The study was performed on 2-month-old male Dark Agouti (DA) rats (Supplementary Table S1). Animals were obtained from the animal facility of the Military Medical Academy, University of Defense, Serbia (RRID: RGD_21409748). All experimental procedures complied with ARRIVE guidelines and were performed in accordance with EU Directive 2010/63/EU. The Ethics Committee for the Care and Use of Laboratory Animals of the Faculty of Biology of the University of Belgrade issued a positive opinion under the number 119-01-4/11/2020-09. All animals were kept under standard conditions with 12-h light/dark cycles at constant ambient temperature (22 ± 2°C) and humidity. All other conditions were kept constant in all experimental units and throughout the study to avoid performance bias.

Experimental autoimmune encephalomyelitis was induced by intradermal injection of 150 μL of an encephalitogenic emulsion into the left hind paw. The encephalitogenic emulsion was prepared by mixing spinal cord homogenate [1 mg tissue/μl phosphate-buffered saline (PBS)] with the same volume of complete Freund’s adjuvant (CFA) supplemented with 4 mg/mL Mycobacterium tuberculosis, as previously described (Dragić et al., 2021a). Prior to immunization, animals were anesthetized by intraperitoneal injection of ketamine (50 mg/kg) and xylazine (10 mg/kg). Nothing was injected into the hindpaw of control rats.

The animals were daily weighed and scored according to the clinical EAE severity scale for motor disability (Bachmann et al., 1999): 1—limp-tailed, 1.5—clumsy, 2—mild monoparesis or paraparesis, 3—hind paw paralysis, 3.5—paraplegia or/and quadriparesis, 4—quadriplegia, 5—moribund state or death. Animals were sacrificed on average at 12–14 dpi when they reached the EAE score of 2.5–3, which corresponded to the state of paraplegia and/or quadriparesis (Figure 1B). The control group consisted of age-matched animals that underwent the same experimental protocol, except for the encephalitogenic injection. Complete epidemiological data are summarized in Supplementary Table S1.

Olfactory performance was analyzed by a buried food test (Kim et al., 2018; Machado et al., 2018). The test was performed daily, starting at −3 dpi until sacrifice. Animals had no access to food during the 17 h preceding the testing (17:00–08:00 h). The body mass was monitored daily to assure that testing did not impact their overall health condition. The feeding regimen resulted in <5% loss of the initial body mass (Supplementary Table S1). Each morning, the animals (n = 10 per experiment repeated three times) were brought into the experimental room, where they were allowed to adapt for 60 min. Each animal was placed in the experimental arena (box 50 cm × 50 cm × 50 cm, 3 cm thick bedding) and allowed to rest for 2 min. The animal was then moved to the clean cage while the pellet was buried 0.5 cm below the bedding in any corner of the test arena. The animal was again placed in the test arena, and activity was recorded for 5 min in dim light (>100 lux) with an ANY-MAZE camera. After each test, the bedding was changed, and the arena was cleaned with 70% ethanol for the next test. Animals that did not show typical exploratory behavior in BFT, which means they did not move from one spot of the arena during the trial at all, were excluded from the study.

The recorded material was analyzed after each measurement. For each animal, the average latency to find buried food, determined from −3 dpi to 0 dpi, was referred to as the baseline latency (s). The time taken by an animal to find buried food from 0 dpi afterwards was expressed relative to its baseline latency, i.e., each animal was a control to itself.

Odor discrimination was tested at 3 dpi and 7 dpi in the odor discrimination task (ODT), with slight modifications (Soffié and Lamberty, 1988). Baseline performance was determined for each animal (n = 10) on the day before immunization (−1 dpi). The rats were placed in the test room and had 60 min to adapt. The test arena consisted of two identical compartments (50 cm × 25 cm × 50 cm) connected by a passageway (10 cm × 10 cm). One compartment contained sawdust from the cage in which the tested animals had spent the last 24 h (familiar odor), while the other compartment contained sawdust from a cage of an unrelated group of animals (unfamiliar odor). The activity of the animal was recorded for 5 min with the ANY-MAZE camera under dim light, and the time (min) the animal spent in each compartment (familiar vs. unfamiliar) was measured. After each trial, the sawdust in the test arena was changed, and the sides of the test arena were cleaned with 70% ethanol. As a parameter for ODT, the discrimination index (DI) represents the difference in exploration time between the two compartments (unfamiliar compartment – familiar compartment) divided by the total amount of exploration in both compartments (unfamiliar compartment + familiar compartment) × 100 (%). DI values can range from −100 to 100%. The DI value close to zero indicates that the animal explores both compartments equally, which indicates loss of odor discrimination ability. The opposite is to be expected if an animal with intact olfactory function prefers to explore a particular compartment. Animals that tend to explore unfamiliar odors have high positive DI values, while animals that tend to spend more time in the compartment with the familiar odor have negative DI values. Data are presented as mean DI (±SD) from one determination in three experimental units.

Spontaneous behavior was analyzed in the open field arena at 7 dpi and was compared to baseline behavior determined at −1 dpi. The animals (n = 12) were habituated to a new environment for 2 h. Each animal was placed in the center of the empty black arena (100 × 100 × 50 cm), and activity was recorded continuously for 10 min using the ANY-MAZE camera under dim light (>100 lux). The sides and floor of the test arena were cleaned with 70% ethanol after each session. The following parameters were extracted by digitally analyzing the recorded video material: number of entries into central/peripheral fields/corners, time spent in central/peripheral fields/corners, distance traveled in central/peripheral fields, latency to enter central field, average speed in central/peripheral fields/corners, maximum speed in central/peripheral fields/corners, mobile time spent in central/peripheral fields/corners, immobile time spent in peripheral fields/corners, number of immobile episodes in peripheral fields/corners, number of freezing episodes in peripheral fields/corners and freezing time spent in peripheral fields/corners (freezing time was calculated if the animal was not moving for more than 10 s, while if the animal was not moving for more than 30 s it was considered as immobility).

