The experimental study protocol was approved by the Institutional Animal Welfare Committee of the Institut de Recerca i Tecnologia Agroalimentàries (CEEA-IRTA, registration number CEEA 384-2023) and the Ethical Commission of Animal Experimentation of the Government of Catalonia (ethical approval n°: CEA-OH/12097/2). The study was conducted by certified staff, and all experiments involving SARS-CoV-2 were performed at the BSL-3 facilities of the Biocontainment Unit of IRTA-CReSA (Barcelona, Spain).

The SARS-CoV-2 isolate used in these studies, identified as Cat02 (GISAID ID EPI_ISL-471472), was isolated from a nasopharyngeal aspirate of a laboratory-confirmed COVID-19 patient in Barcelona, Spain, in 2020. This B.1 lineage isolate has the following point mutations compared to the Wuhan strain: D614G in the Spike protein, K837N in NSP3, and P323L in NSP12.

Production of virus stock (passage number 3), isolation, titration, and virus neutralization assays were conducted using Vero E6 cells (ATCC® repository, CRL-1586™). The infectious virus titer was determined by calculating the 50% Tissue Culture Infectious Dose (TCID₅₀) endpoint, and results were expressed as TCID₅₀/mL (Reed and Muench, 1938).

The preparation of virus and cells was conducted as previously described (Brustolin et al., 2021), using Dulbecco’s modified Eagle Medium (DMEM; Lonza) supplemented with 5% fetal bovine serum (FBS; 35-079-CV, Corning^®), 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM glutamine (all from Gibco Life Technologies).

The experimental design is depicted in Figure 1. A total of 68 GSH (Janvier Labs), aged 6-to-8 weeks, including 34 males and 34 females, were used in this study. Following a one-week acclimatization period, 40 hamsters (20 males and 20 females) were intranasally inoculated under isoflurane anesthesia with 100 μL of SARS-CoV-2 Cat02 isolate (10⁴ TCID₅₀, 50 μL per nostril). An additional 28 hamsters (14 males, 14 females) were intranasally mock-inoculated with the same volume of phosphate-buffered saline (PBS) and served as negative controls.

Clinical signs and body weight were recorded prior to challenge, daily until 10 dpi, three times per week from 11 to 21 dpi, and weekly thereafter until 60 dpi. Oropharyngeal swabs (OS) were collected before challenge and weekly up to 60 dpi for quantitative analysis of SARS-CoV-2 viral shedding.

Neurocognitive function was assessed using four behavioral tests: nest-building test (NBT), burrowing test (BT), open field test (OFT), and elevated zero maze test (EZMT), conducted pre-inoculation and at 25, 39, and 59 dpi. The NBT and BT evaluated innate behaviors that are sensitive to brain lesions, motor deficits, and general welfare (Deacon, 2012), while the OFT and EZMT assessed locomotor activity and anxiety-related behaviors (Shepherd et al., 1994; Seibenhener and Wooten, 2015).

For necropsy and sample collection, four mock-inoculated and four SARS-CoV-2-inoculated hamsters (with equal numbers of males and females) were euthanized via intraperitoneal pentobarbital injection at 4, 7, 14, and 21 dpi. The same procedure was performed at 32, 46, and 60 dpi, using four controls and eight inoculated hamsters at each time point (with equal numbers of males and females).

For immunological and virological analyses, hamsters were sampled for blood, OS, and tissue samples, including the left nasal turbinate, left caudal hemibrain, left caudal lung, heart, small intestine, right hind limb muscle, and spleen.

Blood was collected into 2 mL cryotubes containing 200 μL 0.5 M EDTA, then centrifuged at 1,000 × g for 10 min (min) to separate plasma. Plasma was aliquoted into 1.5 mL cryotubes and stored at −80 °C.

OS and tissue samples (except for the spleen) were placed in 2 mL cryotubes containing 1,000 μL of DMEM (Lonza, Basel, Switzerland) supplemented with 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM glutamine (all from Gibco Life Technologies, Madrid, Spain). OS were resuspended by vortexing for 30 s (s). A 3 mm tungsten bead (69997, QIAGEN, Hilden, Germany) was added to tissue samples except the spleen, which were homogenized at 25 Hz for 30 s using a TissueLyser II (QIAGEN, Hilden, Germany), then centrifuged at 2,000 × g for 2 min. Samples were stored at −80 °C until further analysis.

