Adult male Wistar rats (Harlan Laboratories, Horst, The Netherlands) between 13 and 14 weeks old and with an average weight of 403.5 g (± 18.01) at the time of surgery were used for this study. The study was approved by the University Animal Welfare Committee (CEEAH, protocol number 4848 P1) and was in compliance with the European Community Council Directive (86/609/CEE, 92/65/CEE, 2010/63/UE), Royal Decree 53/201. Rats were individually housed in a controlled environment (21 ± 1 °C; humidity, 60%; lights on from 8:00 a.m. to 8:00 p.m.; food and water available ad libitum).

A total of 27 rats underwent a stereotaxic procedure for intracerebroventricular injection and electrode implantation. The animals were divided into two groups: the vehicle injected control group (Control group, n = 11) and the STZ-injected group (STZ, n = 16). On day 34, the STZ-injected rats were further divided into two groups: those receiving ICSS treatment (STZ + ICSS group, n = 8) and those receiving sham treatment (STZ group, n = 8). Behavioral training in the Morris Water Maze (MWM) was conducted between days 34 and 41 (MWM1) and again between days 139 and 143 (MWM2). The Control group underwent the same behavioral tests in the MWM and sham treatment as the STZ rats. The experimental timeline is shown in Figure 1.

Animals were anesthetized with a mixture of 5% isoflurane and oxygen and subjected to stereotactic surgery. As described in our previous study (Riberas-Sánchez et al., 2024), 8 μL of STZ (2 mg/kg) (Merck Life Science, catalog number S0130-50MG) or vehicle (citrate buffer) were bilaterally injected into the lateral ventricles at a rate of 0.7 μL/min, at DV = −4.00 mm, with the cranial surface as the dorsal reference. Following the i.c.v. injection, a monopolar stainless steel electrode (150 μm in diameter) was unilaterally and chronically implanted into the MFB of the right lateral hypothalamus (AP = −2.3 mm; L = −1.8 mm; DV = −8.8 mm) (Vila-Solés et al., 2022).

Electrical brain stimulation consisted of 0.3-s trains of 50 Hz sinusoidal waves (40–175 μA). For each subject, the optimum current intensity (OI) was established as the mean of the two current intensities that produced the highest response rate in the two shaping sessions (responses/min) (see Vila-Soles et al. for details) (Vila-Solés et al., 2022). Rats in the STZ + ICSS group underwent intracranial self-stimulation (ICSS) treatment, consisting of an initial intensive phase (five post-training daily sessions contingent on each of the acquisition sessions in the MWM), followed by 13 non-training-contingent weekly sessions. In total, the MFB-ICSS treatment lasted 4 months. In each ICSS treatment session, rats were placed in a conventional Skinner box and allowed to press a lever to self-administer 2,500 trains of electrical stimulation at their individualized OI. Rats in the STZ and Control groups were handled and placed in the Skinner box for 30 min, without receiving electrical stimulation (sham treatment).

The MWM task was used to assess spatial learning and memory at the two previously described time points: approximately 1 month (MWM1) and 5 months (MWM2) after STZ administration. Both MWM1 and MWM2 consisted of 5 daily acquisition sessions (A1–A5), each consisting of 4 trials with a mean intertrial interval of 120 s. Seventy-two hours after the last acquisition session in each phase, animals completed a retention test (on days 41 and 146), which involved removing the platform and allowing the animals to swim for 60 s, starting from the East position. In addition, a 4-trial session following a cued procedure was conducted 3 days after the final retention test (day 149) to control for non-spatial factors that could affect the animal’s performance in the task, such as visual and motor abilities as well as their motivation to mount the platform. The principal quantitative outcome measures for acquisition and cued sessions were escape latency or time (seconds) required to locate and mount the platform. For the retention test, the percentage of time spent in the target quadrant (TQ) and in the target annulus (TA), the mean distance to target (MDT), and the Whishaw’s Error (WE) were analyzed. Additional control variables included the percentage of time spent near the walls (Walls), as a measure of thigmotaxis–anxiety, and the swimming speed across each trial and session.

