Taxonomical and functional changes in COVID-19 faecal microbiome are related to SARS-CoV-2 faecal load

Since the beginning of the pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) the gastro-intestinal (GI) tract has emerged as an important organ inuencing the propensity to and potentially severity of the related COVID-19 disease. However, the contribution of the SARS-CoV-2 intestinal infection on COVID-19 pathogenesis remains to be claried. In this exploratory study, we evidenced that alterations in the composition of the gut microbiota depends on the levels of SARS-CoV-2 RNA in the gastrointestinal tract but not on the presence of SARS-CoV-2 in the respiratory tract, COVID-19 severity and GI symptoms. Altered molecular functions in the microbiota proles of high SARS-CoV-2 RNA level faeces as established by metaproteomics highlight mechanisms that may contribute to vicious cycles. Uncovering the role of this gut microbiota dysbiosis could drive the investigation of alternative therapeutic strategies to favour the clearance of the virus and potentially mitigate the effect of SARS-CoV-2 infection. spectrometry-based prole gut microbiota of archaeal, analysed the associated metaproteomic intestinal COVID-19 microbiota found the of SARS-CoV-2 in the respiratory tract, disease severity, and (GI) but correlated with GI levels of SARS-CoV-2 RNA. Examination of the functional composition of the metaproteome provided a shortlist of both microbial and human biomarker candidates indicative intestinal SARS-CoV-2 infection. These biomarkers could be used to monitor Information on how intestinal SARS-CoV-2 affects the microbiota and the host could be useful in the search for alternative therapies promoting viral clearance, with a view to mitigating the impact of SARS-CoV-2 Moulds, accessible through metaproteomics. Our data shows three key results that could be of major importance in the battle against COVID-19: i) faecal SARS-CoV-2 viral load is not correlated to symptoms, ii) an important change in the microbial structure is observed for patients with high faecal SARS-CoV-2 viral load, and iii) a list of microbiome and human markers can be drawn from this study as possible candidates for diagnostic. These results should be incentive for an extensive multi-centric metaproteomics analysis of the gut microbiota of COVID-19 patients.

are secreted from infected gastrointestinal cells. Furthermore, re-emergence of coronavirus disease in several regions was recently associated with processing of frozen food products 13 , and initial contamination via ingestion cannot be excluded. Evidence obtained using organoid models revealed SARS-CoV-2 to exclusively target the apical surface of mature villous enterocytes expressing high levels of three proteins enhancing virus entry into enterocytes: angiotensin conversion enzyme 2 (ACE2) and two related membrane-bound serine proteases, TMPRSS2 and TMPRSS4 14,15 . Although infected enterocytes form syncytia and viral particles are shed into the lumen, experiments examining the effect of simulated human colonic uid on the virus suggest that shed virus is rapidly inactivated as it passes through the colon 16 .
Currently, little is known about the impact of the gastrointestinal presence of SARS-CoV-2 on the course of COVID-19. Evidence that the gut microbiome in uences ACE2 expression has led to several groups hypothesizing a contribution of intestinal microbes to COVID-19, but only very limited data are available on the pro le of the gut microbiome in SARS-CoV-2-infected patients 17,18 . Notably, alterations to the microbiota have mainly been discussed in relation to the detection of SARS-CoV-2 in the respiratory tract and COVID-19 severity. Thus, Zuo et al. 19 investigated transcriptional activity of SARS-CoV-2 and temporal microbiome alterations in faecal samples from patients with COVID-19. They observed that faecal samples showing high SARS-CoV-2 infectivity contained higher abundances of bacterial species Collinsella aerofaciens, Collinsella tanakaei, Streptococcus infantis, and Morganella morganii. Metatranscriptomics analysis revealed higher functional capacity for de novo nucleotide biosynthesis, amino acid biosynthesis, and glycolysis in these samples. In contrast, faecal samples with low-to-no SARS-CoV-2 infectivity were associated with higher abundances of short-chain fatty acid-producing bacteria.
Nevertheless, transcript presence does not necessarily indicate protein synthesis and thus a functional impact. To obtain this type of information, more precise functional information could be obtained by metaproteomics characterization. Pro ling of the human gut microbiome via metaproteomics has proven its value in pathogenesis research in the context of several diseases 20,21 . The potential of these techniques to help guide future clinical diagnosis has also been highlighted 22 .
In this study, we used a mass spectrometry-based approach to pro le the gut microbiota in terms of bacterial, archaeal, yeast, and fungal content, and analysed the associated metaproteomic functions in patients with intestinal COVID-19 infection. Altered microbiota compositions were found to be independent of the presence of SARS-CoV-2 in the respiratory tract, disease severity, and gastro-intestinal (GI) symptoms, but correlated with GI levels of SARS-CoV-2 RNA. Examination of the functional composition of the metaproteome provided a shortlist of both microbial and human biomarker candidates indicative of intestinal SARS-CoV-2 infection. These biomarkers could be used to monitor infection. Information on how intestinal SARS-CoV-2 affects the microbiota and the host could be useful in the search for alternative therapies promoting viral clearance, with a view to mitigating the impact of SARS-CoV-2 infection.

