Commensal microorganisms of the respiratory tract affect the body’s ability to withstand a threat of viral infections (Abt et al., 2012; Lee et al., 2019). This might be due to the modulatory effect of microbes on the immune system (Abt et al., 2012) or direct impact of the microbial metabolites on the viral life cycle (Kamio et al., 2015).

At the same time, certain bacteria can promote viral infections (Bellinghausen et al., 2016), and secondary bacterial infections can lead to serious complications and mortalities (Morens et al., 2008; Lansbury et al., 2020). Viral infections in turn can influence the bacterial community, sometimes promoting bacterial biofilm formation and secondary infections (Hendricks et al., 2016; Hendricks et al., 2021). The relationship between respiratory microbiota and viral infections remains an important research subject (Man et al., 2017; Hendricks et al., 2021).

Until recently, the microbiota of the upper respiratory tract was poorly studied, especially in adults (Dubourg et al., 2019). This situation is now changing due to the current COVID-19 pandemic. Previous study has shown the dynamical nature of microbial community types of the upper respiratory tract during COVID-19 infection (Xu et al., 2021). A number of studies described differences between patients with COVID-19 and healthy people or differences between patients with different levels of disease severity (Rosas-Salazar et al., 2021; Soltani et al., 2021) among others, (see (Yamamoto et al., 2021) for a short review). The results of various studies are not always consistent. It might be due to the phenotypic differences between subjects (e.g. comorbidities), as well as seasonal and geographical factors and different techniques used in data sampling and analysis.

When studying the current COVID-19 samples, researchers use different approaches. In a study performed in Nashville, Tennesy (Rosas-Salazar et al., 2021), the researchers examined 38 COVID-19 patients and 21 uninfected controls by 16S rRNA metagenomics. They obtained amplicon sequence variants (ASVs) using DADA2 method and applied DESeq2 to infer ASVs associated with the infection and also with high viral load. They have noted that ASVs associated with COVID-19 disease and high viral load belonged to Peptoniphilus lacrimalis, Campylobacter hominis, Prevotella copri, and an Anaerococcus unclassified amplicon sequence variant. At the same time, Corynebacterium unclassified, Staphylococcus haemolyticus, Prevotella disiens, and two Corynebacterium_1 unclassified amplicon sequence variants were associated with healthy status and low viral load during COVID-19.

In another study (Ren et al., 2021), the researchers performed metatranscriptome sequencing of 588 oropharyngeal swab samples collected from 192 COVID-19 patients. The authors found that Streptococcus genus was enriched in recovered patients, whereas potential pathogens, including Candida and Enterococcus, were more abundant in patients that have died of infection. They also noted that high abundance of Streptococcus and particularly of S. parasanguinis at the moment of patient’s admission was a strong predictor of fatality. Researchers in India (Devi et al., 2022) analyzed 198 nasopharyngeal and/or throat swabs from COVID-19 patients using WGS sequencing. The patients were categorized into four groups based on disease severity and outcome: mild, moderate, severe and lethal. The researchers discovered significant transcript abundance of Achromobacter xylosoxidans and Bacillus cereus in the lethal group, Leptotrichia buccalis in the severe, Veillonella parvula in the moderate, and Actinomyces meyeri and Halomonas sp. in the mild COVID-19 patients.

Respiratory microbiome has geographic and climatic characteristics, so it is important to study its relationship with COVID-19 disease in different regions. In order to avoid spurious correlations, other diseases and patients’ habits should be also taken into account when analyzing associations between microbiome composition and the course of COVID-19 infection.

In our work, we describe the composition of the respiratory microbiome in 120 inpatients with COVID-19. We performed the 16S rRNA metagenomic analysis of oropharyngeal swabs and reconstructed metabolic pathways using the PICRUST2 tool to find associations with the disease severity.

We aimed to identify microorganisms and KEGG pathways associated with the course of the disease using differential abundance testing with patients’ metadata as covariates. We hypothesize that some of the identified microorganisms might have a protective role in the respiratory tract given the wide geographical and age ranges of COVID-19 patients involved in this study along with the comorbidities considered, we believe that the obtained results can be reproduced in further research.

In our study, we applied the 16S rRNA metagenomic analysis to compare the taxonomic composition of the upper respiratory tract microbiota of COVID-19 patients at the time of admission to the hospital. Clinical samples from the current investigation were handled and sequenced with the similar ones from the other projects ( Supplementary Figure 1 ).