Hamster spleens were collected in RPMI 1640 (11875093, ThermoFisher Scientific), supplemented with 20% FBS (35-079-CV, Corning^®), 1% glutamine, and 1% penicillin/streptomycin (all from Gibco Life Technologies). All steps were performed on ice. Splenocytes were isolated using a combination of mechanical and enzymatic digestion. Briefly, spleens were dissociated by gentle straining through a 40 μm cell strainer (352340, Corning^®). Red blood cells were lysed using a red blood cell lysis buffer (17 mM TRIS, 0.14 M NH₄Cl, adjusted to pH 7.4). Splenocytes were washed and resuspended in FBS (35-079-CV, Corning^®) containing 10% dimethyl sulfoxide (DMSO; D2650, Sigma-Aldrich™) and quickly transferred to cryovials gradually frozen to −80 °C overnight. Samples were then transferred to liquid nitrogen (−196 °C) for long-term storage until further use.

For histopathological examination, the right nasal turbinate, right hemibrain, right lung, heart, small intestine, right posterior limb muscle, and cervical region containing the cervical vagus nerve (CVN) were collected in specific containers (3178–600-01, Biopsafe®, Axlab, Vedbæk, Denmark) and fixed by immersion in 10% buffered formalin for 7 days prior to processing.

Viral RNA was extracted using the IndiMag Pathogen Kit (Indical Bioscience) on a Biosprint 96 workstation (QIAGEN), following the manufacturer’s instructions. Detection of SARS-CoV-2 genomic RNA (gRNA) was performed by reverse transcription quantitative polymerase chain reaction (RT-qPCR) on OS, nasal turbinate, lung, brain, heart, intestine, and muscle samples, as previously described (Corman et al., 2020).

RT-qPCR was conducted using AgPath-ID™ One-Step RT-PCR Reagents (Applied Biosystems, Life Technologies), targeting a segment of the envelope protein gene (UpE; GenBank NC_004718, positions 26,141–26,253). The assay employed the forward primer (5′-ACAGGTACGTTAATAGTTAATAGCGT-3′ [400 nM]), reverse primer (5′-ATATTGCAGCAGTACGCACACA-3′ [400 nM]), and probe (5′-FAM-ACACTAGCCATCCTTA CTGCGCTTCG-TAMRA-3′ [200 nM]). Thermal cycling conditions included reverse transcription at 50 °C for 10 min, initial denaturation at 95 °C for 10 min, followed by 45 cycles of 94 °C for 15 s (s) and 58 °C for 30 s on a 7,500 Fast Real-Time PCR System (Applied Biosystems, Life Technologies).

For absolute quantification, a standard curve was generated using serial 10-fold dilutions of an E gene plasmid (10006896, 2019-nCoV_E_Positive Control, IDT), run in parallel with all PCR assays. Each sample was analyzed in duplicate, with Ct differences of less than one cycle considered acceptable.

Supernatants from lysed samples exhibiting Ct values lower than 36 were analyzed for infectious virus by titration in Vero E6 cells, as previously described (Brustolin et al., 2021). Briefly, each sample was serially diluted 10-fold in duplicate and transferred to a 96-well plate pre-seeded with Vero E6 cells. Following incubation at 37 °C with 5% CO₂ for 5 days, plates were examined for cytopathic effect, and the infectious viral titer was determined by calculating the TCID₅₀, according to the Reed-Muench method (Reed and Muench, 1938).

After fixation, the nasal turbinate, brain, lung, heart, intestine, muscle, and cervical region were routinely processed to obtain 4 μm-thick, formalin-fixed, paraffin-embedded sections for histopathology. Prior to standard processing, nasal turbinate and cervical region samples underwent decalcification in decalcifying solution (10-6553, Casa Álvarez) for 10 and 24 h (h), respectively. Cervical region samples were then sectioned into 2 mm-thick slices, placed in embedding cassettes, and subjected to a second decalcification cycle.

