system (Noldus EthoVisonXT, France). Animals were allowed to freely explore the open-field during 10 min. This test evaluates locomotion-related parameters, such as the total distance travelled or walking speed; exploratory-related parameters, including locomotion, immobility, and number of rearings; anxiety-related indexes, such as latency to start and thigmotaxis (locomotion along the walls); and stereotyped responses, like rearing or grooming.

Spatial Short-Term Memory

As previously reported, the T-maze test was used to rapidly assess cognitive ability in rats, especially the short-term memory deficits when performed in two successive sessions [17^-19^, 21^, 45]. The T-maze consisted of two short arms (A and B), extending from a longer alley (C) and enclosed with high walls. The test involved two trials separated by 1 h. During the training session, one short arm (B) was closed. Rats were placed at the end of the long alley, allowed to visit the maze for 10 min, and then returned into their home cage. During the test session, which was video-tracked (Noldus EthoVisonXT, France), animals were placed in the maze for 2 min, with free access to all arms. The number of visits and time spent in each arm were measured. The results were expressed as ratio of the time spent in the initially closed novel arm, over the time spent in the previous arm and as a ratio of the number of entries into the novel arm over the familiar one. The apparatus was cleaned with diluted ethanol (50%) between animals.

Anxiety Behavior

The anxiety state of rats was measured using their ability to explore open and enclosed arms of an EPM, as previously detailed [21]. The clear plexiglass apparatus consisted of 2 open arms (50 × 10 cm) and 2 enclosed arms (50 × 10 × 45 cm high), extending from a central platform and placed 60 cm above the floor. Each rat was placed at the center of the EPM facing the closed arm and its exploration behavior was recorded for 10 min (Noldus EthoVisonXT, France). The results were expressed as total time spent in the open arms and the total number of entries was counted to verify general motor activity. An entry into an arm was recorded if the animal crossed the line that connected that arm with the central platform with all 4 legs. The apparatus was cleaned with diluted ethanol (50%) between animals.

CORT Assay

Blood samples were collected at the time of sacrifice on 1 mg/mL EDTA (Sigma-Aldrich, France), centrifuged at 4°C, and plasma stored at −20°C until assayed for CORT [18]. Plasma CORT concentrations were assayed using a conventional ELISA kit (Enzo-Life Sciences, Farmingdale, NY, USA) in a 10 µL plasma sample diluted (1:40) with the assay buffer. The assay sensitivity was 27 pg/mL. The intra- and inter-assay coefficients were 6.6 and 7.8%, respectively.

WB Analysis

WB was performed as previously described [18] in the whole hippocampus. All antibodies used are detailed in the online Supplementary Table 2. Briefly, after sacrifice, hippocampi were microdissected, weighed, immediately frozen in liquid nitrogen and stored at −20°C. Tissues were sonicated (VibraCell; Sonics & Materials, Newtown, CT, USA) in a lysis buffer [18] and centrifuged (4°C). Supernatants were collected and the protein concentration was measured using a BCA kit (ThermoFisher Scientific, Illkirch-Graffenstaden, France). Sixty microgram from each sample was taken for WB analysis. Samples were separated in SDS-polyacrylamide gel (12%) and transferred to a PVDF membrane (Merck-Millipore, Molsheim, France). The membrane was incubated overnight (4°C) with the primary antibody, washed and then incubated for 2 h with the appropriate horseradish peroxidase-conjugated secondary antibody. Peroxidase activity was revealed by using enhanced-chemiluminescence reagents (Luminata-Crescendo, Merck-Millipore, Burlington, MA, USA). The intensity of peroxidase activity was quantified using Image-J software (NIH, Bethesda, MA, USA). β-tubulin was used as a loading control for all immunoblotting experiments.

Statistical Analysis

Data are presented as mean ± SEM and analyzed using one-way or 2-way ANOVA followed by a Tukey’s multiple comparisons test (GraphPad-Prism 9.0). In addition, T-maze data are analyzed in comparison to chance level, i.e., maximal deficit, using a one-sample t test (GraphPad-Prism 9.0). p < 0.05 was considered significant. The number of animals in each group is indicated within the columns and was determined by a statistical power analysis based on our previous studies (G*Power software). Before each analysis of variance, the Gaussian distribution was evaluated and validated by a Kolmogorov-Smirnov test (GraphPad-Prism 9.0).

