key: cord-0872471-ux3c3dzf
authors: Zhong, H.; Wang, Y.; Shi, Z.; Zhang, L.; Ren, H.; He, W.; Zhang, Z.; Zhu, A.; Zhao, J.; Xiao, F.; Yang, F.; Liang, T.; Ye, F.; Zhong, B.; Ruan, S.; Gan, M.; Zhu, J.; Li, F.; Wang, D.; Li, J.; Ren, P.; Zhu, S.; Yang, H.; Wang, J.; Kristiansen, K.; Tun, H. M.; Chen, W.; Zhong, N.; Xu, X.; Li, Y.-m.; LI, J.
title: Characterization of Microbial Co-infections in the Respiratory Tract of hospitalized COVID-19 patients
date: 2020-07-05
journal: nan
DOI: 10.1101/2020.07.02.20143032
sha: d32b768e194315e416afd988bf9f02c8296e96f7
doc_id: 872471
cord_uid: ux3c3dzf

Summary Background Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a global pandemic of Coronavirus disease 2019 (COVID-19). However, microbial composition of the respiratory tract and other infected tissues, as well as their possible pathogenic contributions to varying degrees of disease severity in COVID-19 patients remain unclear. Method Between January 27 and February 26, 2020, serial clinical specimens (sputum, nasal and throat swab, anal swab and feces) were collected from a cohort of hospitalized COVID-19 patients, including 8 mildly and 15 severely ill patients (requiring ICU admission and mechanical ventilation), in the Guangdong province, China. Total RNA was extracted and ultra-deep metatranscriptomic sequencing was performed in combination with laboratory diagnostic assays. Co-infection rates, the prevalence and abundance of microbial communities in these COVID-19 patients were determined. Findings Notably, respiratory microbial co-infections were exclusively found in 84.6% of severely ill patients (11/13), among which viral and bacterial co-infections were detected by sequencing in 30.8% (4/13) and 69.2% (9/13) of the patients, respectively. In addition, for 23.1% (3/13) of the patients, bacterial co-infections with Burkholderia cepacia complex (BCC) and Staphylococcus epidermidis were also confirmed by bacterial culture. Further, a time-dependent, secondary infection of B. cenocepacia with expressions of multiple virulence genes in one severely ill patient was demonstrated, which might be the primary cause of his disease deterioration and death one month after ICU admission. Interpretation Our findings identified distinct patterns of co-infections with SARS-CoV-2 and various respiratory pathogenic microbes in hospitalized COVID-19 patients in relation to disease severity. Detection and tracking of BCC-associated nosocomial infections are recommended to improve the pre-emptive treatment regimen and reduce fatal outcomes of hospitalized patients infected with SARS-CoV-2. Funding National Science and Technology Major Project of China, National Major Project for Control and Prevention of Infectious Disease in China, the emergency grants for prevention and control of SARS-CoV-2 of Ministry of Science and Technology and Guangdong province, Guangdong Provincial Key Laboratory of Genome Read and Write, Guangdong Provincial Academician Workstation of BGI Synthetic Genomics, and Shenzhen Engineering Laboratory for Innovative Molecular Diagnostics.

highest number of reads assigned to the targeted species was used for coverage analysis. Reads assigned to a given species were 4 4 aligned against the corresponding reference genome by bowtie2 v2.3.0 (the '-sensitive' mode) 22 . Sequencing depth and genome 4 5 coverage of each reference genome were determined with BEDTools coverage v2.27.1 (genomecov -ibam sort.bam -bg) 23 .

Reliable co-detection with known respiratory viruses was defined when >50% of the genome was covered.

Characterization of non-viral microbial communities in hospitalized patients with SARS-CoV-2 4 8 infection 4 9

Non-viral microbial taxon assignment of the non-human non-rRNA reads was performed using clade-specific marker gene-based 5 0

MetaPhAln2 with the default parameter options for non-viral microbial composition (--ignore-viruses) 24 . The presence 5 1 and relative abundances of non-viral microbial taxa at the phylum, genus and species were estimated (appendix 2 p 7: Table S6 ).