Levels of SARS-CoV-2 RNA in the gut
A total of 39 faecal samples collected from 39 patients has been studied (Table 1). Among them, 32 were included in the COVID group due to a RT-PCR and/or a positive CT-scan (Fig. 1A). The remaining 7 patients were without COVID-19 diagnosis (RT-PCR and CT-scan negative). The median ages and sex ratios of patients with COVID-19 and non-COVID-19 were 76.5 years (34-96) versus 79 (61-99) and 0.89 versus 1.33, respectively. To investigate the effect of intestinal SARS-CoV-2 infection on the composition of the gut microbiota, faecal samples were characterized by applying the pipeline presented in Fig. 1B. The presence of SARS-CoV-2 in the gut was analysed by subjecting stool samples to RT-qPCR. Of the 39 patients, 10 (25.6%) presented a positive RT-qPCR result. Among them, one (C01P001) belonged to the non-COVID group (RT-PCR and CT-scan negative). The median faecal viral load was 3.7 log 10 copies per mg [IQR 4.5-2.7], as estimated by RT-qPCR performed in parallel with standards. A viral load of 1.7 log 10 copies per mg (50 copies/mg) was arbitrarily chosen as the cut off value to group samples into high SARS-CoV-2 RNA levels, low SARS-CoV-2 RNA loads, and negative samples (Fig. 2). No obvious association (Kruskal-Wallis test, p < 0.05) was found between plasma C-reactive protein (CRP) levels and GI tract SARS-CoV-2 levels (p = 0.21). No correlation (Fisher exact test, p < 0.05) between the levels of SARS-CoV-2 RNA in the faecal samples and the positivity of nasopharyngeal swab tests on the rst hand (p = 0.40), and with severity of COVID-19 on the other (p = 1.00), were detected. Signi cantly altered microbiota pro le in patients with a high level of SARS-CoV-2 RNA in the GI tract To gain insight into the relationship between SARS-CoV-2 infection, microbiota, and host, we differentiated samples based on the levels of SARS-CoV-2 RNA detected in the GI tract. The total metaproteomic dataset from the 117 nanoLC-MS/MS runs comprised 7,761,229 MS/MS spectra. With the adjustment procedure of the peptide quantities to inject based on a pre-screen by mass spectrometry, an overall average of 66,335 ± 3,248 MS/MS was obtained per sample with low variation between samples. Peptides were identi ed by searching tandem mass spectra in a two-step cascaded search against a sample-speci c database. This strategy allowed 31 ± 6% of spectra to be assigned per sample.
For each sample, the distribution of the assigned TSMs as a function of their origin -microbial, host, or food-related - (Supplementary Table S1) revealed a higher percentage of host signal in samples positive for SARS-CoV-2 RNA (Fig. 3A). The microbial component of the metaproteomes expressed as a proportion of the average protein biomass at the phylum level was dominated by bacteria (61 ± 19%; mean ± SD), followed by fungi and archaea, which represented 5 ± 2% and 3 ± 1%, respectively. Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes were the predominant bacterial phyla, which combined accounted for an average of 48% of the biomass. Among the archaeal signatures, the Euryarchaeota phylum was the most highly represented (3 ± 1%). A total of 201 genera were identi ed in microbiota pro les, with Clostridium, unclassi ed Lachnospiraceae, Bacteroides, and Lachnoclostridium as dominant taxa in addition to the host. The relative biomass contributions for these groups ranged from less than 1 to 6% for bacterial genera, and up to 58% for Homo sapiens. Among the 13 archaeal genera identi ed, Methanobrevibacter (28,792 TSMs) was the most abundant. Even though it remained a minor component of the microbiota, sequence coverage for this genus was high (6392 taxon-speci c peptides). Ascomycota and Streptophyta tended to be the most abundant Eukaryota phyla, with the notable exception of the Chordata host (Fig. 3B). After ltering out food-related and host signals, dimension reduction by principal component analysis (PCA) revealed distinct microbiota pro les for patients with high intestinal levels of SARS-CoV-2 RNA (viral load > 50 copies per mg of faecal material) compared to pro les for patients with low or no viral RNA (Fig. 3C). Two outliers were observed: sample C01P003 (faecal viral load 1.4 x 10 3 copies/mg), which clustered among negative samples; and sample C01P033 (1 copy/ mg) which had a microbiota composition resembling high SARS-CoV-2 samples.
Because age-related changes to the microbiota have recently been reported 23 , we further examined the alpha diversity of our samples after matching positive (viral load > 50 copies per mg of faecal material) and negative patients by age (88.2 (± 8.7) and 80.5 (± 7.8) years old, respectively). Both Inverse Simpson and Shannon indices indicated that microbial diversity was signi cantly decreased (Tukey's Honest Signi cant Difference) in samples containing SARS-CoV-2 RNA compared to negative samples (Fig. 3D).

