Microbiome analysis revealing microbial interactions and secondary bacterial infections in COVID‐19 patients comorbidly affected by Type 2 diabetes

Abstract

The mortality of coronavirus disease 2019 (COVID‐19) disease is very high among the elderly or individuals having comorbidities such as obesity, cardiovascular diseases, lung infections, hypertension, and/or diabetes. Our study characterizes the metagenomic features in severe acute respiratory syndrome coronavirus‐2 (SARS‐CoV‐2)‐infected patients with or without type 2 diabetes, to identify the microbial interactions associated with its fatal consequences.This study compared the baseline nasopharyngeal microbiome of SARS‐CoV‐2‐infected diabetic and nondiabetic patients with controls adjusted for age and gender. The metagenomics based on next‐generation sequencing was performed using Ion GeneStudio S5 Series and the data were analyzed by the Vegan‐package in R. All three groups possessed significant bacterial diversity and dissimilarity indexes (p < 0.05). Spearman's correlation coefficient network analysis illustrated 183 significant positive correlations and 13 negative correlations of pathogenic bacteria (r = 0.6–1.0, p < 0.05), and 109 positive correlations between normal flora and probiotic bacteria (r > 0.6, p < 0.05). The SARS‐CoV‐2 diabetic group exhibited a significant increase in pathogens and secondary infection‐causing bacteria (p < 0.05) with a simultaneous decrease of normal flora (p < 0.05). The dysbiosis of the bacterial community might be linked with severe consequences of COVID‐19‐infected diabetic patients, although a few probiotic strains inhibited numerous pathogens in the same pathological niches. This study suggested that the promotion of normal flora and probiotics through dietary supplementation and excessive inflammation reduction by preventing secondary infections might lead to a better outcome for those comorbid patients.

1. INTRODUCTION

Severe acute respiratory syndrome coronavirus‐2 (SARS‐CoV‐2) is the etiological root of the coronavirus disease 2019 (COVID‐19) pandemic, which has affected over 605 million people worldwide until August 2022. ¹ The virus expressed itself with highly variable severity, ranging from no outward symptoms to severe respiratory distress. ² , ³ The mortality rate of COVID‐19 was very high among males, the elderly, and individuals having comorbidity such as obesity, cardiovascular diseases, lung infections, hypertension, and/or diabetes. The SARS‐CoV‐2 mortality was 50% in males and 43% in females among intensive care unit (ICU) patients in Lombardy, Italy. ⁴  The mortality rate among SARS‐CoV‐2‐infected diabetic patients was 7.8% compared to 2.7% in noncomorbid patients in China. ⁵ One study found that the risk factors of death increased by 3.2 times from SARS‐CoV‐2 infection when associated with diabetes. ⁶ The mortality is often related to immune homeostasis ⁷ or interaction between intracellular toll‐like receptors (TLRs) and viral RNA which may result in cytokine storms followed by multiorgan failure. However, the pathogenicity, ⁸ carriage information, and the interactions with commensals or opportunistic bacteria of this virus are still unclear. ⁹ , ¹⁰ The number of AEC2 or TLR4 receptors is related to the progression of this disease, ¹¹ although secondary infections may also play a vital role in disease severity like other viral infections. ⁹ Secondary bacterial infections are often followed by respiratory viral infections among comorbidly affected patients. The development of respiratory coinfections is also reported to be associated with COVID‐19 disease severity and fatality in many cases. ¹² , ¹³ , ¹⁴ , ¹⁵

Metagenomic analysis was popularly used to understand the diversity and pathogenesis of microbial populations in a group of subjects. However, metagenomics based on next‐generation sequencing (mNGS) was rarely applied to the clinical samples of SARS‐CoV‐2. ⁵ , ¹⁶ , ¹⁷ This technology could be used to uncover all microbiomes interacting in the same pathological niche, which might lead to ultimate pathophysiological conditions. The present study aims to perform metagenomic in‐depth bioanalysis to characterize SARS‐CoV‐2 diabetic and nondiabetic patients compared with controls to identify the microbial interaction associated with the fatal consequences.

