Menthol (Aldrich Chemical Co., Milwaukee, WI.) was dissolved in distilled water (1:1000 w/v). A closed system prototype was designed to allow the vaporization of fragrance compounds (15). Inhalation of menthol was scheduled for different time periods (from 1 week to months depending on the experiments), with 8 cycles of 15 minutes of inhalation per day (1 cycle of exposure every 3 hours) as previously described (15). Schematics depicting the experimental procedures were created in Biorender.com.

CD11c⁺ dendritic cells (DC) and CD8 T cells were obtained from the spleens of C57/BL6 and OT1 mice respectively, using magnetic separation columns according to the manufacturer’s specifications (Miltenyi Biotech, Germany). The purified CD11c⁺ or CD8⁺ T cells were used for antigen presentation experiments. Briefly, CD11c⁺ cells were incubated 2 hours at 37°C with 10 μg/ml of SIINFEKL peptide (Preprotech (UK). After washing, purified OTI CD8⁺ cells were added to the culture at a ratio 1:4 (CD11c: CD8). The co-culture was maintained for 24 hours to evaluate T cell proliferation (by [methyl-3H] thymidine incorporation) and IFNγ secretion (by ELISPOT) as previously described (15).

Six to eight-week-old BALB/c or C57BL/6 female mice (Envigo, Barcelona, Spain) were used to evaluate the impact of fragrance inhalation.

Male and female APP/PS1 mice were used to test the effect of menthol inhalation on memory. APP/PS1 (a recognized mouse model for AD (16, 17)) were bred and housed in the animal facility of the University of Navarra.

Male and female homozygous APP^NL-G-F mice (a new generation of AD mouse model (18)) were used to test the effect of Treg depletion in memory. The APP knock-in mouse (APP^{NL-G-F/NL-G-F}) carries the humanized APP including three mutations associated with familial AD: the Swedish “NL”, the Iberian “F”, and the Arctic “G” mutation. The levels of pathogenic Aβ in this mouse model are elevated due to the combined effects of these three mutations that promote Aβ toxicity by increasing total Aβ production (Swedish mutation), rising the Aβ42/Aβ40 ratio (Iberian mutation), and promoting Aβ aggregation (Arctic mutation). These mutations lead to Aβ deposition and age-associated cognitive impairment (18). These APP^NL-G-F mice were provided by RIKEN Brain Science Institute (Japan). A colony was maintained and housed in the animal facility of the University of Navarra.

Male and female Foxp3-DTR-GFP transgenic mice (B6.129(Cg)-Foxp3tm3(DTR/GFP)Ayr/J) were obtained from Jackson Laboratory. DTR-eGFP expression is observed in fully functional Foxp3⁺CD4+ T cell populations allowing fluorescent detection of Foxp3⁺ Treg cells. This model allows for specific transient elimination of Foxp3⁺ Treg by treatment with diphtheria toxin. OTI transgenic mice (C57BL/6-Tg(TcraTcrb)1100Mjb/Crl) expressing a transgenic T cell receptor designed to recognize the H2K^b-restricted cytotoxic T cell epitope from ovalbumin (residues 257-264) were purchased by Charles River. NSG mice were kindly provided by Dr. Melero (CIMA). All the experiments using male and female mice were balanced to avoid possible effects of sex. All these animals were inbred and housed in the animal facility of the University of Navarra.

All experiments were performed according to international animal care guidelines and with the approval of our local Ethics Committee for Animal Experimentation (Protocols 060c-19, 046-20, 087c-19, 097-20) and following the European Directive 2010/63/EU.

Contextual fear conditioning test, a quick and reliable method to assess memory in rodents (19), was used to evaluate the mouse cognitive function as previously described (20). Briefly, the test takes 3 days. On day 1 (habituation), mice were placed in the training chamber for 3 min. Twenty-four hours later, day 2 (training phase), mice were placed in the training chamber for 2 min. Subsequently, mice received a footshock (0.3 mA 2 s) and returned to their home cage. Long-term memory was evaluated during the test phase at day 3 after training. Lack of movement was defined as freezing. Freezing scores were expressed as percentages. The procedure was carried out in a StartFear system (Panlab S.L., Barcelona, Spain). The analogical signal is transmitted to the FREEZING and STARTLE software. In T cell depletion experiments, C57BL/6 mice were treated intraperitoneally with 300μg/mouse of anti-CD25 antibody (BioXcell) four days before the fear conditioning test. Treg depletion was confirmed by flow cytometry. Specific depletion of Treg was induced in Foxp3-DTR mice by a single i.p. injection of 1 μg/mouse of diphtheria toxin 4 days prior behavioral studies. Depletion of Treg was confirmed by flow cytometry. Treg inhibition experiments were conducted by i.p injection of the FOXP3 inhibitor peptide CM1315 (produced by Wuxi App tech (China)) as previously described (21).

