A meta-analysis reveals the effectiveness of probiotics and prebiotics against respiratory viral infection

Abstract Experimental experience suggests that microbial agents including probiotics and prebiotics (representative microbial agents) play a critical role in defending against respiratory virus infection. We aim to systematically examine these agents’ effect on respiratory viral infection and encourage research into clinical applications. An electronic literature search was conducted from published data with a combination of a microbial agents search component containing synonyms for microbial agents-related terms and a customized search component for respiratory virus infection. Hazard ratio (HR), risk ratio (RR) and standard deviation (SD) were employed as effect estimates. In 45 preclinical studies, the mortality rates decreased in the respiratory viral infection models that included prebiotics or prebiotics as interventions (HR: 0.70; 95% confidence interval (CI): 0.56–0.87; P=0.002). There was a significant decrease in viral load due to improved gut microbiota (SD: −1.22; 95% CI: −1.50 to −0.94; P<0.001). Concentrations of interferon (IFN)-α (SD: 1.05; 95% CI: 0.33–1.77; P=0.004), IFN-γ (SD: 0.83; 95% CI: 0.01–1.65; P=0.05) and interleukin (IL)-12 (SD: 2.42; 95% CI: 0.32–4.52; P=0.02), IL-1β (SD: 0.01; 95% CI: −0.37 to 0.40; P=0.94) increased, whereas those of TNF-α (SD: −0.58; 95% CI: −1.59 to 0.43; P=0.26) and IL-6 (SD: −0.59; 95% CI: −1.24 to 0.07; P=0.08) decreased. Six clinical studies had lower symptom scores (SD: −0.09; 95% CI: −0.44 to 0.26; P=0.61) and less incidence of infection (RR: 0.80; 95% CI: 0.64–1.01; P=0.06). Our research indicates that probiotics and prebiotics pose a defensive possibility on respiratory viral infection and may encourage the clinical application.

Experimental experience suggests that microbial agents including probiotics and prebiotics (representative microbial agents) play a critical role in defending against respiratory virus infection. We aim to systematically examine these agents' effect on respiratory viral infection and encourage research into clinical applications. An electronic literature search was conducted from published data with a combination of a microbial agents search component containing synonyms for microbial agents-related terms and a customized search component for respiratory virus infection. Hazard ratio (HR), risk ratio (RR) and standard deviation (SD) were employed as effect estimates. In 45 preclinical studies, the mortality rates decreased in the respiratory viral infection models that included prebiotics or prebiotics as interventions (HR: 0.70; 95% confidence interval (CI): 0.56-0.87; P=0.002). There was a significant decrease in viral load due to improved gut microbiota (SD:

Background
Annually, approximately 200 million people experience viral community-acquired pneumonia (CAP) worldwide [1], 24.5% of which were infected with respiratory syncytial virus (RSV), influenza virus and rhinovirus among the most prevalent types [2]. According a review in 2015, an estimated 300-500 million severe cases of the CAP are caused by infection with influenza virus, whereas nearly 14000 in-hospital deaths are related to RSV infection [3]. The emergence of the infectious diseases especially severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS) caused by coronavirus, has led to the life-threatening nature of viral pneumonia.
More recently, novel coronavirus pneumonia (COVID- 19) has become a pandemic catastrophe, with a fatality rate rising to 49% in critical cases [4]. At the time of this analysis, updated by the World Health Organization (WHO) for the COVID-19 pandemic, the pandemic caused by the SARS-CoV-2 coronavirus infection had brought over 1.9 million deaths globally out of 88 million [5]. Depending on the host's immunocompetence, the severity of this respiratory disease can vary from mild symptoms to fatal complications, such as acute respiratory distress and multiple organ dysfunction [6]. However, current strategies to combat SARS-CoV-2 are far from satisfactory. Other than the two neuraminidase inhibitors: acyclovir and ganciclovir, there are few efficacious and efficient antiviral agents.
Preclinical research on the pathology of the viral pneumonia has shown that microbial agents including probiotics (exogenous salutary bacteria like Bifidobacterium) and prebiotics (indigestible substance promoting the growth of salutary bacteria, such as oligosaccharides) may modulate the composition of gastrointestinal flora and provide beneficial effects for patients with respiratory diseases, attributing to the gut-lung axis theory. When mice were manually exposed to a germ-free state through antibiotic treatment, both virus-specific CD4 and CD8 T cells and the influenza-specific antibody levels were markedly decreased, leading to a delay of the invasive virus elimination [7]. It was observed that the population of intestinal Bacteroidetes was significantly increased while the content of Firmicutes was decreased in murine models of RSV. By administering probiotic supplements, the incidence of upper respiratory tract infections was reduced and the disease duration was shortened. According to the microbiota-gut-lung axis theory, the gut microbiota plays the critical role in response to viral lung infections [8].
So far, the studies on probiotics and prebiotics treatments for respiratory viral infections failed to yield the expected results. Thus, in the present study, we proposed and evaluated the existing scientific evidence and credible results from the published preclinical and clinicalstudies, aiming to provide some useful information and suggestions for the future studies of stopping the rampant respiratory virus infection.