2. METHODS

2.1. Patients

Seven SARS‐CoV‐2‐positive samples and four SARS‐CoV‐2‐negative controls were included in this study. All positive cases were selected from the continuous surveillance at the Genome Center, Jashore University of Science and Technology covering four districts of Bangladesh, Jashore, Jhenaidah, Magura, and Narail authorized by the Directorate General of Health Services, Bangladesh, for the screening of COVID‐19. Among those seven positive cases, three patients had a history of Type 2 diabetes with two reported deaths (Table 1).

Table 1.

Demographics and clinical characteristics of the study subjects

ID	Cases	Groups	Sign and symptoms appeared	SARS‐CoV‐2 tests results	Chronic health complications
MHC10	Healthy	Control	No	Antibody negative, rt‐PCR negative	None
MHC15	Healthy	Control	No	Antibody negative, rt‐PCR negative	Diabetes
MOE14	Other unknown etiology	Control	Feverish (99 F)	Antibody negative, rt‐PCR negative	None
MDB11	Death	Control	Fever, dry cough, loss of appetite, pain or swelling of legs, hands	rt‐PCR negative	Hypertension, renal impairment
MMS2	Outpatients	SARS‐CoV‐2 nondiabetic	Dry cough	rt‐PCR positive	None
MICU1	ICU	SARS‐CoV‐2 nondiabetic	Fever; headache; shortness of breath; difficulty breathing; loss of taste; loss of smell.	rt‐PCR positive	None
MICU13	ICU, coma	SARS‐CoV‐2 nondiabetic	Lung coinfection, sore throat, difficulty breathing, loss of appetite, loss of taste, loss of smell	rt‐PCR positive	None
MDN6	Death	SARS‐CoV‐2 nondiabetic	Fever, dry cough, shortness of breath, pain or swelling of legs, hands	rt‐PCR positive	None
MHD7	Hospitalized	SARS‐CoV‐2 diabetic	Fever, headache, running nose, dry cough, sore throat, muscle pain, shortness of breath, loss of taste, loss of smell, diarrhea	rt‐PCR positive	Diabetes, hypertension
MDD8	Death	SARS‐CoV‐2 diabetic	Fever, sore throat, muscle pain, shortness of breath, difficulty breathing, loss of taste, loss of smell	rt‐PCR positive	Diabetes
MDD9	Death	SARS‐CoV‐2 diabetic	Fever, bedsore, broken hip	rt‐PCR positive	Diabetes

Abbreviations: rt‐PCR, reverse transcription‐polymerase chain reaction; SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

2.2. Control group

Four age‐ and sex‐matched subjects were included in this study as healthy (N = 2) controls and unknown etiology controls (N = 2), confirmed as SARS‐CoV‐2 negative by reverse transcription‐polymerase chain reaction (rt‐PCR) (Table 1). Both healthy controls have no history of fever or other illness or uptake of antibiotics in the last 6 months. Among them, one has Type 2 diabetes and the other does not have any chronic health complications. The unknown etiology controls were SARS‐CoV‐2 negative; of these, one has hypertension, renal malfunction, and died. Three of those four were SARS‐CoV‐2 antibody negative, tested by All Check COVID‐19 IgG/IgM Antibody Assay Kit (CALTH Inc.), except the deceased patient.

2.3. mNGS sequencing

Total nucleic acids were extracted from 300 µl of nasopharyngeal samples and eluted with 60 µl sterile RNase‐free water using a commercial kit (Zymo Total Nucleic Acid). The total nucleic acid concentration was assayed by Qubit RNA HS Assay Kit (Thermo Fisher Scientific) using a Qubit 4 Fluorometer. The extracts were enriched and processed for library preparation using Ion Total RNA‐Seq Kit v2.0 (Thermo Fisher Scientific) according to the manufacturer's instructions with a minor modification. In brief, RNA fragmentation was performed using RNase III treatment for 3 min at 37°C following the magnetic bead cleanup to optimize the library size at about 200 bp. After adapter ligation of each sample, complementary DNA (cDNA) was prepared and Ion Xpress™ RNA‐Seq Barcode BC primers (Thermo Fisher Scientific) were added. The amplification of the barcoded cDNA was extended to 18 cycles instead of 16 cycles due to a low concentration of nucleic acid in the samples. The final concentrations of the libraries were diluted into 200 pM instead of 100 M, as suggested in the manufacturer's protocol. One extraction control with sterile water and 11 unknown samples were used for the library preparation. In three mNGS runs, four equimolar libraries were pooled for the preparation of template‐positive Ion Sphere™ Particles (ISPs) using the Ion 520™ and Ion 530™ Kit—OT2 (Thermo Fisher Scientific) on the Ion One Touch™ 2 System (Thermo Fisher Scientifi). Template‐positive ISPs were enriched on the Ion One Touch™ ES system (Thermo Fisher Scientific). The enriched template‐positive ISPs and control ISPs were loaded onto Ion 520/530™ chip and sequenced with the next‐generation sequencing in the Ion S5^TM systems (Thermo Fisher Scientific). The data outputs were analyzed using the automated, streamlined Torrent Suite software (v5.10.0). The primary baseline data were obtained after removing duplicated reads, the average quality scores below Q20, low‐quality 3‐end reads, and adapter sequences.

