Participants were recruited through the University Hospital Medical Information Network-Clinical Trials Registry website, Takanawa Clinic (Tokyo, Japan) website, announcements in an e-mail newsletter and personal contacts. Individuals who met all the following inclusion criteria were enrolled: i) Healthy adults between the ages of 20 and 70; and ii) having negative PCR and IgM/IgG antibodies tests for SARS-CoV-2 at study entry (no previous and current SARS-CoV-2 infection). Chest imaging tests were not used in the present study. Individuals were excluded from this trial if they met any of the following exclusion criteria: i) Pregnant; ii) breastfeeding; iii) duplicate enrollment in other clinical trials; iv) history of infectious disease within 6 months before the enrollment; v) current or past history of chronic inflammatory, immune-related or neoplastic diseases; vi) history of medicinal drug use within 6 months before the enrollment; and vii) underlying conditions associated with higher risk of COVID-19, including hypertension, cardiovascular disease, cerebrovascular disease, diabetes, obesity (body mass index ≥30) (52,53), chronic obstructive pulmonary disease and chronic kidney disease. Therefore, the enrolled subjects had no recorded and reported comorbidities.

In our previous study using QFPD on 18 healthy subjects (51), the effect sizes r obtained were 0.816 (TNF-α), 0.881 (IL-1β), 0.724 (IL-18) and 0.796 (IL-8). The average value of 0.804 was employed as an estimated effect size for the present trial. A priori two-tailed power analysis was conducted with a level of significance α of 0.05, a desired power 1-β of 0.8 and the estimated effect size of 0.804, which suggested a required total sample size of 15 individuals.

Participant recruitment took place between 27 May 2020 and 2 June 2020 at Takanawa Clinic (Tokyo, Japan). A total of 20 volunteers were screened for eligibility, indicated to be eligible and enrolled in the present trial (Fig. 1). In vitro cytokine response assay was performed between 7 and 9 June 2020. LAB were administered to all the enrolled participants between 25 June and 17 August 2020, 2 of whom were excluded from the main analysis due to no visit to Takanawa Clinic following the LAB prescription. Consequently, 18 subjects (1 male and 17 females; age, 28-66 years; mean age ± SD, 44.2±10.1 years) completed the intervention, and the data were subjected to statistical analysis.

The present study comprised two sequential experimental procedures: An in vitro cytokine response assay and a single-arm, double-blind, prospective trial.

The optimal LAB in each subject were determined using co-culture of the peripheral blood with each LAB. QFPD-induced innate cytokine changes were used as an indicator to evaluate the anti-COVID-19 immunomodulatory potential of LAB. The QFPD-induced innate cytokine index (QICI) was defined as follows: QICI=(TNF-α) x (IL-1β) x (IL-18) x (IL-8)/(IL-6), where brackets represent the plasma level of the cytokine in pg/ml. IL-6 is a critical driver of complex immune dysregulation in patients with COVID-19 and was thus adopted as the denominator (54-56).

The LAB with the highest and lowest QICI were used in a subsequent clinical trial to examine whether the ingested LAB could reproduce the in vitro cytokine responses (in vitro QICI). The trial consisted of three consecutive sessions: i) Validation (intervention using the LAB with the highest QICI); ii) washout; and iii) control (intervention using the LAB with the lowest QICI) sessions. The primary outcome measure was the changes in the plasma levels of TNF-α, IL-1β, IL-18, IL-8 and IL-6 and the QICI after each 7-day LAB session compared with those at baseline. The secondary outcome measure was the changes in hematological parameters after each 7-day LAB session compared with those at baseline.

L. plantarum SNK12 [2x10¹² colony-forming unit/g (cfu/g); Bio-Lab Co., Ltd.] is a probiotic strain with potent immunomodulatory, anti-inflammatory and antiviral abilities (57,58). A probiotic B. longum BB536 strain (1.5x10¹¹ cfu/g; Morinaga Milk Industry Co., Ltd.) has been reported to exhibit various clinical benefits, such as anti-allergic effects (59,60), protection against viral infection (41,61) and modulation of gut microbiota (62-64). L. lactis ssp. lactis demonstrates immunomodulatory and antiviral effects through activating plasmacytoid dendritic cells and increasing their ability to produce interferons (IFNs) (42). The commercially available probiotic strain LLL970 (1x10¹¹ cfu/g; Synbio Tech, Inc.) was used in the present study.

