Donepezil may reduce fatigue and depression in long COVID patients with acetylcholine shortage caused by HHV-6B SITH-1

To identify a biomarker capable of stratifying patients for subgroup analysis in a clinical trial, we enrolled a total of 156 patients with post-acute sequelae of SARS-CoV-2 infection (PASC) at Hirahata Clinic (Tokyo), including 45 men and 111 women. All patients were diagnosed according to the World Health Organization (WHO) criteria for post-COVID-19 condition. Symptom severity was assessed for individual PASC symptoms using a numerical rating scale (NRS) ranging from 0 (no symptom) to 10 (worst imaginable severity).

As healthy normal controls (NCs), 36 volunteers (16 men and 20 women) with no history of SARS-CoV-2 infection were recruited by Soiken Inc. through advertisements. To ensure mental and physical health, all NCs were screened using the Beck Depression Inventory (BDI) and the Chalder Fatigue Scale (CFS), and individuals with elevated scores on either scale were excluded. Participants were also excluded if they were currently taking chronic medications or vitamin supplements.

All procedures involving human subjects were conducted in accordance with the Declaration of Helsinki and were approved by the Ethics Committees of the Jikei University School of Medicine, Soiken Inc., and Soiken Clinic. Written informed consent was obtained from all participants prior to their inclusion in the study.

A multicenter, randomized, placebo-controlled, double-blind clinical trial of donepezil for post-COVID-19 fatigue and neuropsychiatric symptoms was conducted in Japan between 14 December 2022, and 31 March 2024. The trial protocol has been published previously (Kawabata et al., 2024). In the original RCT, participants with comorbid psychiatric disorders or concomitant psychiatric medications were excluded at enrollment, as specified in the protocol. Briefly, 120 eligible participants were enrolled and randomly assigned in a 1:1 ratio to receive donepezil or placebo for 3 weeks, and patient-reported outcomes were assessed at 3 and 8 weeks after treatment initiation. Ten individuals withdrew or were lost to follow-up, resulting in 110 patients (55 in the donepezil group and 55 in the placebo group) being included in the final efficacy analysis. Baseline characteristics (e.g., age, sex, symptom severity) did not differ significantly between groups. As reported in the primary analysis (Nakamura et al., 2025), donepezil did not demonstrate a statistically significant effect in the overall cohort, with a mean difference in CFS score at 3 weeks of 0.34 (95% CI, −2.23 to 2.91; P = 0.79).

In the present study, we conducted a retrospective subgroup analysis using serum samples collected during the RCT to investigate the relationship between anti-SITH-1 antibody titers and treatment efficacy. This subgroup analysis was conducted post hoc using stored serum samples from the previously reported RCT and was not pre-specified in the original trial protocol. Accordingly, results from the subgroup analyses should be interpreted as exploratory and hypothesis-generating. Of the 110 participants, 73 patients (35 men and 38 women) were included in the subgroup analysis as PASC cases, based on the WHO concept of symptom persistence for at least 2 months. For this subgroup analysis, PASC was operationally defined as persistent fatigue lasting at least 60 days after SARS-CoV-2 infection. These 73 patients were analyzed according to their original randomized, double-blind treatment assignment to placebo (n = 35; 12 men, 23 women) or donepezil (n = 38; 16 men, 22 women).

Patient-reported outcomes were assessed using validated self-report questionnaires, including the 11-item Chalder Fatigue Scale (CFS), the Hospital Anxiety and Depression Scale (HADS), and the Patient Health Questionnaire-9 (PHQ-9). Assessments were conducted online at 3 and 8 weeks after initiation of treatment. Treatment efficacy was evaluated by calculating the change in total scores from baseline, with more negative values indicating greater symptom improvement. Medication adherence was monitored by counting unused medication and returned packaging, and adherence below 80% was considered a protocol deviation.

Among the 73 patients included in the subgroup analysis, 52 participants (71.2%; 18 men and 34 women) were seropositive for anti-SITH-1 antibodies. Treatment efficacy was compared between seropositive and seronegative groups.

Detection of anti-SITH-1 antibodies was performed using an indirect immunofluorescence assay, as previously described (Kobayashi et al., 2020), with a modified method for titer quantification.

