Evaluation of Cell-Free DNA-Based Next-Generation Sequencing for Identifying Pathogens in Bacteremia Patients

Bacteremia can result in many serious diseases, such as sepsis, septic shock, and multi-organ failure (Laupland et al. 2004). Rapid detection of causative pathogens will be clinically helpful in optimizing accurate antimicrobial therapy in patients with suspected bacteremia. Conventional culture-based diagnostic procedures (e.g., blood cultures) currently remain the most reliable and clinically commonly used method for pathogens detection in patients with suspected bacteremia. However, the limitations of the conventional culture-based strategies for microbial tests in clinical practice have been widely discussed (e.g., it takes over 24 h to get results). Thus, several culture-independent molecular diagnostic procedures (such as polymerase chain reaction (PCR)-based techniques) have been introduced in clinics to identify the causative pathogens in patients with infectious diseases (Rutanga and Nyirahabimana 2016). Mounting evidence has demonstrated the diagnostic values of PCR-based techniques as additional tools to the conventional culture-based methods for identifying some known pathogens of patients with bacteremia and other infectious diseases (Blauwkamp et al. 2019).

Several recent studies have demonstrated the diagnostic values of cell-free DNA metagenomics next-generation sequencing (cfDNA mNGS) as a rapid and accurate method for identifying pathogens in patients with infectious diseases (Simner et al. 2018; Chiu and Miller 2019). The performance of the cfDNA mNGS for the identifying infectious microorganisms is based on the unbiased sequence analyses of the microbiome from the patient’s plasma (human DNA is removed) (Grumaz et al. 2016; Long et al. 2016). The conventional bacterial culture methods are commonly used as a gold standard for identifying bloodstream pathogens in patients with suspected bacteremia. Comparing diagnostic values of cfDNA mNGS and the conventional culture-based methods is seldomly reported in suspected bacteremia patients. In this study, the clinical values of the cfDNA mNGS were compared with the conventional culture-based methods. Moreover, the utility of cfDNA mNGS in suspected bacteremia patients was further evaluated as an alternative detection method.

Clinical information and cfDNA quantification from bacteremia patients’ plasma or serum samples. The microbes-derivative read numbers of the cfDNA extracted from bacteremia patients’ plasma and serum samples were compared. It suggested that more microbes-derivative read numbers could be detected in plasma compared to serum ones. It demonstrated that the plasma was more suitable for detecting cfDNA mNGS (Fig. 1a). Subsequently, 114 patients were divided into two groups according to the blood culture results and the criteria described above. Fifty patients were diagnosed with bacteremia (BC⁺ group) and 64 with suspected bacteremia (BC^– group), and the diagnostic value of cfDNA mNGS in these groups was evaluated. The total cfDNA read numbers/concentrations (copies/ml) by cfDNA mNGS were significantly higher in the plasma samples of the BC⁺ group compared with the BC^– group (Fig. 1b). The microbes-derivative cfDNA reads/total cfDNA reads showed no significant differences between these two groups (Fig. 1c).

Fig. 1

Comparison of the diagnostic value between the conventional blood culture and cfDNA mNGS. The diagnostic values of pathogen detection by cfDNA mNGS and conventional blood culture were compared and are presented in Table I. In this study, 76% of the BC⁺ group (38/50) were positively recognized by cfDNA mNGS, indicating that 72% (36/50) were consistent with the BC results, and only two cases were mismatched between these two methods. Moreover, among the 64 cfDNA mNGS⁺ samples, 26 were BC^–, and 80.7% (21/26) of cfDNA mNGS⁺/BC^– samples were consistent with the clinical manifestation. This observation indicated that the pathogens in BC^– samples were reliable determined by cfDNA mNGS. Thus, to promise the accuracy of cfDNA mNGS results in BC^– groups, performing analysis combined with the clinical manifestation is necessary.

In this study, Escherichia coli and Klebsiella pneumoniae were the most detected pathogens in identified bacteremia patients by cfDNA mNGS (Fig. 2). cfDNA mNGS detected 76.2% (16/21) of E. coli and 92.3% (12/13) of K. pneumoniae in the patients in whom the bacteriemia was detected by conventional blood culture (Table SII).

Fig. 2

In addition to bacterial pathogens, cfDNA mNGS indicated that 54.39% of the blood samples (62/114) came from patients with coexisting viral infections (Table SIII). The Torque Teno virus (TTV) and human herpesvirus 4 (EBV) were the top two species detected in this study. EBV had a significantly higher detection rate in the BC⁺ samples than in the BC^– group. Notably, most viral read numbers were < 20, and the relative abundance of the hepatitis B virus (HBV) in the BC^– group was over 1,000, which nearly equaled the viral copy number determined by quantitative fluorescent PCR (Fig. 2).

The genomic mapping of the microbes detected by cfDNA mNGS in BC⁺ and BC^– groups. Furthermore, the whole genome of individual microbial species, such as E. coli and K. pneumoniae were mapped from the cfDNA reads detected by mNGS, and compared between the BC⁺ and BC^– groups. This study demonstrated that genomic coverage of E. coli and K. pneumoniae was most often detected in BC⁺ group patients, and genomic coverage of Acinetobacter johnsonii and Paucibacter sp. KCTC 42545 was more often detected in BC^– patients (Fig. 3).