Validation of the Chronic CORT Consumption Model

To validate this model (online suppl. Fig. 1a), we assessed relevant parameters associated with chronic CORT treatment in drinking water [26^, 27^, 29]. The daily dose of 10 mg/kg of CORT was the sole dose that significantly increased the plasma concentrations of CORT, while the vehicle (0.45% β-cyclodextrin) and the doses of 1 and 5 mg/kg induced no modification in comparison to control rats (online suppl. Fig. 1b). The dose of 10 mg/kg/day was also the most efficient to atrophy adrenal glands (online suppl. Table 1) and to induce anxiety-like behavior. Indeed, in the open-field test (online suppl. Fig. 1c), we identified several characteristic features of anxiety, such as the increase of the latency to start moving and of the total immobility time. We also observed a decrease of visits in the arena center and the number of grooming and rearing, while locomotion was not affected by the different doses. To confirm these general observations, we performed a well-defined anxiety test, the EPM. Animals treated with the two highest doses of CORT (5 and 10 mg/kg/day) displayed a decrease of the time spent in the open arms, characteristic of an anxious state (online suppl. Fig. 1d).

Impact of Chronic CORT on oAβ Toxicity

We next evaluated the impact of chronic CORT consumption (10 mg/kg/day) on the Aβ toxicity 1 week after the icv injection of oAβ_25–35 (CCoAβ group) (Fig. 1a). Five weeks of chronic CORT consumption in drinking water induced short-term memory deficits (T-maze test) and potentiated significantly the oAβ_25–35-induced alterations of memory performances (Fig. 1b). No effects of scrambled peptide nor vehicle treatment were evidenced (Fig. 1-4). One week after the icv injection of oAβ_25–35, an increase of plasma CORT concentrations was observed, which was partially blunted by the upstream chronic CORT consumption (Fig. 1c). In addition, an additive effect of CCoAβ treatment was evidenced on body weight (Table 1) and on the hippocampal expression of cleaved-caspase 3, while the two treatments alone did not impact the later parameter (Fig. 1d).

Open multimedia modal

Open multimedia modal

Open multimedia modal

Open multimedia modal

Impact on GR Signaling Pathways

As previously reported in the PFC [20], the icv injection of oAβ_25–35 altered GR-associated signaling pathways in the hippocampus (Fig. 2a). Indeed, 1 week after oAβ_25–35, the expression of total and activated GR (phosphorylated on Ser²¹¹) was increased [44] (Fig. 2b–d) and was associated with an increase of heat-shock protein 90/70 (HSP90/70) ratio (Fig. 2b, e), two GR chaperones reflecting the activity of GR and highly involved in AD pathophysiology [46^-48] and an increase of annexin A1 (a potent endogenous anti-inflammatory effector functioning in synergy with GR) [49^, 50] (Fig. 2b, f). We also observed the increased expression of the Src-kinase Fyn (Fig. 2b, g), the active form of GSK3β (phosphorylated on Tyr²¹⁶) (Fig. 2h, i), the total form of Cdk5 (Fig. 2h, j), and the p25/p35 ratio (associated with the activation of Cdk5) [51] (Fig. 2h, k). These three enzymes are particularly relevant since they are involved in GR activation but also in the hyperphosphorylation of Tau [51^-55]. Chronic CORT consumption induced a similar effect as oAβ_25–35 on all of these markers, except on annexin A1 (Fig. 2b, f) and the p25/p35 ratio where we evidenced an additive effect of CCoAβ treatments (Fig. 2h, k).