The CGView Server (http://stothard.afns.ualberta.ca/cgview_server/) was used to construct the circular ring of SARS‐CoV‐2 genome comparison using the blastn ¹⁸ and SARS‐CoV‐2 isolate Wuhan‐Hu‐1 (NC_045512.2) was used as a reference. Average nucleotide identity was calculated using jSpecies  ¹⁹ to compare the SARS‐CoV‐2 genome with the reference genome.

2.4. Bioinformatics processing and taxonomic assignment

The Binary Alignment Map files were transferred to FASTQ format through SAMtools, ²⁰ followed by filtering through BBDuk ²¹ , ²² (with options ftm = 5, k = 21, mink = 6, minlen = 30, ktrim = r, qtrim = rl, trimq = 20, overwrite = true) to remove all low‐quality sequences. On average, 1.34 million reads per sample (maximum = 3.18 million, minimum = 0.45 million) passed the quality control step. The host sequences from the trimmed files were removed by aligning to the human genome (hg38) by using Burrow–Wheeler Aligner ²³ and SAMtools. ²⁰ The taxonomic assignment has been done by Kraken2 ²⁴ with NCBI RefSeq Release 201 database (bacterial, viral, archaeal, and fungal). For the analysis, less than 100 hits were not considered for bacteria and fungus, and less than 10 hits were not considered for viruses and archaea. Data normalization was performed by previously described methods by multiplying the mean with the proportion. ²⁵

2.5. Statistical analyses

The α‐diversity of microbial communities among different groups was compared by calculating the Shannon and Simpson 1‐D diversity indexes, observed, and the Chao‐1 richness index using the “Vegan” package in R. The nonparametric Kruskal–Wallis rank‐sum test was used to evaluate α‐ diversity and pairwise Wilcoxon's rank‐sum test was used to assess pairwise comparison in different groups. α‐Diversity (principal coordinate analysis [PcoA]) was determined using the Bray–Curtis dissimilarity index, using permutational multivariate analysis of variance (PERMANOVA), to estimate a p value for differences among the study groups. Phyloseq and vegan packages were employed for those statistical analyses. ²⁶ Spearman's correlation coefficient and significance tests were calculated using the R package Hmisc. A correlation network was constructed and visualized with Gephi (ver. 0.9.2). A quantitative analysis of comparative RNA‐seq data using shrinkage estimators for dispersion and fold change was employed for differential bacterial species with a statistical significance (q value) <0.01 and absolute value of log 2 (fold change) > 3 using DESeq. 2 (v4.0). The Benjamini–Hochberg correction was used to obtain false discovery rate‐adjusted p values (q values) for multiple hypothesis testing. ²⁷

3. RESULTS

All 11 study subjects were divided into three groups, the control group (N = 4), SARS‐CoV‐2‐positive group without any history of comorbidity (N = 4), and the SARS‐CoV‐2‐positive diabetic group. There was no significant difference in age and body mass index in all groups (p  = 0.20 and 0.49, respectively). The detailed patient demography with symptoms, severity, and outcomes was described in detail in Table 1.

3.1. SARS‐CoV‐2 genomic data analysis and RNA quantification

The raw reads of metagenomic data were submitted to the NCBI SRA BioProject accession number PRJNA733662. The SARS‐CoV‐2 genome sequence data analysis retrieved three complete (GISAID accession ID: EPI_ISL_746318, EPI_ISL_746319, and EPI_ISL_746323) and one partially complete genome sequence of SARS‐CoV‐2 out of seven positive samples (Supporting Information: Figure A1). They belonged to B.1.1.25, B.1.533, and B.1.36.16 lineage.

