Rapid Metagenomic Diagnostics for Suspected Outbreak of Severe Pneumonia

To the Editor: Recent outbreaks of severe pneumonia or acute respiratory distress syndrome (ARDS) have attracted much public interest. Given current awareness levels, clinical personnel and health officials must rapidly and adequately respond to suspected outbreaks to prevent public disturbances. We report a case that highlights the potential of next-generation sequencing (NGS) to complement conventional diagnostics in such scenarios.

To the Editor: Recent outbreaks of severe pneumonia or acute respiratory distress syndrome (ARDS) have attracted much public interest. Given current awareness levels, clinical personnel and health officials must rapidly and adequately respond to suspected outbreaks to prevent public disturbances. We report a case that highlights the potential of next-generation sequencing (NGS) to complement conventional diagnostics in such scenarios.
On March 29, 2013, a police officer (patient 1) was admitted to the emergency department of the University Medical Centre Hamburg-Eppendorf in Hamburg, Germany, because of ARDS. The patient was given mechanical ventilation; all diagnostic test results for pathogens commonly known to cause pneumonia were negative (www.virus-genomics.org/ supplementaries/EID1406.pdf). Although treatment with antimicrobial drugs was immediately initiated, the patient died 6 days later of multiple organ failure.
On April 5, a second police officer (patient 2) from the same county was admitted to the same emergency department because of ARDS. He was transferred to the intensive care unit and given mechanical ventilation. Similar to the situation for patient 1, diagnostic test results were negative, prompting the news media to suspect an outbreak of a novel or mutated virus (1,2). Especially because of simultaneous outbreaks of avian influenza and infections with Middle East respiratory syndrome coronavirus in other parts of the world, these reports caused serious concern among the public and health officials.
After the death of patient 1 and hospitalization of patient 2, we subjected nucleic acids extracted from bronchoalveolar lavage (BAL) specimens from both patients to NGS by using a MiSeq sequencer (www.illumina.com/ systems.ilmn). To enable rapid and unbiased detection of bacterial and viral agents, we did not enrich specific sequences. The entire workflow (www. virus-genomics.org/supplementaries/ EID1406.pdf) was completed within 50 hours.
First-line analysis clearly identified Chlamydophila psittaci in the RNA sample from patient 2, but no sequences of obvious pathogenic origin were detected in samples from patient 1 (Figure). C. psittaci, an intracellular bacterium, can be transmitted by inhaling aerosolized secretions or feces from infected birds (3). Person-toperson transmission of this bacterium is rare (4,5). Ornithosis, a disease characterized by severe pneumonia and influenza-like symptoms, might develop in persons infected with this bacterium. Because of the rarity of the disease, standard diagnostic panels usually do not include C. psittaci. After 11 days of antimicrobial drug treatment, the condition of patient 2 improved, and the patient was transferred to a general hospital ward.
On April 29, a third police officer (patient 3) who had been sharing office space with patient 2, came to the hospital with symptoms of pneumonia. After antimicrobial drug treatment, his condition rapidly improved and the patient was discharged on day 7.
At admission of patient 3, samples from all 3 patients were analyzed at greater read depth by using a HiSeq system. RNA sequencing identified commensal bacteria in all samples, but C. psittaci was present only in samples from patient 2 (Figure). Likewise, subsequent high-throughput 16S rRNA sequencing (6), PCR, and serologic analysis unequivocally identified a C. psittaci infection in patient 2, but not in patients 1 or 3 (www. virus-genomics.org/supplementaries/ EID1406.pdf).
We did not detect viral pathogens in any samples. At the DNA level, most nonhost reads originated from nonpathogenic single-stranded DNA anelloviruses (7). No RNA viruses were found, although influenza A(H3N2) virus was readily identified in a MiSeq analysis of a control BAL sample from a patient with a diagnosis of influenza A (PCR cycle threshold 32) (Figure). Furthermore, pairwise BLAST analysis (http://blast.ncbi.nlm. nih.gov/Blast.cgi) did not reveal the presence of unknown sequence contigs that were shared among the patients, as would be expected in case of infection with a novel viral agent. Together with the confirmed C. psittaci infection in patient 2, the absence of a common pathogen signature strongly suggests that the cases were unrelated.
