Sequencing methods to study the microbiome with antibiotic resistance genes in patients with pulmonary infections

doi:10.21203/rs.3.rs-3142484/v1

Download PDF

Research Article

Sequencing methods to study the microbiome with antibiotic resistance genes in patients with pulmonary infections

https://doi.org/10.21203/rs.3.rs-3142484/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Various antibiotic resistant bacteria are known to induce repeated pulmonary infections and increase morbidity and mortality. A thorough knowledge of the spectrum of bacteria with antibiotic resistance genes (ARGs) can improve the antibiotic treatment efficiency. In this study, we induced metagenomic next-generation sequencing (mNGS) alignment and assembly methods in the bioinformatics analysis pipeline to reveal the profile of bacteria with ARGs (ARB) in samples from patients with pulmonary infections.

Methods

A retrospective analysis of 151 clinical samples from 144 patients with pulmonary infections was undertaken by mNGS and conventional microbiological detection methods. Positive ARB were determined according to the analysis results detected both by the alignment and assembly methods. Co-occurrence analysis of ARG-ARB network was conducted to investigate the attributions between ARGs and microbial taxa. We also evaluated the potential application conditions to predict ARGs using those two approaches.

Results

Compared to that using conventional detection methods, the false-positive detection rate of ARB was significantly higher using mNGS alignment method. The assembly method could assist the determining of the detected pathogens by the alignment method as true ARB and improve the predictive capabilities (46% > 13%). ARG-ARB network revealed the main ARGs in predominant ARB. A total of 361 ARGs were detected, which mostly belonged to the multidrug class and β-lactam antibiotic classes. Specifically, 101 ARGs (existing in two approaches) and 34 ARGs (detected only by assembly method) achieved a clear ARG-bacteria attribution and potentially could optimize the reference antibiotic resistance database. The most prevalent ARB and its corresponding ARG and drug classes were as follows in this study: Acinetobacter baumannii (ADE, multidrug), Pseudomonas aeruginosa (MEX, multidrug), Klebsiella pneumoniae (MDT, aminocoumarin; EMR, fluoroquinolone), Stenotrophomonas maltophilia (SME, multidrug) and Corynebacterium striatum (carA, MLSB).

Conclusion

Collectively, our findings demonstrated the applicability of mNGS alignment and assembly as antibiotic resistant diagnostic methods and uncovered pulmonary infection-associated ARB and ARGs, potentially, as antibiotic treatment targets for the pulmonary infection.

Pulmonary infections

Metagenomic next-generation sequencing

Alignment and assembly methods

Antibiotic resistance genes (ARGs)

Bacteria with ARGs (ARB)

Pulmonary infection is one of the most frequent nosocomial infections associated with significant morbidity and mortality [1, 2]. Substantial evidence supports that inappropriate antibacterial treatment can increase the mortality of pulmonary infection patients [3, 4]. At present, conventional microbial cultures are difficult to satisfy the needs of clinicians, because of the limitations of detection rates, speeds, and available assay targets [5]. Clinicians often prescribe antibiotics empirically to patients with negative etiological test results, which can exacerbate a reinfection and drive the emergence of antimicrobial resistance and multidrug-resistant pathogens [6]. Due to the above reasons, it is necessary to identify a fast and effective method to detect antibiotic resistant bacteria in pulmonary infection patients [7].

Advancements in genome sequencing technologies such as metagenomic next-generation sequencing (mNGS), provide powerful options for overcoming such clinical diagnostic challenges [8]. mNGS, as a revolutionary diagnostic tool, can facilitate rapid identification of potential pathogens and antibiotic resistance genes (ARGs) directly from clinical samples [9]. Previous study highlighted mNGS can be an important complement for traditional culture methods for clinical and surveillance applications and provides opportunities for quick and sensitive resistance determinations [10]. However, there are still many challenges preventing its use in clinical applications. One of the primary challenges is that there is no unified standard for interpreting mNGS results including the judgement of false-positive prediction and ARG-bacteria attribution ( an unknown ARG assigned to a particular bacteria) [11]. Therefore, major efforts have been made to explore the identification of the bacteria with ARGs (ARB) by mNGS assay among pulmonary infection patients.