In all seven SARS‐CoV‐2‐positive samples, the average RNA copies were 231 375 among outpatients, 198 among ICU patients, and 2600 among departed patients. No significant relationship was found between the RNA copies and the severity of the disease (data not shown).

3.2. Microbial diversity and dissimilarity index

The comprehensive assessment of microbial population on the host traits using α‐diversity‐based association analysis found diverse microbial populations in all three groups of samples; however, they were statistically nonsignificant. The microbial diversity of the control, SARS‐CoV‐2 nondiabetic, and SARS‐CoV‐2 diabetic groups were nonsignificant in Shannon (p = 0.18), Simpson (p = 0.23), observed (p = 0.15), or Chao (p = 0.19) (Supporting Information: Figure A2). Visualization of community compositions was observed by the PcoA of Bray–Curtis, indicating a significant dissimilarity index in those groups (PERMANOVA: pseudo‐F = 1.77, p = 0.047). The number of taxonomic units (species) in the control, SARS‐CoV‐2 nondiabetic, and SARS‐CoV‐2 diabetic groups were 134, 120, and 162, respectively. More than 14% were shared by all three groups and more than 18% were overlapped between SARS‐CoV‐2‐positive diabetes and SARS‐CoV‐2‐nondiabetes groups (Supporting Information: Figure A3).

3.3. Bacterial diversity and dissimilarity index

The bacterial populations on the host traits were also assessed by α‐diversity‐based association analysis among the three groups and the Shannon diversity index exhibited significant bacterial diversity (p = 0.05). However, Simpson (p = 0.09), observed (p = 0.18), or Chao (p = 0.18) diversity index differ insignificantly (Figure 1). The PcoA of Bray–Curtis index in the PERMANOVA analysis found significant (PERMANOVA: pseudo‐F = 2.012, p = 0.02) dissimilarities in bacterial species among those groups.

Figure 1.

Bacterial α‐diversity‐based association analysis by (A) Shannon diversity index, (B) Simpson diversity index, (C) observed, and (D) Chao, (E) principal coordinate analysis by Bray–Curtis dissimilarity index among healthy, recovered, and deceased patients. SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

At the phyla level, Firmicutes were the most abundant in all three groups following Bacteroidetes and Proteobacteria. Fusobacteria were abundant only in the control group (Supporting Information: Figure A4). This study identified 207 bacterial species among all cases, of which 22 were pathogens, 30 were opportunistic pathogens, 20 were normal flora, 8 were probiotics, and 127 were commensals. The two‐way ANOVA analysis found that 41% (9/22) of pathogens, 47% (14/30) of opportunistic pathogens, 20% (4/20) of normal flora, 25% (2/8) of probiotics, and 20% (25/127) of commensals were differing significantly between the groups (Supporting Information: Table A1).

3.4. Abundance of pathogens, opportunistic pathogens, normal flora, and probiotics

The most abundant species were Clostridium botulinum, Bacillus cereus, Prevotella melaninogenica, Escherichia coli, Staphylococcus aureus, Prevotella oris, Proteus mirabilis, Pasteurella multocida, Lacrimispora sphenoides, Tennerella forsythia, Salmonella enterica, and Alkalihalobacillus pseudofirmus, which were present in all groups The SARS‐CoV‐2‐positive diabetic group consisted of 41 pathogens/opportunistic pathogens (Figures 2A,B). Of which, Acinetobacter nosocomialis, Shigella flexneri, Bordetella pertussis, Dialister pneumosintes, Streptococcus orlis, Escherichia fergusonia, Achoromobacter sp., Selenomonas sp., Cutibacterium acnes, Dolosigranulum pigrum, Pseudomonas aeruginosa, and Stenotrophomonas maltophilia were present solely in that group. Furthermore, Klebsiella pneumoniae, E. coli O157:H7, Yersinia pestis, Porphyromonas, and Enterobacter were present in both SARS‐CoV‐2‐positive diabetic and nondiabetic groups. In contrast to that, Neisseria meningitidis, Haemophillus pittmaniae, and Streptococcus parasanguinis were present only in the control group. Moreover, 12 out of 20 species of normal flora were solely found in the control group, although they were absent in both SARS‐CoV‐2‐positive diabetic and nondiabetic groups. Only three species of normal flora were common in all groups and four species of normal flora were present in both the control and the SARS‐CoV‐2‐positive diabetic groups, the rest was found only in the SARS‐CoV‐2 diabetic group (Figure 2C). All known probiotic species of Streptococcus, Lactobacillus, Enterococcus, or Bifidobacterium were absent in the control group.