To generate the antigen for staining, a fusion construct comprising SITH-1, a flexible linker ((Gly-Gly-Gly-Gly-Ser)₅), and CAML was cloned into the pFLAG-CMV-5a vector (Sigma-Aldrich). The amino acid sequence of SITH-1 is available under GenBank Accession Number AOC08123.1↗. The resulting plasmid (pCMV-SITH-CAML) was transfected into HEK293T cells cultured on Lab-Tek chamber slides (Nunc) using Lipofectamine LTX (Thermo Fisher Scientific).

Plasma samples were diluted 1:80 in PBS containing 2% BSA and 0.05% Tween 20, and mixed with a 1:200 dilution of anti-CAML mouse monoclonal antibody. The mixture was applied to the slides and incubated at 37 °C for 1 h. After washing with PBS/0.05% Tween 20, cells were incubated with secondary antibodies, Alexa Fluor 488–conjugated goat anti-human IgG (RRID: AB_2534080) and Alexa Fluor 594–conjugated goat anti-mouse IgG (RRID: AB_2534073) (Thermo Fisher Scientific), diluted 1:200 in PBS/2% BSA/0.05% Tween 20 at 37 °C for 30 min. Slides were washed again and mounted with cover glasses.

Fluorescence was visualized using an Olympus BX51 microscope equipped with a DP-73 CCD camera. Image analysis was conducted using ImageJ (NIH). The anti-SITH-1 Antibody titers were calculated using the following

Here, AGFI (average green fluorescence intensity) represents the fluorescence intensity of bound human IgG and reflects the amount of anti-SITH-1 antibodies present in the serum sample. ARFI (average red fluorescence intensity) represents the fluorescence intensity of bound anti-CAML antibody and serves as an indicator of the expression level of the antigen, namely, the SITH-CAML fusion protein, in the transfected cells.

The infecting SARS-CoV-2 strain was estimated based on the predominant circulating variant in Tokyo, the region where participants resided, at the time of their COVID-19 onset. Specifically, infections were attributed to the wild-type (B.1.1) strain until February 2021, the Alpha variant (B.1.1.7) from March to June 2021, the Delta variant (B.1.617.2) from July to September 2021, and the Omicron variant (B.1.1.529 and sublineages) from October 2021 onward.

All animal experiments were conducted using male 8-week-old C57BL/6NCrSlc mice (Sankyo Laboratories, Japan). Mice were housed under standard laboratory conditions with a 12-h light/dark cycle (lights on at 8:00 a.m.), ambient temperature maintained at 24 °C ± 1 °C, and ad libitum access to food and water. All procedures were approved by the Animal Care and Use Committee of the Jikei University School of Medicine and conducted in accordance with institutional guidelines for animal experimentation.

Human embryonic kidney cell lines HEK293A and HEK293T (derived from female fetal tissue) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37 °C in a humidified atmosphere containing 5% CO₂.

The SARS-CoV-2 S1-expressing adenoviral vector was constructed as previously described (Oka et al., 2023), using the Adenovirus Dual Expression Kit (Takara Bio) according to the manufacturer’s instructions. Briefly, the SARS-CoV-2 S1 gene was cloned into the adenoviral cosmid vector pAxCAwtit2 (SARS-CoV-2-S1/pAxCAwtit2), and the recombinant adenovirus was generated by transfecting this construct into HEK293A cells.

The HHV-6B SITH-1-expressing adenoviral vector was generated as previously reported (Kobayashi et al., 2020), using the Adenovirus Expression Vector Kit (Takara Bio). The glial fibrillary acidic protein (GFAP) promoter was obtained from the pGfa2Lac plasmid (kindly provided by Dr. Kazuyoshi Ikuta). The GFAP promoter and the PCR-amplified SITH-1 gene were cloned into an adenoviral cosmid to construct Ad-GFAP-SITH1. HEK293A cells were co-transfected with Ad-GFAP-SITH1 and the helper cosmid (pAxcwit2) to produce recombinant virus.

All recombinant adenoviruses were propagated in HEK293A cells and purified using the Adeno-X Virus Purification Kit (Takara Bio). Viral titers were quantified using the Adeno-X Rapid Titer Kit (Takara Bio).

