This study is a single-center, randomized, double-blind, placebo-controlled trial (NCT05874089) designed to evaluate the efficacy of VSL#3®, a consortium of mixed probiotics at high concentration, in reducing fatigue and improving different features of patients’ wellbeing, microbiota composition and function, and immunological status.

Patients were consecutively recruited from the Long COVID Outpatient Clinic at IRCCS Fondazione Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy, between November 2022 and November 2023. Individuals were required to have a previous SARS-CoV-2 infection documented by molecular or antigenic nasopharyngeal swab. Demographic information, including age and sex, and history of COVID-19, including severity according to WHO classification, were obtained from hospital records. Signs and symptoms of long COVID persisting for at least 3 months after SARS-CoV-2 infection were evaluated for study inclusion. Participants were randomly assigned (1:1) to receive either VSL#3® or placebo for 4 weeks using REDCap software with block randomization stratified by sex. Double blinding was achieved by packaging supplementation and placebo in identical sealed and consecutively numbered sachets similar in appearance, smell, and taste.

The study was conducted according to the Declaration of Helsinki and Good Clinical Practice guidelines after approval by the Ethics Committee of IRCCS Fondazione Ca’ Granda Ospedale Maggiore Policlinico di Milano (approval number S62043). Written informed consent was obtained from all participants before enrolment.

Adult patients aged between 18 and 65 years with previous SARS-CoV-2 infection documented by molecular or antigenic nasopharyngeal swab, Chalder Fatigue Scale (CFS) score ≥4/11 in the dichotomous scoring system, and persistent long COVID symptoms for at least 3 months after infection were eligible for inclusion. Long COVID manifestations included fatigue, sleep disturbances, cognitive deficits (brain fog, impaired concentration and memory, anxiety, depression), strength deficits, arthralgias and myalgias, and gastrointestinal symptoms including reduced appetite, nausea, changes in bowel habits, and abdominal pain.

Exclusion criteria included use of antibiotics, probiotics, immunosuppressive/immunomodulatory drugs, opioids, or antidepressants within 30 days before enrolment; cardiovascular and pulmonary diseases with moderately severe organ dysfunction; decompensated endocrine or metabolic diseases; documented diagnosis of fibromyalgia, myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), irritable bowel syndrome, neurological disorders, psychiatric diseases, or chronic musculoskeletal pathologies prior to SARS-CoV-2 infection; pregnancy or breastfeeding; alcohol or drug abuse; and participation in other clinical trials within 30 days before enrolment. Use of probiotics, antibiotics, immunosuppressive/immunomodulatory drugs, opioids, or antidepressants during the study was considered prohibited medication. Major dietary modifications during the trial were considered protocol violations.

VSL#3® (Actial Farmaceutica) is a consortium of eight live freeze-dried lactic acid bacteria and bifidobacteria (Streptococcus thermophilus BT01; Bifidobacterium breve BB02; Bifidobacterium animalis subsp. lactis BL03; Bifidobacterium animalis subsp. lactis BI04; Lactobacillus acidophilus BA05; Lactobacillus plantarum BP06; Lactobacillus paracasei BP07; and Lactobacillus helveticus BD08) at high concentration (450 billion CFU/sachet). Placebo sachets were identical in appearance and composition except for the absence of probiotics.

After screening evaluation, study procedures were performed at baseline (t0), after 4 weeks of supplementation (t4), and after an additional 4-week follow-up period (t8). Participants received two sachets/day for 28 days. Compliance was assessed by sachet counting and considered adequate when ≥80% of sachets were consumed. Fecal samples were collected for microbiome profiling and metabolomic analyses, including short-chain fatty acids (SCFAs) and neuromodulators. Blood samples were collected for immune phenotyping by multiparametric flow cytometry and serum cytokine profiling by multiplex Luminex assay.