Fig. 3

Proteobacteria was the most common phylum in both BC⁺ and BC^– patients (Fig. 4a). Gammaproteobacteria had the highest relative abundance ratio at the class level, and was more abundant in the BC⁺ patients (p = 0.03), whereas Alphaproteobacteria (p < 0.001) and Betaproteobacteria (p < 0.001) were more abundant in the BC^– samples (Fig. 4b). At the order level, Enterobacterales (p < 0.001) was significantly abundant in the BC⁺ samples, whereas Rhizobiales (p = 0.001) and Pseudomonadales (p = 0.01) were more abundant in the BC^– samples. At the family level, Enterobacteriaceae (p < 0.001) was significantly abundant in the BC⁺ samples, whereas Moraxellaceae (p = 0.01) and Caulobacteraceae (p < 0.001) were more abundant in the BC^– group. At the genus level, Escherichia (p < 0.001) was relatively higher in the BC⁺ samples, whereas Acinetobacter (p = 0.01) was relatively higher in the BC^– samples (Fig. 5).

Fig. 4

Fig. 5

Dynamics of microbial derivative cfDNA read numbers determined by mNGS during antibiotic treatment. The read numbers of cfDNA from the plasma samples collected from three patients at different time points were analyzed to investigate the dynamic changes of the microbial derivative cfDNA read numbers during antimicrobial treatment (Fig. 6). In patients of A47 and A45, the cfDNA read numbers of K. pneumoniae and E. coli significantly decreased after of the effective antimicrobial treatment. It showed that the dynamics of the bacterial read numbers in these two cases reflected the efficacy of the antibiotic treatment. However, the cfDNA read numbers of Staphylococcus aureus in patient A48 remained positive with a transient increase of cfDNA read number after eight days of antibiotic treatment with remission of clinical symptoms. Subsequently, with persistent antibiotic treatment, the S. aureus cfDNA read numbers decreased and became negative (Fig. 6). With the decrease of cfDNA read numbers, the fever symptoms gradually improved, and while blood cells, C-reactive protein, and procalcitonin were gradually reduced (Fig. 7).

Fig. 6

Fig. 7

An ideal diagnostic test would enable the simultaneous detection and identification of pathogens from various clinical specimens, including rare or unculturable ones. cfDNA-based mNGS has been widely reported as a promising diagnostic tool for detecting microorganism from various clinical samples, such as bronchoalveolar lavage fluid and cerebrospinal fluid (Horiba et al. 2018; Miao et al. 2018; Miller et al. 2019). However, the evaluation of the diagnostic values of cfDNA mNGS for microbial testing of bacteremia patients’ blood samples has rarely been reported. Here, our data showed that cfDNA mNGS could detect pathogens in plasma samples independently of BC tests, demonstrating that 72% of the mNGS-positive samples were consistent with the gold standard approach of BC.

One of the most urgent needs in clinics is to evaluate the clinical significance of cfDNA mNGS for determining microbial species in the BC^– group. i.e., patients with suspected bacteremia (Horiba et al. 2018). Our data indicated that cfDNA mNGS yielded accurate microbial results in 32.8% of the BC^– group of patients with suspected bacteremia, supporting the future application of this approach in clinics. Although the cost of cfDNA NGS is five to six times that of conventional blood culture method, it takes only 24 hours from blood sample collection to NGS detection, bioinformatics analysis, and final results. Especially for those patients with severe bloodstream infections, it might be helpful to obtain a clinical prognosis in the shortest time and start timely anti-infection treatment.

Previous studies have indicated that herpes viruses (cytomegalovirus, Epstein-Barr virus, herpes simplex virus types 1, and human herpesviruse types 6), polyomaviruses (JC and BK), and an anellovirus (Torque Teno virus) are significant and commonly found in plasma samples of septic patients (Béland et al. 2014; Walton et al. 2014). Whether the increased propensity for infections with relatively weakly pathogenic organisms results from viral-mediated effects to impair immunity and whether viral reactivation occurs more readily in more profoundly immunosuppressed septic patients remains unknown. Applying cfDNA-based NGS may facilitate monitoring serum viral titers and identifying viral reactivation or flare in serum samples of patients with identified or suspected bacteremia. Therefore, discovering occult DNA viruses in host plasma samples may be an advantage in the future.

The antimicrobial susceptibility of traditional microbial culture depends on the microbial isolation and identification. Some molecular diagnostic methods, such as PCR, can both analyze the microbial species and determine the antimicrobial-resistance genes to provide the additional guidance for the antibiotic treatments (Gu et al. 2019). Previous researches have indicated that mNGS detection of clinical samples can reveal information about both the microbial species and the antimicrobial gene distribution, which can facilitate undertaking treatment choices for bacteremia patients (Grumaz et al. 2016; Horiba et al. 2018; Charalampous et al. 2019). In this study, blood culture results showed that these bacteria were not resistant to commonly used antibiotics, so we did not conduct in-depth analysis of resistance genes in cfDNA mNGS sequencing results. This study was probably limited by its single-center nature, and small number of patients. Thus, to some extent, it may affect the accuracy and reliability of the results.

In conclusion, cfDNA mNGS may be an alternative detection method for pathogen identification in the patients with bacteremia. Applying cfDNA mNGS, especially combined with BC and clinical manifestation, can significantly enhance the detection rate and accuracy of cfDNA microbial detection in patients with bacteremia. High read numbers of microbial cfDNA will facilitate using cfDNA mNGS to determine the antimicrobial susceptibility and dynamical detection of the cfDNA-derivative microbial read numbers. The achieve the accurate diagnosis by cfDNA mNGS one should avoid the operational contamination and exclude the interference from the extra or occult microbial species. Additional clinical trials are needed to determine the sensitivity and specificity of this NGS method.