Characterizing microbial communities from low-biomass samples like oropharyngeal swabs can be challenging due to the risk of contamination from exogenous DNA during sample collection and processing, which affects assay results (Salter et al., 2014). In our study, we used the decontam tool, but there are also a number of other recommendations and approaches for controlling contaminant sequences in low-biomass communities (Eisenhofer et al., 2019; Claassen-Weitz et al., 2020). We decontaminated the array of sequences from multiple projects simultaneously, enabling a more comprehensive cleaning of the array from contaminated sequences. We also excluded the identified contaminant sequences and introduced the variable “contamination” in the sample to account for this effect in the models. Our analysis focused solely on samples with complete metadata, including age, city, severity of lung damage based on CT scan, the season of sample collection, additional O2 requirements, hypertension, CAD, diabetes, and obesity. We employed these variables as covariates in our statistical analysis, although this approach did reduce our sample size, it ultimately allowed for a more robust statistical analysis. Finally, the 174 samples from 120 COVID-19 inpatients were left for analysis.

All the patients were divided into mild and severe groups according to their CT severity score (CT 1, 2, 3, or 4), which reflected the lung damage level at the time of hospitalization. This approach allowed us to consider the observed pneumonia as a community-acquired one and to avoid the finding of hospital-associated pathogens in the metagenomic samples.

We deliberately did not consider the option of comparing the oropharyngeal microbiome of healthy and diseased groups, as it was already performed in multiple studies (Xiong et al., 2021), (Iebba et al., 2021; Ren et al., 2021). It occurs that SARS-CoV-2 can also be found among healthy people that can be asymptomatic carriers of the virus. So, using asymptomatic (“healthy”) people as a control group can introduce biases into the analysis. Moreover, according to the principle of Anna Karenina, which has recently been frequently discussed in the context of metagenomic research, the microbiome of healthy people is less variable and more homogeneous than the microbiota of people with different diseases (Ma, 2020). Therefore, we decided to compare the upper respiratory tract microbiota between patients with different disease severity of SARS-CoV-2.

The oropharyngeal swabs were collected from COVID-19 patients during the first and the second waves of the epidemic, which were dominated by the “Wuhan’’ variant of SARS-CoV-2 (Awadasseid et al., 2021). Some papers discuss the effect of seasonality on the upper respiratory tract microbiota (Camarinha-Silva et al., 2012; Schoos et al., 2020; Zhao et al., 2022). In our study, the period of sample collection affected all seasons, which allows us to use seasonality as a covariate ( Figure 1 ). The taxonomic composition of oropharyngeal swabs of COVID-19 patients is consistent with the results of studies about human oropharyngeal microbiota (Zaura et al., 2009; Bach et al., 2020). The taxonomic profile of our samples were found to be homogeneous, as confirmed by the heatmaps of Supplementary Figure 3 and Supplementary Figure 4 , where we observed the absence of clear clustering by such parameters as sex, age, city, seasonality, and batch effect.

Strikingly, the compared patient groups did not differ in the need for supplemental oxygen despite differences in the severity of lung damage. They also did not differ in the suffer of obesity, known to be a potential predictor of severe COVID-19 (Simonnet et al., 2020; Wang et al., 2022). Meanwhile, we observed the higher prevalence of hypertension and diabetes in the severe patient group. However, there is no conclusive evidence yet that hypertension is more common in patients with COVID-19 or that it might increase the risk of SARS-CoV-2 infection. Although hypertension is common among severe COVID-19 patients, it is more likely to be attributed to the vulnerability of middle-aged and older individuals to SARS-CoV-2 infection, according to this literature review (Shibata et al., 2020).

When comparing the alpha diversity of the upper respiratory tract microbiota of mild and severe patients, we observed no significant differences between these two groups. This is consistent with the results of (Hernández-Terán et al., 2021), but at the same time contradicts the results of the following study (Shilts et al., 2021), where the difference in the alpha diversity between patients with mild and severe disease was significant. The greatest difference was observed between CT4 and CT1-3 groups, where patients with СT4 had the lowest alpha diversity indices. Theoretically, this strong decrease in microbiota diversity could be related to inflammation in the airways and more frequent medication treatment of patients with severe pneumonia (Hong et al., 2021). Nevertheless, due to the restricted number of samples (n = 3) and the lack of information on drug treatment, we can neither confirm nor deny this assumption.