The primary endpoint was reduction in fatigue after 4 weeks of supplementation as measured by the Chalder Fatigue Scale (CFS). Secondary outcomes included variations in Hospital Anxiety and Depression Scale (HADS), Short Form-36 Health Survey (SF-36), Structured Assessment of Gastrointestinal Symptoms (SAGIS), Symptoms Check List-12 (SCL-12), Karnofsky Performance Status (KPS), and Visual Analogue Scale (VAS) scores. Additional exploratory endpoints included modulation of gut microbiota composition, circulating immune parameters, short-chain fatty acids (SCFAs), and neuromodulators.

Structured questionnaires were completed at baseline (t0), after supplementation (t4), and at follow-up (t8), and included CFS, HADS, SF-36, SCL-12, KPS, VAS, and SAGIS questionnaires. Detailed description of questionnaires and scoring systems is reported in the Supplementary Methods.

Assuming a placebo effect of 30% on fatigue reduction and a probiotic efficacy of 60% on the primary endpoint, 23 patients per arm were estimated to provide 80% statistical power with a two-sided significance level of 0.05. Considering an estimated dropout rate of 15%, at least 53 patients were planned for recruitment.

Multiplex analysis of circulating cytokines and chemokines was performed using a 45-plex Human XL Cytokine Discovery Luminex panel (R&D Systems) according to the manufacturer’s instructions. Samples were acquired on a Luminex 200SD instrument and analyzed using xPONENT software v4.2. Luminex analyses were performed on a selected subset of patients (n = 28; placebo n = 15; VSL#3® n = 13). The complete list of analytes included in the panel is summarized in Supplementary Table S1.

Peripheral blood mononuclear cells (PBMCs) collected at baseline and t4 were isolated by Ficoll density-gradient centrifugation. Cells were stained using fluorochrome-conjugated monoclonal antibodies listed in Supplementary Table S2. For intracellular cytokine staining, cells were stimulated for 3 h at 37 °C with PMA (50 ng/mL), ionomycin (1 μg/mL), and Brefeldin A (10 μg/mL), then fixed and permeabilized using Cytofix/Cytoperm (BD Biosciences). Samples were acquired on a FACSLyrics flow cytometer (BD Biosciences) and analyzed using FlowJo software v10.8 after exclusion of doublets and non-viable cells. Flow cytometric analyses were performed on the entire study cohort.

Fecal samples collected at t0 and t4 were stored at −80 °C until DNA extraction. DNA extraction, 16S rRNA gene amplification, purification, library preparation and pair-end sequencing on the Illumina MiSeq platform were performed as previously described (V3-V4 regions, 300 bp paired end) [25]. Reads were pre-processed using the MICCA pipeline (v.1.7.2) (https://micca.readthedocs.io/en/latest/#) [26]. Forward and reverse primers trimming and quality filtering were performed using micca trim and micca filter, respectively. Filtered sequences were denoised using the UNOISE [27] algorithm implemented in micca otu to determine true biological sequences at the single nucleotide resolution by generating amplicon sequence variants (ASV). Bacterial ASVs were taxonomically classified using micca classify and the Ribosomal Database Project (RDP) Classifier v2.13 [28]. Multiple sequence alignment (MSA) of 16S rRNA gene sequences was performed using the Nearest Alignment Space Termination (NAST) algorithm implemented in micca msa [29]. Phylogenetic trees were inferred using micca tree [29, 30]. Sampling heterogeneity was reduced rarefying samples at the depth of the less abundant sample using micca tablerare. Alpha (within-sample richness) and beta-diversity (between-sample dissimilarity) estimates were computed using the phyloseq R package [31]. Permutational multivariate analysis of variance (PERMANOVA) test was performed using the adonis2 function in the vegan R package with 999 permutations. ASVs differential abundance testing was carried out using the R package DESeq2 using the non-rarefied data [31, 32].

Fecal short-chain fatty acids (SCFAs), fecal neuromodulators, and serum neuromodulators were quantified using targeted chromatographic approaches. Metabolomic analyses were performed on a subset of patients, selected on the basis of the material availability to perform the analyses (n = 30; placebo n = 16; VSL#3® n = 14).