Hematoxylin and eosin (H&E) staining was performed on all tissue sections to assess inflammation resulting from SARS-CoV-2 infection using an optical microscope. Inflammation severity was scored with a semi-quantitative system (0–3), based on previously published criteria (Brustolin et al., 2021; Vidal et al., 2022): 0 = absent, 1 = mild, 2 = moderate, and 3 = severe.

Immunohistochemistry (IHC) was conducted on sections of nasal turbinate, lung, brain, heart, intestine, and muscle to detect SARS-CoV-2 nucleocapsid (N) protein using a rabbit monoclonal antibody (40143-R019, Sino Biological) at a 1:10000 dilution (Rockx et al., 2020). Viral antigen levels were assessed using a semi-quantitative scale (0–3), based on previously described criteria (Brustolin et al., 2021; Vidal et al., 2022): 0 = absent, 1 = low, 2 = moderate, and 3 = high.

H&E-stained sections from the cervical region of animals euthanized at 60 dpi were digitized using a slide scanner (Aperio, Leica Biosystems) and assessed blindly with Aperio ImageScope software (v12.4.6.5003, Leica Biosystems). The CVN was identified based on its anatomical location: dorsolateral to the carotid artery, medial to the jugular vein, and adjacent to the sympathetic trunk (Planitzer et al., 2017). At a standardized anatomical level, defined by the presence of thymic tissue, the diameter of the CVN was measured perpendicularly to the orientation of the nerve fibers (Supplementary Figure S1). Two animals, male identification (ID) 56 and female ID 77, were excluded from analysis due to the absence of visible thymic tissue in their sections, which would have compromised the consistency of diameter measurements.

To ensure cytokine quantification was conducted within the required protein concentration (1 mg/mL), total protein concentrations in lung and brain homogenates were measured prior to the Luminex assay using the Pierce™ Bicinchoninic Acid (BCA) Protein Assay kit (23227, ThermoFisher Scientific), with bovine serum albumin as the standard. Absorbance was measured at 562 nm using a PowerWave XS Microplate spectrophotometer (BioTek).

A 9-Plex Multiplex Luminex Assay (Hamster Cytokine Panel-1, K22608, Ampersand Biosciences) was used to quantify a panel of cytokines and chemokines – IL-2, IL-4, IL-10, IL-6, IFN-γ, MCP-1, MIP-1α, TNF-α, and VEGF – in plasma, lung, and brain samples collected at 4, 32, and 60 dpi. Plasma samples were diluted 1:5 in sample dilution buffer, while brain and lung homogenates were diluted 1:10 and 1:15 in Mili-Q H₂O, respectively, to achieve an estimated final protein concentration of 1 mg/mL. All samples were analyzed in triplicate according to the manufacturer’s instructions, and plates were read using the Luminex MagPix system (Luminex Corporation), equipped with RBM Plate Viewer software (v4.1.0.16979, Rules Based Medicine).

A hamster-specific ELISA kit (HMFI0011, Assay Genie) was used to quantify IL-1β in the same sample sets. Samples were diluted 1:2 in sample dilution buffer and analyzed in triplicate. Absorbance was measured at 450 nm on a PowerWave XS Microplate spectrophotometer (BioTek).

Results were normalized to total protein content and expressed as pg./mg protein for tissue homogenates and pg./mL for plasma samples.

Neutralizing activity in plasma was assessed in duplicate using the cPass SARS-CoV-2 Neutralization Antibody Detection Kit (L00847; GenScript), following the manufacturer’s instructions. Optical density (OD) at 450 nm was measured using a PowerWave XS Microplate spectrophotometer (BioTek). The assay included positive and negative controls provided by the manufacturer, which met all quality control criteria, as well as a SARS-CoV-2 neutralizing antibody calibrator (L00487; GenScript) for standard curve generation. Neutralizing activity was calculated as % inhibition = (1 − OD sample / OD negative control) × 100%, with values <30% considered negative, and antibody levels reported as semi-quantitative units/mL based on the standard curve. This assay was performed in all animals from both the negative control and SARS-CoV-2 groups at 4, 32, and 60 dpi, with the exception of one male in the SARS-CoV-2 group at 4 dpi due to insufficient plasma volume.