The impact of neuroinflammation on general cognitive ability was tested with novel object recognition test (NORT) at 7 dpi. Each animal (n = 12) was placed in the center of the testing arena at the position equidistant from two identical rectangular objects (uniformly yellow-colored, diameter 5 × 5 × 20 cm). The animal was left to freely explore objects for 10 min and then returned to their home cages (sampling phase). For the testing phase, one object was replaced with a new one (uniformly red-colored conical object, 15 × 20 cm, d × h). The animal was returned to the test arena after 3 h and left to freely explore for another 10 min (testing phase). Animal activity was recorded with ANY-MAZE camera under dim light (>100 lux). After each session, both arena and objects were carefully cleaned with 70% alcohol to ensure the complete removal of any olfactory cues which could interfere with the testing. Percentage of time spent exploring the novel object relative to the total time spent exploring both objects is expressed as recognition index (RI) and was used to assess novel object recognition (Antunes and Biala, 2012).

The animals (n = 4/group) were sacrificed by decapitation. Brains were carefully dissected, fixed in 4% paraformaldehyde/0.1 phosphate buffer (pH 7.4) for 24 h, and cryoprotected/dehydrated in 10, 20, and 30% sucrose–phosphate buffer. Serial 25-μm coronal brain sections at stereotaxic coordinates Bregma 7.56–6.60 mm (Paxinos and Watson, 2013) were cut on a cryostat, mounted on SuperFrost^™ Plus adhesion microscope slides (Epredia Inc., United States), air-dried at RT for 1–2 h, and stored at −20°C until use.

For immunohistochemical staining, sections were left at RT for 30 min, rehydrated in PBS (pH 7.4) for 2 × 5 min, and incubated with 0.3% H₂O₂/methanol at RT for 20 min to inhibit endogenous peroxidase activity. After washing in PBS for 2 × 5 min, the sections were blocked with 5% normal donkey serum (NDS)/PBS for 1 h at RT. The sections were then incubated overnight with the primary antibodies at 4°C (Supplementary Table S2). After washing in PBS for 3 × 5 min, sections were incubated for 2 h at RT with HRP-conjugated secondary antibodies (Supplementary Table S2). 3,3-S-diaminobenzidine tetrachloride (DAB kit, Abcam, UK) was used as a chromogen for HRP-conjugated secondary antibodies. The reaction was stopped in PBS. Sections were incubated with thionin for 20 min at RT to counterstain the tissue. Sections were dehydrated in 50, 70, 95, and 100% ethanol for 5 min each, cleared in xylene for 2 × 5 min, and mounted with DPX medium (Sigma Aldrich, United States). Digital images (2,088 × 1,550 pixels) were acquired using LEITZ DM RB light microscope (Leica Mikroskopie & Systems GmbH, Wetzlar, Germany), LEICA DFC320 CCD camera (Leica Microsystems Ltd., Heerbrugg, Switzerland), and LEICA DFC Twain software (Leica, Germany).

For immunofluorescence staining, sections incubated with primary antibodies were incubated with fluorophore-labeled secondary antibodies (Supplementary Table S2) for 2 h at RT in a dark chamber. Sections were washed in PBS for 3 × 5 min and mounted with Mowiol (Sigma Aldrich, USA). Images (1,024 × 1,024) were captured using a confocal laser scanning microscope (LSM 510, Zeiss) with an argon laser (488 nm) and two helium–neon lasers (534 nm and 633 nm) at 40× or 63× magnification.

A number of Iba1⁺ and TH⁺ cells were counted in low-power micrographic images (40×) using the ImageJ Cell Counter plugin.¹ The number of immunopositive cells was counted in the corresponding OB layers in n = 7 sections per experimental group and expressed as a number of cells/mm². High-resolution images (1,024 × 1,024) were analyzed for integrated density in the JACoP plugin (Bolte and Cordelieres, 2006), and the degree of overlap between multiple signals was assessed by calculating the Pearson correlation coefficient (Dunn et al., 2011). The Pearson correlation coefficient (PSS) is a statistical parameter that reflects the co-occurrence of pixels from different channels, regardless of the intensity levels and signal offset. Seven to nine images/animal (four animals/experimental group) were acquired with the same laser gain and exposure. PCC analysis was performed at 40× or 63× magnification. Whole images were used for analysis. Prior to colocalization analysis, confocal images were adjusted by background correction (Landmann and Marbet, 2004).

Total RNA was extracted from OB tissue using TRIzol reagent (Invitrogen, Carlsbad, CA, United States), according to the manufacturer’s instructions. The concentration and purity of RNA were determined using OD260 and OD260/OD280, respectively. Complementary DNA (cDNA) was synthesized using the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, Waltham, MA, United States) according to the manufacturer’s instructions. Quantitative real-time PCR was performed by the ABI Prism 7000 Sequence Detection System (Applied Biosystems, United States) by mixing cDNA with primers (Supplementary Table S3) and Fast EvaGreen qPCR Mastermix (Applied Biological Materials Inc., Richmond, BC, Canada), under the following conditions: 10 min enzyme activation at 95°C, 40 cycles of 15 s denaturation at 95°C, 30 s annealing at 60°C, 30 s amplification at 72°C, and 5 s fluorescence measurement at 72°C. Samples from five animals/experimental group were run in duplicate. The relative expression of target genes was quantified using the 2^−ΔCt method relative to CycA, which was used as a reference gene. Data were expressed as fold-change of gene expression in EAE compared to control. At the end of each experiment, a melting curve analysis was performed to confirm the formation of a single PCR product.

Crude membrane fraction (P2) was isolated from OB tissue by using a previously established protocol (Gray and Whittaker, 1962). Briefly, OB tissue (g) was homogenated in nine volumes (ml) of ice-cold isolation buffer (0.32 M sucrose in 5 mM HEPES pH 7.4) supplemented with 0.5% w/v protease inhibitor cocktail (ab65621, Abcam, UK). The crude nuclear fraction (P1) was pelleted by centrifugation at 1,000 × g for 10 min, while centrifugation of the resulting supernatant at 16,000 × g for 30 min yielded the plasma membrane fraction (P2). The P2 pellet was resuspended in HEPES buffer, and the samples were stored at −20°C until use. Protein content was determined using the Micro BCA Protein Assay Kit (Thermo Fisher Scientific, Rockford, United States). Samples were aliquoted and stored at −80°C until use.