Search strategy
We conducted a meta-analysis following the recommended Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. We searched for the published studies up to December 2020 in PubMed, the Cochrane Library, Embase, the China National Knowledge Infrastructure, the Chinese Biomedical Literature Database and VIP databases, using combinations of a microbial agents search component containing synonyms for microbial agents-related terms and a customized search component for viral pneumonia. Retrieved articles were exported to an EndNote file. After removing duplicates, we screened the titles, abstracts and full texts according to the study selection criteria. There was no restriction for language, and a consensus was reached by mutual discussion.

Study selection
We included only the randomized controlled studies that investigated directly the effects of microbial agents on the respiratory viral infections. We excluded the articles that reported only the preventive effects on respiratory infection events which also contained bacteria-related events but did not specifically refer to the viral infection. We also excluded the comments, letters and other similar articles from which no relevant data could be extracted. There were no restrictions for specific variables such as different virus species and types of microbial agents.

Quality assessment and data extraction
To assess the quality for the studies included, we applied the Cochrane Risk Assessment Scale to the clinical studies, and used SYRCLE's risk of bias tool (RoB) for animal studies. The following general information from the studies were extracted: first author, publication year, age, sex, animal species, sample size, condition at baseline, respiratory virus pathogens, intervention (type, dose, duration), control, samples collected for outcome evaluation and effect estimates. Disagreements between the authors during the data abstraction were resolved by referring to the original article.

Statistical analysis
We evaluated the heterogeneity between study-specific estimates using the I 2 statistic and Cochran's Q test, for which an I 2 value >50% or a P-value <0.10 was considered as significant heterogeneity. Considering the significant differences in study design among the selected studies, we pooled data using a random-effects model. In cases with significant heterogeneity, we carried out subgroup and sensitivity analyses if permitted. We also estimated the publication bias using Egger's test. The analyses for preclinical and clinical trials were conducted separately. We extracted data separately for evaluation from various sampling locations, microbial agents types and viral species in the individual studies. We employed the software applications GetData 2.20, RevMan 5.3 and Stata 12.0 for the data extraction and synthesis. For all statistical procedures (except for heterogeneity), we defined a P-value of <0.05 as statistically significant.

Literature search
As shown in Figure 1A,B, the broad database searches resulted in 9634 preclinical and 10661 clinical hits, which we reduced to 8282 and 8438, respectively, after removing duplicates. After screening separately the titles and abstracts meeting with the protocol eligibility criteria, we included preclinical 104 and clinical 65 articles, respectively. After examining the full text, only 45 preclinical and 6 clinical studies were eligible for the data extraction and final evaluation.

Study characteristics Preclinical studies
Among the 45 selected studies ( Table 1 ), most of them (42) were conducted using mice model; only one using chickens, one using preweaned dairy calves [9] and one using pigs [10][11][12]. Among the challenge viruses, influenza A had the highest frequency, followed by RSV. In terms of microbial agents, the included animal studies mainly determined the effects of administering prebiotics and probiotics on respiratory viral infection, in which different soluble oligosaccharides were regarded as prebiotics, and live or heat-inactivated lactic acid bacteria were selected as probiotics. The datasets from two studies that used ribavirin (a broad-spectrum antiviral drug) were combined as an additional control group [13,14], distinguishing them from the other studies just employing phosphate saline buffer or saline.