Animals were perfused with PBS and prefrontal cortex and choroid plexus were dissected and frozen. RNA was purified and after the reverse transcription, cDNA amplification was performed on an Applied Biosystems CFX96 RT System using sense and antisense specific primers ( Supplementary Table 1 ) as previously described (15). Relative expression of target genes was determined using the formula 2^ΔCt, where ΔCt indicates the difference in the threshold cycle between the housekeeping gene (Cyclophilin A) and target genes.

Brain Aβ₄₂ levels (soluble and insoluble) were measured by using a sensitive sandwich ELISA kit (Invitrogen). Tissue (prefrontal cortex) was homogenized in a buffer containing 5 M guanidine HCl and 50 mM Tris-HCl, pH 8, protease inhibitors (Complete Protease Inhibitor Cocktail, Roche, Barcelona, Spain) and phosphatase inhibitors (0.1 mM Na3VO4, 1 mM NaF) and the assay was performed according to the manufacturer’s instructions.

Normality was assessed with Shapiro-Wilk W test. Statistical analyses were performed using parametric (Student´s t test and one-way ANOVA) and non-parametric (Mann-Whitney U and Kruskal-Wallis) tests. For all tests a p value <0.05 was considered statistically significant. Descriptive data for continuous variables are reported as means±SEM. GraphPad Prism for Windows was used for statistical analysis.

To confirm the immunostimulatory properties of exposure to menthol (15), C57BL/6 mice (6 weeks old) were immunized with OVA mixed with poly I:C. Then, mice were housed in vaporization cages and exposed to cycles of 15 min of menthol or water vapor (air control) every 3 h during 7 days (8 cycles per day, 15 min/cycle, n=12-15 mice per group). Ten days after immunization, mice were evaluated in their cognitive capacity by using the fear conditioning test ( Figure 1A ). After this analysis, mice were sacrificed to evaluate the immune response against the OVA cytotoxic T cell epitope SIINFEKL measuring the number of IFN-γ-producing cells by ELISPOT ( Figure 1B ). Mice exposed to menthol had significantly higher numbers of IFN-γ producing cells specific for SIINFEKL peptide. Notably, mice exposed to menthol exhibited significantly more freezing than control mice, suggesting an improvement in their cognitive capacity ( Figure 1C ). Interestingly, mRNA expression levels in the prefrontal cortex indicated a significant reduction of CD3 as well as IL-6 and IL-1β in mice exposed to menthol compared to control mice ( Figure 1D ), two cytokines that have been associated with cognitive decline in humans (9, 22, 23).

In an attempt to go deeper into this complex interaction, we studied the effect of anosmia on the immunomodulatory/cognitive effects of menthol ( Figure 2A ). It was reported that methimazole (MTZ) administration produces extensive degenerative changes in olfactory epithelium and a severe deficit in odor detection in rats or mice (24–26). We treated the mice with MTZ or with saline (n=20-24 mice per group) and ten days later, the olfactory epithelium was isolated and analyzed by H&E staining. As previously described, cilia atrophy (black arrows) and epithelial thickening (white arrows) in the olfactory epithelium was observed ( Figure 2B ). When a fear conditioning test was carried out in mice 10 days after a single injection of MTZ, a reduction in the percentage of freezing values was observed in anosmic mice, suggesting impairment in their cognitive capacity ( Figure 2C ). This impairment in memory capacity was associated with an increase in CD3, IL-6 and IL-1β mRNA expression in the prefrontal cortex of the mice (measured in a subgroup of 6-8 randomly selected mice, Figure 2D ).

We then studied the potential impact of MTZ on the immune system by measuring its effects in mice after immunization with OVA (n=12-15 mice per group, Figure 2E ). Although this immunization induced high numbers of IFN-γ producing cells specific for the cytotoxic T cell epitope SIINFEKL, this effect was dramatically impaired when mice were treated with a single dose of MTZ after immunization with OVA. Moreover, exposure to menthol odor in MTZ treated mice did not restore the immunogenicity of OVA previously observed ( Figure 2F ). Importantly, immune inhibition induced by MTZ treatment was accompanied by an impaired cognitive capacity, which was not recovered by exposure to menthol ( Figure 2G ). Freezing values in control mice in the experiments plotted in Figures 2C, G are different. This is probably due to the differences in the protocol and the environment to which the animals were subjected in both settings (27, 28). Despite these differences in both control groups, it can be concluded in both experiments that MTZ treatment negatively affects cognitive ability measured in the fear conditioning experiment. These data may suggest that anosmia causes immunosuppression, and has a deleterious effect on memory capacity.