Swim paths were recorded using a closed-circuit video camera and the Smart Video Tracking System (Version 3.0, Panlab). Swimming trajectories were analyzed during the fifth acquisition session of both MWM1 and MWM2. For each animal, the predominant search strategy across the four trials of the session was classified according to categorization by Rogers et al. (2017), as Non-spatial (Random swimming, Thigmotaxis), Semi-spatial (Scanning, Circling), or Spatial (Chaining, Goal-directed search, Direct swimming).

Serum samples were collected from all rats on days 29 (Ser0), 73 (Ser1) and 136 (Ser2) via lateral tail vein extraction, using Microvette tubes (Microvette® CB 300, SARSTEDT Sau). The tubes were kept at room temperature for 45 min and centrifuged at 3,000 rpm for 10 min to collect supernatant and obtain serum. Serum samples were stored at −80 °C until the isolation of total RNA using the mirVana PARIS Kit (Invitrogen™, ThermoFisher Scientific™). Tissue samples were obtained after the animals’ sacrifice with pentobarbital (200 mg/kg body weight, i.p.) on day 150. Ipsilateral brain hemispheres were fixed in 4% formaldehyde solution in phosphate buffered saline (PBS) for 24 h at 4 °C, post-fixed for 48 h and cryopreserved in 15 and 30% sucrose PBS solution subsequently. Finally, coronal sections of 35 μm and 12 μm thickness were obtained using a cryostat and stored at −80 °C until their use.

The human neuroblastoma SH-SY5Y cell line was purchased from the American Type Culture Collection (ATCC, CRL-2266, deposited by JL Biedler) and cultured in DMEM/F12 GlutaMAX™ (Dulbecco’s Modified Eagle Medium: Nutrient Mixture F12, Gibco™, ThermoFisher Scientific™) supplemented with 10% of heat-inactivated fetal serum bovine (FBS) (Gibco™, ThermoFisher Scientific™) and 1% penicillin–streptomycin (P/S) (Corning®, Cultek) in a humidified atmosphere containing 5% CO2 at 37 °C following manufacturer instructions.

For transfection experiments, SH-SY5Y cells were seeded in a 6-well plate (1×10^6 cells/well) and transfected with 30 nM of the miRNA-495 mimic (MC11526), miRNA-495 inhibitor (MH11526), the negative control mimic NEG-1 (NC mimic) (4464058) or negative control inhibitor (NC inhibitor) (4464076) (Invitrogen™, ThermoFisher Scientific™) using Lipofectamine™ RNAiMAX (Invitrogen™, ThermoFisher Scientific™) for 48 h according to the Forward protocol described by Invitrogen™. Each condition was tested in triplicate in three independent experiments.

For STZ administration, SH-SY5Y cells were seeded into 6-well plates and treated with 1 mM of STZ (Sigma-Aldrich, Merck) or vehicle (citrate buffer) during 24 h according to the protocol described by Plaschke et al. for human neuroblastoma cells (Plaschke and Kopitz, 2015). Each condition was tested in triplicate in three independent experiments.

Total RNA and protein from the SH-SY5Y cells were isolated using the mirVana PARIS Kit (Invitrogen™, ThermoFisher Scientific™), previous mechanical detachment using a scraper and 200 μL of Cell Disruption Buffer. RNA and protein samples were quantified using a NanoDrop 1,000 Spectrophotometer (ThermoFisher Scientific™) or Pierce BCA Protein Assay kit (ThermoFisher Scientific™) and a Synergy 4 spectrophotometer (BioTek), respectively.

From total RNAs, cDNAs were synthetized and preamplified with the TaqMan® Advanced miRNA cDNA Synthesis Kit (Applied Biosystems) in an AB Veriti 96-well thermocycler (Applied Biosystems). PCRs were run on an AB QuantStudio 7, using the TaqMan Advanced miRNA qPCR assays rno-miR-16-5p, hsa-miR-30-5p, rno-miR-132-3p, rno-miR-146a-5p, hsa-miR-181a-5p, rno-miR-181c-5p, hsa-miR-191-5p, hsa-miR-196a-5p and mmu-miR-495-3p or hsa-miR-495-3p (Applied Biosystems). Based on the most stable endogenous candidate identified by the NormFinder algorithm, hsa-miR-let-7a-5p (Applied Biosystems) was used as the normalizer to estimate the relative quantity of each target miRNA, which was calculated as 2^⁻ΔΔCт (ΔΔCт = ΔCt sample − ΔCt reference sample; ΔCt = Ct target − Ct normalizer). The mean level of Ser 0 in the Control group was used as the reference sample and set to 1 in the serum sample. In cell samples, the mean level in the Control group was used as the reference sample.