Functional composition of the metaproteome reveals potential biomarkers of SARS-CoV-2 infection
To retrieve functional information from the metaproteome for different sample groups, microbiota and host proteins identi ed with an FDR of 1% were annotated. A total of 88,135 proteins and 60,179 protein groups were listed in the dataset, functions of which were assigned to 887 KO (KEGG Orthology) entries (Supplementary Table S2). The relative abundance of each functional term was calculated at phylum level to reduce loss of information when peptides were unambiguously assigned to the different taxa at a ner resolution. To investigate dissimilarities in microbiota-derived KO functions (n = 664), unsupervised PCA was performed on age-matched SARS-CoV-2-positive (faecal viral load > 50 copies/mg) and negative samples (Fig. 4A). This PCA analysis revealed distinct clustering based on the presence of SARS-CoV-2, with only SARS-CoV-2-positive C01P003 sitting closer to the negative samples. Comparative analysis of the functional pro les (Wilcoxon test, FDR-adjusted p < 0.05) revealed 341 KOs to be signi cantly differentially abundant between the two groups of samples. Of these, 21 were increased in SARS-CoV-2positive samples ( Fig. 4B/C). These KOs are included in 67 KEGG pathways, the most populated of which were metabolic pathways (10), biosynthesis of secondary metabolites (5), glycolysis / gluconeogenesis (3), and microbial metabolism in diverse environments (3). Interestingly, the identi cation of molecular functions related to citrulline ux (KO names OTC and arcA) suggested that Firmicutes species were still adapting to cope with stress and gain an energetic advantage. Evidence for this ongoing adaptation was also provided by the increase in abundance of polypeptides belonging to two-component systems (mcp and yesN) and cobalamin production (cobS -cobV, Firmicutes). Similarly, KOs for drug exporter pump (K06994), NADPH:quinone reductase (qor, CRYZ), and NTE family proteins associated with Actinobacteria suggesting the implementation of mechanisms conferring a competitive advantage in stressful environments, such as the SARS-CoV-2-infected gut. Modules involved in sulfur (soxD, Bacteroidetes) and glutathione metabolism (pepN, Actinobacteria) were also signi cantly increased.
Another marker of the host response, K06856 (IGH, immunoglobulin heavy chain), was increased in association with both Actinobacteria and Firmicutes. Some fungi-associated molecular functions were also altered (PDC, AdhP, SET2). Among these markers, AdhP is known to be involved in retinol metabolism, and the histone modi cation protein SET2 plays a key role in mucosal immune responses, and could be critically involved in integrating a variety of external signals driving fungal expansion. This fungal expansion would in turn in uence the host immune response. Among the 21 altered KOs identi ed, 9 (K00134, K00344, K01256, K07001, K01568, K22622, K07720, K11686, K16703) were linked to 14 pathways. Levels of these pathways were also signi cantly increased relative to age-matched samples in which lower levels of SARS-CoV-2 RNA were detected (viral load < 50 copies/mg) ( Fig. 4B/C).
To investigate alterations in host molecular functions, the relative abundance of host-associated KOs was also analysed. The abundance of a total of 72 out of 187 KOs was altered in samples containing high levels of SARS-CoV-2-RNA (viral load > 50 copies/mg) compared to negative samples (Supplementary Table S2). Of these KOs, 42, represented in 117 pathways, were increased in SARS-CoV-2positive samples with a high viral load (Fig. 4D). These KOs included host molecular functions involved in the ACE2 signalling network (Renin-angiotensin system) such as peptidyl-dipeptidase A (ACE), aminopeptidases (ANPEP/CD13), glutamyl aminopeptidase (ENPEP), and neprilysin (MME). In