In recent years, the bioinformatics approaches including mNGS alignment and assembly methods have been studied for detecting genetic determinants of antibiotic resistance [12, 13]. Alignment method enable the identification of ARGs from low-abundance organisms present in complex communities, which may be missed by assembly methods owing to incomplete or poor assemblies [14]. Therefore, it had gained traction in recent years, especially in clinical diagnostics where conducting real-time sequencing-based resistance prediction is crucial [15]. However, mapping reads directly to large datasets was likely to be with false-positive predictions, as reads derived from protein coding sequences may spuriously align to other genes as a result of local sequence homology [11, 16]. Meanwhile, alignment method depend largely on curated antibiotic resistance databases that are comprehensive and contain all variants of the reference genes information [17]. However, the assembly method, according to the contigs facilitates, can identify ARGs that are more divergent from and lack homology to known sequences in the reference databases [11]. Thus, to some extent, it potentially could assist the alignment method to determine the positive predictions and address the issue of ARG-bacteria attribution. Given the above, we explored the performance of ARGs detection through a combination of assembly and alignment method using our in-house sequencing data.

In this study, firstly, we compared the alignment and assembly methods with culture test for studying ARB from the clinical pulmonary infection samples. And then, we performed a comprehensive analysis of mNGS alignment and assembly methods to identify the distribution of ARGs and ARB and in clinical samples. Finally, the two approaches are used for comprehensive evaluation, including computational strategies and resources for resistance gene identification in metagenomic samples. Our study demonstrates that comprehensive application of two approaches perform better assess the antibiotic resistance than the conventional methods. The potential ARG-ARB network was also analyzed to provide bioinformatics-oriented evidence for antibiotic drug selection. Collectively, our study provided a profile of the predominant ARB and ARGs from pulmonary infection patients by combination with alignment and assembly method to improve clinical diagnostic effectiveness and potentially better treat these infections.

Sample preparation

In this retrospective study, we consecutively enrolled 151 clinical specimens from 144 patients with pulmonary infections, including bronchoalveolar lavage fluid (BALF) and sputum samples, were collected for clinical mNGS testing at Huayin Medical Center (Guangzhou, China) from January 2021 to March 2022. All patients were diagnosed with pulmonary infections judged by the board-certified physician on the basis of clinical manifestations and diagnostic criteria. Basic information, including age, sex, department, and comorbidities, was extracted from their medical records.

mNGS sequencing

Clinical samples were collected from each patient by following the standards of aseptic processing procedures [18]. (1) Nucleic acid extraction was performed by using a TIANamp Micro DNA Kit (DP316, Tiangen Biotech Co., Beijing, China). (2) DNA libraries were constructed using the VAHTS Universal Plus DNA Library Prep Kit for Illumina (ND617-C2, Vazyme Biotech Co., Nanjing, China). Agilent 2100 was used for quality control of the DNA libraries. Qualified libraries were sequenced by the Illumina NextSeq 550 platform. (3) High-quality sequencing data were generated by removing low-quality reads, followed by computational subtraction of human host sequences mapped to the human reference genome (hg38) using Kraken2 and BWA, and the low-quality reads and short reads (length < 50 bp) were removed by fastp 0.20.1. The remaining data were used to analyze pathogens or antibiotic resistance genes by alignment or assembly methods. The classification reference databases were downloaded from the National Center for Biotechnology Information (NCBI) (ftp://ftp.ncbi.nlm.nih.gov/genomes/).

Alignment method Antibiotic resistance genes in a sample can be detected by aligning reads to the reference databases using pairwise alignment tools such as Bowtie2 [19].

(i) For pathogens, the filtered data were aligned to the Pathogenic Microbial Genome Databases by Bowtie2. The filter conditions were unique reads > 50 and coverage length > 3000 bp.

(ii) For ARGs, the filtered data were aligned to the Comprehensive Antibiotic Resistance Database (CARD) by Bowtie2. The filter conditions used were as follows: unique reads > 1, unique ratio (Uratio) > 10%, unique read coverage ratio > 30%, and unique read coverage index (Uindex) > 30%.

$$\text{U}\text{r}\text{a}\text{t}\text{i}\text{o} = \left(\frac{\text{u}\text{N}}{\text{a}\text{N}}\right)\text{*} 100\text{%}$$

Uratio: unique read ratio (how many aligned reads are unique reads in X species)

uN: number of unique reads mapped to X species.

aN: number of all reads mapped to X species.

$$\text{U}\text{i}\text{n}\text{d}\text{e}\text{x} = \left(\frac{\text{s}\text{i}\text{z}\text{e}\text{R}}{\text{s}\text{i}\text{z}\text{e}\text{C}}\right)\text{*} 100\text{%}$$

sizeR: real coverage of unique reads.

sizeC: coverage of unique reads (when calculated coverage < genome size: sizeC = num of unique reads * 50 then coverage > = genome size: sizeC = genome size).