P2 samples adjusted to 1 mg/mL in 6× Laemmli buffer [4% sodium dodecyl sulfate (SDS), 0.02% bromophenol blue, 20% glycerol, and 125 mmol/L Tris–HCl] were boiled at 95°C for 5 min. About 20-μl aliquots were separated on 10% SDS-polyacrylamide gels and electrotransferred to 0.45 μm PVDF support membranes (Millipore, Germany) using the Trans-Blot Turbo Transfer System (Bio-Rad, Hercules, CA, United States). The support membranes were blocked for 1 h at RT with 5% non-fat dry milk (SERVA, Germany) in Tris-buffered saline containing 0.1% Tween-20 (TBST) and then incubated overnight at 4°C with the primary antibodies diluted in TBST (Supplementary Table S2). After washing in TBST for 3 × 5 min, the immunoreactivity profile was detected by incubating the support membranes at RT for 1 h with HRP-conjugated secondary antibodies diluted in TBST (Supplementary Table S2) and rinsing in TBST for 3 × 5 min at 4°C using the SmartBlot device. The chemiluminescent signal was detected with ECL solution (Bio-Rad, Hercules, CA, United States) using the ChemiDoc-It Imager (Ultra-Violet Products Ltd., Cambridge, United Kingdom). Primary and secondary antibodies were removed using a mild stripping protocol (0.2 mmol/L glycine, 0.1% SDS, and 1% Tween-20, pH 2.2) to blot the GAPDH, which was used as a loading control. The abundance of the target protein was quantified by densitometry using ImageJ software (NIH, Bethesda, MA, United States). The optical density (OD) of the target protein band was normalized to the OD of the GAPDH band in the same lane, and the ratio was expressed relative to the control sample (defined as 100%). Data were expressed as mean target protein abundance ± SD (%) of four samples/group blotted two times. Cropped PVDF support membranes presented in Figure 2 are presented integrally in Supplementary Figure S2.

Ectonucleotidase activities were determined by measuring the amount of inorganic phosphates (Pi) released by enzymatic hydrolysis of ATP, ADP, or AMP (Nedeljkovic et al., 2003), using the malachite green protocol as previously described (Grković et al., 2019). About 10-μl aliquots of the P2 sample (1 mg/mL protein concentration) were pre-incubated at 37°C in the assay buffer (10 mM HEPES pH 7.4 and 1 mM MgCl₂) for 10 min (for ATP and ADP hydrolysis) or 30 min (for AMP hydrolysis). The reaction was started by adding 1 mM nucleotide solution to the assay buffer. The reaction was stopped by adding 22.5 μL of 3 M perchloric acid. 20 μL of malachite green working solution (0.1% malachite green in 20% H₂SO₄, 7.5% ammonium molybdate, and 11% Tween-20 in a ratio of 10:2.5:0.2) was added to 80 μL of the reaction mixture and incubated at RT for 30 min to form a green-colored complex whose absorbance was measured at 620 nm. The amount of Pi released was calculated using the calibration curve generated with KH₂PO₄ as standard. The results were expressed as mean activity (nmol Pi/mg/min) ± SD, determined in four different P2 samples per experimental group, performed two times in duplicate.

The activity of adenosine deaminase (ADA) was determined by measuring the amount of NH₄⁺ ions released by the enzyme reaction (Dragic et al., 2022). Aliquots containing 20 μg of the P2 sample proteins were pre-incubated for 10 min at 37°C and the reaction was started by adding 1 mM adenosine. The reaction was carried out for 2 h at 37°C and stopped by adding 500 μL phenol nitroprusside reagent and 500 μL sodium hypochlorite reagent. After 15 min at 37°C, the absorbance, which is directly proportional to the NH₄⁺ concentration, was measured at 635 nm. The amount of NH₄⁺ was calculated from the calibration curve using NH₄Cl as a standard. The results were expressed as mean activity (nmol NH₄⁺/mg/h) ± SD, determined on four different P2 preparations per group, performed two times in duplicate.

A total of 90 animals were divided into 22 experimental units (4–5 animals/cage), of which 17 experimental units (66 animals) were randomly assigned to the EAE group, while the remaining five (18 animals) formed the control group (Supplementary Table S1). The sample size was estimated using G*Power analysis (Faul et al., 2007). All animals were weighed daily and examined for motor signs of EAE. Each animal was sacrificed on the day it had an EAE score of 2.5–3, which was an average at 12 dpi. In addition to weighing and EAE scoring, each experimental unit was used only for one particular behavioral test prior to sacrifice and preparation of OB samples (microscopic sections, mRNA extracts, and P2 membrane fractions). Samples from animals belonging to one experimental unit were not pooled and used only for one particular analysis. Each behavioral test and molecular analysis was repeated minimum of two times with different animals.

Quantitative data obtained within the study fell into the category of continuous data, with the exception of neurological scores, which were discrete. Data were analyzed using descriptive statistics and expressed as mean ± SD or SEM, as indicated in figure legends. The normality of data was tested using the Shapiro–Wilk test, and then appropriate parametric or non-parametric tests were used. For EAE to control group comparisons (gene expression level, WB, hydrolysis rate, PCC), the two-tailed Student’s t-test was used or Wilcoxon signed rank test if the normality condition was not met. Values of p < 0.05 were considered statistically significant. Neurological assessments (BFT, ODT, OFT, and NORT) were performed as repeated measurements in the EAE group only, and the collected data were compared within-subject and between-group. Differences were assessed using a repeated-measures ANOVA followed by a Dunnett’s post-hoc test, with the significance level set at 0.05 or with Mann–Whitney U test if the normality condition was not met, while for between-group comparison, the two-tailed Student’s t-test was used.