The ELIspot assay was performed using splenocytes collected at 60 dpi with the hamster IFN-γ kit (3102-4APW-2, Mabtech) according to the manufacturer’s instructions. Briefly, cryopreserved splenocytes were quickly thawed and resuspended in RPMI 1640 (11875093, ThermoFisher Scientific) supplemented with 10% FBS (35-079-CV, Corning^®), 1% glutamine, and 1% penicillin/streptomycin (all from Gibco Life Technologies). Cells were counted and plated at a concentration of approximately 2 × 10⁵ cells/100 μL/well in 96-well ELISpot plates. Cells were stimulated with either SARS-CoV-2 Spike protein (10 μg/mL; 40589-V08B1, Sino Biological), Nucleocapsid protein (10 μg/mL; 40588-V08B, Sino Biological), Concanavalin A (ConA, 7 μg/mL; L7647, Sigma-Aldrich™) as a positive control, and supplemented RPMI 1640 medium alone as a negative control, all in triplicate. Following 24 h of incubation at 37 °C, cells were removed, and membranes were developed according to the kit protocol to detect IFNγ spots. Spots were quantified using an AID iSpot Reader and AID ELISpot Software version 8.0 (Autoimmun Diagnostika GmbH). Assay validity was confirmed by a minimum of 100 spots per well in the ConA-stimulated positive control, with responses at least threefold higher than in unstimulated wells. Due to poor splenocyte isolation, the assay was successfully completed in 5 of 8 inoculated animals and 3 of 4 mock-inoculated animals euthanized at 60 dpi. ELISpot results were background-corrected by subtracting spot counts from paired unstimulated wells.

Four neurocognitive tests—NBT, BT, OFT, and EZMT—were conducted at baseline (pre-inoculation) and during the post-acute phase of infection. These tests were chosen for their objective, quantifiable readouts, and demonstrated consistency across repeated trials (Tucker and McCabe, 2017). Tests were distributed over two consecutive days, progressing from least to most stressful. On day 1, animals underwent the BT in the morning, followed by the provision of a pressed cotton square (nestlet) for the NBT. On day 2, nests were scored in the morning, followed by the OFT, and, subsequently, the EZMT in the afternoon.

Due to the time-intensive nature of the behavioral tests and scheduling limitations within the BSL-3 facility, mock-inoculated controls (n = 4, two males and two females) were assessed on 17–18, 31–32, and 44–45 dpi, while SARS-CoV-2–inoculated hamsters (n = 8, four males and four females) were tested on 24–25, 38–39, and 58–59 dpi. Two-week intervals between sessions were implemented to minimize habituation and learning effects from repeated exposure (Deacon, 2012).

Hamsters were individually housed on minimal bedding (~2 cm high) in the procedure room 72 h prior to testing to facilitate acclimatization. All assays were performed under 70% dimmed lighting by the same observers. Following completion of testing on day 2, animals were returned to group housing. To minimize odors, the equipment was wiped with 5% ethanol between subjects and disinfected with PeraSafe™ between sessions.

For the BT, a PVC tube (68 mm in diameter, 200 mm in length, sealed at one end, with the open end raised 3 cm) containing 120 g of clay balls (8424084237824, Massó Garden) was placed inside each cage (Supplementary Video S1). After 30 min, the remaining material inside the tube was weighed following established protocols. This test assesses the natural burrowing behavior of rodents, which can be impaired by neurodegenerative deficit (Deacon, 2012).

In the NBT, a standard nestlet (NES3600, Ancare) was provided for 24 h without additional enrichment material (Supplementary Figure S2). Nests were scored on a five-point scale: 1 = untouched; 2 = < 50% shredded; 3 = 50–90% shredded without a distinct nest; 4 = > 90% shredded with material gathered into a flat nest; 5 = > 90% shredded forming a crater-shaped nest, as previously described. This test evaluates the rodents’ capacity to build nests, a behavior sensitive to neurodegenerative lesions (Deacon, 2012).

