The Volatile and Heterogenous Gut Microbiota Shifts of COVID-19 Patients Over The Course of A Probiotics-Assisted Therapy

Chunyan Wu Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital https://orcid.org/00000003-3192-3126 Qian Xu Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Zhan Cao Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Dengdeng Pan Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Ying Zhu Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Sheng Wang Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Danping Liu Shanghai Public Health Clinical Center Zhigang Song Shanghai Public Health Clinical Center Wei Jiang Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital Yumeng Ruan Reabio Genomics Institute Nan Qin (  qinnan001@126.com ) Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital https://orcid.org/00000002-9128-9414 Hongzhou Lu Shanghai Public Health Clinical Center Huanlong Qin Tongji University Tenth People's Hospital: Shanghai Tenth People's Hospital

Given that the patients may manifest severe gastrointestinal symptoms and microbiota abnormalities, inclusion of probiotics into the medication may improve their treatment outcome. Here we employed a probiotics-assisted therapy to treat a group of COVID-19 patients, and the dynamics of gut microbiota and clinical manifestations was monitored over the course of treatment. We showed that the COVID-19 patients exhibited prominent gut microbiota shifts that were characterized by decreased microbial diversity, extensive taxonomic and transcriptional alterations, and great interpersonal heterogeneity and intertimepoint uctuations. Our ndings suggested a lack of relative stability in the gut microbiota of COVID-19 patients, which may help formulate strategies to improve the treatment outcome and posttreatment rehabilitation.

Results
Treatment outcome of the patients with COVID- 19 We recruited 13 COVID-19 patients, 15 non-COVID-19 pneumonia controls, and 15 healthy controls. The median ages of the three groups were 66, 51, and 64 years old, respectively (Table S1). All COVID-19 patients displayed respiratory symptoms (Table S1) and several of them had gastrointestinal manifestations, including one individual with diarrhea. Two patients (COV06 and COV13) showed positive results for SARS-CoV-2 in stool based on viral RNA metagenomics.
We next examined the strain-level variations in the COVID-19 patients. To this end, we applied StrainPhlAn to build the phylogenetic trees based on single nucleotide variants (SNVs) in species-specific marker genes, which revealed considerable strain-level heterogeneity of Escherichia coli in the patients with COVID-19 (Fig. S2). Speci cally, the dominant strains were phylogenetically close to E. coli UMN026 in patient COV11, to E. coli UMEA_3212-1 in patient COV13, and to E. coli KTE196 in patients COV01, COV06, COV09 and COV12, respectively. Hence, our analyses revealed that COVID-19 was associated with taxonomic and transcriptional shifts in the gut and/or airway microbiotas and that there was great heterogeneity among individual patients.

