Patients
From September 2018 to December 2020, all immunocompromised patients with suspected pneumonia admitted to the First Affiliated Hospital of Wenzhou Medical University were retrospectively investigated.
The inclusion criteria of patients were as follows: (A) at least eighteen years old; (B) had an immunocompromised condition; (C) admitted to the Department of Pulmonary and Critical Care Medicine due to suspected pneumonia; (D) had mNGS of BALF within 48 h after admission, and other relevant samples available for the conventional microbiological tests in standard procedures;
Immunocompromised status [7] (≥ 1 of the following risk factors were found): (A) hematologic cancer; (B) chemotherapy during the last three month; (C) chronic steroid (> 0.3 mg/kg/d of prednisone-equivalent for ≥ 3 weeks) or biologic drug use for autoimmune diseases or other immunosuppressive therapy; (D) solid-organ transplant receipt during the last six months; (E) neutropenia; (F) acquired or inherited severe immunodeficiency.
Patients with suspected pneumonia meet the both criteria below: (A) new-onset fever, cough, expectoration or dyspnea; (B) new-onset abnormal chest imaging manifestations.
Data collection
The following data were recorded: age, sex, underlying diseases and clinical manifestations, illness severity (pneumonia severity index (PSI) score), chest radiology, results of laboratory examination including microbiological testing and patient outcomes (regular clinic follow-up or telephone conversation).
Microbiological tests
All patients underwent the operation of bronchoalveolar lavage following a standard safety protocol [8]. The BALF specimens were divided into aliquots. One aliquot was set for bacterial and fungal smear (including Gomori methenamine silver (GMS) staining and acid-fast stain) and culture. One aliquot was submitted for the detection of galactomannan antigen and cryptococcus capsular polysaccharide antigen. X-pert MTB/RIF detection of DNA sequences specific for Mycobacterium Tuberculosis (MTB) and real-time polymerase chain reaction (PCR) for cytomegalovirus (CMV) and influenza A/B virus. The other aliquots (5 mL, settled in a sterile sputum container, stored at − 20 °C) were sent to BGI-Huada Genomics Institute (Shenzhen, China) for detection.
For the other samples, urine was submitted for the detection of antigen of Legionella pneumophila and Streptococcus pneumoniae. Peripheral blood samples were sent for the detection of the galactomannan antigen to Aspergillus, cryptococcus capsular polysaccharide antigen to Cryptococcus and immunoglobulin G and M antibodies to parainfluenza, adenovirus, CMV, Chlamydia pneumoniae and Mycoplasma pneumoniae using commercial enzyme-linked immunosorbent assays (Ani Labsystems) according to the manufacturer’s instructions.
mNGS of BALF
The procedure of mNGS for BALF samples consists of nucleic acid extraction, library construction, sequencing and bioinformatic analyses.
Nucleic acid extraction
A 1.5-mL microcentrifuge tube with 0.6 mL of BALF sample taken from patient, Lyticase (Tiangen, Beijing, China) and 1 g of 0.5 mm glass bead (BioSpec Products, OK, USA) was attached to a horizontal platform on Vortex-Genie2 vortex mixer (Scientific Industries, USA) and agitated vigorously (2800–3200 rpm for 30 min). After agitation, 0.3 mL sample was separated into a new 1.5-mL microcentrifuge tube, and DNA was extracted in 40 μL elution volume by the TIANamp Micro DNA Kit (DP316 Tiangen Biotech, Beijing, China).
Library construction
DNA libraries (DNA engaged in library preparation was 100–400 ng) were constructed by DNA fragmentation, end repair, adapter-ligation, and PCR amplification using the MGIEasy DNA Library Prep Kit (MGI, Wuhan, China). Agilent 2100 Bioanalyzer Instrument and Qubit 3.0 (Thermo Fisher Scientific, USA) platform were used for library quality control. Quality qualified libraries were sequenced on MGISEQ-2000 platform [9].
Sequencing and bioinformatic analyses
Initially, low-quality reads were removed for obtaining high-quality sequencing data. Then, human host sequences were recognized and excluded by mapping against the human reference genome (hg19) using Burrows-Wheeler Alignment [10]. Finally, the remaining sequence data by removal of low-complexity reads were aligned to 4 Microbial Genome Databases (including viruses, bacteria, fungi, and parasites), which were downloaded from NCBI (ftp://ftp.ncbi.nlm.nih.gov/genomes/). RefSeq contains 4,945 whole-genome sequences of viral taxa, 6,350 bacterial genomes or scaffolds, 1064 fungi related to human infection, and 234 parasites which were associated with human diseases.
Criteria for mNGS result
Oral commensals, which were normally parasitic in the human oropharynx and usually without clinically significant, were easily to be detected and always will be carefully considered in combination with a clinical analysis in clinical practice. So, we use a mix of Miao’s [11] and Peng’s [6] diagnostic criteria in our diagnostic criteria, which was more close to clinical practice.
(1) Bacteria (mycobacteria excluded), virus and parasites: A microbe was considered clinically significant microbes (CSMs) when its coverage rate scored tenfold greater than that of any other microbes according to Miao’s study [11]. (2) Fungi: A microbe was considered CSMs when its coverage rate scored fivefold greater than that of any other fungus due to its low biomass in DNA extraction [12, 13]. (3) Mycobacteria: MTB was defined as positive when at least one read was mapped to either the species or genus level because of the low possibility for contamination and difficulty of DNA extraction [14]. Nontuberculous mycobacteria (NTM) were considered when the mapping read number was in the top 10 in the bacteria list because of the low possibility for contamination [15] and low yield rate [16]. Regardless of coverage rate, oral commensals were not defined as CSMs unless they were deemed to be significant by the physicians or proven otherwise [6]. Coverage rate was regarded as the measurement parameter in our study because of the consideration of confounding factors such as pathogen genome size, nucleic acid contamination and total number of sequencing reads [12].
Clinical composite diagnosis as the reference standard
The medical records such as clinical features, laboratory examination, microbiological tests (including mNGS and CMTs), chest imaging and therapeutic response, were reviewed independently by the two physicians who specialize in the management of infection in immunocompromised hosts to determine whether the patients had infectious etiology or not and identify the pathogens (definite or probable). An in-depth discussion was performed when there was any disagreement between the two physicians and another senior physician was consulted if consensus could not be reached.
Statistical analysis
SPSS 22.0 (IBM Corporation) was used for analysis. Continuous variables were reported as the mean ± standard deviation (SD) or the median (25th, 75th percentiles) depends on they were normally distributed or non-normal distribution. Categorical variables were presented as numbers (percentages). Determination of microbiological etiology and clinical composite diagnosis were used as the reference standard. The chi-square test, McNemar’s test or Fisher’s exact test was used to compare the diagnostic performance of mNGS and CMTs. Test concordance was assessed using the kappa (κ) statistic. All tests were two-tailed and P < 0.05 was considered significant.