Impact on APP Metabolism

Then, we characterized by Western blot the different pathways of APP maturation, through the assessment of the main cellular elements involved in the amyloidogenic and non-amyloidogenic pathways (Fig. 3a). One week after oAβ_25–35 injection and as previously reported [16^, 18], we evidenced the induction of the amyloidogenic pathway in the hippocampus (Fig. 3j), consistent with an Aβ_1–42 increase, as previously reported by our group [20^, 56^, 57]. We observed an increased expression of both full-length APP (precursor of amyloid proteins) (Fig. 3b, c), C99 (precursor of Aβ peptides) (Fig. 3b, d), BACE1 (β-APP cleaving enzyme, β-secretase) (Fig. 3b, e), and PS1 (presenilin-1, γ-secretase subunit) (Fig. 3b, f). This induction is associated with a concomitant inhibition of non-amyloidogenic pathways (Fig. 3j), characterized by a decrease of sAPPα (α-secretase-cleaved soluble APP ecto-domain) (Fig. 3b, g), α-secretase ADAM10 (A disintegrin and metalloprotease domain-containing protein 10) (Fig. 3b, f), and IDE (insulin-degrading enzyme, involved in the clearance of Aβ) (Fig. 3b, g). Interestingly, chronic CORT consumption induced the same modifications as those observed 1 week after the icv injection of oAβ_25–35 (Fig. 3), and CCoAβ treatment induced no supplementary effects, except for BACE1 where we showed a significant additive effect (Fig. 3b, e).

Impact on Synaptic Deficits and Neuroinflammation

Decreased memory performances and GR signaling pathway alterations observed 1 week after the icv injection of oAβ_25–35 were associated in the hippocampus with pre- (synaptotagmine) and postsynaptic deficits (Fig. 4a–c). These deficits were potentiated by the prior overexposure to exogenous CORT in animals injected with oAβ_25–35 (Fig. 4a–c). Moreover, and as previously reported [19^, 20], synaptic deficits induced by the Aβ toxicity were associated with important modifications of hippocampal inflammatory processes. Indeed, after 1 week, the icv injection of oAβ_25–35 provoked both the induction of pro-inflammatory mechanisms (activation of astrocytes [GFAP] and recruitment of microglia [Iba1]) (Fig. 4a, d, e) and the induction of anti-inflammatory components, as interleukin-10 (IL-10) (Fig. 4a, e, f). By contrast, while chronic CORT favored anti-inflammatory pathways (Fig. 4g) by inhibiting astrocytes activation (Fig. 4a, d), microglia recruitment (Fig. 4a, e), and IL-10 induction (Fig. 4a, f), the CCoAβ group displayed exacerbated neuroinflammation (Fig. 4g). Indeed, co-treatment increased the activation of astrocytes and the recruitment of microglia (Fig. 4a, d, e) and inhibited IL-10 (Fig. 4a, f). In addition, ADAM17, another α-secretase (also known as tumor necrosis factor-α converting enzyme or TACE), highly activated in neuroinflammatory processes and promoting pro-inflammatory cytokines maturation in turn (for review, see [58]), was upregulated after the icv injection of oAβ_25–35 (Fig. 4a, g). And although chronic CORT inhibited the expression of this metalloprotease, as previously observed with ADAM10 (Fig. 3b, f), the CCoAβ treatment re-established the hippocampal increase of ADAM17 (Fig. 4a, g).

Model of Chronic GC Consumption

First, we aimed to validate our chronic GC consumption model, well known to be a model of chronic anxiety/depression [26^, 27] associated with memory deficits and hippocampal alterations [31^, 59]. According to the literature, in our study, we focused on the anxious state displayed in animals treated for 4 weeks with CORT in drinking water. At the optimal dose determined (i.e., 10 mg/kg/day), animals developed many anxiety-like behaviors in the open field and EPM tasks. At this dose, we also evidenced an increase of GC plasma levels associated with an atrophy of adrenal glands. These results are in accordance with studies showing that chronic CORT administration suppressed the inhibitory feedback of GC on the HPA axis activity [29^, 60], characteristic of mood disorders such as generalized anxiety or major depression [61^, 62]. In accordance to previous works conducted by others groups, we validated this model as a chronic model of GC overexposure, displaying HPA axis impairment and anxiety-like behaviors.