Relevant biases were identified, and methods were applied to reduce them. Confounding factors were mitigated by randomization to experimental units with equal and predetermined sample sizes sufficient to ensure statistical power without compromising the data quality or robustness of the study. Performance biases were attenuated by standardization and constant conditions in the experimental groups. The bias resulting from repeated measurements was mitigated by testing each animal at the same time of day each day. Neurological assessment was performed by two investigators who had no knowledge of the experimental groups, and other phases of behavioral assessment and data analysis were performed blindly. The OD and image analyses were performed using ImageJ.

All statistical analysis and graphing were performed using the GraphPad Prism 9.0 software package (San Diego, CA).

Animals were daily scored by EAE neurological scale, which reflects the severity of motor neurological symptoms (Figure 2B). In most animals, the first discrete motor signs appeared between 7 dpi and 9 dpi and then progressed fast in the next few days. Accordingly, the disease progression up to 8 dpi was referred to as the premotor phase, while the subsequent phase was referred to as the motor phase of EAE. Each animal was sacrificed when it reached the neurological score of 2.5–3, which corresponds to paraplegia and/or quadriparesis. At the time of sacrifice, EAE animals lost no more than 5.0% of their initial body mass (Figure 2B).

In parallel, animals were tested in a buried food test, which measures the time required by animals to find buried food. Animals were tested 3 days before EAE induction to establish their individual baseline latency (s). Each animal’s performance in BFT from 0 dpi afterwards was expressed as relative to the baseline latency (fold change) and the average daily latency ± (SD) was plotted in Figure 2B. A transient decline in olfactory performance was observed at ~2 dpi, as evidenced by more than 3-fold increase in latency to find buried food compared to the baseline latency before immunization (Figure 2B). After a brief recovery, a second, more pronounced increase in relative latency occurred at ~8 dpi, just with the appearance of discrete signs of motor impairment. Thereafter, relative latency in BFT increased in parallel with EAE score and development of motor impairments (Figure 2B). A one-way ANOVA with repeated measurements and Dunnett’s post-hoc comparison of the data revealed a statistically significant increase in latency to detect buried food at 3 dpi (t = 2.043, p < 0.021) and 7 dpi (t = 3.097, p < 0.0069) and thereafter compared to −1 dpi (i.e., before immunization).

Next, we examined the ability of animals to discriminate between familiar and non-familiar social odors (ODT) by determining the discrimination index (DI) (Figures 2C,D). Discrimination index of ~70% at −1 dpi indicated strong animals’ preference to explore the compartment with non-familiar odor. One-way ANOVA [F (1, 95, 10, 74) = 9.452, p < 0.0044] and Dunnett’s post-hoc tests showed a significant decrease in the DI values at 3 dpi (16.0 ± 41.3, p < 0.0132) and 7 dpi (24.9 ± 29.8, p < 0.0314), which indicated an reduced ability to discriminate social odors due to impaired olfactory ability.

The open field test (OFT) was used to evaluate general locomotion and anxiety-related behavior in EAE rats at 7 dpi, which reflect animal’s capacity and willingness to move. Parameters reflecting general locomotor activity, including distance traveled, average, and maximal speed, did not vary at 7 dpi with respect to pre-immunized animals (Table 1). Significant changes were observed in the amount of time spent in central and peripheral areas of the chamber and tendency of animals to remain close to walls, which strongly implied anxiety-like behavior (Figures 2E–I). EAE animals displayed less number of entries in peripheral fields (Figure 2F, t = 8.259, df = 9, p < 0.0001), increased time spent in peripheral fields (Figure 2G, W = 55, p = 0.002), and increased number of immobile episodes (Figure 2H, p < 0.01) and total immobile time spent in peripheral fields (Figure 2I, W = 45, p = 0.0039) compared to their performance at −1 dpi. All parameters obtained in OFT are summarized in Supplementary Table S4.

Novel object recognition test (NORT) was conducted to assess cognitive performance in EAE animals (Figure 2J). Intact animals have natural tendency to explore novel unfamiliar objects. At −1 dpi, animals spent greater amount of time with the novel object, while at 7 dpi, they did not show a preference for the novel object, demonstrated by significantly lower decrease in recognition index (RI) at 7 dpi compared to −1 dpi (t = 2.227, df = 9, p = 0.0488). Parameters obtained in NORT are summarized in Supplementary Table S5.

The observed decrease in olfactory performance and odor-related behavior in EAE prompted us to investigate the histopathological status of OB. It has been previously postulated that neuroinflammation decreased plasticity and renewal of interneurons in the OB are the underlying mechanisms of olfactory dysfunction in EAE (Tepavčević et al., 2011). The presence of infiltrated CD4⁺ cells in OB was investigated by CD4-directed immunohistochemistry (Figure 3). CD4⁺ cells were not present in healthy OB tissue; only a few CD4⁺ cells were restricted to the meningeal surface in the control OB (Figures 3A–C). In EAE sections, numerous CD4⁺ cells infiltrated the OB in the EAE (Figures 3D–F). CD4⁺ cells were accumulated along the axons of the olfactory nerve in the ONL, GL, and EPL (Figure 3E) and around the blood vessels (Figure 3F). Since massive infiltrates were seen on the EAE sections, the images are a qualitative confirmation of a disturbed blood–brain barrier and infiltration with peripheral cells.

Iba1 immunohistochemistry was performed to detect infiltrated monocytes/macrophages and resident microglia cells. In control OB, Iba1-ir labeled quiescent microglial cells, with small cell bodies and thin radial processes, which were evenly distributed throughout the OB (Figures 4A–C). In EAE, Iba1-ir was much more prominent throughout the OB, with distinct patterns of Iba1-ir in the superficial and deeper OB layers (Figure 4D). Specifically, in the ONL, Iba1-ir labeled numerous monocytes/macrophages and highly reactive ameboid microglia (Figures 4D,E), which were more abundant on the lateral side than on the medial side of the OB. In the GL, clusters of highly reactive microglial cells lost their domain organization (Figure 4F). However, in the EPL and GCL, especially in the parts that did not show infiltrates (Figures 4G–I), microglial cells showed slightly reactive morphology but preserved their domain organization (Figures 4G–I). Iba1-directed immunofluorescence confirmed differences in microglial shape/reactivity from the superficial to inner OB layers (Figures 4J–L). Quantification of Iba1-ir signal intensity shows the spatial gradient of Iba1⁺ cell density from the ONL to the innermost OB layer (Figure 4M), i.e., the density of Iba1⁺ cells observed gradually decreases from the ONL to the center of the OB.