Ten μL of RNA from cell extracts, treated with TURBO DNA-free (ThermoFisher Scientific™), were used for cDNA synthesis with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems™, ThermoFisher Scientific™) and an AB Veriti 96 well thermocycler (Applied Biosystems™, ThermoFisher Scientific™). Quantitative reverse transcription PCR (qRT-PCR) was run on an AB QuantStudio 7, using the TaqMan Gene Expression Assays (Fatty acid elongase 7) (ELOVL7) (Hs00405151_m1); NRF2 (Hs00232352_m1); insulin like growth factor 1 receptor (IGF1R) (Hs00609566_m1); Tumor necrosis factor receptor superfamily, member 1B (TNFRSF1B) (Hs00961750_m1) and hypoxanthine phosphoribosyltransferase 1 (HPRT1) (Hs02800695_m1) (Applied Biosystems™, ThermoFisher Scientific™). The relative quantification of each mRNA was determined as 2^−ΔΔCт, where ΔΔCт it is calculated as ΔCt sample − ΔCt reference sample and ΔCt is defined as Ct target − Ct normalizer. The mean of the Control group was used as a reference sample, and HPRT mRNA levels served as endogenous normalizer.

A total of 15 μg of protein was loaded onto Criterion TGX Stain-Free PreCast Gels with 18-well comb (Bio-Rad) under reducing conditions, and subsequently electrotransferred to PVDF membranes. After 1 h of blocking with 5% Bovine Serum Albumin (Sigma-Aldrich, Merck) in TBS-T (tris-buffered saline [100mMNaCl, 10mMTris-HCl pH 7.5] containing 0.1% Tween-20), membrane was incubated with the primary antibody NRF2 (1:500, sc-365949, Santa Cruz Biotechnology) at 4 °C overnight. Membranes were then incubated for 1 h at room temperature with a peroxidase-conjugated goat anti-mouse secondary antibody (1:20,000, no. 115–035-044, Jackson ImmunoResearch). Intensities of antibody reactive bands were detected using Immobilon Western Chemiluminescent HRP Substrate (EMD Millipore) in a FluorChem luminometer, and quantified by densitometry using FluorChem SP software (AlphaEaseFC™). Protein relative intensities were normalized by total protein lanes as loading controls according to Aldridge et al. (2008).

Histological sections (12 μm thick) were stained using a filtered 0.5% cresyl violet solution for 3.5 min, rinsed and washed in distilled water for 5 min, and subsequently dehydrated and cover-slipped in Entellan mounting medium (Sigma-Aldrich, Merck). Photomicrographs of the corpus callosum region between Bregma coordinates −3.4 to −3.6 were acquired with a Vanox-T AH-2 microscope (Olympus). Quantification of corpus callosum thickness was performed in a blinded manner by two independent investigators using CellSens Standard imaging software. For each subject, corpus callosum thickness was calculated as the mean of the values obtained from three histological sections.

To assess NRF2 levels in the CA1 region of the hippocampus, free-floating immunohistochemistry was performed. Coronal sections of 35 μm in thickness were placed in 12-well plates containing Tris Buffer Saline (TBS) and fixed with 2% formaldehyde TBS for 20 min at room temperature (RT). Samples were incubated in 0.1% H₂O₂ TBS for 15 min followed by 7% Normal Goat Serum for 1 h at RT. Sections were then incubated with primary antibody anti-NRF2 (PA5-27882, Invitrogen™, ThermoFisher Scientific™) diluted 1:500 in TBS-T 1% BSA for 48 h. After washing with TBS-T, sections were incubated with a biotinylated secondary antibody (Goat anti-rabbit IgG, 111–066-144, Jackson ImmunoResearch) diluted 1:800 in TBS-T 1% BSA for 1.5 h under the same conditions. Subsequently, samples were incubated with ABC solution (Vectastain Elite ABC-Peroxidase kit, PK-6100, Vector Laboratories) for 40 min at RT followed by a 2-min incubation with DAB. Finally, sections were mounted onto gelatin-coated slides, dehydrated and cover-slipped with Histomount mounting medium (HS-103, National Diagnostics).