Figure 2.

Comparison of bacterial species with relative abundance among the groups of cases. (A) Presence of known pathogenic species with the relative intensity of bacterial genome. (B) Presence of opportunistic pathogen among different groups with relative abundances. (C) The relative abundance of normal flora and known probiotic species (red highlights). (D) The significant difference of bacterial species and phyla by DESeq.2 analysis with log₂ fold changes between control and SARS‐CoV‐2 diabetic group (p < 0.05). (E) The significant difference of bacterial species by DESeq.2 analysis with log₂ fold changes between control and SARS‐CoV‐2 nondiabetic group (p < 0.05). (F) The significant difference of bacterial species by DESeq.2 analysis with log₂ fold changes between SARS‐CoV‐2 diabetic group and SARS‐CoV‐2 nondiabetic group (p < 0.05). SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

The DESeq.2 RNA sequence data analysis illustrated the difference in bacterial species between the groups. Several pathogens, opportunistic pathogens, and normal flora differ significantly in the SARS‐CoV‐2 diabetic and nondiabetic groups compared to the control group (Benjamini–Hochberg corrected, p < 0.05) (Figure 2D–F).

The top 26 highly abundant species were compared with pairwise Wilcoxon's rank‐sum test among the three groups (Supporting Information: Figure A5). E. coli O157:H7, K. pneumoniae, P. aeruginosa, S. enterica, Clostridium sphenoides, and Prevotella intermedia found significantly higher abundance among the SARS‐CoV‐2‐positive diabetic group (p < 0.05).

3.5. Networking of bacteria

Spearman's correlation coefficient analyses illustrated 183 significant positive correlations with an r value ranging from 0.6 to 1 (p < 0.05) among all pathogens and opportunistic pathogens. P. aeruginosa positively associated with 14 other pathogenic bacteria including D. pneumosintes, E. coli O157:H7, P. intermedia, A. nosocomialis, and synergistically correlated with C. botulinum. D. pneumosintes was positively associated with 12 other pathogenic bacteria. The increasing abundance of T. forsythia in the SARS‐CoV‐2 diabetic group compared to the control group was also associated with 11 other pathogenic bacteria. K. pneumoniae was positively associated with Y. pestis, E. coli O157:H7, Enterobacter sp., S. enterica, and S. oralis. H. parainfluenzae was positively associated with H. pittmaniae, N. meningitidis, Alloprevotella sp., and Tennerella sp. A total of 13 significant negative correlations (p < 0.05) were observed associated with C. botulinum ranging from −0.67 to −0.77 correlation (r) value (Figure 3A).

Figure 3.

(A) Network analysis shows the co‐occurrence patterns of pathogen opportunistic pathogens. Positive Spearman's correlation represents (r > 0.6) with significant (p < 0.05) correlation and negative Spearman's correlation (r) range from −0.67 to −0.77 with significant (p < 0.05) correlation. The node size is proportional to the mean abundance of the species. (B) Network analysis shows the co‐occurrence patterns of probiotics and normal flora. Positive Spearman's correlation represents (r > 0.6) with significant (p < 0.05) correlation and no negative correlation was found here. The node size is proportional to the mean abundance of the species. SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

Correlation analysis (p < 0.05) with probiotics and normal flora was found to have 109 positive correlations with r > 0.6. The control group featured a cluster of 15 normal flora, mostly of Leptotrichia and Bacteroides, positively associated with each other. In the SARS‐CoV‐2 diabetic group, six normal flora and six probiotics showed significant positive associations (Figure 3B).