Astrocyte reactivity in the inflamed OB was analyzed by GFAP immunohistochemistry (Figure 5). In the control OB, GFAP-ir labeled a network of OEC processes in the ONL and typical astrocytes in successive OB layers (Figure 5B). In the center of the OB, GFAP-ir moderately labeled glial tubes bordering the RMS (Figure 5C). In EAE, astrocytes showed signs of hypertrophy, reflected as moderately increased overall protein abundance of GFAP (Supplementary Figure S1), although the cells retained the non-overlapping arrangement (Figure 5D). In the ONL, GFAP-ir labeled ticked processes of OEC and slightly activated astrocytes (Figures 5E,F), while in the successive layers, astrocyte hypertrophy gradually decreased (Figures 5G,H), so that astrocytes in the GL showed a similar morphology to those of controls (Figure 5I, arrow).

As a consequence of the massive infiltration of immune cells and a strong activation of glial cells, we expected further signs of neuroinflammation. Quantitative RT-PCR was used to assess expression of the proinflammatory mediators of glial activation in EAE. Student’s t-test showed several-fold induction of genes encoding for TNF-α, IL-1β, IL-6, complement component C3, and lipocalin-2 (Lcn) (Table 1).

The plasticity of neuronal networks in OB was analyzed by immunohistochemistry directed to tyrosine hydroxylase (TH), doublecortin (DCX), and NTPDase2. Tyrosine hydroxylase was used as a marker for dopaminergic neurons. In the control OB, very strong TH-ir labeled small to medium cell bodies in the GL (Figure 6A) and the dense network of their neurites extending between the glomeruli (Figure 6B). TH-ir also stained sparse medium-sized neurons in EPL (Figure 6C) and GCL (Figure 6D). Although the overall intensity of TH-ir apparently decreased in EAE (Figures 6E–H; Supplementary Figure S3A), particularly in periglomerular neurites (Figure 6F), the number of TH-ir cell bodies increased in EAE in respect to control (p < 0.0001) (Supplementary Figure S3B).

Doublecortin (DCX) was used as a marker for actively dividing neuroblasts. In the control OB, DCX-ir labeled migrating type A cells in the RMS (Figures 7I–L), which are responsible for the continual replacement of granular and periglomerular interneurons. In EAE, the density of DCX-ir cells decreased significantly both in the RMS (Figure 7M) and in GCL and GL (Figures 7N,O). In addition to the reduced number, the DCX-ir cells in the GCL also changed their spatial arrangement (Figure 7N), while their processes showed conspicuous signs of atrophy (Figure 7O). Furthermore, DCX-ir cells in the RMS lost their bipolar shape and shortened processes (Figure 7P), suggesting a reduced replacement of GL and GCL interneurons and overall impaired plasticity in EAE (Table 2).

NTPDase2 immunohistochemistry was used to label type B cells, which are astrocyte-like type B cells that form glial tubes as a scaffold for the migration of type A neuroblasts in the RMS. In control OB, strong NTPDase2-ir labeled OEC cell bodies and processes in the ONL (Figures 7Q,R) and RMS (Figure 7T). In EAE, NTPDase2-ir was slightly less intense overall (Figure 7U), although the pattern of NTPDase2-ir in the ONL (Figure 7V) was comparable to that in the control OB. In the EPL, however, only slight somatic labeling was observed (Figure 7Z). The difference in NTPDase2-ir was most pronounced in the RMS, where NTPDase2-ir cells were less numerous and less conspicuous (Figure 7ZZZ).

We first determined the purinome expression in OB under control conditions and in EAE (Table 1). Significant changes in gene expression in EAE compared to control were observed for P2rx4, P2ry12, Adora1, Adora3, Entpd1, Entpd2, and Nt5e. Consistent changes at protein level were observed for eN/CD73 (Figure 7H, t = 4.40, df = 6, p = 0.0004), A₁R (Figure 7D, t = 6.80, df = 6, p < 0.01), A₃R (Figure 7G, t = 3.98, df = 6, p = 0.0073), and P2Y₁₂R (Figure 7C, t = 5.42, df = 6, p = 0.001). The abundance of P2X₇R (Figure 7A, t = 5.70, df = 6, p = 0.0013), P2Y₁R (Figure 7B, t = 4.47, df = 6, p = 0.0002), and A_2BR (Figure 7F, t = 5.15, df = 6, p = 0.0021) significantly increased at the protein level only. Representative PVDF membranes are presented in Supplementary Figure S2.

The activities of ectonucleotidase enzymes were further determined in OB. The rate of ATP hydrolysis (Figure 7I), indicative of both NTPDase2 and NTPDase1, significantly decreased in EAE (151.2 ± 19.4 nmol Pi/mg/min, t = 2.75, df = 7, p = 0.0285) compared to control (199.7 ± 33.33 nmol Pi/mg/min). On the other hand, the rate of ADP hydrolysis (Figure 7J), indicative of NTPDase1, increased in EAE (130.4 ± 9.916 nmol Pi/mg/min, t = 4.27, df = 7, p = 0.0037) compared to control (99.06 ± 12.2 nmol Pi/mg/min). The rates of AMP hydrolysis (Figure 7K, t = 0.56, df = 7, p = 0.5643) and adenosine deamination (Figure 7L, t = 0.66, df = 7, p = 0.4569) catalyzed by eN/CD73 and ADA, respectively, remained steady in EAE.