The OFT consisted in placing each hamster individually in the center of a white, opaque, square methacrylate arena (100 cm × 100 cm × 40 cm) and recording spontaneous behavior for 10 min with a high-resolution video camera (1080p 36 MP, IXNAIQY EU) (Supplementary Video S2). Videos were analyzed using Ethovision XT Basic Software (v17.0, Noldus Information Technology). This test assesses rodents’ natural aversion to large open spaces alongside their instinct to explore novel environments, providing insight into their emotional state and locomotor activity. Anxiety-related measures included the percentage of time spent in the center of the arena and the center-to-total distance ratio. Locomotor activity was evaluated by total distance moved (cm) and average velocity (cm/s) (Seibenhener and Wooten, 2015).

The EZMT involved placing each hamster individually in one of the open quadrants of a 100 cm-diameter annular platform elevated 50 cm above the floor, comprising two opposing enclosed (anxiolytic) and two opposing open (anxiogenic) quadrants (Panlab 760680). Animals were recorded for 10 min using a high-resolution video camera (1080p 36 MP, IXNAIQY EU) (Supplementary Video S3), and the videos were analyzed as described above. This test assesses rodents’ innate fear of heights and open spaces alongside their motivation to explore novel environments, providing insight primarily into their emotionality. Anxiety-related parameters were assessed based on the percentage of time spent and the distance moved (cm) within the open areas (Shepherd et al., 1994).

Delta (Δ) scores for each behavioral parameter were calculated by subtracting the pre-challenge (baseline) value from the post-challenge measurement; therefore, negative Δ values indicated a decline from baseline. This within-subject normalization accounted for inter-individual variability and allowed individual evaluation following SARS-CoV-2 inoculation. Additionally, z-scores were calculated to standardize Δ values by the negative control group mean and standard deviation, as previously described (Guilloux et al., 2011). For the BT, z-scores were normalized by sex but pooled across time points due to statistically significant sex differences without temporal effects, resulting in 6 control observations per sex. In contrast, the NBT, OFT, and EZMT Δ values were standardized using pooled control data across sexes and time points, as neither factor significantly affected performance, yielding 12 control observations per test.

Data distributions were first assessed using the Shapiro–Wilk test for normality and F-test for homogeneity of variances. Body weight changes between groups were analyzed by a mixed-effects model with Šídák’s correction to account for repeated measures and group effects. Viral shedding data, which did not meet normality assumptions, were compared between sexes at each time point using the Kruskal-Wallis test with Dunn’s multiple-comparisons correction. The mean cross-sectional area (CSA) of the CVN was compared between groups using a two-tailed unpaired t-test, as the data were normally distributed with equal variances. Cytokine and behavioral data were analyzed by two-tailed unpaired t-tests when normally distributed with equal variances; Welch’s correction was applied if variances were unequal, and Mann–Whitney tests were conducted if normality was not met. For behavioral tests, z-scores for each parameter were calculated as previously described (Guilloux et al., 2011), with statistical significance determined at a threshold of ±1.96, assuming a normal distribution. All statistical analyses were conducted using GraphPad Prism v10.1.2, with significance set at α < 0.05.

SARS-CoV-2 inoculated hamsters (n = 40) exhibited gradual weight loss, reaching a mean reduction of 11.6% ± 3.8% standard deviation (SD) at 6 dpi. Weight recovery began at 7 dpi, with no significant differences between inoculated and control groups by 17 dpi (Figure 2a). When stratified by sex, weight loss peaked on average at 13.2% in males (n = 20) at 7 dpi, while females (n = 20) reached an average loss of 11.1% by 6 dpi, initiating recovery one day earlier. However, these sex-based differences were not statistically significant.

Heightened inter-individual weight loss variability was observed in the SARS-CoV-2 group during the acute phase (4–10 dpi) compared to controls, as indicated by higher coefficients of variation (CV). At 6 dpi, CV was 4.1% in inoculated males versus 2.2% in control males; at 7 dpi, 5.0% in inoculated females versus 2.9% in control females. By 60 dpi, variability converged across all groups. Notably, one male (ID 57) exhibited delayed recovery, consistent with a greater weight loss during the acute phase than other infected animals (Figure 2b).

In OS, SARS-CoV-2 gRNA loads peaked at 4 and 7 dpi, declined by 9 dpi, and became undetectable from 21 dpi (Figure 3a). All OS samples remained below the limit of quantification (<10^1.8 TCID₅₀/mL) in viral titration assays using Vero E6 cells at all time points.