Mono-dominance of a given microbial taxa (genus or species) was defined if a taxon had a relative abundance >60% in one 5 3 sample as suggested by a previous study 25 . Most of the RNA reads of the two predominant bacterial genera Burkholderia and

Parabacteroides, respectively, identified in the respiratory and gastrointestinal tract of severe cases could hardly be assigned to 5 5 species level by MetaPhAln2, which might reflect that the two genera contain closely related species that are difficult to 5 6 differentiate by marker genes. In order to determine which species and how abundant species were in samples mono-dominated by 5 7

Burkholderia or Parabacteroides, we downloaded four reference genome sequences of the most frequently isolated BCC species 5 8 (B. cenocepacia J2315, B. multivorans ATCC BAA-247, B. cepacia ATCC 25416 and B. dolosa AU0158) and two gut 5 9

Parabacteroides species (P. distasonis ATCC 8503 and P. merdae NCTC13052). For each sample, reads were mapped against 6 0 corresponding references by bowtie2 v2.3.0, and the sequencing depth and genome coverage were estimated by BEDTools 6 1 coverage v2.27.1 as described above. The summary of coverage and depth of reference genomes for selected samples are 6 2 presented in appendix 2 p 8: Table S7 . Likewise, metatranscriptomics-detected bacterial or fungal respiratory co-infections were 6 3 defined if the respiratory specimens (at least one sample) of severe patients were mono-dominated (relative abundance >60%) by 6 4 pathogenic microbes known to cause nosocomial infections (appendix 2 p 9: TableS8).

To identify the presence and expression patterns of potential virulence factors in B. cenocepacia identified in P01, we collected 6 7 multiple functional categories of virulence genes previously studied and verified by gene mutation analysis in B. cenocepacia 6 8 strains as well as corresponding gene ID in the annotated J2315 genome 26 , including 1) resistance to stress conditions, 2) 6 9 antimicrobial resistance, 3) quorum sensing, 4) iron uptake, 5) flagella and cable pilus, 6) lipopolysaccharide and 7) 7 0 exopolysaccharide. In addition, a pathogenicity island identified on chromosome 2 (BCAM0233-BCAM0281) of B. cenocepacia 7 1 J2315 by using comparative genomics was included 27 . Non-human non-rRNA microbial reads of all samples from P01 were 7 2 mapped against the reference genome of B. cenocepacia J2315 using bowtie2 v2.3.0 as described above and identified by the gene 7 3

IDs in the J2315 genome. For each sample, only virulence genes with more than 10 mapped reads were retained. A total of 50 7 4 expressed virulence genes were identified in clinical samples collected from P01 and presented in appendix 2 p 10: TableS9.

To compare the expression levels between different genes, we performed normalization of target gene expression levels among all 7 6 detected virulence genes using the following equation:

where i (1,2, … k) refers to a given virulence gene identified in B. cenocepacia J2315; Lg i is the length of gene i; N i is the reads 7 9

number that mapped to gene i.

Non-metric multidimensional scaling (NMDS) ordination of respiratory microbial community was conducted using the Manhattan 8 2 distances based on a presence/absence matrix of genus profiles of 47 respiratory specimens (7 from mild cases and 40 from severe 8 3 cases) (R version 3.6.1, vegan package). Kruskal-Wallis test and Wilcoxon rank-sum test were performed to compare differences 8 4 in SARS-CoV-2 viral loads among five different types of clinical specimens (throat swab, nasal swab, sputum, anal swab and 8 5 feces) and between overall respiratory specimens from mild and severe cases, respectively (R version 3.6.1, coin package).

Demographic information of patients and clinical specimens used in the study 8 8 23 patients with COVID-19 hospitalized in the period January 10-March 31, 2020 in four hospitals in the Guangdong Province,

China, were enrolled into this study. 15 infected patients (40-80 years) admitted to the ICU and recived mechanical ventilation 9 0 were defined as having severe COVID-19, and the remaining 8 patients (2-65 years) were mild cases (appendix 2 p 2: TableS1).

60 % (9 out of 15) of severely ill patients received invasive mechanical ventilation (appendix 2 p 2: TableS1), and 95.7% of the 9 2 patients (22 out of 23) received antiviral medications. To prevent and control nosocomial infections, all severe cases received 9 3 broad-spectrum antibiotics, and simultaneously 93.35% (14 out of 15) received antifungal agents (appendix 2 p 2: TableS1). By

contrast, none of the mild cases were treated with antibacterial or antifugal drugs. Up to March 31, 2020, 53.3% (8 out of 15) of 9 5 the severely ill patients had been transferred out of ICU or discharged from hospitals, and all mild cases had been discharged,

whereas an elderly male patient died one-month after admission to ICU (P01) (appendix 2 p 2: TableS2). Sixty-seven serial 9 7 clinical specimens from the respiratory tract (RT) (n=47, sputum, nasal and throat swab) and gastrointestinal tract (GIT) (n=20, 9 8 anal swab and feces) of these patients were obtained during the same above period for comprehensive assessment of microbial non-rRNA transcripts) varied between different types of specimens, constituting a relatively high fraction of total high-quality 0 6 reads among RT specimens and a low fraction among GIT specimens (appendix 1 p 1: Supplementary figure 1 and appendix 2 0 7 p 4: TableS3). After removing host data, SortMeRNA was applied 21 to filter microbial rRNA from the metatranscriptomic data.