SARS-CoV-2 infection strongly affects the human immune system, causing a cytokine storm (Jose et al., 2020), activation of monocytes and macrophages, which in turn affects the human upper respiratory tract microbiome (Merad et al., 2020). COVID-19 can cause changes in the airway epithelium, increase local inflammation, and promote adhesion of respiratory pathogens. In our study, we identified bacterial drivers of patients with mild (CT1) and more severe (CT >= 2) lung damages, taking into account the metadata described above as covariates.

We found that ASVs identified as Streptococcus salivarius, Capnocytophaga sputigena, and Haemophilus parahaemolyticus were significantly more prevalent in patients with mild lung damage. These three bacteria belong to genera that are widely distributed in the human oropharyngeal microbiome.

Haemophilus parahaemolyticus is considered as a natural oropharyngeal commensal in humans, which can be pathogenic in certain cases. In a number of studies (Ren et al., 2021; Soffritti et al., 2021) that compared the microbiomes of healthy people and patients infected with SARS-CoV-2 virus, the genus Haemophilus and in particular Haemophilus parahaemolyticus species was associated with the healthy subjects.

Capnocytophaga is a commensal genus that is a part of the normal bacterial flora of the human oral microbiome. Capnocytophaga spp. and Capnocytophaga sputigena in particular often act as opportunistic pathogens associated with poor oral hygiene and periodontitis. At the same time, they occur in groups of healthy people and are closely associated with other bacteria (Idate et al., 2020; Jolivet-Gougeon et al., 2021; Ma et al., 2022). In the article (Soffritti et al., 2021) have reported an increased abundance of Capnocytophaga sputigena in SARS-CoV-2 patients as compared to healthy controls. In addition, Capnocytophaga spp. and other oral opportunists have been found in the BALF of the COVID-19 patients (Bassis et al., 2015; Bao et al., 2020; Wu et al., 2020).

Streptococcus salivarius is also a part of the commensal microflora of the upper respiratory tract. At the same time, some strains of this species are able to regulate the acid-alkaline balance of the oral cavity through the production of alkali. This property can have an antagonistic effect on a number of opportunistic species like Streptococcus pyogenes, Streptococcus pneumoniae, Moraxella catarrhalis and Haemophilus influenzae that prefer a more acidic environment and can cause a number of respiratory infections(Aebi, 2011; Zupancic et al., 2017; Abranches et al., 2018; Di Pierro, 2020; Wen et al., 2020).

Speaking about the patients with severe lung damage (CT2-4), we found Neisseria, Veillonella and Lautropia genera to be overrepresented in their oropharyngeal samples. All the three genera are widespread in the human oropharynx and belong to the commensal microflora (Zaura et al., 2009; Bach et al., 2020).

The Neisseria genus is one of the most common genera in the human oropharynx. Non-pathogenic Neisseria may have protective properties against some pathogens by producing secondary antimicrobial metabolites (Aho et al., 2020; Baerentsen et al., 2022). Some Neisseria species are also important for the development of a T-cell-independent polyclonal IgM response (Dorey et al., 2019). Many studies that compare the oropharyngeal microbiome of SARS-CoV-2-infected patients and healthy controls show a significant decrease in the representation of the genus Neisseria in the group of infected individuals (Li et al., 2022). The high abundance of nonpathogenic Neisseria in the oropharyngeal microbiome of SARS-CoV-2-infected patients is considered as a predictor of successful recovery (de Castilhos et al., 2022).

In our study, we observe overrepresentation of Neisseria mucosa and Neisseria oralis species in the group of severe patients. Our results are similar to the work where Neisseria mucosa was found to be associated with COVID-19 (Soffritti et al., 2021). Interestingly, we observed an opposite dynamic in the relative abundance of Neisseria oralis and Neisseria mucosa between the first and second time points in the mild group of patients, which may reflect a change of species at the same ecological niche. Moreover, Picrust2 analysis showed these species contribute significantly to the differentially represented metabolic pathways of the Type IV secretion system associated with the severe course of the disease. The type IV secretion system VirB/VirD4 is a major virulence determinant for subversion of human endothelial cell (HEC) function. (Schmid et al., 2004; Costa et al., 2021). According to the literature, the contribution of non-pathogenic Neisseria spp. to the prevalence of metabolic pathways associated with the type IV secretion system indicates that these genes, including virulence genes, are necessary for them to survive in the niche, but not for pathogenicity (Calder et al., 2020). Such genes are common to commensal and pathogenic Neisseria species and are necessary for the adhesion and invasion of bacterial cells to host cells and play an important role in the struggle for colonization within a given genus (Wörmann et al., 2016; Calder et al., 2020). In addition, members of the Neisseria genus are naturally competent, which increases the adaptability of the genus to changing environmental conditions such as inflammatory processes in the host organism and the use of antibiotics(Snyder et al., 2007; Goytia et al., 2021).