In tissues, SARS-CoV-2 gRNA was detected across all samples at 4 dpi, with the highest concentrations in the nasal turbinate and lung, exceeding the levels in extra-respiratory tissues (intestine, muscle, heart, and brain) by more than 4 log units. By 7 dpi, gRNA remained detectable in both upper and lower respiratory tracts, as well as in the intestine, and brain. Viral gRNA persisted in respiratory tissues through 60 dpi, showing a gradual decline over time (Figure 3b). No significant differences in viral load were observed between males and females.

The highest infectious viral titers, measured by Vero E6 titration assays, were observed at 4 dpi in the nasal turbinate (log₁₀ 5.93 TCID₅₀/mL) and lung (log₁₀ 4.51 TCID₅₀/mL). At 7 dpi, only one nasal turbinate sample was positive, and no infectious virus was isolated from lung homogenates (Supplementary Figure S3). Infectious virus was not detected in extra-respiratory tissues at any time point (below the limit of quantification in all samples).

In the nasal turbinate, mild to moderate (scores 1–2) inflammatory lesions were observed at 4 and 7 dpi, characterized by multifocal to diffuse muco-suppurative rhinitis (Supplementary Figure S4A). SARS-CoV-2 N protein levels were moderate to high (scores 2–3) at 4 dpi, and absent to minimal (scores 0–1) by 7 dpi (Supplementary Figure S4B). From 14 dpi onward, neither lesions nor N protein were detected.

In the lung, inflammatory lesions ranged from mild (n = 1) to moderate (n = 2) and severe (n = 1) at 4 dpi, characterized by multifocal peribronchial and peribronchiolar lymphohistiocytic infiltration, with type 2 pneumocyte hyperplasia and hypertrophy. Broncho-interstitial pneumonia peaked (score = 3) at 7 dpi, remitted to mild (score = 2) by 14 dpi, and resolved by 21 dpi. One animal displayed very mild pneumonia at 46 dpi, but this was not observed in samples examined at later time points (Supplementary Figure S4C). SARS-CoV-2 N protein levels in the lung were variable at 4 dpi, low at 7 dpi, and undetectable from 14 dpi onward (Supplementary Figure S4D).

No histopathological lesions or SARS-CoV-2 N protein were detected in extra-respiratory tissues (brain, heart, intestine, muscle) at any time post-inoculation.

At 60 dpi, one SARS-CoV-2 inoculated female (ID 76) exhibited a markedly increased CVN CSA. Although the SARS-CoV-2 group showed greater variability (25,300–39,600 μm²; CV = 18%) compared to controls (25,400–34,300 μm²; CV of 12%) (Figure 4), group comparison did not reveal statistically significant differences.

Cytokine profiling revealed pronounced time- and sex-dependent changes in the lung, brain, and plasma of SARS-CoV-2-inoculated hamsters. Across all compartments, infected animals displayed a sharp early increase in both pro- and anti-inflammatory mediators that gradually normalized but maintained notable individual and sex-specific variability.

In the lung, both pro- and anti-inflammatory mediators increased significantly in infected animals by 4 dpi, including IL-10, IL-2, IL-4, MCP-1, MIP-1α, TNFα, IL-1β, and IFNγ. Notably, females exhibited IFNγ concentrations up to sixfold higher than males (Supplementary Figure S5). By 32 dpi, lung levels of IFNγ, IL-2, IL-4, MIP-1α, and IL-1β had fallen below control values, although two infected males and one female exhibited IL-10 levels threefold higher than controls. At 60 dpi, lung cytokine levels largely returned to baseline, except for elevated IL-10 and IL-1β and decreased IL-2 and IL-4 (Supplementary Figure S5). Specifically, IL-10 was twofold higher in two females (IDs 76, 77) and twofold lower in one male (ID 57) compared to controls. IL-1β was also elevated in two females (IDs 75, 77). IL-4 was reduced in one male (ID 57) and three females (IDs 75, 77, 78), with the latter additionally exhibiting decreased IL-2 (Figure 5).