Clinical studies
Only six clinical studies were eligible under the predetermined criteria (Table 2), and all were the randomized, double-blind, placebo-controlled trials. Rhinovirus was chosen as the final infectious target in the studies, except for two studies choosing human influenza virus [15] and SARS-CoV-2 [16]. Various probiotics or prebiotics were used. Turner [15,19]. Luoto et al. used a mixture of two prebiotics (galacto-oligosaccharide and polydextrose at a 1:1 ratio) and also a probiotic (L. rhamnosus GG) as the intervention factors in their study. They wanted to determine the prophylactic effect in 94 preterm infants with the gestational ages greater than 32 + 0 or less than 36 + 6 weeks and birth weight >1500 g [20]. In study from Ettorre et al., participants were 70 hospitalized patients positive for COVID-19. Oral bacteriotherapy using a multistrain formulation, which contained various probiotics such as Streptococcus thermophilus and L. acidophilus, were selected as intervention.    SCFA, short-chain fatty acid; SPF, specific-pathogen-free. ↑, the effect in intervention group was greater than control group; ↓, the effect in intervention group was smaller than control group. Quality assessment according to the Cochrane Risk Assessment Scale. Abbreviations: IFN, interferon; TCID, tissue culture infective dose. ↑, the effect in intervention group was greater than control group; ↓, the effect in intervention group was smaller than control group.

Figure 2. Survival analysis in preclinical studies
We performed a forest plot of the survival analysis in preclinical studies using RevMan 5.

Survival analysis
Twenty studies comprising 24 subgroups were included in the meta-analysis using the random effects model. As a result, a pooled hazard ratio (HR) of 0.70, with a 95% confidence interval (CI) of 0.56-0.87 was obtained, which corresponded to a higher survival rate for the microbial intervention group than that for the control group after the virus challenge, without heterogeneity (I 2 = 0.0%, P=0.88) ( Figure 2). Notably, none of the pooled effect estimate in each selected study had statistical significance, probably due to the small sample size except for the data from one study by Maruo et al. [21]. We also evaluated the publication bias using Stata software 12.0, with no quantitative publication bias exists in Egger's test (P=0.10).

Viral load
Results showed a pooled standard deviation (SD) of −1.22 and 95% CI of −1.50 to −0.94 (P<0.001), revealing that the consumption of probiotics or prebiotics alleviated the viral load after a respiratory viral infection (Figure 3). Due to the significant heterogeneity (I 2 = 71.3%, P<0.001), we performed a subgroup analysis and sensitivity analysis ( Table 3). We observed that the heterogeneity would be significantly affected if the eligible studies were grouped into 'SD' (I 2 = 79%, P<0.001) and 'standard error of the mean (SEM)' (I 2 = 22.3%, P=0.24) based on the effect estimates adopted by the individual authors. We sequentially analyzed the studies in which 'SD' was employed as the effect estimate. Heterogeneity was significantly affected when the studies were grouped into 'H1N1' (I 2 = 73%, P<0.001) and 'others' (I 2 = 0.0%, P=0.53) according to the specific virus challenge but was not affected when the studies were divided into 'probiotics' (I 2 = 84%, P<0.001) and 'prebiotics' (I 2 = 90%, P<0.001) according to the types of microbial agents, indicating that the differences between the specific virus species might cause the final heterogeneity.
Considering that there was no clear decrease in heterogeneity in the subgroup analyses by microbial agents' types, we conducted a sensitivity analysis and discovered that the study performed by Maramatsu et al. was the main source of Figure 3.

Analysis of viral load in preclinical studies
We performed a forest plot of viral load in preclinical studies using RevMan 5.3. We included studies using SD and SEM, transforming SEM to SD for better construction. The data were pooled using a random effects model and expressed as SD with 95% CIs.  Data were analyzed using a random-effects model. We analyzed the Effect estimates group for the included preclinical studies regarding viral load. We analyzed the viral type and microbial agent groups in the preclinical studies using SD as the effect estimate. Abbreviations: Ctr, control group; Exp, experimental group.
heterogeneity. We also evaluated the publication bias using Stata data analysis and determined a distinct publication bias based on Egger's test (P<0.001).

Adverse events
In most animal studies, the adverse events of probiotics and prebiotics are not recorded for lack of observable consequences. The reported aspiration pneumonia could be induced by either intranasally administrating high doses of live L. rhamnosus or the same dose of dead L. rhamnosus [22].

Infection rate
Three studies that consisted of five subgroups including a highly heterogenetic subgroup focused on prebiotics were pooled for analysis using the random-effects model, As a consequence, an overall risk ratio (RR) and 95% CI of 0.80 and 0.64-1.01 (P=0.06) were shown in Figure 6. The results suggested that a decreased viral morbidity was due to the treatment of probiotics and prebiotics. If larger sample size were used for the analysis, the observed treatment effects could be statistically significant [20].