Photomicrographs of the hippocampal CA1 region between Bregma coordinates −3.2 to −3.8 were acquired using an Olympus Vanox-T AH-2 microscope. NRF2 immunoreactivity was quantified using Fiji (ImageJ, NIH), by measuring the mean gray value within circular ROIs manually placed over the pyramidal layer by two independent investigators in a blinded manner. For each animal, three ROIs were analyzed per section across three hippocampal sections.

Statistical analyses were performed using IBM SPSS Statistics 25. Performance in the MWM was analyzed using a 3 × 5 two-way mixed ANOVA (GROUP × SESSION) for the acquisition sessions, with the average score of four trials per session. Variables in the retention test, cued session (average score of the four trials), serum miRNAs and hippocampal CA1 NRF2 expression levels were analyzed using one-way ANOVA for parametric data or the Kruskal–Wallis test for nonparametric data (according to normality analysis results), followed by the appropriate post-hoc tests. A one-sample t-test against a constant was used to determine whether the percentage of time spent in the TQ differed from chance (25%) for each group. The distribution of swimming strategies was analyzed using the Chi-square test to assess group differences in categorical data. Serum miRNAs levels were longitudinally compared across the three time-points (Ser0, Ser1 and Ser2) and the experimental groups using a 3 × 3 two-way mixed ANOVA (GROUP × TIME). NRF2 expression levels in transfected SH-SY5Y cells were compared to the respective controls using an independent samples t-test, for parametric comparisons, or a Mann–Whitney U test, for nonparametric comparisons. Correlations between variables were assessed using Spearman’s bivariate correlation test. Results were considered to be statistically significant at p < 0.05, with a 95% confidence interval. The rank-biserial correlation (r), Hedges’s g (g), partial Eta squared (η_p²), Epsilon squared (ε²) or Cohen’s d (d) were used as effect size estimates according to the used test.

We analyzed the acquisition and retention of the spatial MWM task in a SAD-like rat model at two time points following STZ administration: 1 month (MWM1) and 5 months (MWM2). In MWM1, we also assessed the effects of ICSS administered post-training during the acquisition sessions (5 sessions), while MWM2 was conducted after completing 13 additional ICSS sessions, administered once per week and non-contingent on training.

As shown in Figure 2A, the acquisition results in MWM1 revealed significant differences between the three experimental groups (GROUP: F_2,23 = 8.01; p = 0.002, η_p² = 0.41), regardless of the session (GROUP × SESSION: F_8,92 = 0.28; p = 0.97, η_p² = 0.02). STZ rats, independent of ICSS treatment, displayed an impaired learning curve compared to Control rats (STZ: p = 0.001; STZ+ICSS: p = 0.01). No significant differences were detected between the STZ and STZ+ICSS groups (p = 0.30). However, intra-group analysis revealed that only the STZ group did not show a clear significant reduction in escape latencies between the first and the last training session (Control: p = 0.02; STZ: p = 0.08; STZ+ICSS: p = 0.03). Regarding the time spent near the walls, although all groups exhibited a linear downward trend over the 5 sessions (F_1,23 = 15.96; p = 0.001, η_p² = 0.41; GROUP × SESSION F_2,23 = 0.13; p = 0.88, η_p² = 0.01), there were differences between the groups (F_2,23 = 4.47; p = 0.02, η_p² = 0.28), with both STZ groups spending more time near the walls compared to the Control group (STZ: p = 0.03; STZ+ICSS: p = 0.01) (Figure 2B).

In the probe test conducted 72 h after the MWM1 last acquisition session, differences between groups were observed in the TQ (F_2,25 = 4.88, p = 0.02, η_p² = 0.30), TA (F_2,25 = 5.80, p = 0.009, η_p² = 0.34) and MDT (F_2,25 = 3.83, p = 0.04, η_p² = 0.25). Specifically, both STZ groups spent less time in the target quadrant (STZ: p = 0.02; STZ+ICSS: p = 0.01) and in the target annulus (STZ: p = 0.01; STZ+ICSS: p = 0.006), and had a greater mean distance to the target (STZ: p = 0.05; STZ+ICSS: p = 0.02) compared to the Control group. In the TQ test, only the Control group performed above the level of chance (25%) (Control: t₁₀ = 3.69, p = 0.004; STZ: t₆ = 0.41, p = 0.7; STZ+ICSS: t₇ = 0.51, p = 0.63, t-test) (Figure 2E). No significant differences were observed between groups in either the EW or the Walls variables (Figure 2F).