Functional shift of gut microbiota in the patients with COVID-19
Having revealed the clear taxonomic alterations of gut microbiota in the COVID-19 patients, we next sought to examine the microbial functional characteristics using the transcriptomic data. Of the 5,125 genes functionally annotated to Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologous groups (KOs), 1,982 exhibited differential transcriptional activities between the HCs and COVID-19 patients, including 866 decreased and 1,116 increased in COVID-19. At pathway level, the COVID-19 patients were characterized by enrichment of 11 MetaCyc pathways including beta-Lactam resistance, Bio lm formation-Escherichia coli, and Bacterial invasion of epithelial cells as well as depletion of 86 MetaCyc pathways, most of which belonged to the categories of amino acid metabolism, lipid metabolism, and carbohydrate metabolism (Table S5). In addition, the expression level of several virulence factors in the COVID-19 patients was signi cantly higher than both the healthy controls and pneumonia controls (Fig.   1d). The top 5 transcriptionally active virulence factors in the COVID-19 patients were Enterobactin (VF0228), ECP (VF0404), T2SS (VF0333), Type 1 mbriae (VF0221), and Yersiniabactin (VF0136), collectively accounting for 61.9% of the total transcriptomic abundance (Fig. 1e, Table S6). In addition, the top 5 contributing species of virulence factors in the COVID-19 patients were Escherichia coli, Shigella dysenteries, Yersinia pestis, Yersinia enterocolitica, and Salmonella enterica, collectively accounting for 98.8% of the total transcriptomic abundance (Fig. 1f, Table S7). The prominent presence of virulence factors and their contributing pathogens or pathobionts illustrated the deteriorating gut ora in the COVID-19 patients.
We next investigated the pro le of antibiotic resistance genes in gut microbiota of the participants. Genes potentially conferring resistance to 35 different antibiotics were identi ed in the COVID-19 patients, of which Macrolide, Tetracycline, Penam, Diaminopyrimidine, and Phenicol were the most transcriptionally active (Fig. S3a, Table S8). The expression of antibiotic resistance genes (ARGs) was signi cantly increased in COVID-19 ( Fig. 1d), with the most abundant types being resistance-nodulation-cell division (RND) antibiotic e ux pump, Erm 23S ribosomal RNA methyltransferase, major facilitator superfamily (MFS) antibiotic e ux pump, tetracycline-resistant ribosomal protection protein, and pmr phosphoethanolamine transferase (Fig. S3b, Table S9). Similar to our observation in taxonomic analyses ( Fig. 1b and Fig. S1), there were great interpersonal differences in ARGs. For example, the most transcriptionally active ARGs were Erm 23S ribosomal RNA methyltransferase in COV04, COV10, and COV11, RND antibiotic e ux pump in COV05, COV06, COV07, COV08, COV09, COV12, and COV13, tetracycline-resistant ribosomal protection protein in COV01, COV03, and COV13, and CfxA betalactamase in COV02, respectively. Overall, our ndings revealed extensive alterations of gut microbiota in COVID-19 in both taxonomic and functional aspects, which featured prominent variations among individual patients.
Association between gut microbiome and disease severity of COVID-19 We next investigate the correlation of gut or upper airway microbes with a set of clinical indicators (Fig.   2a). Among upper airway taxa, Bacteroides, Akkermansia, and Enterococcus showed negative association with CD3 and CD4. Among gut taxa, Rhodococcus and Acinetobacter showed negative association with CD3, CD4, CD45, haemoglobin (Hb) concentrations, and red blood cells (RBC); Bacteroides and Veillonella showed positive association with haemoglobins, RBC, and CD3, respectively; Enterococcus showed positive association with plasma concentration of carbon dioxide.
We also detected signi cant correlations between microbial functional components and clinical indicators. The expression level of bacteria virulence factors (e.g., Yersiniabactin, Salmochelin, Shu, and SgrA) were negatively correlated with CD8 (Fig. 2b), whereas multiple antibiotic resistance genes (e.g.,beta lactamase, chloramphenicol acetyltransferase (CAT), undecaprenyl pyrophosphate related proteins, and multidrug and toxic compound extrusion (MATE) transporter) were negatively associated with CD3 (Fig. 2c). These ndings indicated that gut and airway microbiota shifts were linked to the clinical manifestations of in ammation, possibly related to COVID-19.
Convalescence coincided with the recovery of dysbiosis in the COVID-19 patients To investigate the dynamics of the gut microbiome in the COVID-19 patients during the process of treatment, we compared the taxonomic data among the baseline (T0, rst sampling date prior to the probiotics-assisted therapy), 7-day posttreatment (T1), and 14-day posttreatment (T2) for each patient. Among the 12 recruited COVID-19 patients, stool samples both before and after the therapy could only be collected from 8 individuals (Fig. 1b). At the end of the treatment, the 8 COVID-19 patients showed a partial rise in the microbial diversity (Wilcoxon rank-sum test, P < 0.05) and 6 out of the 8 COVID-19 patients (75%) showed a partial microbial compositional "recovery", evidenced by a decrease of BC (Bray-Curtis) distance to the healthy group (Fig. 3a). This alteration was accompanied by substantial taxonomic shift, such as COVID-19-increase genera Enterococcus and Rhodococcus registered at least 2fold reduction in 3 patients (COV03, COV05, and COV12) and 4 patients (COV05, COV08, COV09, and COV11), respectively, and COVID-19-depleted genera Faecalibacterium and Roseburia exhibited at least 2fold increase in 3 patients (COV11, and COV12) and 4 patients (COV03, COV06, COV09, and COV12), respectively (Table S10). This partial compositional "recovery" was accompanied by a similar trend in the microbial transcriptome. After treatment, 23%-63% of the species that had COVID-19-associated transcriptional elevation exhibited at least 2-fold reduction, whereas 7%-67% of the species that had COVID-19-associated transcriptional reduction exhibited at least 2-fold increase (Table S11). For example, the major gut commensal bacterium Faecalibacterium prausnitzii was increased in patients COV05, COV08, COV09, COV11, and COV12; butyrate-producing anaerobic bacterium Roseburia hominis was increased in patient COV05, COV09, COV11, and COV12. On the contrary, opportunistic pathogen Escherichia coli was decreased in all patients except COV03; Salmonella enterica was decreased in COV05, COV06, COV08, COV09, and COV12; Providencia alcalifaciens was decreased in all patients except COV03 and COV11; Staphylococcus auricularis was decreased in COV03, COV07, COV11, and COV12; Klebsiella pneumoniae was decreased in all patients except COV03 and COV05 (Fig. 3b). The ndings suggested that the treatment correlated with a noticeable recovery of the gut microbiota perturbations in the COVID-19 patients.
An important nding of the longitudinal 16S data was that there were great intertimepoint compositional variations of gut microbiome in most COVID-19 patients (Fig. 1b). In patients COV03, COV05, COV06, and COV08, the most dominant genus at T0 exhibited a great reduction in relative abundance at the end of treatment, whereas the most dominant genus at the end of treatment was barely detectable at T0. For example, the top taxa of patient COV03 were Bacteroides and Lactobacillus before and after the treatment, respectively; the top taxa of patient COV05, COV06, and COV08 was Escherichia/Shigella prior to the treatment, as were Bacteroides, Enterococcus, and Veillonella at the end of treatment. The prominent microbiota changes were also evident in patient COV11, whose shares of Akkermansia and Bacteroides were 54% and 29%, 75% and 5% or 19% and 44% at T0, T1 or T2. For the airway microbes, the drastic intertimepoint changes were present in only 3 patients. Speci cally, the top genera of patients COV01, COV03, and COV03 were Ralstonia, Clostridium sensu stricto, and Pseudomonas at T0 and Lactobacillus, Lactobacillus, and Corynebacterium at T2, respectively (Fig. S1).
The drastic compositional shift over the course of treatment was mirrored by the bacterial transcriptinal activities. In patients COV03, COV05, COV06, COV09, and COV12, the most transcriptionally active species differed between the rst and last timepoints. For example, the top species of patient COV03 were Bacteroides fragilis before the treatment and Yersinia enterocolitica afterwards. Of note, there was apparent divergence between the most abundant taxa and most transcriptionally active microbes. For example, in patient COV01, the most abundant taxon was Enterococcus, whereas the most transcriptionally active species was Bacillus cereus; in patient COV12, Fecaelibacterium prausnitzii was transcriptionally active, although the genus was barely detected in 16S rRNA gene-based analysis.
However, noticeable coincidence can be found for Escherichia/Shigella and E. coli in patients COV05, COV06, and COV09, Akkermansia and A. muciniphila in patient COV11, Klebsiella and K. pneumoniae in patient COV06, and Veillonella and V. parvula in patient COV08. Overall, our longitudinal analyses revealed that over the course of treatment, the COVID-19 patients exhibited apparent recovery in the disease-associated microbial abnormalities and that the gut microbiota of the COVID-19 patients showed drastic intertimepoint variations.