To discard potential unspecific immunosuppressive effects of MTZ, we tested its effect on T cells and DCs in vitro and in vivo. Splenocytes isolated from naïve mice were stimulated with anti-CD3 antibodies in the presence/absence of 50, 10 or 1 μM of MTZ and T cell proliferation was measured. No significant changes in proliferation were observed upon MTZ treatment, suggesting that MTZ does not have a direct detrimental effect on T cells ( Supplementary Figure 1A ). We then treated naïve mice with 75 mg/kg of MTZ or saline and, one week later, splenocytes were stimulated in vitro with anti-CD3 antibodies to measure T cell proliferation. No significant differences were observed in the proliferative capacity of T cells ( Supplementary Figure 1B ). Similarly, we could not detect any inhibitory effect of MTZ on DC in vitro or in vivo. DCs isolated from naïve C57BL/6 mice and treated in vitro with 100 or 500 μM of MTZ and then pulsed with SIINFEKL peptide, stimulated OT1 T cell proliferation and IFN-γ production similarly to untreated control DC ( Supplementary Figure 1C ). Likewise, DCs purified from MTZ-treated mice and then pulsed with SIINFEKL peptide showed the same antigen presentation capacity as DC from untreated mice ( Supplementary Figure 1D ). DCs purified from MTZ or saline-treated mice responded similarly to LPS stimulation (not shown). These data indicate that MTZ has no direct impact on immune cells.

Once the link between the olfactory and the immune system was established, we aimed to study the effect of the immune system on memory capacity. First, we evaluated the fear memory of the highly immunodeficient NSG mice (NOD scid gamma mouse strain, with a mixed background between BALB/c and C57BL/6) lacking T cells, B cells and natural killer cells. BALB/c and C57BL/6 showed a similar % of freezing in the fear conditioning test ( Supplementary Figure 2A ). While BALB/c mice had normal freezing values that were improved by menthol exposure, NSG mice had a dramatic memory impairment that was not affected by menthol vaporization (n=10-15 mice per group, Figure 3A ). These data suggest that immune cells, either those absent (T, B or NK cells) or those defective (monocytes or macrophages) in NSG mice could have a role in learning function. Indeed, there is accumulating evidences on the implication of cytokines produced by immune cells in the behavioral response to stimuli, cognition and learning but also in the context of neurodegenerative diseases (reviewed in (13)). To evaluate the role of immune T cell activation, we carried out in vivo experiments of depletion or inhibition of T regulatory cells, a key specific T cell subpopulation maintaining immune tolerance by controlling the activation of T cells (29). In fact, it was suggested that the depletion of Treg cells might mitigate the neuroinflammatory response in murine models of AD and reverse the cognitive decline observed in these animals (30). Thus, C57BL/6 mice were treated with anti-CD25 antibodies four days before the fear conditioning experiment (n=16 mice per group). Interestingly, Treg depletion significantly improved mice’s cognitive capacity ( Figure 3B ). Similar results were found when using Foxp3DTR knock-in mice treated with diphtheria toxin to deplete Foxp3⁺ Treg cells (n=25 mice per group, Figure 3C ) or when mice were treated with the Treg inhibitory peptide CM1315 (31) (n=20 mice per group, 100 μgr/mouse per day, Figure 3D ). Diphtheria toxin administration in C57BL/6 wt mice does not affect cognitive capacity ( Supplementary Figure 2B ). Baruch et al. showed that transient Treg depletion affects the brain’s choroid plexus, a selective gateway for immune cell trafficking to the CNS (30). Thus, we studied the levels of expression of CD3, IL-6, and IL-1β mRNA expression in the choroid plexus (measured in a subgroup of 4-6 randomly selected mice). Interestingly, we found that CD25 cell depletion resulted in higher CD3 mRNA levels, suggesting an increase in T cell infiltration into the brain. Moreover, we found a reduction in IL-1β and an increase in IL-6 mRNA levels ( Figure 3F ). Similar results were found in mice treated with the Treg inhibitory peptide CM1315 ( Figure 3G ). These data suggest that IL-1β can have a detrimental effect on a fear-motivated learning task, results in agreement with previous observations in different murine models (32). Notably, we observed that intraperitoneal administration of the IL-1β inhibitor anakinra for 6 consecutive days (50 mg/kg per day) significantly improved the freezing time in our system (n=10 mice per group, Figure 3E ), suggesting a beneficial role of IL-1β blockade in learning capacity.

