4. Discussions
Currently, culture-based techniques that routinely employed to isolate pulmonary pathogens often using selective culture media designed for specific groups of microorganisms. The culture conditions are biased towards known, previously encountered microorganisms, and some novel and rare microorganisms might be missed. So the detection rate of microorganisms in conventional culture is low. Miao et al.’s study reported the sensitivity of culture for diagnosing infectious disease was 35.2% [22]. Jain et. al reported that no bacterial pathogen was ever isolated by culture in up to 75% of pneumonia cases [4], which was similar to the culture-negative rate seen in another study [3]. The present study found that the culture-positive rate for the diagnosis of lung infections was even lower, as low as 16.3%. The possible reasons for the lower culture-positive rate are as follows, first of all, all of the patients in the present study received empiric broad spectrum antibiotics prior to sample collection. Detection of microorganisms from routine culture is limited due to the early administration of broad-spectrum or prophylactic antimicrobial drugs that could have been sufficient to affect culture results but not to eradicate infection [23, 27]. Secondly, 74% patients were immune-impaired in the present study, microorganisms infecting the immune-impaired host can be fastidious to grow or non-cultivable [3, 28, 29].
The mNGS allows for unbiased detection of virtually any pathogen present in a given sample through direct sequencing of the sample’s extracted DNA [30, 31]. We conducted a literature review and found that there are currently several studies on the use of mNGS for the diagnosis of pulmonary infections. Although mNGS had different sensitivities for detecting pathogens in these studies, and the gold standards used to calculate sensitivity vary, some were based traditional methods [19 ~ 21] and some were based on the final diagnosis [22, 23] as the gold standard. However, a similar conclusion is reached that mNGS is more sensitive and more advantageous than traditional methods in identifying pathogenic microorganisms. Li et al. applied mNGS to detect the microbial pathogens in CT-guided puncture lung biopsy tissues, they reported that the sensitivity and specificity were 100.0% and 76.5% for bacteria, 57.1% and 61.5% for fungi when compared to culture [19]. Langelier et al.’s study applied mNGS of BALF to detect microbial pathogens in hematopoietic cell transplant patients with acute respiratory illnesses, which reported the sensitivity of mNGS for detecting respiratory microbes (human metapneumovirus, respiratory syncytial virus, Stenotrophomonas maltophilia, human herpesvirus 6 and cytomegalovirus) was 100% when compared to standard testing [20]. Zhang et al.’s study reported 13 cases of Pneumocystis pneumonia (PCP) identified through mNGS of BALF or sputum or blood. Pneumocystis jirovecii was detected by mNGS in all samples and by conventional methods in 5 out of 13 samples, respectively. They concluded that mNGS showed satisfying Pneumocystis pneumonia detection rate compared to conventional methods [21]. Pan et al.’s study explored the application of mNGS of BALF in the microbiologies diagnosis of community acquired pneumonia in immune-impaired patients. They reported standard procedures identified pathogens in 6 out of 13 patients, while mNGS detected pathogens in 12 out of 13 patients [32]. Miao et al.’s study reported the sensitivity of mNGS for diagnosing infectious disease was 50.7%. However, in their study, the study was conducted on patients with various types of infectious diseases, not just lung infections and specimens were not limited to respiratory specimens [22]. In the present study, the gold standard used to calculate sensitivity and specificity was the final diagnosis. We found that the sensitivity of mNGS of BALF to detect pathogenic microorganisms was significantly higher than that of traditional culture of BALF, regardless of bacterial pneumonia (85.7% versus 42.9%), fungal pneumonia (71.4 % versus 4.8%), or generalized pneumonia (81.4% versus 16.3%).