Eight-week-old male C57BL/6 mice were anesthetized with isoflurane prior to intranasal administration. Each adenoviral vector was diluted in sterile water (rather than isotonic buffer). A 20-μL drop of the viral solution, containing 1–2 × 10⁹ plaque-forming units (pfu)/mL, was placed at the entrance of each mouse’s nasal cavity. The solution was inhaled passively through spontaneous respiration.

The acetylcholinesterase inhibitor donepezil (FUJIFILM Wako) was administered via drinking water at a concentration of 40 mg/L, corresponding to an approximate dose of 4 mg/kg/day. Treatment began on the same day as adenovirus inoculation and continued throughout the experimental period.

Seven days after intranasal inoculation with S1-Ad or Vector-Ad, mice underwent behavioral testing to evaluate fatigue- and depression-like symptoms.

In the tail suspension test (TST), immobility time over a 10-min period was recorded and analyzed using TailSuspScan software (CleverSys Inc.). In the Y-maze test, the number of arm entries during an 8-min session was measured using TopScan software (CleverSys Inc.).

Twenty-four hours after completion of the behavioral tests, mice were euthanized, and brain tissues were collected for real-time PCR and immunohistochemical analysis.

Total RNA was extracted from mouse tissues using the RNeasy Mini QIAcube Kit (Qiagen), and reverse transcription was performed using the PrimeScript RT Reagent Kit (Takara Bio) to synthesize complementary DNA (cDNA). Quantitative real-time PCR was carried out using Premix Ex Taq (Perfect Real Time) (Takara Bio) and the QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific).

The thermal cycling conditions were as follows: initial denaturation at 95 °C for 30 s, followed by 45 cycles of denaturation at 95 °C for 5 s and annealing/extension at 60 °C for 31 s. Data were analyzed using QuantStudio Design and Analysis Software (Thermo Fisher Scientific).

TaqMan Gene Expression Assays (Thermo Fisher Scientific) were used to quantify mRNA levels of the following genes: Mouse Il1b (IL-1β): Assay ID Mm00434228_m1, Mouse Chat (Choline acetyltransferase): Assay ID Mm01221880_m1, Mouse Fgf2 (basic FGF): Assay ID Mm01285715_m1, Eukaryotic 18S rRNA (endogenous control): Assay ID Hs99999901_s1.

Relative gene expression levels were normalized to 18S rRNA using the ΔΔCt method.

For immunohistochemical analysis, mice were euthanized and transcardially perfused with physiological saline followed by 10% neutral-buffered formalin (pH 7.4). The brains were harvested, fixed in formalin, embedded in paraffin, and sectioned. For immunofluorescence staining, paraffin-embedded sections were deparaffinized, subjected to antigen retrieval, and blocked with Image-iT FX Signal Enhancer (Thermo Fisher Scientific) for 30 min at room temperature.

Primary antibodies included anti–active caspase-3 (RRID: AB_443014), anti–doublecortin (DCX) (RRID: AB_732011), and anti–choline acetyltransferase (ChAT) (RRID: AB_2721842), all obtained from Abcam. Secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG [H+L]) (RRID: AB_2576217) was obtained from Thermo Fisher Scientific. All antibodies were diluted using Can Get Signal® Immunostain Solution A (TOYOBO).

After staining, sections were mounted with coverslips and observed using a fluorescence microscope (BZ-9000, Keyence).

Normality of data distribution was assessed using the Shapiro–Wilk test. For comparisons between two groups, the nonparametric Mann–Whitney U test was used. For comparisons among multiple groups, one-way analysis of variance (ANOVA) followed by Fisher’s least significant difference (LSD) post hoc test was used for parametric data, while the Kruskal–Wallis test followed by Dunn’s post hoc test was used for nonparametric data. In the figures, red horizontal lines indicate either the median or mean value. A p-value of <0.05 was considered statistically significant.

The RCT subgroup analyses were post hoc and exploratory, and p values are reported as nominal. To address multiple testing, p values from placebo versus donepezil comparisons across three strata (overall, anti-SITH-1 positive, anti-SITH-1 negative) and two time points (weeks 3 and 8) were additionally adjusted using the Benjamini–Hochberg false discovery rate procedure within each questionnaire family. FDR-adjusted q values are provided in Supplementary Table S1.