Discussion
In this study, we characterized the dynamic composition and transcriptomics of gut microbiota in patients with COVID-19 over the course of treatment. The patients with COVID-19 showed prominent gut microbiome shifts from healthy controls and non-COVID-19 that were characterized by taxonomic and transcriptional differences as well as great interpersonal heterogeneity and intertimepoint variations. This was accompanied by microbial alterations in the upper airway. The taxonomic shifts included decreased relative abundances of Faecalibacterium, Roseburia, and Clostridium XlVa as well as increased proportions of Enterococcus, Rhodococcus, and Acinetobacter. The transcriptional alterations in the gut microbiota included the augmented presence of opportunistic pathogens or pathobionts (e.g., Escherichia coli, Salmonella enterica, Staphylococcus auricularis, and Klebsiella pneumoniae), virulence factors, and ARGs as well as diminished presence of Faecalibacterium prausnitzii, a major gut commensal. These changes indicated the deteriorating gut ecosystem in the patients. In addition, correlations were detected between microbial components (e.g., gut microbes, bacteria virulence factors, antibiotic resistance genes, and airway microbes) and in ammatory indicators (e.g., CD3, CD4, haemoglobin concentrations, and red blood cells). After treatment, the COVID-19 patients exhibited a partial recovery of the gut microbiome perturbations, evidenced by the increased microbial diversity as well as a partial microbiota recovery: that is the "approaching" of the COVID-19-associated taxonomic and transcriptional pro les to those in healthy controls. Hence, our ndings illustrated the complex changes occurring in the gut and upper airway microbial communities of COVID-19 patients that are pivotal to understand the disease and its therapy.
Intestinal ora is essential for the maintenance of host immunity and homeostasis. The microbial community may regulate host immune and in ammatory responses along the gut-lung axis via microbial metabolites and the mucosal immune system, thereby playing a pivotal role in host health 11 . An important feature of the human gut microbiota is its relative stability, which may be sustained for one year 12 or even decades 13 . In contrast, the COVID-19 patients exhibited great intertimepoint uctuation. Such fragile, unstable gut microbiota in COVID-19 patients is likely attributed to the reduced microbial diversity as well as decreased presence of some major commensals (e.g., Faecalibacterium prausnitzii).
In light of this, measures targeting the COVID-19-associated microbiota abnormalities could be a crucial avenue of multi-pronged therapies.
In severe COVID-19 cases, the infection can trigger a series of immune responses or cytokine storms that correlate with many of the symptoms. In our study, the probiotics-assisted therapy led to clinical improvements in COVID-19 patients, which was characterized by the reduced in ammation (e.g., tumor necrosis factor (TNF)-α, interleukin-1β, interleukin-4, and interleukin-12P70). It worth noting that a high dosage of probiotics (i.e., ca. 200 billion colony forming units of bacterial cells daily) was prescribed for the COVID-19 patients, who showed no adverse reactions (e.g., diarrhea) linked to the auxiliary therapeutics. The results suggested that the probiotics supplementation is safe and is amenable for inclusion into other medications for COVID-19 patients who may have gut microbiota abnormalities.