Live LAB cells were disrupted by vibrating them twice at 4,600 rpm (the number of the figure-8-shaped movement of sample tubes per minute) for 3 min at room temperature with 0.5 g of 3.0-mm zirconia beads in a commercial bead vibrator (PS-2000; Kurabo Industries Ltd.) followed by heating at 60˚C for 20 min. Dead LAB were suspended at a 100 µg/ml concentration in Dulbecco's PBS (without Ca²⁺ and Mg²⁺).

The ability of the LAB to stimulate QFPD-like cytokine production by peripheral blood immune cells was evaluated as described in previous studies (65-67). Briefly, heparinized peripheral blood (0.4 ml) from each subject was co-cultured with each killed LAB suspension (4 µg) in 1 ml RPMI-1640 medium (DS Pharma Biomedical Co., Ltd.) supplemented with 10% fetal bovine serum (FBS; Sigma-Aldrich; Merck KGaA) in a humidified incubator with 5% CO₂ at 37˚C for 48 h. Concentrations of TNF-α, IL-1β, IL-8 and IL-6 in culture supernatants were measured using the V-PLEX Proinflammatory Panel 1 Human kit (cat. no. K15049D-1; Meso Scale Diagnostics, LLC) and the concentration of IL-18 was measured using the Human IL-18 ELISA kit (cat. no. ab215539; Abcam) according to the manufacturers' protocols. The data were used to calculate the QICI value for the characterization of the species with the highest or lowest QICI.

The present trial was a single-arm, double-blind, prospective trial. Each subject was instructed to orally ingest the live LAB with the highest QICI (1x10¹¹ cfu/day) in the in vitro cytokine response assay twice daily in the morning and evening between meals for 7 days (days 1-7). After a 7-day washout period (days 8-14), a negative control trial was conducted, in which the LAB with the lowest QICI in the in vitro cytokine response assay (1x10¹¹ cfu/day) was orally administered twice daily in the morning and evening between meals for 7 days (days 15-21). Peripheral blood samples were obtained from each subject on days 0, 8 and 22. Neither the subjects nor physicians in charge were aware of the results of the in vitro LAB assessment, prescribed LAB or their QICI properties until the final blood sampling was completed. Concentrations of plasma TNF-α, IL-1β, IL-18, IL-8 and IL-6 were quantified as aforementioned. Hematological and blood biochemical tests (parameters as listed in Table SII) were outsourced to SRL, Inc.

A total of ~9 months after the completion of the trial, 10 healthy subjects were randomly selected from the 18 trial participants and randomly assigned to either the L. plantarum group (n=5) or the B. longum group (n=5) through simple randomization. Ingestion of L. plantarum and B. longum and blood sampling were conducted with the same protocol as that of the trial (1x10¹¹ cfu/day; twice daily for 7 days). After a 7-day ingestion, innate immune cell activity was measured using standard methods as described below in detail.

Measurement of the phagocytic activity of neutrophils was outsourced to BML, Inc. Briefly, heparinized peripheral blood (0.1 ml) from each subject was mixed with 40 µl fluorescent microbeads (Fluoresbrite^® YG Carboxylate Microspheres 1.75 µm; Polysciences, Inc.) diluted 4-fold with Dulbecco's PBS [-(without Ca²⁺ and Mg²⁺)] and incubated with gentle agitation at 37˚C for 30 min. The samples were treated with 2 ml 10X FACS lysing solution (BD Biosciences) at 4˚C for 15 min to lyse erythrocytes under gentle hypotonic conditions, followed by flow cytometric analysis using FACSCalibur™ flow cytometer (BD Biosciences). Granulocytes were characterized as medium-sized cells with high granularity and separated by setting a medium forward scatter (FSC)/high side scatter (SSC) gating. The percentage of fluorescence-positive granulocytes (granulocytes that phagocytosed the fluorescent microbeads) to the total count of granulocytes was calculated using the BD CellQuest™ Pro software version 6.0 (BD Biosciences).