It is worth noting that this study demonstrates that the advantages of mNGS in the field of fungus testing are more prominent. Culture indentified pathogens in only one fungal pneumonia patients (1/21), while mNGS detected pathogens in 19 fungal pneumonia patients (19/21). The two patients missed by mNGS were a patient with cryptococcal pneumonia and a patient with Aspergillus pneumonia. The culture and GM test of the patients with Aspergillus pneumonia missed by mNGS happen to be positive. Although the culture-positive rate is low, it is very necessary to combine culture, GM test and mNGS for the diagnosis of fungal pneumonia, so as to avoid missed diagnosis to the greatest extent. It should be emphasized that mNGS is not omnipotent. For cryptococcal pneumonia, mNGS does not have an advantage. As shown in this study, the capsular polysaccharide antigen was positive in all four cases of cryptococcal pneumonia, but one case was missed by mNGS. ROSE also played an important role in the diagnosis of cryptococcal pneumonia. We saw granulomas in TBLB specimens of all 4 patients by ROSE for the first time, and Cryptococcus was found in TBLB specimens in 3 of them by ROSE. Therefore, our recommendation for the diagnosis of cryptococcal pneumonia is that further detection of capsular polysaccharide antigen is of great significance, while mNGS is not necessary, if cryptococcal pneumonia is highly suspected based on the patient’s exposure history, clinical manifestations, imaging findings and ROSE results.
In terms of detecting bacteria, mNGS still has advantages over culture, although this advantage is not as prominent in fungal detection. Pseudomonas aeruginosa, Klebsiella pneumoniae and Escherichia coli are easily detected by culture. However, some bacterial pathogens detected by mNGS, including Nocardia, Mycobacterium tuberculosis, Pneumocystis jirovecii, Haemophilus parainfluenzae, Pyramidobacter piscolens and Prevotella, were not detected by culture. These bacteria are either fastidious microbes (such as anaerobe) or require long incubation times (such as Mycobacterium tuberculosis). In P4, two obligate anaerobic bacteria (Pyramidobacter piscolens and Prevotella) were identified, in P25, Pyramidobacter piscolens were identified. Pyramidobacter piscolens and Prevotella were usually isolated from the oral cavity of patients with dental pulp disease, periodontal infection or alveolar abscess and healthy people. They may be a potential pathogen of pulp disease and periodontal disease [33, 34]. Study also reported that Prevotella induces severe bacteremic pneumococcal pneumonia in mice [35].
In the present study, 74% patients were immune-impaired and 23% of patients with pulmonary infection were confirmed to be mixed infections by mNGS. Conventional culture is powerless in identifying mixed infections, while mNGS exhibited its remarkable advantages in detecting pathogens of mixed pulmonary infections in immune-impaired patients. Therefore, mNGS might be more likely to benefit immune-impaired patients who are susceptible population of various pathogens.
Using mNGS, we further explored the characteristics of flora composition in the three respiratory specimens (including the relative abundance of pathogens and the relative abundance and richness of common oropharyngeal microbiota). Surprisingly, we found that TBLB and BB samples were similar in flora composition except for the richness of common oropharyngeal microbiota, while the relative abundance and richness of common oropharyngeal microbiota in BALF were higher than TBLB.
In our study, in comparison to BALB and BALF, BB gave a higher number of true positive, in comparison to BALF, BB gave much lesser number of false positive and false negative cases, showing its superiority in diagnosing infective PPLs. So regardless of the diagnostic efficacy, or the relative abundance of the pathogenic microorganisms and the contamination of common oropharyngeal microbiota, BB is no worse than TBLB. In the present study, BB yielded a better diagnostic performance most likely because it allowed cells and microorganisms to be collected from a larger area. This suggests that in future, high throughput sequencing of the BB samples might be an alternative choice for patients with infectious PPLs but couldn’t tolerate more invasive TBLB procedures, such as patients with hematological diseases who cannot tolerate TBLB because of thrombocytopenia or poor platelet function.
The current study has several limitations: the first limitation is a small sample size, the number of patients who had mNGS in all three samples is limited due to the relatively expensive cost of mNGS sequencing. The second limitation is using a retrospective design. The third limitation is that this study did not make a statistical comparison of specificity differences. Because ROSE was performed before the mNGS specimens were taken, if the clinical information, CT results and ROSE performance were all indicative of infectious lesions, then the sample was further sampled for mNGS, so the number of negative cases was small. In this case, the calculated specificity is not reliable, so the specificity is not compared between groups.