For comparisons of baseline characteristics in Tables 1, 2, continuous variables, including age and post-onset period, were analyzed using the Mann–Whitney U test, and the categorical variable sex was analyzed using Fisher’s exact test. All statistical analyses were performed using Prism 8 (GraphPad Software) and BellCurve for Excel (Social Survey Research Information Co., Ltd.).

Because HHV-6B SITH-1 expression can be inferred from anti-SITH-1 antibody titers in peripheral blood (Kobayashi et al., 2020), we compared these titers between patients with PASC and normal controls (NCs). A total of 156 individuals with PASC were evaluated alongside 36 NCs with no history of SARS-CoV-2 infection (Table 1). Anti-SITH-1 antibody titers were significantly higher in the PASC group (Figure 1A). Titers were normalized to the median value observed in NCs, which was set to 1. The Receiver operating characteristic (ROC) analysis yielded an AUC of 0.819 (95% CI, 0.759–0.879). Using the threshold of 1.43, sensitivity was 62.8% (95% CI, 55.0%–70.0%) and specificity was 91.7% (95% CI, 78.2%–97.1%).

Receiver operating characteristic (ROC) analysis revealed that, when a threshold of 1.43 was applied, 62.8% of PASC patients tested positive for anti-SITH-1 antibodies, compared to 5.6% of NCs (Figure 1B). To explore the possible impact of different SARS-CoV-2 variants, PASC patients were categorized according to the presumed infecting strain (wild type, α, δ, or ο), based on the period of infection. No significant differences in anti-SITH-1 antibody titers were observed among these groups (Figure 1C).

Furthermore, among the most prevalent symptoms reported after COVID-19, fatigue and depression were significantly more severe in patients who were seropositive for anti-SITH-1 antibodies compared to those who were seronegative (Figure 1D). The severity of symptoms was assessed using the highest numerical rating scale (NRS) score during the observation period.

To investigate the mechanisms underlying HHV-6B reactivation and the exacerbation of neuropsychiatric symptoms observed in PASC patients, we employed a mouse model. SITH-1 is known to be expressed during HHV-6B reactivation in latently infected astrocytes of the olfactory bulb (Kobayashi et al., 2020), and previous studies have identified interleukin-1β (IL-1β) and basic fibroblast growth factor (bFGF) as reactivation-inducing factors in these cells (Tamai et al., 2017). In a mouse model in which the SARS-CoV-2 spike (S1) protein was transiently expressed in the nasal cavity, mimicking viral exposure in human COVID-19 (Figure 2A), we observed upregulated expression of IL-1β (Figure 2B) and bFGF (Figure 2C) in the olfactory bulb. These findings suggest that SARS-CoV-2 S1 protein may contribute to HHV-6B reactivation and subsequent SITH-1 expression in the olfactory bulb of PASC patients.

To further explore the relationship between HHV-6B SITH-1 and PASC-associated neuropsychiatric symptoms, we created a mouse model in which SITH-1 was transiently expressed in olfactory bulb astrocytes via intranasal administration of an adenoviral vector encoding the HHV-6B SITH-1 gene (hereafter referred to as the SITH-1 mouse) (Figure 2D). Consistent with previous reports, these mice exhibited olfactory bulb apoptosis and depression-like behaviors (Kobayashi et al., 2020). Given that nasal expression of the SARS-CoV-2 S1 protein in mice has been shown to reduce acetylcholine production and induce neuropsychiatric abnormalities (Oka et al., 2023), we examined whether SITH-1 expression induces similar cholinergic deficits. Immunohistochemical analysis revealed a marked reduction in choline acetyltransferase (ChAT), the key enzyme for acetylcholine synthesis, in the medial septum (MS) and diagonal band of Broca (DBB) (Figures 2E,F), along with decreased ChAT expression in the whole brain (Figure 2G). These findings suggest that SITH-1 expressed as a consequence of HHV-6B reactivation induced by SARS-CoV-2 infection, similar to the SARS-CoV-2 S1 protein, reduces acetylcholine production and thereby contributes to the development of neuropsychiatric symptoms.

The efficacy of the acetylcholinesterase inhibitor donepezil in ameliorating neuropsychiatric symptoms induced by SITH-1 was evaluated using the SITH-1 mouse model. In the SITH-1 mouse model, reduced hippocampal neurogenesis was observed. This reduction was reversed following donepezil administration (Figures 3A,B), suggesting its potential to mitigate SITH-1–induced impairments in memory and depressive behavior.