Analysis of NK cell activity using chromium-51 (⁵¹Cr) release assay was outsourced to SRL, Inc. Lymphocytes were isolated from 5 ml peripheral blood using density gradient centrifugation (Lymphosepar I; Immuno-Biological Laboratories Co., Ltd.) according to the manufacturer's instructions. The lymphocytes were washed twice with Dulbecco's PBS (-) and resuspended at 1x10⁶ cells/ml in RPMI-1640 supplemented with 10% FBS. A total of 200-µl aliquots (effector cells; 2x10⁵) were mixed with human chronic myelogenous leukemia K562 cells (target cells; 1x10⁴ cells/10 µl; cat. no. CCL-243; American Type Culture Collection) radiolabeled with ⁵¹Cr (PerkinElmer, Inc.) and incubated at 37˚C for 3.5 h in a 5% CO₂ incubator. The cells were collected by centrifugation, and the remaining ⁵¹Cr radioactivity was measured using WIZARD^® Automatic Gamma Counter (PerkinElmer, Inc.). ⁵¹Cr-loaded K562 cells treated with effector-free culture medium (RPMI-1640; 10% FBS) were used for the quantification of spontaneously released ⁵¹Cr.

The serum level of neopterin, an activation marker produced primarily by IFN-γ-stimulated monocytes and macrophages (68,69), was assessed for macrophage activity. Determination of serum neopterin was outsourced to SRL, Inc. Serum (0.3 ml) was analyzed by a reverse-phase high-performance liquid chromatography column-switching method (LC-2000Plus; JASCO Corporation) (70) using Wakosil GP-N6 4.6x150 mm as a pretreatment column and Wakosil-II 5C18 HG 4.6x250 mm as an analysis column (FUJIFILM Wako Pure Chemical Corporation). The neopterin level was determined by detecting its native fluorescence (excitation, 353 nm; emission, 438 nm) with a fluorescence detector (FP-2025; JASCO Corporation).

For the in vitro cytokine response assay, when IL-1β was undetectable in negative control samples (14 of 20 enrolled subjects; Table SI), 0.5x lower limit of detection (0.05 pg/ml) was used to calculate the QICI value.

In the analysis of the trial data, the interquartile range (IQR) method was used to identify outliers; any values that fell below Q1-1.5x IQR or above Q3 + 1.5x IQR (Q1, first quartile; Q3, third quartile) were considered outliers and removed from the statistical analysis (71,72). In order to perform statistical tests of matched pairs, the paired values of the outliers were removed, even if they fell into the non-outlier range. When calculating QICI scores, outliers and zero values were handled as follows: i) The outliers were replaced with mean values that were calculated from the non-outliers to avoid unreasonable reduction of QICI scores by simple removal of the outliers; and ii) zero values in the measurement of IL-1β and IL-18 were replaced with the values of 0.5x lower limit of detection (IL-1β, 0.05 pg/ml; IL-18, 8.30 pg/ml) (73,74).

The normality of the data was firstly examined using the normal quantile-quantile plots and the Shapiro-Wilk test. On the basis of the results from these normality tests, the Friedman test was used, followed by the Nemenyi post hoc test for the data from the clinical trial; a two-tailed paired Student's t-test was used for the data from the assays of innate immune cell activity.

All statistical analyses were performed with EZR v1.53 (Saitama Medical Center, Jichi Medical University), which is a graphical user interface for R (R Foundation for Statistical Computing; https://www.R-project.org/) (75). A priori sample size calculation and post hoc power analysis were performed using G*Power v3.1.9.2 (Department of Experimental Psychology, Heinrich Heine University Düsseldorf; https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower) (76). Spearman's rank correlation coefficients for the post hoc power analysis were calculated with EZR v1.53(75). P<0.05 was considered to indicate a statistically significant difference.

Firstly, the species with the highest and lowest QICI among the three probiotic LAB (L. plantarum, B. longum and L. lactis ssp. lactis) were determined in each of the 20 study subjects, using in vitro cytokine response assay. L. plantarum demonstrated the highest QICI in all subjects, whereas B. longum demonstrated the lowest QICI (Table SI). L. plantarum had a 52-7,210-fold (mean ± SD, 1,350±1,870; 95% CI, 458-2,250) higher QICI than B. longum and a 3-188-fold (mean ± SD, 33±44; 95% CI, 12-54) higher QICI than L. lactis ssp. lactis. The present results indicated that L. plantarum had a superior QFPD-like ability to stimulate innate cytokine production by blood immune cells. Therefore, L. plantarum and B. longum were selected in all subjects for the subsequent clinical trial.