The final remaining non-human non-rRNA data (ranged from 386Mb to 145Gb) were then used to assess viral and non-viral 0 9 microbial composition by Kraken2X 20 and MetaPhlan2 24 , respectively(Methods). Detailed data statistics for each processing step 1 0 are provided in appendix 2 p 4: TableS3.

1 1

First, viral RNA reads belonging to family Coronaviridae were defined as SARS-CoV-2-like reads. As expected, Coronaviridae 

Except for Coronaviridae, RNA-seq analysis also revealed a great diversity of viral composition in clinical samples from 2 4 infected patients. Natural hosts of the highly abundant viruses differed, including but not limited to animals (e.g., Picornaviridae,

Pneumoviridae and Herpesviridae), bacteria (e.g., Podoviridae, Siphoviridae and Myoviridae) and plants (Virgaviridae) (figure 2 6 1B and appendix 2 p 5: Table S4 ). The co-infections of known human respiratory viruses were further confirmed (viral 2 7 genome coverage>50%) in four out of thirteen severely ill patients with metatranscriptomic data of respiratory samples (30.8%), 

We next evaluated the occurrence of bacterial/fungal co-infections that have been shown to be associated with worse clinical 

Next, the non-viral microbial RNA composition of all 67 clinical specimens was analyzed to fully assess possible nosocomial 5 0

infections. As none of mild cases was admitted to ICU or received antibacterial/antifungal agents, we compared the respiratory 5 1 microbial communities between mild (n=7) and severe (n=13) cases to determine microbial dysbiosis, co-infection and their 5 2 associations with clinical outcomes. Remarkable microbial differences in RT specimens between mild and severe cases were occurrence >80% and mean relative abundance>5%) ( figure 2B,C) , which is consistent with common microbial communities 6 0 found in the human nasal and oral cavity 43 . However, except Veillonella, each of the latter three genera enriched in mild cases was 6 1 only detected in few severe cases (n≤3, figure 2A) and had a mean abundance of less than 0.05% (figure 2C).

Of note, RT microbial features of severe cases were identified to be patient-specific. Among 40 respiratory samples from severe 6 3 patients, over 60% were mono-dominated (relative abundance>60%) by bacterial genus Burkholderia (11 samples from P01, P04 6 4 and P20), Staphylococcus (6 samples from P10 and P19) or Mycoplasma (7 samples from P05, P06, P14, and P18) (figure 2B,C 6 5 and appendix 1 p 5: Supplementary figure 4D) . Each genus was detected in 69.2% of (9 out of 13) severe patients and 92.3% Table S7 ). All

Staphylococcus RNA reads of RT samples from P06 (who also had positive S. epidermidis culture), P10 and P19 were also 7 1 assigned to S. epidermidis ( figure 2B and appendix 1 p 6: Supplementary figure 5B) . However, S. aureus, a major 7 2 hospital-acquired pathogen 44 , was not detected in metatranscriptomic data of any of the sequenced RT samples. Mycoplasma orale 7 3 and M. hominis, rather than M. pneumoniae, were the two identified pathogenic Mycoplasma members (figure 2B and appendix 1 7 4 p 6: Supplementary figure 5C). Propionibacterium and Escherichia were also frequently detected in RT samples of severe cases 7 5 (individual occurrence >80%) but were less abundant than the former three genera (mean relative abundance<3%) (figure 2C).

All the five dominant bacterial genera in severe cases have been frequently associated with nosocomial infections, while they 7 7

were not detected or present in extremely low abundance in mild cases (relative abundance<0.15%) (figure 2A,C) . monodominance (relative abundance>60%), we determined that a metatranscriptomics-based bacterial co-infection rate among 1 mono-dominated the gut microbial transcripts in five samples from four severe cases and displayed a much higher abundance in 0 2 severe cases than that in mild cases (appendix 1 p 7: Supplementary figure 6B-D and appendix 2 p 8: Table S7 ). The extreme 0 3 bloom of P. distasonis, a low abundant but common taxa in the human gut, has been reported after beta-lactam ceftriaxone 0 4 treatment 25 . Taken together, metatranscriptomic findings have not only complemented and enhanced the laboratory-diagnosed 0 5 microbial co-infections but also provided comprehensive information of microbial dysbiosis on SARS-CoV-2 infected patients. of B. cenocepacia to non-viral transcripts gradually increased from 9.5% to 64% (from January 29 to February 07) (figure 3A).