In the course of the research, we focused on the components of adenosine signaling whose expression is altered in EAE. Cell distribution was analyzed by immunohistochemistry and confocal immunofluorescence. In the control OB, strong eN/CD73-ir labeled the ONL parenchyma (Figures 8A,B), whereas the pattern of eN/CD73-ir in the GL (Figure 8B), EPL, and GCL (Figure 8C) indicated dominant astrocyte localization. Intense eN/CD73-ir labeled the astrocyte-like cells lining the glial tubes in the RMS (Figure 8A). The overall intensity of eN/CD73-ir fluorescence increased in EAE compared to control (Figure 8D). The intensity of parenchymal labeling increased in the ONL, although it was difficult to detect individual cell bodies and processes of OECs (Figure 8E). The eN/CD73-ir was more prominent in the GL, EPL, and GCL astrocytes (Figure 8F). At higher magnification, sporadic perisomatic labeling of tufted cells was also observed in the EPL (Figure 8H, arrows), whereas the immunoreaction was less pronounced in the RMS compared to the control (Figure 8I). Double immunofluorescence confirmed a significant overlap between eN/CD73-ir and GFAP-ir (Figure 8J), confirming an increase in astrocytic expression of eN/CD73. However, a substantial portion of eN/CD73-ir also overlapped with Iba1-ir, indicating expression in infiltrated immune cells and/or reactive microglia (Figures 8K,L). A distinct eN/CD73-ir pattern was observed around the bodies and proximal dendrites of granule cells, indicating a possible presynaptic localization in glutamatergic dendritic inputs of mitral and tufted cells (Figure 8M). Pearson’s correlation coefficient (PCC) was determined to assess the degree of overlap between two fluorescent signals (Figure 8N). A 2-fold increase in PCC value for the overlap of eN/CD73-GFAP in EAE compared to control (t = 3.33, df = 37, p = 0.0059) and the 3-fold increase in PCC value for the overlap of eN/CD73 and Iba1 (t = 3.19, df = 36, p = 0.0096) confirmed that the increase in eN/CD73 expression occurred in both reactive astrocytes and microglia.

As for adenosine receptors, significant basal expression of A₁R was detected in OB, especially in GL, EPL, and MCL (Figure 9A). At higher magnification, A₁R-ir was localized in the cell bodies of periglomerular neurons in the GL, tufted cells (Figure 9B), and mitral cells (Figure 9C). The intensity of A₁R-ir increased throughout the OB in EAE (Figure 9D). The more intense A₁R-ir was observed in the ONL parenchyma (Figure 9D) and in the cell bodies of tufted, mitral, and granule neurons (Figures 9E,F). The increase in the somatic localization of A₁R was confirmed by double immunofluorescence directed against A₁R and NeuN (green) (Figures 9G–J). The increase in A₁R fluorescence intensity was also observed in mitral cells, which are among the rare neurons that do not express NeuN and therefore remained unlabeled in the microscopic sections (Figure 9I, asterisks). Nevertheless, a significant part of extrasomatic A₁R fluorescence was observed in the EPL (Figure 9I), which harbors dendro-dendritic and dendrosomatic synapses between mitral/tufted cells and granule cells (Figure 9J, arrow). Alteration in presynaptic A₁R in EAE was examined with double immunofluorescence directed to A₁R and VGlut1, the latter being used as a marker for glutamatergic synapses. VGlut1 and A₁R immunofluorescence signals were closely localized in the EPL and IPL (Figures 9K,L), whereas a significant increase in PCC value for the overlap of A₁R-VGlut1 signals in EAE indicated an increase in synaptic expression of A₁R (Figure 9M, t = 3.42, df = 42, p = 0.0019).

A_2AR immunohistochemistry in OB showed similar patterns in control and EAE (Supplementary Figure S4). The pattern of A_2BR-ir largely overlapped with the pattern of GFAP-ir in control OB (Figures 10A,B), indicating a dominant astrocytic localization in GL (Figure 10C), EPL (Figure 10D), and GCL (Figure 10E). A pronounced A_2BR-ir was observed in the RMS (Figure 10F). In EAE, a slightly pronounced A_2BR-ir was observed in the ONL (Figure 10G), where it labeled OECs (Figure 10H) and infiltrated cells (Figure 10I). The pattern of A_2BR-ir in GL and EPL was similar to the control (Figures 10J–L). Slightly more pronounced A_2BR-ir was observed in the mitral cells (Figure 10L) and in the RMS (Figure 10M). Double immunofluorescence directed against A_2BR and GFAP confirmed a clear overlap between the two fluorescent signals (Figures 10N,O). The same was true for the overlap between A_2BR and Iba1 (Figures 10P,Q) in EAE. The high PCC value for the A_2BR-GFAP signal overlap and the low PCC value for the A_2BR/Iba1 signal overlap confirm the dominant astrocytic expression of A_2BR in the control OB (Figure 10R). However, the unchanged PCC value for A_2BR-GFAP overlap and the increase in PCC value for A_2BR/Iba1 signal overlap in EAE strongly suggested that the increase in A_2BR expression in EAE occurred in infiltrated cells and/or reactive microglia (Figure 10R, GFAP/A_2BR: t = 0.21, df = 39, p = 0.2172; Iba1/A_2BR: t = 2.27, df = 37, p = 0.0377).

With respect to A₃R, the A₃R-ir labeled periglomerular interneurons and granule cells (Figures 11A–C), whereas in the EPL, the apical dendrites of mitral and tufted cells extending perpendicular to the GL were most intensely stained (Figure 11D). The marked increase in A₃R-ir in EAE (Figure 11E) was consistent with the mRNA and protein expression data. However, in contrast to control, numerous A₃R-ir immune cells were seen in the ONL (Figure 11F). Periglomerular interneurons showed conspicuous A₃R-ir (Figure 11G), while the most conspicuous A₃R-ir was observed in GCL and RMS (Figure 11H). Double A₃R/Iba1 immunofluorescence confirmed the dominant expression of A₃R-ir in peripheral infiltrates and ameboid microglia in the ONL (Figure 11I), which was largely responsible for the observed A₃R upregulation, while ramified microglia in the EPL lacked the A₃R-ir (Figure 11J). Significant expression of A₃R-ir was detected in granule cells in the GCL (Figure 11M). Double A₃R/GFAP immunofluorescence showed low overlap between the two signals in ONL (Figure 11K) and EPL (Figure 11L). PCC value for A₃R and Iba1 signal overlap was significantly increased in EAE compared to control (Figure 11N, t = 5.12, df = 37, p = 0.0006), confirming that infiltrated cells accounted for the increase in A₃R-ir in EAE. On the other hand, the low PCC values for the signal overlap between A₃R and GFAP indicate sparse localization and unaltered expression of A₃R in astrocytes in EAE.