Synergistic Effects

Second, we compared the effects of chronic GC consumption alone or in combination with the icv injection of oAβ_25–35. CORT treatment caused short-term memory and synaptic deficits in the hippocampus, with no modification on caspase 3 levels. These alterations following GC overexposure are widely described in the literature. It was shown that chronic stress or GC administration altered synaptic terminal structure, neurogenesis, and induced neuronal atrophy, more particularly in the hippocampus [13^, 29^, 63^-65]. Since CORT treatment also induced anxiety-like behaviors in the EPM, we could not totally exclude this emotional aspect in the T-maze performance. However, as reported by others, CORT treatment induced many hippocampal alterations (dendritic retraction, decreased cell proliferation, and global volume) [31^, 58] sufficient to induce such short-term memory deficits. In addition, Han et al. [16] also reported an aggravation of cognitive impairment in APP/PS1 mice submitted to chronic unpredictable stress. By contrast, it is interesting to note that CCoAβ group displayed exacerbated memory and synaptic deficits. Furthermore, we also observed a synergistic effect of this co-treatment that induced caspase 3 expression in the hippocampus, whereas the two treatments alone did not impact it. This suggested that the potentiation of synaptic and memory deficits could be due to the induction of apoptosis processes.

Additive Effects

Moreover, we evidenced some additive effects of the exogenous GC treatment and the icv injection of oAβ_25–35. Indeed, we showed that the induction of BACE1 was potentiated. This enzyme was identified as the major β-secretase-like protein, highly involved in the amyloidogenic cleavage of APP, and thus in Aβ peptides production [66]. These findings are consistent with the study conducted by Green et al. [14]. The authors have shown that 1-week treatment of dexamethasone in 4-month-old 3xTg mice significantly increased BACE1 expression and higher levels of Aβ_1–40 and Aβ_1–42. Our result was in accordance with the literature since BACE1 production could be triggered by chronic stress as well as by Aβ. In fact, BACE1 transcription is under a direct control of GC since a glucocorticoid response element has been described in the promoter region of BACE1 [67]. In addition, BACE1 production could be also indirectly regulated by stress and Aβ through the activation of other intracellular mediators. (i) Both chronic GC overexposure [68^, 69] and Aβ toxicity [2^, 70] induce the c-jun N-terminal kinase/activator protein 1 (JNK) pathway (or stress-activated kinase), one of the main pathways favoring BACE1 production [71]. (ii) The p25/Cdk5 pathway, upregulated in our study, was also known to trigger BACE1 transcription [66]. This complex of enzymes is highly involved in AD pathology, particularly for its action on Tau hyperphosphorylation [2^, 51], but also on GR activation via phosphorylation on Ser211 in rodent (Ser224 in human) [72]. Some studies demonstrated that chronic unpredictable stress [73^, 74], as well as Aβ pathology [51^, 75] increased p25/Cdk5 expression and activity. (iii) These mechanisms could be due to an increased activity of calpain-1, a member of cysteine proteases family showing aberrant activity in AD [76]. This enzyme is particularly involved in the activation of GSK-3β [77] and the maturation of p35 to p25 [78]. Interestingly, we demonstrated recently in the oAβ_25–35 model that all these enzymes (i.e., GSK3-β, p25/Cdk5, and calpain-1) were directly or indirectly regulated by GC and GR in the prefrontal cortex [20]. Our results suggest that calpain-1 activity may also be potentiated by chronic GC overexposure since the hippocampal expression of 2 of its principal substrates was increased. So, by these different mechanisms, it was not surprising that combined toxicity of chronic CORT consumption and oAβ_25–35 injection had additive effects on the expressions and/or activities of BACE1 and p25/Cdk5. Since these 2 enzymes are highly involved in amyloid and tau pathologies, it brings new data on how chronic GC consumption is detrimental in AD development. However, further specific investigations are needed to characterize precisely tau system, in particular to assess the levels of tau hyperphosphorylation epitopes, its eventual aggregation, its localization and its staining pattern.