Targeted quantification of dopamine and serotonin in serum and fecal samples was performed by ultra-high-performance liquid chromatography coupled with tandem mass spectrometry (UHPLC-MS/MS). Briefly, serum proteins were precipitated using ice-cold acetonitrile, whereas fecal samples were extracted using 50% methanol solution in the presence of deuterated internal standards (dopamine-d4 and serotonin-d4). UHPLC-MS/MS analyses were carried out on a Shimadzu Nexera UHPLC system coupled to a triple quadrupole LCMS-8050 mass spectrometer equipped with an electrospray ionization source. Chromatographic separation was performed on an HSS T3® C18 column (150 × 2.1 mm, 1.8 μm; Waters®) using water and acetonitrile containing 0.1% formic acid as mobile phases. Analyses were conducted in scheduled multiple reaction monitoring (sMRM) mode.

SCFA quantification was performed by gas chromatography-mass spectrometry (GC-MS) according to a previously validated protocol with minor modifications based on Zhu et al [33]. Briefly, freeze-dried fecal samples were acidified, extracted with methyl tert-butyl ether (MTBE), and supplemented with deuterated butyric acid as internal standard. GC-MS analyses were performed using an Agilent 8890 GC system coupled with a 5977B mass selective detector operating in selected ion monitoring (SIM) mode. Chromatographic separation was achieved using a J&W DB-WAX GC column (30 m × 0.25 mm × 0.25 μm; Agilent Technologies).

Data acquisition and processing were performed using MassHunter Qualitative Analysis software (Agilent Technologies). Correlation analyses integrating metabolomic, microbiota, immune, and clinical datasets were performed using Spearman’s correlation tests in R environment (v4.4.1).

All statistical analyses were performed using R v4.4.1 and GraphPad Prism v10.2.3. Parametric and non-parametric tests were applied according to data distribution, including ANOVA, t-test, Wilcoxon signed-rank test, Mann–Whitney U test, and Spearman’s correlation analyses. Primary endpoint analyses were performed both in the intention-to-treat and per-protocol populations.

16S rRNA gene sequencing data are available in the European Nucleotide Archive (https://www.ebi.ac.uk/ena) under accession number PRJEB78610.

Between November 2022 and November 2023, a total of 347 individuals with a history of SARS-CoV-2 infection and persistent symptoms were screened for eligibility at the Long COVID outpatient clinic of Policlinico Hospital (Milan, Italy). Among these, 53 patients met the inclusion criteria and were enrolled in the study (Figures 1A,B). Overall, 294 patients were excluded after screening. The main reasons for exclusion were relevant comorbidities (47.5%), absence of clinically relevant fatigue (35.8%), ongoing antidepressant treatment (7.9%), refusal to participate (8.3%), and concomitant probiotic use (0.4%). Among comorbidities, the most frequent causes of exclusion were cardiovascular diseases, psychiatric disorders, oncological diseases, musculoskeletal disorders, neurological disorders, pulmonary diseases, metabolic diseases, immunological diseases, immunosuppression, and the presence of multiple concomitant conditions. All enrolled participants reported persistent symptoms for at least 3 months after acute infection, consistent with current definitions of post-acute sequelae of SARS-CoV-2 infection. Following randomization, 28 patients were allocated to the placebo group and 25 to the VSL#3® group. Five participants were excluded due to low adherence to the study protocol, resulting in a final study population of 48 individuals (25 placebo, 23 VSL#3®) (Figures 1A,B). Baseline demographic and clinical characteristics of the study population are reported in Table 1. The cohort included 24 females and 24 males, with a median age of 54 years and a median body mass index (BMI) of 25. No significant differences were observed between the placebo and VSL#3® groups in terms of age, sex distribution, BMI, or clinical history (Table 1).