Prompted by the positive effects of Treg depletion in memory tasks, we tested the effect of Treg depletion in the behavior of 5-month-old APP^NL-G-F mice, already described to suffer an age-dependent memory impairment (33) ( Supplementary Figure 2C ). APP^NL-G-F mice were treated with saline or with anti-CD25 antibodies at months 3, 4 and 5 (one administration of 300 μg/mouse per month, n= 9-10 mice per group). Treg depletion was analyzed in the peripheral blood 6 days after each antibody administration. One week after the last anti-CD25 treatment at month 5, mice were evaluated in the freezing conditioning paradigm ( Figure 4A ). Interestingly, it was found that mice treated with the anti-CD25 antibody had a significantly higher percentage of freezing, suggesting an improvement in their cognitive capacity ( Figure 4B ). This improvement was associated with a significant increase in CD3 and a significant decrease of IL-1β and IL-6 mRNA in the prefrontal cortex ( Figure 4C ). This decrease of IL-1β and IL-6 mRNA was also observed in the choroid plexus ( Supplementary Figure 3 ). We then explored the effect of anti-CD25 on Aβ pathology in APP^NL-G-F mice. Total Aβ42 levels (soluble and insoluble) in the prefrontal cortex of these animals were determined by a sandwich ELISA. As shown in Supplementary Figure 4A , no differences were observed in Aβ42 levels in APP^NL-G-F mice treated with anti-CD25 compared with vehicle-treated mice.

We analyzed the effect of anakinra in a new cohort of APP^NL-G-F, (n=5-6 mice per group) but this time we used 9-month-old mice (having a very significant cognitive decline). Mice were treated with 15 consecutive doses of anakinra (one administration of 50 mg/kg per day) ( Figure 4D ) and evaluated later in the fear conditioning assay. Unfortunately, we could not see a beneficial effect of the IL-1β inhibition in this experimental setting ( Figure 4E ).

Once observed the effects of the immune cells on the memory of both young mice and the APP^NL-G-F mice and considering the effect of menthol on the immune system and cognitive capacity of healthy naïve mice, we aimed to study the effect of menthol inhalation in a different AD mouse model. In this case, we used the APP/PS1 mice that develop beta-amyloid deposits in the brain at 4 months of age, having a slower progression of the disease than the APP^NL-G-F mice but having also a cognitive decline at this age ( Supplementary Figure 2D ). In this 6-month-long experiment, APP/PS1 mice were exposed to menthol or water vapor for 6 weeks in total, 1 week per month, with 8 cycles of 15 minutes of inhalation per day (n=15-16 mice per group). A fear conditioning experiment was performed at the end of the experiment to compare the effect of menthol versus water vapor inhalation ( Figure 4F ). As previously described, APP/PS1 mice treated with water vapor (air control) showed a poor performance in fear-conditioning. However, 1-week exposure to menthol per month, during these 6 months was able to significantly prevent the cognitive impairment observed in the APP/PS1 mice treated with water vapor reaching freezing values similar to wild-type control mice ( Figure 4G ). Interestingly, the memory improvement was again associated with a significant reduction of IL-1β mRNA in the brain cortex of the animals exposed to menthol (measured in a subgroup of 6-8 randomly selected mice, Figure 4H ). We also analyzed the effect of anakinra in a new cohort of 9-month-old APP/PS1 mice. Mice were treated with 15 consecutive doses of anakinra (one administration per day) (n=5 mice per group, Figure 4I ) and evaluated later in the fear conditioning assay. Surprisingly, the blockade of IL-1β with anakinra in the APP/PS1 model achieved a very significant improvement in cognitive capacity in this experimental setting ( Figure 4F ).

In order to determine whether memory recovery observed in menthol APP/PS1 treated mice was associated with an amelioration of amyloid pathology, total levels of Aβ42 were determined in the prefrontal cortex of these animals. As shown in Supplementary Figure 4B , exposure to menthol did not reduce Aβ42 levels, indicating that the mechanism of improvement in cognitive ability is independent of this parameter.