(iii) For ARB, when ARGs were detected in a given sample and the sample had one or more bacteria in the CARD (CARD Prevalence, Resistomes, & Variants data 3.1.0) for prevalence of the ARG, the bacteria were defined as ARB in that sample, and the converse was defined as non-ARB.

Assembly method A fundamental goal of DNA sequencing is to generate large, continuous regions of an identifiable gene DNA sequence. The desired DNA sequences are generally longer than the individual sequencing read lengths using some suitable mapping approaches. One of the most widely used approaches is the random fragmentation of DNA samples prior to sequencing, namely, ‘shotgun sequencing’. The process of computational reconstruction of the original DNA sequences is described as ‘assembly’.

(i) For pathogens, the filtered data were assembled by Megahit. The filtering conditions for contiguous fragments (contigs) were length > 200 bp and coverage > 2. Based on the contig long sequences, species annotation was performed by the Kraken2 and NCBI databases. We retained the results only of contig species that could be annotated to the species level using Kraken2, which specified sequence homology.

(ii) For ARGs, ARGs detected in contigs were processed by Abricate 0.8 software and the CARD. The filter condition was ≥ 75% identity and ≥ 50% coverage length for gene annotations.

(iii) For ARB, the contigs were annotated with species information when they were detected with ARGs. The bacteria with those contigs were defined as positive ARB in that sample.

Statistical analysis. SPSS Statistics 21.0 was used for clinical data analysis. Continuous measurement data that followed a normal distribution are shown as the means ± standard deviations (x ± s), and the Pearson chi-squared (χ2) test or Fisher’s exact test was used for the comparison of frequencies of categorical data. Data with a nonnormal distribution are shown as the medians (interquartile ranges). A P value of 0.05 was considered statistically significant.

Patient characteristics and clinical samples.

In this study, the clinical information of 144 patients with newly diagnosed pulmonary infections was retrospectively analyzed. The mean age of all patients was 67.1 years, ranging from 15 to 95. Among them, 77.0% of patients were male. 61.1% patients had underlying comorbidities, including respiratory failure (32, 22.2%), disorders of consciousness (14, 9.7%), cardiovascular disease (9, 6.3%), malignancies (11, 7.6%), and diabetes (6, 4.2%). The bacteria with ARGs (ARB) positive results were associated with the department (P = 0.001) but gender, age, antibiotic history and comorbidities (Table 1).

Table 1

Basic clinical characteristics of 144 patients with suspected pulmonary infections
Characteristics	N (%)	ARB detected (Alignment)		ARB detected (Assembly)
Characteristics	N (%)	Yes	P	Yes	P
Demographics
Age
Mean age (years)	66.6
<=50	27 (18.75)	25	0.172	7	0.894
> 50	117(81.25)	98	0.172	44	0.894
Sex
Male	108 (77)	93	0.817	42	0.119
Female	36 (25)	30	0.817	9	0.119
Antibiotic history
Yes	119(82.63)	98	0.694	43	0.889
No	25(17.36)	25		8
Department
ICU	94 (65.27)	76	0.001***	28	0.001***
PD	13 (9.03)	13		10
PCCM	23 (15.97)	21		3
Others	14 (9.72)	13		10
Comorbidities
Respiratory failure	32 (22.22)	10	0.633	4	0.149
Disorders of consciousness	14 (9.72)	10		4
Cardiovascular disease	9 (6.25)	8		5
Malignancies	11 (7.64)	10		7
Diabetes	6 (4.17)	5		3
Others	17 (11.80)	15		6
None	55 (38.19)	65		22
ICU: intensive care unit; PD: Pneumology Department; PCCM: Pneumology Critical Care Medicine. *** P < 0.001

A total of 151 clinical specimens were collected, which included BALF (n = 134, 88.7%) and sputum (n = 17, 11.3%). The number of effective reads ranged from 10³ to 10⁷ counts (Fig. 1A). Primary mNGS analysis were performed for 151 samples, of which 114 samples were regarded as ARB positive by the alignment method, and 52 samples were assessed positive by the assembly method (Table S1). To better identify the ARB, the mNGS assay 52 samples with positive ARGs by both approaches are used for comprehensive evaluation. And then, 14 culture positive samples were performed the consistency ratio analysis combined with the mNGS alignment and assembly analysis. Finally, the predominant ARB-ARGs and the reference application conditions were assessed (Fig. 1B).