Discussion
The gut microbiota plays multiple critical roles not only in nutrition through food processing, but also by maintaining human health as a result of both local and systemic effects, such as limiting pathogen colonization, helping maintain the intestinal barrier function, and training the immune system. Importantly, in the context of SARS-CoV-2 infection, the GI tract has emerged as an organ signi cantly in uencing the propensity to develop, the ensuing disease, COVID-19, and potentially predict its severity. Studies based on taxonomical molecular biology approaches have demonstrated that respiratory infections are associated with changes in the composition of the gut microbiota 4 , but the correlation between respiratory disease and the amount of virions present in the gut has received less attention 19 .
Here, we provide a broad pro le of gut microbiota obtained by a mass spectrometry-based detecting bacteria, archaea, yeasts, and fungi. Organisms were proteotyped and their respective biomass contributions directly compared. We also investigated whether the presence of SARS-CoV-2 in the gastrointestinal tract was associated with changes to the composition of the microbiota.
In this article, we report that the presence of SARS-CoV-2 RNA in the GI tract is not directly related to the detection of the virus in the respiratory tract or to COVID-19 severity at the time of detection. This result is in agreement with the detection of SARS-CoV-2 in tissues throughout the GI tract, and virus shedding in stools in a signi cant proportion of patients. GI shedding often continues for prolonged periods following virus clearance from the respiratory tract 17 . In addition, we provide evidence of changes in the composition of the gut microbiota as a function of the abundance of SARS-CoV-2 RNA in the intestine. In particular, our results show that the microbiota was signi cantly different in patients with a faecal viral load greater than 50 copies per mg, whereas microbiota from patients with a lower or negative viral load tended to be more similar. Closer examination revealed a relatively high level of SARS-CoV-2 RNA to be associated with a signi cantly higher abundance of several genera belonging to the Euryarchaeota and Ascomycota phyla, as well as bacterial genera from the Actinobacteria order, Chitinophaga, Paenibacillus, Sphingomonas, and Bacillus. The overgrowth of fungi is in itself indicative of a disruption of commensal communities. In addition, fungal overgrowth in the gut can cause macrophage polarization which has been linked to increased in ltration of in ammatory cells in allergic airways 24 . Interestingly, bacterial genera from the Bacteroidetes phylum -known to be associated with suppression of colonic expression of ACE2 in the murine gut 25  Our peptide-based functional metaproteome analysis con rmed the differences observed between samples containing high levels of SARS-CoV-2 RNA and negative samples, and the existence of a complex interplay between the gut and SARS-CoV2 infection. In particular, the alterations to host functions observed in faeces containing high levels of SARS-CoV-2 revealed an in amed GI tract characterized by activation of the immune response, as re ected by the molecular alterations governing the host's antiviral defence system. In addition to serving as a receptor for SARS-CoV-2, ANPEP/CD13 expression is known to be dysregulated in in ammatory diseases. Its detection in wide numbers of gut samples led to the persistent intestinal in ammation hypothesis 30 . In our samples, expression levels for this marker were consistent with increased latexin and SMPDL3, a sphingomyelinase-related protein abundantly expressed on macrophages and dendritic cells 31 . Both of these proteins are upregulated by in ammatory stimuli. In parallel, functions described by KOs such as dihydrolipoamide