In the brain, SARS-CoV-2-inoculated hamsters exhibited elevated levels of both pro- and anti-inflammatory cytokines at 4 dpi, including IFNγ, IL-10, IL-2, IL-4, IL-6, MCP-1, MIP-1α, TNFα, and IL-1β, compared to controls (Supplementary Figure S6). Notably, IFNγ, IL-10, IL-2, and IL-4 exhibited significantly high variability within the infected group, with females showing higher concentrations than males. By 32 dpi, IFNγ, IL-2, IL-4, MCP-1, MIP-1α, TNFα had declined in the SARS-CoV-2 group, while IL-10, VEGF, and IL-1β levels were comparable to controls. Additionally, sex-specific patterns were observed, with three females exhibiting the lowest IFNγ concentrations and consistently higher IL-1β concentrations compared to males (Supplementary Figure S6). At 60 dpi, mean cytokine levels in the SARS-CoV-2 group were generally comparable to controls, although IFNγ, TNFα, IL-10, and MIP-1α remained highly variable. Specifically, IFNγ levels were twofold higher in one male (ID 57) and one female (ID 78), while a twofold reduction was observed in another male (ID 56) and three females (IDs 75, 76, 77) (Figure 5).

Overall, cytokine concentrations in the brain were approximately one log lower than in the lung for all cytokines except for IL-1β, which was higher in the brain. Notably, at 4 dpi, MCP-1 was nearly 10,000-fold higher in the lung than in the brain, although this disparity diminished over time (Supplementary Figures S5, S6).

In plasma, cytokine levels were elevated in SARS-CoV-2-inoculated hamsters compared to controls at 4 dpi, including IFNγ, IL-10, IL-2, IL-4, IL-6, MCP-1, MIP-1α, TNFα, and IL-1β (Supplementary Figure S7). IFNγ and IL-1β levels were significantly elevated across all inoculated animals, while IL-10, IL-4, IL-6, and TNFα showed significant variability, with one male exhibiting the highest concentrations of all four cytokines. By 32 dpi, IL-2, IL-4, and IL-6 concentrations declined in the SARS-CoV-2 group, though individual variability remained high (Supplementary Figure S7). In contrast, MCP-1 and VEGF were elevated on average, with one male and three females showing marked increases, contributing to the heightened variability in the SARS-CoV-2 group (Supplementary Figure S7). At 60 dpi, plasma cytokine levels in inoculated hamsters were generally comparable to controls. However, four females (IDs 75–78) and one male (ID 58) exhibited IL-1β levels 10-fold higher than both controls and the other infected animals. Notably, one of these females (ID 78) also exhibited markedly elevated levels of IFNγ, IL-2, IL-4, IL-6, MIP-1α, and TNFα (Supplementary Figure S8).

At 4 dpi, all animals showed low SARS-CoV-2 neutralizing activity, with females (n = 2) exhibiting higher levels (>100 U/mL) than males (<100 U/mL). By 32 dpi, robust neutralizing activity was observed in all animals, which was maintained up to 60 dpi. At these later time points, neutralizing antibody levels exceeded the assay’s upper limit of quantification (600 U/mL), indicating a durable and robust humoral immune response (Supplementary Figure S9A). No SARS-CoV-2 neutralizing antibodies were detected in the negative control at any time point (signal inhibition < 30%).

At 60 dpi, inoculated animals exhibited variable antigen-specific cellular immune responses, with IFNγ spot counts in response to Spike stimulation generally exceeding those elicited by Nucleocapsid. Female ID 76 showed the strongest response, with 109 spots/10⁶ cells for Spike and 51 spots/10⁶ cells for Nucleocapsid. Female ID 78 and male ID 57 mounted moderate Spike-biased responses, while male ID 58 and female ID 77 demonstrated the lowest IFNγ spot counts (Supplementary Figure S9B).

Analyses from all behavioral tests focused primarily on individual variability, given that group-level differences were not expected. To ensure consistency and comparability across subjects, results were normalized considering intra-individual variability, calculating individual changes from baseline (Δ scores). Δ scores were then standardized relative to the mean and standard deviation of the negative control group, expressed as z-scores. This approach allowed for a robust assessment of individual behavioral changes while minimizing the impact of baseline differences and inter-subject variability.