At 5 months, the group factor was found to be significant again (F_2,23 = 8.30, p = 0.002, η_p² = 0.42), regardless of session (GROUP × SESSION: F_8,92 = 0.97, p = 0.46, η_p² = 0.08) (Figure 2C) during the acquisition phase. STZ animals had significantly longer latencies than those in the Control group (p < 0.001). In contrast to the findings in MWM1, the STZ+ICSS group, after 13 additional ICSS sessions, performed better than the STZ group (p = 0.03). Intra-group analysis of latency progression across acquisition sessions showed that, once again, the STZ group does not improve its performance between the first and the last session (p = 0.13), whereas the Control (p = 0.02) and STZ+ICSS groups (p = 0.002) did. Regarding the time spent near the walls, all groups displayed similar durations, which remained consistent throughout the sessions (SESSION: F_4,92 = 1.00, p = 0.41, η_p² = 0.04; GROUP × SESSION: F_8,92 = 0.93, p = 0.50 η_p² = 0.07; GROUP: F_2,23 = 1.85, p = 0.18, η_p² = 0.14) (Figure 2D).

In the MWM2 retention, STZ rats continued to show poorer retention 72 h after the last acquisition session compared to Control rats. Group differences in retention test variables trended toward significance (TQ: F_2,25 = 2.38, p = 0.12, η_p² = 0.17; TA: F_2,25 = 2.63, p = 0.09, η_p² = 0.19; EW: F_2,25 = 3.10, p = 0.07, η_p² = 0.23), the STZ group spent less time in the TQ (p = 0.04) and had a lower WE (p = 0.03) compared to the Control group. Nevertheless, the STZ+ICSS group did not differ from the Control group. The Control and the STZ+ICSS groups performed above chance level (25%) in the TQ (Control: t₁₀ = 4.95, p = 0.001; STZ: t₆ = 2.16, p = 0.07; STZ+ICSS: t₇ = 3.30, p = 0.01) (Figure 2E). No differences were observed between groups in the time spent near the walls during this test (Figure 2F).

In the Cued session, no differences were observed between groups in either swim latencies (Figure 2G) or time spent near the walls (Figure 2H). Finally, no differences in swim speed were observed in any session, either during the spatial protocol or the Cued session.

As shown in Figure 3, trajectory analysis revealed differences between groups both at 1 and 5 months post-STZ administration (χ²₄ = 9.618, p = 0.047; χ²₄ = 10.803, p = 0.029, respectively). The STZ rats predominantly exhibited non-spatial or semi-spatial exploratory behaviors, whereas control rats consistently relied on hippocampal-dependent spatial strategies. In STZ rats receiving ICSS, a shift was observed from predominantly semi-spatial strategies in MWM1 to spatial strategies in MWM2.

We analyzed the serum expression of 9 miRNAs related to neural plasticity, postulated as potential biomarkers of AD, at different times post-STZ or vehicle administration. As shown in Figure 4, levels of miR-16, miR-30c, miR-181a, miR-191 and miR-196a were significantly upregulated in all STZ rats (both STZ or STZ+ICSS groups, as ICSS treatment had not yet been applied) compared to the Control group (U = 19, p = 0.007, r =0.55; U =13, p = 0.001, r = 0.63; t₂₂ = 2.15, p = 0.04, g = 0.88; U = 25, p = 0.02, r = 0.47; U = 16, p = 0.003, r = 0.61, respectively) at 29 days post-STZ (Ser0). However, no significant changes on the levels of miR-132, miR-146a, miR-181c and miR-495 were observed between groups at 29 days post-STZ.

No significant differences in serum miRNAs levels were observed among the experimental groups at 73 days (Ser1) or 136 days (Ser2) post-STZ (Supplementary Figure S1).