Overall, 50% of participants had a history of hospitalization during the acute phase of COVID-19, and a substantial proportion reported gastrointestinal symptoms both during the acute infection and at the time of Long COVID evaluation (Table 1). The median duration of persistent symptoms was 30 months, with comparable distribution between the two study groups (Table 1). The spectrum of Long COVID symptoms among the participants included cognitive impairment (“brain fog”) (95.8%), reduced concentration (91.7%), memory loss (72.9%), musculoskeletal pain (68.8%), gastrointestinal symptoms (52.1%), dyspnea (50%), dizziness (37.5%), hair loss (22.9%), poor appetite (18.8%), chest discomfort (12.5%), chronic cough (8.3%), and rhinitis (4.2%) (Table 1). All participants reported the presence of more than four symptoms in addition to fatigue. Analyses of clinical outcomes were performed on the full study population. For microbiota, immune, and metabolomic analyses, subsets of patients were included based on sample availability; the number of samples analyzed for each assay is indicated in the corresponding figures. No clinically relevant adverse events were reported during the study period.

The primary endpoint of the study was the change in fatigue, assessed by the Chalder Fatigue Scale (CFS), after 4 weeks of supplementation. A significant reduction in fatigue was observed in patients receiving VSL#3® compared to placebo from baseline (t0) to the end of treatment (t4) (Figures 2A,B; Table 2). In line with this, the change in CFS score was greater in the VSL#3® group compared to placebo (24.24 vs. 6.06, p = 0.037) (Figure 2A). This improvement was maintained at follow-up (t8) in the VSL#3® group, whereas no significant changes were observed in the placebo group (Figure 2B). In parallel, the proportion of responders was higher among VSL#3®-treated patients compared to placebo (68% vs. 35.7%, p = 0.019) (Figure 2C).

Patient-reported outcomes showed patterns consistent with the primary endpoint. Improvements in quality of life, assessed by SF-36, were observed in the VSL#3® group at week 4; however, these changes did not reach statistical significance (Figure 2D; Table 2). Longitudinal analysis of SF-36 fatigue scores showed an improvement from baseline to week 4 that was maintained at follow-up (t8) in the VSL#3® group, whereas no consistent changes were observed in the placebo group (Figure 2E). Significant improvements were observed in specific symptom domains, including epigastric pain and gastrointestinal disturbances, as well as in measures of emotional wellbeing (adjusted p-values <0.05; Supplementary Figures S1–S3). No statistically significant differences were detected across other clinical questionnaires, although numerical improvements were observed in the VSL#3® group across several measures.

Overall, these findings demonstrate that VSL#3® supplementation significantly improved fatigue in long COVID patients, with benefits persisting up to 4 weeks after treatment discontinuation, accompanied by improvements in selected patient-reported outcomes.

Gut microbiota alterations have been reported in patients with COVID-19 and may persist after resolution of the acute infection [10, 11]. In the present study, no significant differences in microbial diversity were observed between placebo and VSL#3®-treated patients at baseline (t0) or at the end of treatment (t4), as assessed by alpha diversity (Shannon index) and beta diversity (Bray-Curtis dissimilarity) (Figures 3A,B).

Despite the absence of global changes in community structure, differential abundance analysis identified specific taxa enriched in VSL#3®-treated patients at t4, including members of the Bifidobacterium, Lactobacillus, Streptococcus, and Barnesiella genera (Figures 3C,D). Several of these taxa correspond to strains contained in the VSL#3® formulation (Supplementary Table S3).

Within the VSL#3® group, clinical responders exhibited distinct microbial signatures compared to non-responders, with enrichment of taxa including Bacteroides, Ruminococcus, Coprococcus, Phascolarctobacterium, Sutterella, Prevotella, Holdemanella, Akkermansia, and Clostridium cluster IV (Figure 3E).

Functional prediction analysis suggested differences in microbial metabolic pathways following VSL#3® supplementation, including enrichment of glycine, threonine, and methionine metabolism (KEGG pathway ko00260), a pathway utilized by commensal bacteria and potentially linked to microbial metabolic activity³⁶ (Supplementary Figure S4). Conversely, no consistent enrichment of microbiome-related pathways was observed in placebo-treated individuals (Supplementary Figure S4).