Adverse events
Among the studies were pooled for analysis, only Turner et al. reported gastrointestinal adverse events after treatment with probiotics, but no details were described [18].

Quality assessment
For clinical studies, all targets included for evaluation had low risk of selection bias and performance bias because they were randomized, double-blind, placebo-controlled trials. Most of the studies had low risk of reporting bias and other bias except for one study that had high risk of attrition bias [18] (Table 2). Quality evaluating in preclinical studies was detailed in Supplementary Table S1.

Discussion
With the exploration of microbial agents, their properties that favor successful defense against respiratory virus infection have gaining a mass of interests. Olaimat et al. presented clinical fruits of the use of probiotic supplementation to prevent or treat respiratory tract infections [23] while Shinde et al. determined to identify evidence relating to potential mechanisms [24]. However, they just elaborated this subject qualitatively. Our study aimed to evaluate the effectiveness of on probiotics and prebiotics for viral pneumonia quantitatively, through the published data from 45 preclinical and 6 clinical studies. The probiotics included the live probiotics such as L. rhamnosus [25][26][27][28], Bifidobacterium [29,30] and a heat-inactivated one Enterococcus faecalis [13,31]. The prebiotics were commonly used such as inulin [32], polysaccharide [13,[33][34][35] and oligosaccharide [36]. As a result, we found that both probiotics and prebiotics played a crucial role in treating viral pneumonia. In viral pneumonia, inflammatory cascades could be commonly observed. Our results showed that probiotics or prebiotics helped to reduce the viral load leading to increase the overall survival through the up-regulation of the antiviral cytokines, IFN-α, IFN-γ, IL-1β and IL-12 and the down-regulation of the other cytokines IL-6 and TNF-α.