dehydrogenase (DLD), SOD2, Cu/Zn superoxide dismutase (SOD1), and ferritin heavy chain (FTH1) could re ect mitochondrial dysfunction and interplay between in ammation and oxidative stress. Interestingly, expression of tissue-nonspeci c alkaline phosphatase in the colon was also previously reported to be upregulated during in ammatory episodes as a consequence of in ammation-driven tissue-in ltration by neutrophils 32 . The identi cation of elevated levels of KOs linked to functions located in the enterocyte brush border supports in ammation-induced enterocyte damage and increased intestinal permeability.
This possibility is supported by the over-detection in faecal samples containing SARS-CoV-2 RNA of trefoil factor family peptides. These essential proteins are involved in protection and repair of the gastrointestinal tract 33 . This response suggests that expression of TFF variants could be used to predict prognosis, or to monitor therapeutic e cacy. Similarly, the increased abundance of several leaky-gutrelated functions like those associated to antileukoproteinase (SLPI) 34 ; bleomycin hydrolase (BLMH), a cytosolic aminopeptidase thought to contribute to MHC class I peptide presentation 35  Although the COVID-19 pandemic has led to the launch of a large number of clinical studies 42 , to our knowledge, this is the rst time the composition and functionality of gut microbiota have been analysed by differential metaproteomics, with samples distinguished based on the presence of SARS-CoV-2 RNA in the intestinal tract. Interestingly, even in stool specimens containing high levels of viral genetic material, no SARS-CoV-2-derived peptides were detected in our discovery metaproteomics approach. This negative result could be ascribed to the complexity of the samples and a lack of sensitivity of our approach. Indeed, although targeted proteomics can successfully detect SARS-CoV-2 proteins in nasopharyngeal swabs and gargle samples [43][44][45] , viral proteins have never been detected by discovery proteomics approaches applied to complex matrices 42 . Alternatively, these samples may only contain RNA fragments as previously reported for faeces collected at later time points following disease onset 10 .
Whatever the case, and despite the relatively modest number of samples analysed and the fact that extrapulmonary detection of viral RNA does not constitute proof that infectious virus particles are or were present, our results con rm the clinical relevance of testing for viral RNA in faeces because of its direct correlation with altered gut microbiota.
The association between COVID-19 and the presence of SARS-CoV-2 in the intestine remains to be further Some of these cases could be linked to continuous infection of the gut epithelium or severe alterations of the gut microbiota. Ultimately, mechanistic studies will be required to examine how composition of the microbiota affects how SARS-CoV-2 infects the GI tract, and to advance in the search for novel therapies to reduce the severity of COVID-19.
Until now, the impact of SARS-CoV-2 infection on the gut microbiota has been scarcely studied. Here, we assessed on a cohort of 39 patients the gastrointestinal SARS-CoV-2 viral load and correlated it with their full-range microbiota, including Bacteria, Archaea, Fungi and Moulds, accessible through metaproteomics. Our data shows three key results that could be of major importance in the battle against COVID-19: i) faecal SARS-CoV-2 viral load is not correlated to symptoms, ii) an important change in the microbial structure is observed for patients with high faecal SARS-CoV-2 viral load, and iii) a list of microbiome and human markers can be drawn from this study as possible candidates for diagnostic.
These results should be incentive for an extensive multi-centric metaproteomics analysis of the gut microbiota of COVID-19 patients.