We used a comprehensive metagenomic approach to resolve cases suspected of representing an ongoing outbreak. The method used enabled diagnosis of a C. psittaci infection within a reasonable timeframe to allow for timely clinical intervention. These findings strongly suggest that NGS methods can complement conventional diagnostics (8)(9)(10) and also highlight their potential to aid clinical personnel and health agencies in making appropriate decisions during suspected outbreaks. Clearly, however, NGSbased methods will have to be further standardized and validated before their full potential in diagnostic settings can be realized.
The absence of pathogenic sequences in patients 1 and 3 might suggest that their clinical symptoms had noninfectious causes. However, although samples were collected during the acute phase of clinical symptoms, and despite our ability to detect an influenza A infection in controls, we cannot fully exclude the possibility that a potentially causative pathogen present at low levels might have evaded detection. Thus, systematic and correlative studies evaluating the sensitivity of NGS-based detection methods in different diagnostic entities are urgently needed.
This study was supported in part by a project grant from the German Center for Infection Research to N.F. and A.G. EVs are classified into 12 species according to the molecular and antigenic properties of their viral capsid protein (VP1). To date, 7 species are known to infect humans, including EV-A to EV-D and rhinovirus A, B, and C (www.picornastudygroup.com/taxa/ species/species.htm) EV-C117 was a newly found EV-C genotype. It was identified in a nasopharyngeal sample from a hospitalized child, 3 years and 9 months of age, with community-acquired pneumonia in Lithuania in 2012 (2,3). However, aside from this case, little is known about the prevalence and clinical significance of EV-C117. Here, we report the detection of EV-C117 in children in China and Mongolia with respiratory tract infections (RTIs).
During March 2007-March 2013, we screened for EV-C117 in respiratory samples from patients with RTIs in China and Mongolia, including nasopharyngeal aspirates collected from 3,108 children in China who had lower respiratory tract infections when they were admitted to Beijing Children's Hospital (4) and swab samples from 2,516 patients in Mongolia with influenza-like illness (online Technical Appendix Table 1, wwwnc. cdc.gov/EID/article/20/6/13-1596-Techapp1.pdf). Respiratory viruses in samples from China were screened by using multiplex PCR and single PCR assays as described (4). Samples from Mongolia were screened by using the FTD Respiratory Pathogens Multiplex Assay Kit (Fast-track Diagnostics, Luxembourg City, Luxembourg). EVpositive samples were further genotyped by using reverse transcription PCR (RT-PCR) and primers sequentially targeting the VP1 region (5), the 5′-untranslated region (5′-UTR)/VP4/ VP2 region (6) and the 5′-UTR (7). A 394-nt amplicon corresponding to the 5′-UTR of EVs was obtained from 10 children in China; a 598-nt amplicon corresponding to the 5′-UTR/ VP4/VP2 region was obtained by RT-PCR from 5 children in Mongolia. Blastn analysis (www.blast.ncbi. nlm.nih.gov/Blast.cgi) of PCR amplicons showed that only amplicons detected in 2 children from China (patients BCH096A and BCH104A) and 2 children from Mongolia (patients MGL126 and MGL208) had the highest similarity (95%-98%) to the EV-C117 prototype strain LIT22.
Virus isolation for EV-C117 by using Vero and H1-HeLa cells was unsuccessful. Through blastn and phylogenetic analyses, we also found that the previously identified EV-C strain HC90835 (EU697831, from Nepal) (8), EV-C104 strain CL-C22 (EU840734, EU840744, and EU840749, from Switzerland) (9) and a rhinovirus strain SE-10-028 (JQ417886, from South Korea), also belong to EV-C117 ( Figure, panel  A), indicating that EV-C117 is widely distributed geographically. Because a large proportion of EV infections are subclinical or mild (1), the prevalence of EV-C117 should be further estimated by using serologic investigations in general populations.
The VP1 sequences of the EV-C117 strains isolated in China and Mongolia were 89.9%-95.6% (nt) and 95.2%-98.3% (aa) identical to the EV-C117 prototype strain LIT22 (patient JX262382). Alignment analysis of amino acid sequences showed differences between strains isolated in this study and LIT22, i.e., Ser 15 (strains