Effects on APP Metabolism and GR Signaling Pathways

On the principal proteins associated with GR signaling pathways (i.e., GR, p[Ser211]GR, HSP90/HSP70, GSK3-β, Cdk5, and Fyn) and APP metabolism (i.e., APP, C99, IDE, sAPPα, ADAM10, and PS1) assessed in our study, there was no additive effect of chronic CORT on the Aβ toxicity. Interestingly, in 3xTg mice, 1-week treatment with dexamethasone was able to induce an increase of APP and C99 levels [14]. However, the authors used 4-month-old 3xTg animals before the establishment of important AD-like symptoms. Thus, dexamethasone could potentiate the expression of these markers. In our study, the modifications induced by each treatment (chronic CORT or oAβ_25–35) were closely similar. This suggests common regulatory mechanisms and supports the important role played by GR in the pathophysiology of AD (for review, see [3^, 4^, 6^, 24]). Indeed, we recently demonstrated in this acute model of AD (oAβ_25–35) that these pathways were impaired [20] and associated with a strong and long-lasting disruption of the HPA axis activity and functionality [21]. Then, to evaluate the contribution of GR in the Aβ toxicity, we tried to restore their functionality using new selective GR modulators (sGRm) [19^, 20]. This family of compounds has the advantage to selectively abrogate pathological effects of GR overactivation, while retaining their physiological function [79^, 80]. By restoring their functionality with sGRm, we counteracted all parameters measured and induced by the Aβ toxicity, highlighting the central role played by GR in the development of AD [20]. Taken together, all of these results highly suggest that a part of Aβ toxicity could be a consequence of the GC overexposure and GR signaling impairment.

However, the 2 α-secretases ADAM10 and ADAM17 are differentially regulated in the hippocampus by chronic CORT consumption and oAβ_25–35. We found that Aβ toxicity increased ADAM17 levels while inhibiting ADAM10 expression. By contrast, chronic CORT decreased the two α-secretases. These proteases are neuroprotective by triggering the non-amyloidogenic pathway and the generation of sAPPα. But the precise role of ADAMs in AD pathogenesis is not fully understood. ADAM10 seems to constitutively cleave APP while fine regulation is mediated by ADAM17 [81]. In the CSF of AD patients, ADAM10 is decreased [82], which is correlated with the inhibition of the non-amyloidogenic pathway. Inversely, ADAM17 activity is increased both in early and advanced phases of AD [83] that evidences more complex regulatory mechanisms. In fact, besides its neuroprotective role, ADAM17 is also highly activated by neuroinflammatory processes and promotes pro-inflammatory cytokines maturation in turn (for review, see [58]). These findings are relevant with our results in oAβ and CCoAβ groups. In these animals, we found decreased ADAM10 and sAPPα expressions and increased ADAM17 expression concomitantly with enhanced neuroinflammatory processes (Fig. 4h). So, we could imagine that ADAM10 participated to APP non-amyloidogenic cleavage more than ADAM17 that could be regulated by pro-inflammatory processes. To summarize, we found that both oAβ and chronic CORT consumption altered APP processing balance by increasing the amyloidogenic (C99, BACE1, and PS1) and inhibiting the non-amyloidogenic (ADAM10 and IDE) pathways (Fig. 3j). It is interesting to note that the main difference observed between chronic CORT and oAβ groups is their inflammatory phenotypes. Such difference could participate to explain the exacerbation of AD-related markers in CCoAβ animals (Fig. 4h).