The present study provided evidence for increased expression of A₁R in EAE, mostly in mitral and tufted cells, where the receptor plays a functional role in neuronal circuits involved in the processing of odor information (Rotermund et al., 2019; Schubert et al., 2022; Wendlandt et al., 2023). Previous electrophysiological data show that tonic A₁R activation in mitral cells is responsible for slow hyperpolarizing oscillations of mitral cell membrane potential (Roux et al., 2015), which control glutamate release at their dendro-dendritic synapses with granule cells (Rotermund et al., 2018; Schulz et al., 2018). As a result, the granule cells reduce GABA release in the reciprocal synapses and attenuate recurrent mitral cell inhibition (Rotermund et al., 2018). Therefore, tonic A₁R activity provides a high output signal-to-noise ratio during olfactory stimulation (Rotermund et al., 2018). Therefore, it can be assumed that any deviation from baseline A₁R activity impacts the entire network via modulation of reciprocal synapses between mitral and granule cells. This assumption is supported by data showing that stimulation of granule cells enhances odor discrimination, while their inhibition impairs it (Abraham et al., 2010; Gschwend et al., 2015). Accordingly, A₁R-deficient mice have shown better olfactory performance compared to their wild-type counterparts (Schulz et al., 2018). On the other hand, A₁Rs have no effect on mitral cell bursting activity upon synaptic input from olfactory sensory axons (Rotermund et al., 2018, 2019). Instead, glomerular synaptic arrangements between olfactory sensory axons and mitral cell dendrites are regulated by ATP and adenosine via P2Y₁R and A_2AR, respectively, which are mainly expressed in periglomerular astrocytes and interneurons (Johansson et al., 1997; Kaelin-Lang et al., 1998). Our data showed no changes in gene expression for Adora2a or P2y1r, confirming the recent finding that they are not involved in EAE-induced impairment of olfaction (Wendlandt et al., 2023).

Our study also found the reduction in total TH-immunohistochemical labeling in the GL. About 10–16% of all GL interneurons express TH, and among them, the most numerous are axonless periglomerular interneurons with small cell body and dendrites extending to the neighboring glomerulus, and short axon surface (sSA) cells, which have a long axon extending to a variable number of glomeruli. Determination of the integrated TH-ir density showed that the decrease in TH-ir is due to the apparently weaker reactivity of neurites extending within and between glomeruli. DA neurons play an important role in odor processing as they inhibit excitatory transmission to OB excitatory (or principal) neurons (Borisovska et al., 2013; Banerjee et al., 2015; Liu et al., 2016; Capsoni et al., 2021). However, counting the number of TH⁺ cell bodies unexpectedly revealed an increased density of TH⁺ cell bodies in the GL. Therefore, it can be assumed that the reduced TH⁺ reactivity of DA neurites and the higher number of TH⁺ cells probably influence the resting activity of mitral cells in EAE animals. In healthy animals, periglomerular cells are continuously replaced by type A cells arriving from the SVZ via the RMS to OB. However, a previous study performed in EAE indicated that impaired migration of neuroblasts in the RMS may lead to reduced replacement of periglomerular cells in the OB (Tepavčević et al., 2011). Indeed, in our study, a reduced number of type A and type B neuroblasts was found in the RMS, along with a marked atrophy of granule cells in the GCL. Therefore, our data do not suggest an increased incorporation of new periglomerular cells in the GL, but rather some other compensatory mechanism(s), including cell type-specific enhancement of TH⁺ cell neurogenesis or altered fate specification. DA neurons are involved in substantial activity-dependent plasticity, which impacts the cell composition of DA population in OB through modulation of adult neurogenesis (Bovetti et al., 2009). Furthermore, the level of TH expression and dopamine release depend on the sensory input (for review, see Bonzano et al., 2016). While the molecular mechanisms of the activity-dependent plasticity of DA neurons remain to be elucidated, it is possible that peripheral infiltrates and reactive microglia release inflammatory mediators that affect newcomers arriving in the GL during EAE. Among other inflammatory mediators, it was shown that IL-6, which increased significantly in EAE in OB, plays a crucial role in adult DA neurogenesis and differentiation in OB (Storer et al., 2018). Taken together, the data suggest that the enhancement of A₁R-mediated signaling in principal output OB neurons and the altered connectivity of DA neurons likely influence resting network activity and contribute to impaired odor processing in EAE.

Regarding neuroinflammation, our study showed massive infiltration of the OB with immune cells and marked microgliosis in EAE, which are otherwise recognized as the cause of the early inflammatory lesions in MS/EAE (Kaskow and Baecher-Allan, 2018). A previous comprehensive assessment of the spatiotemporal pattern of CNS infiltration has shown that the initial recruitment of CD4⁺ T cells in EAE occurs via the meninges and choroid plexus rather than the vasculature (Brown and Sawchenko, 2007). While our data confirmed the presence of CD4⁺ T cells and monocytes/macrophages in the superficial layers of the OB in contact with the subarachnoid space, the presence of infiltrates in the subependymal zone and RMS could not be detected. The spatial arrangement of reactive microglia mirrored that of CD4⁺ T cells, with a decreasing gradient from the superficial to the innermost layers of the OB. Our additional observation that infiltration occurred more frequently on the lateral side than on the medial side of the OB supports the conclusion that the subarachnoid space was the main route for immune cell entry in the OB, at least at this stage of EAE. A recent study reported a structural feature that may predispose the OB as one of the earliest targets in MS/EAE (Fatuzzo et al., 2023; Spera et al., 2023). In particular, anatomical discontinuity of the arachnoid barrier at the level of the cribriform plate on the CNS side could be the cause of early infiltration of the OB with immune cells (Spera et al., 2023).