Behavioral testing revealed that SITH-1 mice exhibited prolonged immobility time in the tail suspension test, indicative of depression-like behavior, and a reduced frequency of entries in the Y-maze, reflecting decreased spontaneous activity. Both phenotypes were significantly improved by donepezil treatment (Figures 3C,D). In addition, olfactory bulb apoptosis, which likely contributes to decreased central acetylcholine production in this model, was also ameliorated by donepezil administration (Figure 3E). These findings collectively support the therapeutic potential of donepezil in alleviating SITH-1–induced neuropsychiatric symptoms.

Based on the preclinical findings, we hypothesized that donepezil may exert therapeutic effects in PASC patients with elevated anti-SITH-1 antibody titers. To test this, we conducted a subgroup analysis within a previously reported randomized, double-blind clinical trial of donepezil in post-COVID-19 patients (Kawabata et al., 2024; Nakamura et al., 2025), stratifying participants based on their anti-SITH-1 antibody status. The threshold for antibody positivity was defined according to the ROC analysis shown in Figures 1A,B.

From among the 110 participants enrolled in the original clinical trial (Nakamura et al., 2025), we identified a subgroup of 73 patients who met the WHO criteria for PASC, defined as persistent fatigue lasting at least 2 months following SARS-CoV-2 infection (Figure 4A). Baseline characteristics, including age, sex, and post-onset period, did not differ significantly between the donepezil and placebo groups in the overall cohort, nor when stratified by anti-SITH-1 antibody status (Table 2). Fatigue was assessed using the 11-item Chalder Fatigue Scale (CFS) (Morriss et al., 1998; Jing et al., 2016), while depressive symptoms were evaluated using the Patient Health Questionnaire-9 (PHQ-9) (Kroenke et al., 2010; Costantini et al., 2021) and the Hospital Anxiety and Depression Scale (HADS) (Fernández-de-Las-Peñas et al., 2022). Each scale consists of multiple items scored from 0 to 3, with total scores used for analysis. Treatment efficacy was quantified by calculating the change in total scores before and after the 3-week treatment period, with more negative values indicating greater improvement. Clinical assessments were conducted at 3 and 8 weeks after treatment initiation.

In anti-SITH-1 antibody positive PASC patients, donepezil significantly reduced HADS depression scores at week 3 compared to placebo (Figure 4B) and CFS total scores at week 8 (Figure 4C). Item-level changes in the HADS depression subscale at week 3 and the CFS at week 8 are shown in Supplementary Figures S1, S2, respectively. No significant improvement was observed in PHQ-9 scores. In addition to the outcomes shown in Figure 4, results for the remaining questionnaire outcomes and time points are presented in Supplementary Figure S3. Nominal p-values and BH-FDR-adjusted q-values for all placebo versus donepezil comparisons across questionnaires, strata, and time points are summarized in Supplementary Table S1. Across clinical sites, the proportion of patients positive for anti-SITH-1 antibodies ranged from 63.2% to 100%, with an overall positivity rate of 71.7% (Figure 4D). Together, these findings suggest that donepezil is effective in alleviating neuropsychiatric symptoms in PASC patients who are positive for anti-SITH-1 antibodies.

Anti-SITH-1 antibodies were not measured prior to developing COVID-19 and associations with HHV-6B reactivation and SITH-1 expression in patients developing PASC were not tracked over time. Also, since almost all adults are latently infected with HHV-6B, we could not perform a study on a population not infected by HHV-6B. Because the subgroup analysis was exploratory and not pre-specified, the findings require confirmation in a prospective biomarker-stratified trial. Accordingly, the subgroup findings should be interpreted as hypothesis-generating, with full outcome results and BH-FDR-adjusted q-values provided in;. The ROC-derived threshold was developed in the present cohort and has not been externally validated; therefore, its generalizability requires confirmation. Vaccination status was not systematically available for the RCT cohort. In the observational cohort, data on acute COVID-19 severity, vaccination status, comorbid psychiatric conditions, and concomitant medications were not uniformly collected, which may allow residual confounding. Supplementary Figure S3 Supplementary Table S1

The original contributions presented in the study are included in the article/, further inquiries can be directed to the corresponding author. Supplementary Material