Figure 4. Evaluation of cytokines in preclinical studies
We evaluated the IFN-α, IFN-γ, IL-12, TNF-α, IL-1β and IL-6 concentrations compared with the control group through a randomeffects models using RevMan 5.3. The pooled data are expressed as SD with 95% CIs. I 2 and P-values represent the heterogeneity among the studies, while an I 2 value >50% and a P-value <0.1 indicate considerable heterogeneity. The 95% CI of the result intersecting with the solid vertical line represents no statistical significance.
The results from the cohort clinical studies on the influenza virus are generally consistent with those from preclinical studies. Patients treated with probiotics or prebiotics had lower disease severity and fewer infections. However, these findings warrant further studies to understand the effectiveness of probiotics and prebiotics in preventing viral pneumonia becuase of currently insufficient clinical studies and coherent indicators.  Interferon is a crucial antiviral factor that has a vital role in assessing host immunity. Unfortunately, there was no measure valid for the benefit of interferon in clinic yet. Our analysis revealed that all the probiotics or prebiotics tested could notably increase the interferon levels, consistent with a previous study that healthy athletes increased interferon secretion after a 1-month course of L. acidophilus. through a mechanism of engaging Toll-like receptors on the surface of antigen-presenting cells, which would, in turn, affect the subsequent cytokine secretion pattern [37,38], thereby indicating that probiotics and prebiotics work energetically by enhancing the host's antiviral capabilities. Toll-like receptors (TLRs) play crucial roles in the innate immune system by recognizing pathogen-associated molecular patterns derived from various microbes. It has been reported that both neutrophil granulocytes and regulatory T cells put a halt to the reaction by releasing quantities of anti-inflammatory chemicals when an inflammatory response occurs within a capable immune system [14,39]. However, when the host is immunocompromised, effective responses break down. Accordingly, viral duplication that exacerbates tissue damages is attributed to a burst of inflammatory cytokines induced. Typically, the critical COVID-19 patients reported with higher plasma concentrations of IL-7, IL-8, IL-9, basic fibroblast growth factor, granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor were in high fatality [6].
A well-documented postbiotic is called short-chain fatty acids (SCFAs) is that are produced by bacterial fermentation of indigestible fibers. The most abundant SCFAs refer to propionate, acetate and butyrate. SCFAs act on G protein-coupled receptors (GPRs) 41 and GPR43, [40,41] or as histone deacetylase inhibitors [42] to down-regulate proinflammatory chemokine and combat cytokine cascades. Our findings demonstrated that immune cells took advantage of bacterial metabolites to enhance antiviral response. The mucosal immune system, specifically the lymphoid tissues on the mucosal surface, was also involved in modulating anti-virus immunity [43,44]. Microfold cells could absorb bacterial metabolites into the circulation, with a bond of mucosal immune system where they would stimulate immune cells and rapidly recruit them [45].
IL-12 and IL-6 are representative activators to Th1 and Th2, respectively. In severe influenza infections, there was a marked Th polarization shift from Th2 to Th1 [46]. However, during probiotics or prebiotics treatment, this inflammation-oriented polarization could be reversed through up-regulation of IL-12 and down-regulation of IL-6. Furthermore, Chen et al. and other researchers discovered that production of IL-10 from Treg cells was increased [13,25]. IL-10 functions to limit the host immune response to pathogens, thereby preventing damage to the host and maintaining normal tissue homeostasis. Thus, we can speculate that probiotics and prebiotics could recruit Treg cells and up-regulate IL-10 concentrations to achieve an antiviral effect by preventing immoderate inflammatory responses through inhibiting production of the inflammatory cytokines, TNF-α and IL-6.
Concerning secondary infections incurred by enterogenous endotoxemia during severe viral pneumonia, the gut microbiota could notably defend it. The gut mucosal barrier is composed of mucus, symbiotic flora, tight junctions between intestinal epithelial cells and mucosal immune cells, making it difficult for opportunistic infections to take root. Once the virus fiercely strikes at the respiratory tissues, it would possibly bring about systematic hypoxia, where the intestinal epithelial cells would dysfunction and act to weaken the mucosal barrier [14]. In addition, the misuse or overuse of broad-spectrum antibiotics could invariably result in dysbiosis, blemishing intestinal permeability and endamaging gut mucosal barrier during antiviral treatment [47].
Probiotic supplements were considered to be an optimal approach to restore the gut mucosal barrier function in viral pneumonia. Probiotics can bind to Toll-like receptor-4, whose population could increase with the help of inactivated L. salivarius and fructo-oligosaccharides [48], thereby competing against harmful bacteria. In addition, probiotics and its metabolic profiles including bacteriocin, hydrogen peroxide, antimicrobial peptides and defensin, help to modulate the local immunity and drive enterocyte and goblet cells to secret mucus as a consequence to strengthen the mucosal barrier at length [49,50].
For COVID-19 patients, SARS-CoV-2 binds its spike proteins to angiotensin-converting enzyme 2 (ACE2). ACE2 is highly expressed in the bronchi and gastrointestinal tract to facilitate to viral invading and replication [51,52]. Since the invasive bindings to ACE2, ACE2 located in the gut might not function effectively, potentially altering the symbiotic flora and undermining the intestinal barrier, leading to patients prone to secondary infections. A published study reported that treatment of an irritant, compared with wildtype littermates, caused gut microbiota alteration to promote profoundly inflammatory reaction in ACE2 mutation mice, which could be directly regulated by microbial agents [53]. Encouragingly, probiotics and prebiotics treatment has been incorporated as adjuvant therapy for critical patients to prevent secondary infections in the fourth Trial Edition of COVID-19 Diagnosis and Treatment Plan by the National Health Commission of China, [54].
In summary, our study suggested that probiotics and prebiotics could be an inspiration for healthcare givers when treating viral pneumonia. This therapy could limit inflammatory responses, stimulate both innate and adaptive immune cells to defend against the viral attacks and preventing secondary infections. Our findings implied a promising target and encourage probiotics or prebiotics to be incorporated into regular treatments for patients infected with respiratory virus, particularly for the patients with severe viral pneumonia.

Limitations
The present study has several limitations. Firstly, most of the clinical studies related to the topic were not included because they focused generally on the respiratory tract infections but not particularly on respiratory viral infections. Thus, only a small number of clinical studies was eligible for our analysis, thereby influencing the extrapolation of outcomes. Secondly, considering the differences in experimental designs, we did not conduct a direct comparison of the merits of individual microecological agents tested. We also did not confirm optimal dosage, dosage form and duration, which need further investigation. Thirdly, despite remarkable functions showed in applying probiotics and prebiotics to treat viral pneumonia, the effects only limit to a certain amount of bacteria species and their products. Therefore, it is appropriate to specify the individual probiotics or prebiotics with more explorations.

Data Availability
All data generated or analyzed during the present study are included in this published article.