Our data show that infiltrated monocytes/macrophages abundantly express A₃R. Induction of Adora3 and A₃R has been previously demonstrated in peripheral lymphoid (Wei et al., 2013) and inflamed spinal cord tissue during EAE (Wei et al., 2013; Jakovljevic et al., 2017), as well as during ischemia, hypoxia, and subarachnoid hemorrhage (Coppi et al., 2023). Since A₃R controls neutrophils and T-cell activity (Butler et al., 2012; Gessi et al., 2013) and M1/M2 polarization in macrophages (Levy et al., 2006; Li et al., 2020), it may be also responsible for functional phenotype of activated microglia. Specifically, A₃R-mediated signaling controls the release of inflammatory cytokines, including TNF-α via p68/STAT6 pathway, thus contributing to development of functional phenotype of macrophages and microglia (Coppi et al., 2022).

The pattern of peripheral immune cell infiltration overlapped with the pattern of microgliosis. A significantly higher number of microglial cells was found in GL, EPL, and GCL, while their reactivity decreased with increasing distance from the lateral side of the OB. Microgliosis was induced by the cytokines of CD4⁺ T cells, i.e., IL-17 and IFN-γ, which promote inflammation by inducing local expression of proinflammatory cytokines by reactive microglia (Dogan and Karpus, 2004; Sindern, 2004; Murphy et al., 2010). Indeed, the present study reported several-fold induction of TNF-α, IL-1β, and IL-6, which are likely responsible for the observed changes in purinergic signaling (Adzic Bukvic et al., 2024). Several previous studies showed a significant increase in eN/CD73 (Lavrnja et al., 2015; Jakovljevic et al., 2017), A_2AR (Jakovljevic et al., 2017), A_2BR (Wei et al., 2013), and A₃R (Wei et al., 2013; Jakovljevic et al., 2017) in inflamed spinal cord tissue in EAE. The present study shows no changes in A_2AR expression in the OB, which is consistent with stable A_2AR signaling in the OB in EAE (Wendlandt et al., 2023). However, we have shown that highly reactive microglial cells express eN/CD73, A_2BR, and A₃R, which are not present in quiescent microglial cells under physiological conditions (Langer et al., 2008; Franco et al., 2021). Under basal conditions, eN/CD73 is localized in astrocytes and ependymal cells, while under neuroinflammatory conditions, it is expressed in ameboid microglia overexpressing arginase 1 and CD68 (Dragić et al., 2021b), characterizing the phagocytic microglial state (Franco and Fernández-Suárez, 2015). As for A_2BR, it is localized in astrocytes, oligodendrocytes, and neurons under physiological conditions, while it is expressed in reactive microglia during neuroinflammation. A_2BR stimulates the production of IL-6 in peripheral immune cells and plays pathogenic roles in MS/EAE by enhancing Th17 differentiation (Wei et al., 2013; Merighi et al., 2017) and functional polarization of microglial cells (Koscso et al., 2012; Franco et al., 2021). As for A₃R, activation in reactive microglia alters the expression of numerous immune-related genes, including the attenuation of TNF-α (Lee et al., 2006) and the increase in IL-10 production (Haskó et al., 1996, 2008). Our data also showed transcriptional upregulation of P2xr4 in EAE, which is the key determinant in the control of microglial activation and induction of microglial cell death (Vázquez-Villoldo et al., 2014). The A_2BR and A₃R are coupled to Gs and Gi/q, respectively, and therefore influence adenylate cyclase and cAMP levels in opposite ways (Fredholm et al., 2011). However, both receptors mobilize Ca²⁺ from intracellular stores via phospholipase C (Fredholm et al., 2011). Stimulation of P2X₄R in microglia also leads to an increase in intracellular Ca²⁺ and activation of p38-MAPK (Trang et al., 2009). Therefore, the simultaneous activation of A_2BR, A₃R, and P2X₄R in cells such as reactive microglia may not be contradictory from a physiological point of view. Overall, these data, together with the finding of increased expression of C3, involved in phagocytizing complement-opsonized synapses (Hammond et al., 2020), suggest that OB microglial cells in EAE develop a functional phenotype of phagocytizing microglia that promotes tissue repair and homeostasis. In conclusion, the exceptionally high abundance of purinergic receptors and nucleotide-degrading enzymes in OB underscores the importance of purinergic signaling for the olfactory system, and the present study adds to the existing knowledge on the contribution of adenosine signaling to the impairment of olfaction in EAE. Our results suggest enhanced adenosine signaling via A₁R in principal OB neurons, which may alter resting network activity in EAE. Our data further show enhanced adenosine signaling in microglial cells via the eN/CD73/A_2BR/A₃R molecular pathway(s) and induction of inflammatory mediators that may be responsible for the impairment of olfaction in EAE and likely in other CNS disorders. It is worth noting, however, that the present study did not address the causal relationship between these phenomena. However, not only does adenosine signaling contribute to the causative processes that lead to the main neurological and motor symptoms in MS/EAE, but the signaling has been linked to development of neuropsychiatric comorbidities characteristic of MS, such as anxiety and depression (Duarte-Silva et al., 2019; Gomes et al., 2021). Therefore, some degree of overlap and dysfunction in the adenosine signaling pathway would be expected as the mechanism responsible for the apparent neuroinflammation, olfactory dysfunction, and anxiety-related behavioral changes. However, further functional studies in genetic models are required for a complete assessment. Further investigation of olfactory dysfunction in MS/EAE may provide a valuable window for monitoring the disease onset and progression, while investigation of the functional role of purinergic signaling may point to potential molecular targets for therapeutic intervention in MS.