In the BT, SARS-CoV-2 inoculated GSH showed increased burrowing at 25 dpi compared to baseline and controls (Figure 6a). This increase was driven entirely by males, all with positive z-scores, whereas females scored at or below zero (Figure 6b). By 39 dpi, burrowing declined significantly (p < 0.05) compared to 25 dpi, returning to or dropping below baseline and control levels (Figure 6a), although one male (ID 56) maintained increased activity (Figure 6b). By 59 dpi, most females (IDs 75–77) trended toward recovery, except one (ID 78) that declined significantly. Conversely, most males (IDs 55, 57, 58) sustained reduced burrowing with negative z-scores (Figure 6b).

In the NBT, both SARS-CoV-2–inoculated and control animals exhibited high variability (Supplementary Figure S10A), with z-scores near zero at all time points (Supplementary Figure S10B).

In the OFT, anxiety-related behavior was assessed in terms of time spent in the center of the arena (Figures 7a,b) and center-to-total distance ratio (Figures 7c,d). At 25 dpi, SARS-CoV-2-inoculated animals spent less time in the center of the arena (Figure 7a) and reduced the center-to-total distance ratio (Figure 7c) compared to controls. This effect was driven entirely by females, whereas males showed no evident change from baseline (Figures 7b,d). By 39 dpi, group averages converged for both parameters; however, at 59 dpi, greater variability was observed within the SARS-CoV-2 group (Figures 7a,c). Notably, female ID 77 consistently spent less time in the center across all time points (p < 0.05, Figure 7b) and exhibited a reduced center-to-total distance ratio at 25 and 59 dpi (p < 0.05, Figure 7d).

Locomotor activity was evaluated based on total distance moved (Figures 7e,f) and velocity (Figures 7g,h). The SARS-CoV-2 group showed a progressive decline in distance traveled over time, with a statistically significant decrease between 25 and 59 dpi (p < 0.05, Figure 7e). Velocity remained comparable between SARS-CoV-2 and control groups (Figure 7g). Notably, male ID 57 exhibited a significant reduction in distance traveled at 59 dpi (p < 0.05, Figure 7f) and in velocity at both 39 and 59 dpi (p < 0.05, Figure 7h).

In the EZMT, anxiety-like behavior was assessed by measuring time spent in the open arms (Figures 8a,b) and distance moved within this area (Figures 8c,d). At 25 dpi, infected animals generally increased open-zone exploration (Figure 8a), except for two males (IDs 57 and 58), which had negative z-scores (Figure 8b). Most infected animals maintained baseline activity (z-scores ≈ 0) for distance moved, except for male ID 57, who showed a significant reduction (Figure 8d). At 39 dpi, males IDs 57 and 58 again demonstrated reduced open-zone time and movement (Figures 8b,d). By 59 dpi, male ID 58 returned to baseline, while male ID 57 and female ID 76 underperformed on both measures (Figures 8b,d).

A comprehensive analysis of all behavioral tests using z-score standardized data is presented in Supplementary Figures S11, S12. At 25 dpi, male ID 57 showed reduced performance in both the OFT and EZMT (Supplementary Figure S11A), while all four females (IDs 75–78) underperformed in the OFT (Supplementary Figure S12A). By 39 dpi, male performance declined further, with three out of four individuals (IDs 55, 57, 58) exhibiting deficits across multiple tests (Supplementary Figure 11B). All females (IDs 75–78) continued to underperform in the OFT and began to reduce performance in the BT (Supplementary Figure S12B). At 59 dpi, male ID 57 again displayed the lowest scores across nearly all behavioral measures (Supplementary Figure S11C), and males IDs 55 and 58 continued to underperform in the BT. Female outcomes varied across the four tests, but their behavior in the OFT was consistently compromised, with scores lower than those of the control group (Supplementary Figure S12C).

Overall, males exhibited greater impairments in the BT and EZMT, while females demonstrated consistent deficits in the OFT, suggesting potential sex-specific patterns of behavioral dysfunction. Notably, male ID 57 consistently underperformed in both the EZMT and OFT across all time points, as well as in the BT from 39 dpi onward (Supplementary Figure S11). Similarly, female ID 77 displayed persistent deficits in the OFT throughout the study period (Supplementary Figure S12).