Conclusions
We identi ed substantial COVID-19-associated gut and upper airway microbiota shifts, including alterations in taxonomic composition and transcriptional activities. In addition to resolving the respiratory symptoms in COVID-19 patients, a probiotics-assisted therapy correlated with partial recovery of the microbiota perturbations. These ndings provided new insights into the gut and airway microbiome characteristics of COVID-19 that may have crucial clinical implications.

Recruitment of participants
This study recruited 13 hospitalized COVID-19 patients, 15 hospitalized patients with community acquired pneumonia (pneumonia controls) and 15 healthy healthy controls (Table S1). SARS-CoV-2 infection was detected by dual RT-PCR test targeting two different regions of the RdRp gene. Non-COVID-19 pneumonia controls were patients admitted with community-acquired pneumonia who were tested negative for SARS-CoV-2. Healthy controls were individuals with no past medical history or history of antibiotic intake in the past 3 months and tested negative for SARS-CoV-2. All participants were admitted to Shanghai Public Health Center (SHPHC, Shanghai, China). Written informed consent was obtained from all participants or their families. All procedures were performed in compliance with the Declaration of Helsinki.
Epidemiological, clinical, radiological characteristics, laboratory, and treatment data were obtained from the electronic medical records. Clinical information included demographic data, medical history, exposure history, underlying comorbidities, symptoms, signs, laboratory ndings, chest computed tomographic (CT) scans, and treatment measures (ie, antiviral therapy, antibiotics therapy, respiratory support).

Statistics
To detect signi cant differences in relative abundance of taxon features, the nonparametric Wilcoxon test (wilcox.test in R) was performed with false discovery rate (FDR) <0.05 (Benjamini-Hochberg), and the enrichment group was then determined according to the higher rank-sum.

Declarations
Ethics approval and consent to participate This study was approved by the institutional review boards at Shanghai Public Health Center (SHPHC, Shanghai, China). Written informed consent was obtained from all participants or their families.

Consent for publication
Not applicable.

Availability of data and material
Sequencing dataset was deposited to the NCBI Sequence Read Archive under BioProject accession number PRJNA649161.

Competing interests
The authors declare that they have no competing interests.   Shifts of gut microbiome in COVID-19 before and after the treatment. a. Dissimilarity of the gut microbiota of COVID-19 patients to that of healthy controls before and after treatment. The microbiota dissimilarity was calculated as bray-curtis dissimilarity. b. The alteration of COVID-19 enriched species (c) and depleted species of transcriptional activity after treatment.