3.6. Interactive modeling of pathogens and probiotics

An interactive model analysis found positive associations in E. coli O157:H7, K. pneumoniae, Pasturella, Prevotella, and S. enterica with all eight probiotic strains in COVID‐19 patients. A negative association was observed in B. pertussis, C. botulinum, Klebsiella oxytoca, Klebsiella quasipneumoniae, Tennerella, and A. nosocomialis with most of the probiotics. However, Salmonella, Shigella, Porphyromonas, and Tenerella showed negative associations with lactobacilli (Figure 4).

Figure 4.

Pairwise Spearman's correlation with pathogenic bacteria probiotic microbial community in SARS‐CoV‐2 patients. The numbers display Spearman's correlation coefficient (r). Blue and red indicate positive and negative correlations, respectively. The color density, ellipse size, and numbers reflect the scale of correlation. Significance levels: *p  <  0.05; **p  <  0.01; ***p  <  0.001. SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

3.7. Secondary bacterial infections

E. coli O157:H7, H. influenzae, K. pneumoniae, N. meningitidis, P. aeroginosa, and S. aureus were known to cause secondary infections, followed by respiratory viral infections. E. coli O157:H7, K. pneumoniae, S. aureus, and P. aeroginosa had significantly higher abundance among the diabetic group (p < 0.05) compared to the control group (Figure 5). However, H. influenzae and N. meningitidis were found only among the control group.

Figure 5.

Comparison of relative abundance of seven secondary infection‐causing bacteria in SARS‐CoV‐2 diabetic and SARS‐CoV‐2 nondiabetic group compared to control group. Green, orange, and ash color represent SARS‐CoV‐2 diabetic, SARS‐CoV‐2 nondiabetic, and control groups, respectively. The diversity for each species is plotted on boxplots and comparisons are made with pairwise Wilcoxon's rank‐sum test. Significance level (p value) 0.05 and 0.1 is represented by the symbols “*” and “NS,” respectively. SARS‐CoV‐2, severe acute respiratory syndrome coronavirus‐2.

4. DISCUSSION

Respiratory viral infections are known to enhance the growth of opportunistic pathogens by immune suppression, altered immune response, or dysbiosis of the respiratory tract. ⁹ Secondary bacterial infections often take the lead after the subsequent viral infections, especially among comorbidly affected patients. The SARS‐CoV‐2 virus may cause similar effects with a devastating outcome of 6.5 million deaths so far. Most of them are people with old age, obesity, Type 2 diabetes, hypertension, and cardiovascular diseases. ²⁸ Microbiome analysis of comorbid patients is thus imperative to understand the influence of microbiota on immune processes in SARS‐CoV‐2 infections. ²⁹ Our study analyzed the microbial interactions of SARS‐CoV‐2‐infected diabetic patients compared to controls adjusted with age and gender. Our α‐diversity analysis provided an insight into the disparity in species to understand microbial communities and diversities on a host, while β‐diversity provides the index of variation in species composition among different habitats. The microbiome analysis observed significant dominance of pathogenic bacteria among diabetic patients. The dysbiosis of the bacterial community, also described in other studies, ³⁰ , ³¹ were evident in SARS‐CoV‐2 diabetic and nondiabetic groups compared to the control group (Figure 2A); however, the SARS‐CoV‐2 diabetic group was enriched with opportunistic pathogens compared to others (Figure 2B). The Kruskal–Wallis significance test of variance demonstrated that the SARS‐CoV‐2 diabetic patients possess a significantly increased species of pathogens (p < 0.05) and opportunistic pathogens (p < 0.05) compared to the control and SARS‐CoV‐2 nondiabetic groups. The taxonomic unit (species) of pathogenic bacteria (both pathogens and opportunistic pathogens) in this study were 41 in the SARS‐CoV‐2 diabetic group compared to 26 in the control group and 24 in SARS‐CoV‐2 nondiabetic group (Figure 2A,B). A similar finding was reported in a comparative cross‐sectional study by Rodriguez et al., ¹³ which demonstrated that the SARS‐CoV‐2‐infected ICU patients harbored more pathogenic bacteria and viruses. Other recent studies ¹⁰ , ¹⁵ , ³² also reported a significantly higher abundance of opportunistic pathogens in SARS‐CoV‐2‐infected patients, such as Streptococcus, Rothia, Veillonella, and Actinomyces, and a lower relative abundance of beneficial symbionts compared with healthy human. In our study, the controls were enriched with numerous species of normal flora compared to both SARS‐CoV‐2‐positive groups (Figure 2C). The control group contained 19 species of normal flora, which was reduced to eight in the SARS‐CoV‐2 diabetic group and only three in the SARS‐CoV‐2 nondiabetic group. A group of researchers reported that the specific intestinal microbiota of COVID‐19 patients ³³ could suppress the SARS‐CoV‐2 attachment. The severely infected patients might feature dysbiosis, where the normal microbiota is replaced by pathogenic bacteria. ³⁴ Hyperglycemia, inflammation, and severe oxidative stress in a patient's physique may alter the oral microbiome. ³⁵ Other evidence also suggested an association of dysbiosis of the normal microbiota due to diabetes. ³⁶ , ³⁷ Several studies found that opportunistic pathogens were the most common cause of secondary infections in the viral epidemic. ³⁸ , ³⁹ In our bacterial abundance comparison study (Supporting Information: Figure A5), we found a significantly higher abundance of E. coli O157:H7, K. pneumoniae, P. aeruginosa, S. enterica, C. sphenoides, and P. intermedia; most of those species were known to cause secondary infections after viral respiratory tract infection. ⁹ Abundance of secondary bacterial infection analysis exposed that P. aeruginosa was present in only the diabetic patients (p < 0.05). S. aureus, E. coli, and K. pneumoniae were evident in the respiratory tract among all SARS‐CoV‐2 diabetic patients (p < 0.05) and some nondiabetic patients compared to none in the controls (Figure 5). One interesting finding in our study was the absence of H. influenzae and N. meningitidis in the SARS‐CoV‐2 diabetic groups, unlike other secondary infection‐causing bacteria (Figure 5). However, those bacteria significantly enriched the control groups in our study, which was also evident by Qin et al. ³⁷ and his research team. During an active infection, those bacteria may move down from the upper respiratory tract and could not be detected from the nasopharyngeal samples.