The correlation analysis between modulated serum miRNAs expression levels at 29 days after STZ administration, and behavioral performance in MWM2 for the experimental groups is summarized in Table 1. Ser0 levels of miR-16, miR-30c, miR-181a, and miR-191 and miR-196a were significantly associated with worse performance in MWM2 in the STZ+ICSS group. Specifically, Ser0 levels of these miRNAs positively correlated with latencies during acquisition sessions (all ρ > 0.7). Negative correlations were also observed between Ser0 levels of miR-16 and miR-181a, and the TA and TQ variables from the retention test of MWM2 (Figure 5). miR-181a negatively correlated with WE, while miR-16, miR-181a and miR-191 positively correlated with MDT. In the STZ group, Ser0 levels of miR-196a correlated with the TA variable. No correlations were observed in the Control group.

Correlations between Ser1 or Ser2 miRNA levels and behavioral performance during acquisition or retention of MWM2 are summarized in Tables 2, 3.

The analysis of serum miRNA expression levels across the three time points (Ser0, Ser1 and Ser2) showed that the 5-month time course of miR-146a and miR-495 were significantly different between the experimental groups (F_1,391 = 6.09, p = 0.01; F_1,137 = 5.88, p = 0.02, respectively) (Figure 6). Specifically, miR-146a and miR-495 serum levels increased from Ser1 to Ser2 in the STZ group (p = 0.03 and p = 0.04 respectively), an increase not observed in the STZ+ICCS and Control groups.

To further determine the potential gene targets of miR-495, transfection assays were performed in SH-SY5Y cells. Transfection efficiency was confirmed by a robust increase in miR-495 expression in cells transfected with the miR-495 mimic (Figure 7A).

Based on the bioinformatic analysis data from the TargetScanHuman (McGeary et al., 2019) and miRDB target (Chen and Wang, 2020) prediction algorithms, 4 candidate target genes of miR-495, ELOVL7, IGF1R, TNFRSF1B and NRF2, were selected for experimental validation due to their previously reported involvement in pathways related to the AD pathophysiology. Transfection with either miR-495 mimic or inhibitor did not alter mRNA levels of ELOVL7, IGF1R and TNFRSF1B. Nonetheless, mRNA levels of NRF2 were significantly reduced in cells transfected with miR-495 mimic (t₄ = 2.58, p = 0.03, d = 2.10) compared to control (Figure 7B). Levels of NRF2 protein were also significantly downregulated following miR-495 mimic transfection (t₄ = 3.30, p = 0.03, d = 2.70), and upregulated after miR-495 inhibitor transfection (t₄ = −3.07, p = 0.04, d = −2.50) in SH-SY5Y cells (Figures 8A,B).

To analyze the effects of STZ administration on miR-495 and NRF2 expression, SH-SY5Y cells were exposed to STZ and compared to vehicle-exposed controls. STZ-exposed cells presented a marked increase in miR-495 expression (t_8.593 = −3.534, p = 0.007, g = 1.43) (Figure 7C), accompanied by a significant downregulation of NRF2 protein levels (t_6,922 = 2.746, p = 0.03, g = 1.42) (Figures 8C,D).

Immunohistochemical analysis of NRF2 expression in the CA1 region of the hippocampus revealed significant differences between the three experimental groups (H₂ = 10.11, ε² = 0.35, p = 0.006). In STZ animals, NRF2 levels were downregulated compared to the Control group (p = 0.002). Notably, STZ+ICSS rats exhibited significantly higher NRF2 levels than STZ rats (p = 0.045) (Figure 9A).

Correlation analysis between NRF2 levels in CA1 and serum miRNAs levels at 5 months post-STZ injection showed a positive correlation with miR-495 (ρ = 0.83, p = 0.02) within the STZ group, with no significant correlations observed for the other serum miRNAs.

Nissl staining showed a significant reduction in corpus callosum thickness in STZ rats compared to the Control group (t₁₅ = 1.88, p = 0.04, g = 0.88). Within the STZ group, corpus callosum thickness correlated positively with TQ (ρ = 0.83, p = 0.04) and negatively with MDT (ρ = −0.83, p = 0.04). STZ rats treated with MFB-ICSS showed a significant increase in corpus callosum thickness compared to untreated STZ rats (t₁₃ = 1.71, p = 0.05, g = 0.83) (Figure 9B).