To investigate whether L. plantarum could reproduce in vivo the QFPD-like immunomodulatory activity observed in vitro, a single-arm, double-blind, prospective trial that included three consecutive sessions was conducted: i) A validation session using L. plantarum in the first 7 days; ii) a 7-day washout period; and iii) a control session using B. longum in the last 7 days. The peripheral blood samples that were obtained before (day 0) and after (day 8) L. plantarum ingestion and after B. longum ingestion (day 22) were evaluated for plasma TNF-α, IL-1β, IL-18, IL-8 and IL-6, as well as the QICI.

As indicated in Table I, oral intake of L. plantarum significantly increased plasma IL-1β [median (IQR), 0.000 (0.000-0.000) vs. 0.134 (0.092-0.292) pg/ml; P=0.0000310 after Friedman test; P=0.00284 after Nemenyi post hoc test] and decreased plasma IL-6 [median (IQR), 1.180 (0.812-2.130) vs. 0.495 (0.425-0.775) pg/ml; P=0.0131 after Friedman test; P=0.0128 after Nemenyi post hoc test]. There were no significant differences in the plasma levels of TNF-α, IL-18 and IL-8. The QICI value was significantly increased [mean fold change, 17-fold; median (IQR), 1,760 (680-3,550) vs. 12,300 (5,440-42,200); P=0.0173 after Friedman test; P=0.0138 after Nemenyi post hoc test]. The QICI values were increased in 16 of 18 subjects (88.9%) with considerable individual differences in the fold change (1-128-fold; mean fold change, 19-fold; Fig. 2A), suggesting that there are large variations in responsiveness to L. plantarum among individuals.

By contrast, oral intake of B. longum induced a significant decrease in plasma IL-1β [median (IQR), 0.134 (0.092-0.292) vs. 0.000 (0.000-0.000) pg/ml; P=0.0000310 after Friedman test; P=0.000706 after Nemenyi post hoc test]; however, the QICI value did not change significantly [mean fold change, 2-fold; median (IQR), 12,300 (5,440-42,200) vs. 2,780 (940-9,820); P=0.0173 after Friedman test; P=0.474 after Nemenyi post hoc test]. The QICI values increased in 8 of 18 subjects (44.4%), but the fold changes were markedly lower (1-9-fold; mean fold change, 3-fold; Fig. 2B) than those obtained during the L. plantarum session. The present results suggested that orally administered L. plantarum induced an in vivo cytokine change similar to that induced by oral QFPD in a previous experiment (51).

L. plantarum ingestion also caused a minor but significant change the mean corpuscular hemoglobin concentration [median (IQR), 33.3 (32.4-33.9) vs. 32.6 (32.1-33.1)%; P=0.00456 after Friedman test; P=0.00836 after Nemenyi post hoc test] (Table SII). No significant changes in the hematological parameters were observed during the B. longum session (Table SII).

The post hoc two-tailed power analysis revealed that satisfactory effect sizes were obtained [0.845 (QICI in the L. plantarum session)-1.20 (IL-1β in the B. longum session)], as well as statistical powers [0.908 (QICI in the L. plantarum session)-0.996 (IL-1β in the B. longum session)] after completion of the trial (Table SIII).

Subsequently, the present study examined whether the L. plantarum-induced cytokine changes (increase in the QICI score) led to the increased activity of innate immune cells. L. plantarum ingestion significantly enhanced the activity of NK cells, which are key effectors of antiviral innate immunity that directly attack virus-infected host cells (Table II; Fig. 3A) (77,78). By contrast, B. longum ingestion significantly promoted the phagocytic activity of neutrophils (Table II; Fig. 3B). Neither L. plantarum nor B. longum ingestion activated macrophages, as assessed by the serum neopterin levels (Table II; Fig. 3C). The present results further supported the immunological benefits of L. plantarum and B. longum against viral infection.