The time-dependent dynamics of transcript levels of B. cenocepacia suggested that transfer from the upper respiratory tract to the 1 5 lower gastrointestinal tract had caused a secondary systemic infection in P01. Our findings were also consistent with the patient's 1 6 death certificate record indicating septic shock as the cause of death (appendix 2 p 2: Table S1 ). Indeed, several studies have 1 7 pointed out that among BCC-infected CF patients, infection with B. cenocepacia, rather than other common isolated BCC 1 8 members (such as B. multivorans and B. cepacia), constituted the highest risk factor of death 26,40 .

Next, virulence factors (VF) expressed by B. cenocepacia in P01 were analyzed in order to better understand the pathogenic 2 0 mechanisms of this possible lethal pathogen in this severe COVID-19 case (Methods). The gene rpoE (a member of 2 1 the extracytoplasmic function subfamily of sigma factors) was the most abundantly expressed VF during the entire sampling 2 2 period in P01 with SARS-CoV-2 infection ( figure 3B) . RpoE, as a stress response regulator, has been demonstrated to be essential 2 3

for the growth of B. cenocepacia and the delay of phagolysosomal fusion in macrophages during infection 48 .

A delay in phagolysosomal fusion has also been reported to be an important host immune escape strategy for several bacterial 2 5 pathogens 49 . Other VFs in response to oxidative stress conditions in the host environment, such as those encoding superoxide 2 6 dismutase, peroxidase or catalase (sodC and katB), were also expressed (figure 3B, appendix 2 p 10: Table S9) 

In this study, ultra-deep metatranscriptomic sequencing in combination with clinical laboratory diagnosis, including cultures, and 3 5 colorimetric assays provided a characteristic spectrum of co-detected respiratory pathogenic microbes in a Guangdong 3 6 hospitalized patient cohort infected with SARS-CoV-2, which is distinct from prior results on the most common bacterial 3 7 co-infections identified in previous coronavirus outbreak and influenza pandemics 4-8 or common pathogens associated with 3 8 nosocomial respiratory infections in China 52,53 .

In particular, respiratory co-infections with BCC were detected in 23.1% of severe cases with evidence from both laboratory and a subsequently impaired mucociliary clearance 54-57 , which could promote the adherence and colonization of mucin-degrading 4 7 pathogens in the respiratory tract. Furthermore, BCC bacteria, a major threat to hospitalized CF patients, were found to be 4 8 predominantly localized in the phagocytes and mucus of CF patients 58 . Even though the small sample size in our study limited our 4 9 ability to assess the general prevalence of BCC infections among severely ill SARS-CoV-2 patients, our findings highlight the 5 0 need for monitoring and controlling nosocomial BCC/SARS-CoV-2 co-infection by using rapid diagnostic technologies such as 5 1 PCR-based multilocus sequence typing and enzyme-linked immunosorbent assay. Metatranscriptomic data also revealed that 5 2 30.8% of severely ill patients had respiratory tract co-infections associated with Mycoplasma spp. (including M. hominis and M.

orale), which lack a cell wall and have inherent resistance to commonly administered beta-lactam antibiotics.

Although Mycoplasma usually causes mild illness, it has also been associated with serious infections in seniors 5 5 and immunocompromised individuals 59,60 , who have a high risk for developing severe symptoms from COVID-19. Moreover, we 5 6 did observe super-high expression levels of M. orale in the RT of two severely ill patients (P14 and P18) with prolonged ICU stay 5 7 (>30 days). Thus, our results suggest that the possibility of Mycoplasma-associated co-infection in severely ill COVID-19 patients 5 8 warrants increased attention.

The SARS-CoV-2 infection is often accompanied by GI symptoms such as vomiting, diarrhea and abdominal pain 61,62 , and the 

Med. 2020. DOI:10.1056/NEJMc2001737.

29 He X, Lau EHY, Wu P, et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat Med 2020.

DOI:10.1038/s41591-020-0869-5. 

MetaPhlAn2 for enhanced metagenomic taxonomic profiling

Antibiotics-induced monodominance of a novel gut bacterial order