Moreover, studies on SARS viral epidemic demonstrated that coinfection was one of the major complications in prolonged hospitalization and mechanical ventilation. Pathogenic bacteria like E. faecalis, K. pneumonia, A. baumannii, and S. maltophilia inhabiting the oral cavity can also cause nosocomial infections. ⁴⁰ Legionella pneumophila, ⁴¹ N. meningitidis, and Moraxella catarrhalis ⁴² were known to be associated with influenza coinfection. Porphyromonas gingivalis, also found in our study, was an important cause of periodontitis. ⁴³ The use of antibiotics to prevent those secondary infections may also lead to the loss of normal flora and probiotics causing dysbiosis in patients with COVID‐19.

An interesting feature in our study was the significant reduction of normal flora in the SARS‐CoV‐2 patient groups such as Leptotrichia, Bacteroides, Fusobacterium, Chorynebacterium, and Bernesiella spp., indicating the imbalance of microbiota (Figure 2D–F). Moore et al. ⁴⁴ also found a significant reduction of Fusobacterium periodonticum in the nasopharynx during SARS‐CoV‐2 infections. Those bacterial flora found in the oral cavity may inhibit pathogenic bacteria by producing antimicrobial substances such as bacteriocins, lactic acid, and hydrogen peroxide, which might create a hostile condition for the pathogenic bacteria. ³³ The presence of probiotic species with several pathogens and opportunistic pathogens in the SARS‐CoV‐2 diabetic group revealed that those probiotics might assist the host by inhibiting those pathogenic bacteria.

The correlation coefficient network analysis in this study found significant positive associations and few negative associations among the pathogenic bacteria in SARS‐CoV‐2‐infected diabetic patients (Figure 3A). This analysis indicated various patterns of pathogenic networks in the SARS‐CoV‐2 diabetic group, especially among enteric pathogens, nosocomial bacteria, and other opportunistic pathogens. However, this analysis found no correlation in the most abundant species between the control group and SARS‐CoV‐2‐positive nondiabetic groups.