Methods
Ethics approval, consent to participate, and study population did not have to provide written informed consent to take part in the study. Stool samples received in the Department of Microbiology from the Department of Infectious Diseases (University Hospital Nîmes, France) from March 17, 2020 to May 11, 2020 were included. During this period, a nationwide lockdown was applied with an emergency state due to the context of the COVID-19 health crisis. The hospital admitted exclusively patients with acute health problems.
The owchart of the study is presented in Fig. 1A. Among the 41 stools received during the indicated period, two were excluded from the analysis due to the low quantity. Data including demographics, laboratory results and imaging results were extracted from the electronic medical records of the University Hospital management system (Clinicom®, Intersystems SAS, France).
Collecting and processing faecal specimens

Proteome extraction and digestion
Following sample homogenization, three 50 mg aliquots of each faecal sample were transferred to fresh tubes to create three technical replicates for each sample. Aliquots were lysed in 100 µL LDS 1X (Lithium dodecyl sulfate) sample buffer (Invitrogen™, Thermo Fisher) supplemented with 5% betamercaptoethanol (vol/vol). Samples were sonicated for 5 min in an ultrasonic water bath (VWR ultrasonic cleaner) and incubated at 99°C for 5 min before transfer to 2-mL Screw Cap microtubes (Sarstedt, Marnay, France) containing 200 mg ceramic beads, as previously described 49 . Cell disruption was performed on a Precellys Evolution instrument (Bertin Technologies, Aix en Provence, France) operated at 10,000 rpm for ten 30-s cycles, with 30 s rest between cycles. After lysis, samples were centrifuged at 16,000 × g for 3 min. For each sample, the resulting supernatant (25 µL) was loaded onto a NuPAGE 4-12% Bis-Tris gel, and proteins were subjected to short ( Preliminary quantitation of peptides extracted from faecal samples by high-resolution mass spectrometry survey For each faecal sample, 1 µL of the extracted peptide mixture was injected for analysis on an LTQ-Orbitrap XL (Thermo Fisher Scienti c, Waltham, USA) tandem mass spectrometer coupled to an Ultimate 3000 nano LC system (Thermo Fisher Scienti c). The proteolyzed products were desalted online on a reverse-phase PepMap 100 C18 µ-precolumn (5 µm, 100 Å, 300 µm id × 5 mm, ThermoFisher) and resolved on a nanoscale PepMap 100 C18 nanoLC column (3 µm, 100 Å, 75 µm id × 50 cm, Thermo Fisher) at a ow rate of 0.3 µL/min prior to injection into the mass spectrometer. A linear chromatographic gradient of mobile phase A (0.1% HCOOH/100% H2O) and phase B (0.1% HCOOH/80% CH3CN) was applied from 5 to 40% B in 30 min. Full-scan mass spectra were measured from m/z 350 to 1500 in data-dependent mode using a Top5 strategy. Brie y, a scan cycle was initiated by a full high mass-accuracy scan in the Orbitrap analyser, operated at 30,000 resolution, followed by MS/MS scans in the linear ion trap on the ve most abundant precursor ions. A 10-s dynamic-exclusion window was applied to previously selected ions. Precursor ions were isolated using a 3-m/z isolation window and activated with 35% normalized collision energy.

NanoLC-MS/MS characterization of peptides extracted from faecal samples
To normalize the peptide amount injected for each faecal sample, the total ion current (TIC) chromatogram obtained with the LTQ-Orbitrap XL instrument was used to calculate the exact volume (µL) to be injected on a Q-Exactive HF mass spectrometer (Thermo) by dividing 16.1E7 (the target optimal TIC previously established) by the TIC value obtained with the LTQ-Orbitrap XL. The high-eld Orbitrap instrument was used in combination with an UltiMate 3000 LC system (Dionex-LC) and operated in datadependent mode, as previously described 51 . The appropriate volumes of peptides were injected for each aliquot (technical triplicate for each faecal sample), desalted on an Acclaim PepMap100 C18 precolumn (5 µm, 100 Å, 300 µm id × 5 mm), and then resolved on a nanoscale Acclaim PepMap 100 C18 column (3 µm, 100 Å, 75 µm id × 50 cm) with a 120-min gradient at a ow rate of 0.2 µL/min. The gradient was developed from 4 to 25% of [CH3CN, 0.1% formic acid] over 100 min, and then from 25 to 40% over 20 min. Peptides were analysed during scan cycles initiated by a full scan of peptide ions in the ultra-higheld Orbitrap analyser, followed by high-energy collisional dissociation and MS/MS scans on the 20 most abundant precursor ions (Top20 method). Full-scan mass spectra were acquired from m/z 350 to 1500 at a resolution of 60,000 with internal calibration activated on the m/z 445.12002 signal. During ion selection for MS/MS fragmentation and measurement, a 10-s dynamic-exclusion window was applied with an intensity threshold of 1.7 × 10 4 . Only ions with positive charges 2 + and 3 + were considered.
Assigning peptides and analysing metaproteomics data

Statistical analyses
Count values from both taxonomic (number of TSMs) and functional data (spectrum counts) were scaled relative to their sum total in the sample. The categorical and continuous variables were compared among the patients using the Fisher exact test and the Kruskal-Wallis test, respectively. Principal component analysis was performed as previously described 55 . The R package metacoder 56 was used to represent taxonomic abundance as a differential heat tree. Univariate differential analysis of taxon and KO abundances between conditions was performed by applying non-parametric Wilcoxon tests corrected for multiple comparisons (Benjamini-Hochberg adjustment). Pairwise alpha-diversity indices were calculated using the vegdist function from the vegan package in R.

Declarations
Data availability The mass spectrometry and proteomics datasets acquired on faecal samples are available through the

Competing interests
The authors declare that they have no competing interests.  Figure 1 representation of the experimental analytical work ow. Stool samples were analysed in parallel by RT-qPCR to detect the presence of SARS-CoV-2 RNA in the gut, and by shotgun tandem mass spectrometry to investigate the taxonomical and functional composition of the microbiota.   were also signi cantly increased when compared to levels detected in age-matched samples containing