Effects on Inflammatory Processes

Neuroinflammation is a key feature of AD pathology. Activated microglia and astrocytes trigger the overproduction of pro-inflammatory cytokines, which promotes Aβ accumulation and thus neurodegeneration [2]. As we already observed [17^-20], we confirmed in this study that oAβ_25–35 induced neuroinflammation, since it increased GFAP and Iba1, but also ADAM17, which is highly activated by neuroinflammatory processes and, as discussed above, promotes pro-inflammatory cytokines maturation in turn (for review, see [58]) (Fig. 4h). Inversely, chronic CORT consumption favored the anti-inflammatory pathways (Fig. 4h). It was well known that stress and GC influence substantially the level and the quality of the immune response. The activation of GR-annexin A1 pathway inhibits the pro-inflammatory response, thus mediating the major immunosuppressive effects of GC [50^, 84]. Here, we observed the decrease of GFAP and Iba1 levels in the hippocampus, associated with the increase of GR, p(S211)GR, annexin A1 and the anti-inflammatory cytokine IL-10 expressions. Same results were obtained in recent research that investigated the effects of chronic stress on astroglial cells. Both chronic variable stress and chronic restraint stress increased GC plasma levels while decreasing GFAP expression by around 20% in astrocytes [85^-87]. Such variation in astrocytes was sufficient to trigger depressive and anxiety-like behaviors [85^-87]. The authors found that this modification is not due to glial cell loss, but could be due to an important remodeling of astrocytes morphology and network. This decrease of astrocyte plasticity led to a loss of their trophic functions on neurons, particularly on synaptic formation, LTP induction, and growth factor expression [85^, 87^, 88]. Based on our results, these alterations could be linked to synaptic and memory deficits induced by the chronic CORT consumption. Moreover, a negative glucocorticoid-responsive element on the promoter of GFAP was identified, indicating that GC inhibits directly the transcription of GFAP [89]. So, it is not surprising that in these conditions, chronic CORT seemed to be highly immunosuppressive (Fig. 4h). To reinforce this view, these results were also correlated with ADAM17 decreased expression. As mentioned, ADAM17 expression and activity are under the control of pro-inflammatory processes [58].

Otherwise, we observed attractive results in CCoAβ animals. We found in this group that GC plasma levels were higher than in control group, but not as well as chronic CORT consumption or oAβ groups. Such result could be explained by the atrophy of adrenals and the impairment of HPA axis, indicating that this endocrine system was exhausted in CCoAβ animals. In accordance with our results, other groups demonstrated that chronic GC overexposure might blunt stress-induced CORT secretion [29^, 90]. These findings were obtained with animals pretreated with GC and then submitted to behavioral stressors. This observation could be extended to cellular stressors such as oAβ_25–35 in our study. These results in CCoAβ animals could also explain the incapacity of GC to restrain inflammatory processes induced by oAβ. Indeed, in these animals, we observed a potentiation of pro-inflammatory (GFAP, Iba1, ADAM17) and an inhibition of anti-inflammatory (IL-10) markers (Fig. 4h). This incapacity of GC to regulate neuroinflammatory processes could partly explain the exacerbation of AD hallmarks evidenced here. Such phenomenon was already described in chronically stressed animals injected with lipopolysaccharide [91^-94]. Thus, in accordance with our results, it appears that chronic stress, as well as chronic CORT consumption, primes the neuroinflammatory response to a subsequent pro-inflammatory stimulus as oAβ_25–35 or lipopolysaccharide. Therefore, the neuroimmune context is more responsive to inflammation, also favoring GC insensitivity or reducing the HPA response [94^-96].

Furthermore, additional effects of chronic CORT consumption and oAβ_25–35 injection were evidenced in the hippocampus on annexin A1. As previously mentioned this protein is activated by pro-inflammatory mediators and is an important anti-inflammatory actor mediating GC and GR effects [49^, 50]. However, no study identified a glucocorticoid response element on the promoter region of annexin A1 gene, and its transcription seems to be mediated by pro-inflammatory cytokines and oxidative stress [97^, 98]. Here, surprisingly, even if GR signaling pathways were impaired, annexin A1 levels were highly increased in CCoAβ animals. Taken together, we can envisage that this increase of annexin A1 could be a compensatory mechanism to counteract exacerbated pro-inflammatory processes. Further investigations are needed to assess if the GR-annexin A1 axis is still functional, even in a context of GR overactivation.

We focused on GR as we previously observed and established its central role in AD pathophysiology [3^, 19^, 20]. In stress conditions, and a fortiori in pathological conditions, GR seem to be more involved than MR [10]. However, a recent work demonstrated that hippocampal MR are overexpressed and overactivated by oxidative stress. This leads to microglia activation, pro-inflammatory mediators’ secretion, and downregulation of anti-inflammatory factors [99]. Thus, we cannot totally exclude the involvement of MR in our study. Further investigations are needed to decipher the precise role of MR between chronic stress, high levels of GC, and the pathophysiology of AD.