Overall, these findings indicate that VSL#3® supplementation is associated with selective compositional changes in the gut microbiota, particularly among clinical responders, in the absence of major shifts in overall microbial diversity.

The gut microbiota has been shown to influence host immune responses [34]. To investigate the impact of VSL#3® supplementation on systemic immune parameters, serum cytokine and chemokine profiles were assessed by multiplex analysis. Overall, circulating cytokine and chemokine levels were comparable between placebo and VSL#3®-treated patients (Figures 4A,B). Within the VSL#3® group, reductions in IL1α, IL6, IL7, IFNα, and G-CSF levels were observed at the end of treatment (t4) compared with baseline (Figure 4C), suggesting a partial attenuation of persistent low-grade immune activation associated with long COVID. However, no significant differences between treatment groups were detected at t4. Therefore, these findings should be considered exploratory and interpreted with caution.

In parallel, immune cell phenotyping revealed changes in cellular immune populations following VSL#3® supplementation, including increased proportions of CD8⁺ T cells and CD68⁺ macrophages together with reduced CD66b+ granulocytes at t4 compared to baseline (Figure 4D), suggesting a remodelling of innate and adaptive immune responses following supplementation.

To further explore the relationship between microbiota composition and immune responses, correlation analysis was performed. Distinct association patterns between microbial taxa and immune parameters were observed in VSL#3®-treated patients compared to placebo (Figure 4E). In particular, taxa enriched in the VSL#3® group were associated with members of the Lachnospiraceae family and with IL1α levels, a cytokine involved in cellular stress and inflammatory signaling [35], whereas in placebo-treated individuals, Prevotella spp. were associated with IFNα, a cytokine involved in antiviral and pro-inflammatory immune responses [36]. In addition, Coprococcus, a commensal bacterium [37], showed positive correlations with IL10 and IL7 in responder patients, suggesting a potential link between commensal microbial signatures and pathways involved in immune regulation and persistence of low-grade inflammation [25, 38, 39]. Conversely, in non-responders, Akkermansia showed negative correlations with IL17 and G-CSF (Figure 4F), suggesting distinct microbiota-immune interaction patterns associated with altered inflammatory signaling.

Collectively, these findings indicate that VSL#3® supplementation is associated with remodeling of inflammatory and immune-related pathways, accompanied by distinct microbiota-immune interaction patterns in long COVID patients.

To further investigate the relationship between gut microbiota remodeling and clinical improvement, fecal SCFAs together with fecal and serum neuromodulators were evaluated by targeted metabolomic analyses. Overall, fecal SCFA and neuromodulator levels showed high inter-individual variability and no major differences between placebo and VSL#3®-treated patients over time (Figure 5A). Similarly, fecal dopamine and serotonin concentrations did not show consistent differences between treatment groups (Figure 5B). Serum dopamine and serotonin levels also showed high variability across patients, without significant differences between placebo and VSL#3®-treated groups over time (Figure 5C).

However, exploratory analyses in responder patients identified associations between specific microbiota-derived metabolites and bacterial taxa enriched following supplementation. In particular, correlation analyses performed in responder patients revealed positive associations between isovaleric acid and Coprococcus, as well as between valeric acid and Ruminococcus (Figure 5D), two commensal bacterial genera associated with microbial metabolic activity and gut ecosystem balance [37, 40]. Furthermore, exploratory correlation analyses integrating metabolite levels, cytokine concentrations, ASVs differentially enriched between responder and non-responder groups, and questionnaire scores at T4 revealed coordinated microbiota-immune-metabolic patterns associated with clinical outcomes (Figure 5E).

Collectively, these findings suggest that clinical improvement following VSL#3® supplementation is associated with coordinated microbiota-immune-metabolic interaction patterns potentially involved in gut-brain communication.

Long COVID is frequently associated with persistent fatigue and impaired quality of life.

VSL#3® significantly improved fatigue compared with placebo in patients with long COVID.

This work represents an advance in biomedical science because it provides randomized placebo-controlled evidence supporting microbiota-targeted interventions for long COVID and identifies biological pathways associated with clinical improvement.