According to several studies, the members of the genus Veillonella strongly predominate in the oral microbiota of patients with Covid-19 compared to healthy controls (Iebba et al., 2021; Ma et al., 2021; Soffritti et al., 2021). Similar to Capnocytophaga spp., some Veillonella species associated with periodontitis have been overrepresented not only in the oral microbiome, but also in the BALF of the COVID-19 patients with pneumonia (Wu et al., 2020).

Lautropia mirabilis is normally isolated from human dental plaque and also from a supragingival or subgingival biofilm (Ikeda et al., 2020). To date, we have not found any information on the relationship of Lautropia mirabilis with COVID-19.

There are many studies worldwide describing the features of respiratory biotope in COVID-19 patients. Nevertheless, our study has several important strengths, such as the inclusion of patients from more than one geographic region, whose oropharyngeal swabs were collected during all seasons. In further comparison analysis we included only hospitalized patients whose swabs were taken at admission; therefore, we were more likely to capture the early upper respiratory tract microbiome associated with the development of severe COVID-19 rather than the one acquired after a hospital stay. We also had an opportunity to analyze samples collected from the same patients on the day of release. To our knowledge, this is the first study of the upper respiratory tract microbiome of COVID-19 patients from Russia.

However, we are aware of the objective limitations of our study leading to the sequencing batches (Wang et al., 2020; de Goffau et al., 2021). During this study we did everything to avoid it at sample processing level - all samples were processed the same way and sequenced simultaneously. To account for batch effects, we incorporated a categorical variable (batch) when calculating differential abundance of taxa and functions. Instead of using batch correction or adjustment methods, we chose this approach because our primary variable of interest (severity) was evenly distributed across batches, and we wanted to explicitly include this information in the model as recommended in the literature (Nygaard et al., 2016). Anyway, the samples were collected in different medical institutions, by different staff, using different consumables. Unfortunately, we were unable to control the sample acquisition procedure at this stage due to the height of the pandemic and the limitations involved. We suspect that this factor may have had the most significant impact, but it is not feasible to trace its effect.

Although experimental validation is needed, in our study we hypothesize that environmental and host-related factors could be affecting the respiratory microbiota prior to viral infection, potentially compromising the immune response of the host against disease. Also, the upper respiratory tract microbiome acts as a reservoir of opportunistic pathogens, which descend to the lower parts of the respiratory tract causing the elevated inflammation and lung damage. According to our data, we could suspect Veillonella acting this way. As far as nonpathogenic Neisseria are concerned, their over-representation in a group of severe COVID-19 patients can be considered as a predictor for further successful recovery in accordance with (de Castilhos et al., 2022). We can also assume that the microbiome of patients from the severe group recovers more slowly than from the mild one. It is evidenced by an increased relative abundance of opportunistic microorganisms, such as Veilonella sp., against an increased relative abundance of commensal flora, such as Capnocytophaga sputigena and Haemophilus parahaemolyticus, at the time of discharge from hospital.

It is worth mentioning that the comparative analysis of the relative abundances of certain bacteria in the oropharyngeal microbiome of mild and severe COVID-19 patients is rather ambiguous due to the complexity of the mechanisms leading to the observed trends. We believe that the findings of this work contribute to our understanding of the role of the upper respiratory tract microbiome in the development of COVID-19. Anyway, further studies that would shed light on the specific mechanisms through which the bacteria from respiratory microbiota are involved in host response against viral infections are required.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA751478.

The study was approved by the ethical committee of RCPCM. All patients gave written informed consent for sample collection and personal data processing. The patients/participants provided their written informed consent to participate in this study.

JG, DF, AM, DKo, DKr, performed bioinformatics analyses. EI, AP, IG and VG conceived and supervised the study. ES, JG, EI wrote the manuscript. VB, KK, VV, MM and RG obtained samples and epidemiological data from patients and performed sample sequencing. All authors contributed to the article and approved the submitted version.