The co‐occurrences of the network among the probiotic and normal flora identified several significant positive associations (N = 109) but no synergistic correlations. There was a separate cluster of 15 normal flora in the control group with 90 significant positive associations, which were absent in SARS‐CoV‐2‐positive diabetic and nondiabetic groups (Figure 3B). The decrease of normal flora in the later groups indicated that they were outnumbered by the pathogenic species mentioned above. The increase in the pathogenic environment results in a nonproductive hyperactive immune response in the host, which ultimately suppresses the adaptive immune response against SARS‐CoV‐2. Differential immunogenicity increases the favorable environment for secondary infections among diabetic patients and brings poor outcomes and fatalities. Nevertheless, increased species of probiotic strains in the SARS‐CoV‐2 diabetic group compared to the control group and SARS‐CoV‐2 nondiabetic group indicated an inadequate resistance against highly abundant pathogenic bacteria ¹⁰ (Supporting Information: Figure A6). Another study ⁴⁵ demonstrated that immunomodulatory probiotics, Rothia mucilaginosa, K. oxytoca, Enterobacter kobei, B. cereus, Faecalibacterium prausnitzii, and so forth, were enriched in the COVID‐19‐positive patients. Our interactive model analysis found negative associations in B. pertussis, C. botulinum, K. oxytoca, K. quasipneumoniae, Tennerella, and A. nosocomialis with probiotic strains in those patients (Figure 4). However, positive associations were also observed in E. coli O157:H7, K. pneumoniae, Pasturella, Prevotella, and S. enterica. Our study had a limitation in that the microbiome analyses were performed with a small number of individuals studied. In developing countries, this limitation is quite common because of paradoxical situations: the doubled price of metagenome reagents in developing countries, the extended delivery time with a short period of expiry, and unavailability of reagents during the peak times of COVID‐19 infections. Therefore, there are very little data reported from those regions where the highest number of patients are having comorbidity. However, our preliminary observations and hypothesis were supported by appropriate statistical methods and the results are compared with suitable controls.

5. CONCLUSIONS

The SARS‐CoV‐2‐positive diabetic patients were possessed by significantly increased pathogenic species compared to the control and SARS‐CoV‐2 nondiabetic groups. In both groups, the normal‐flora strains were replaced by secondary bacterial infections that might correlate with the severity and outcome of complications. Those dysbiosis suppressed the adaptive immune response against SARS‐CoV‐2 because of the induced immune response against those pathogenic bacteria. The presence of few probiotic species among the SARS‐CoV‐2 diabetic patients indicated that those probiotics were inhibiting the pathogens as observed in our study. However, the numbers might not be competitive enough to provide successful protection as seen within deceased patients. One approach for maintaining a healthy microbiome in SARS‐CoV‐2 diabetic patients might include promoting probiotics and normal flora by dietary changes and reducing proinflammatory states. Relocation of the microbial balance with normal flora and sufficient probiotics may prevent secondary infections that might enhance the adaptive immune response, leading to better outcomes for SARS‐CoV‐2 diabetic patients.

AUTHORS CONTRIBUTIONS

Hassan M. Al‐Emran: Conceptualization; methodology; validation; formal analysis; investigation; visualization; writing original draft; funding acquisition. Shaminur Rahman: Software; formal analysis; data curation. Md. Shazid Hasan: Investigation; resources; validation. Rubayet Ul Alam: Writing review and editing; resources. Ovinu Kibria Islam: Writing review and editing; resources. Ajwad Anwar: Writing review and editing. Md. Iqbal Kabir Jahid: Formal analysis; validation; writing review and editing; resources; supervision; project administration. Anwar Hossain: Writing review and editing; supervision; project administration; funding acquisition.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

ETHICS STATEMENT

Ethical approval to conduct this metagenomic study was granted by the Ethical Review Committee of the Jashore University of Science and Technology (ERC no: ERC/FBST/JUST/2020‐41). Informed consent was taken from all COVID‐19‐positive patients and healthy volunteers.

Supporting information

Supplementary information.

Al‐Emran HM, Rahman S, Hasan MS, et al. Microbiome analysis revealing microbial interactions and secondary bacterial infections in COVID‐19 patients comorbidly affected by Type 2 diabetes. J Med Virol. 2022;95:e28234. 10.1002/jmv.28234

DATA AVAILABILITY STATEMENT

The raw reads of metagenomic data are available in the NCBI SRA BioProject with Accession no.: PRJNA733662. The SARS‐CoV‐2 genome sequence data were deposited in GISAID with the Accession nos: EPI_ISL_746318, EPI_ISL_746319, and EPI_ISL_746323.