Evaluating the performance of mNGS alignment and assembly analysis with conventional method for the identification of ARB

52 out of 151 samples with positive ARB by both mNGS alignment and mNGS assembly methods are used for comprehensive evaluation. A total of 165 ARB was obtained by these two approaches, of which 142 ARB detected by the alignment method (red bar) were much more than the 48 by the assembly method (blue bar), respectively (Fig. 2A). Among them, 25 ARB was obtained in both methods (Fig. 2B, Table S2 ). The alignment method detected much more ARB than the assembly method (142 > 48). However, many congeneric species in the same genus, such as Acinetobacter baumannii, Acinetobacter pittii and Acinetobacter nosocomialis were detected as ARB by alignment, only A. baumannii in Acinetobacter was detected as ARB by the assembly method. This may be caused by the false-positive ARB due to mapping reads directly to large data sets can inflate false-positive predictions, as reads derived from protein-coding sequences may spuriously align to other genes as a result of local sequence homology [11, 16]. In addition, it was found that substantial non-ARB (gray bar) evaluated by the mNGS assembly method were classified as ARB by the alignment method. This may be caused by the over-reliant on the reference antibiotic resistance database by the alignment method, which requires the existence of relationship between bacteria and its corresponding ARG (ARG-bacteria attribution). However, mNGS assembly method in longer sequence (contigs) has the advantages of facilitating more accurate judgment of the issue of ARG-bacteria attribution as described previously [11]. To evaluate whether the above susception were the major cause, we then compared the mNGS analysis and the culture test results.

We assessed the detection efficiency analysis of the mNGS method and the conventional method. Among the 52 patients, 14 of them showed the culture positive results (Table S3). Among the 14 double-positive patients’ samples, mNGS alignment method (87%, 36/42 vs. 14%, 6/42) and the assembly method (81%, 25/31 vs. 19%, 6/31) yields a higher positive detection efficiency for pathogens than compared with the culture method (Fig. 2C, culture, pathogens). The detected ARB of mNGS alignment in 12 patients and 85.7% matched with were the culture and drug susceptibility tests results (DST), and 64.2% (9/14) matching by the assembly method (Fig. 2C, *CD). In total, mNGS yields a higher positive detection efficiency for pathogens and ARB compared with the conventional method.

We further assessed the consistency analysis according to the number of times of detected pathogens to distinguish between true- and false-positive ARB. Comparing with the DST results, mNGS assembly method indicated consistency results with a matching accuracy of 46% (12/26), which was much higher than the mNGS alignment method with 13% (19/141) (Fig. 2D, 2E). Those results revealed that comparing to conventional detection methods, the false-positive detection rate of ARB was significantly higher using mNGS alignment method. The assembly method could assist the determining of the detected pathogens by the alignment method as true ARB and improve the predictive capabilities. Intrigued by our finding, we further performed ARGs evaluation through a combination of mNGS alignment and assembly methods to explore relationship between the bacteria and its corresponding ARG (ARG-bacteria attribution).

Analysis of the ARG-ARB network based on the alignment and assembly methods.

To gain an insight into the potential interplay of ARG-bacteria attribution, we performed an ARG-ARB network analysis by the mNGS alignment and assembly methods. A total of 361 ARGs were detected, of which 141 were detected by the assembly method and 352 were detected by the alignment method from 52 clinical samples. According to the antibiotic categories, these ARGs mainly comprised multidrug resistance (n = 91), macrolide-lincosamide-streptogramin B (MLSB) (n = 11), tetracycline (n = 14), aminoglycoside (n = 36), β-lactam (n = 145), aminocoumarin (n = 3), phenicol (n = 7), peptide (n = 9), fluoroquinolone (n = 11), disinfecting agents and antiseptics (n = 10) (Table S4), which mostly belonged to the multidrug class and β-lactam antibiotic classes.

Specifically, we found substantial associations between a ARG and multiple bacteria, especially complex network results displayed by the alignment method (Fig. 3, alignment). Nevertheless, the assembly method results showed a ARG was attributed a particular bacterium (Fig. 3, assembly). For example, among the 101ARGs existing in both two approaches, correlations between the ADE family (adea, adeb, adec, adef, adeg, adeh, adel, aden, ader, ades) and multiple bacteria (A. baumannii, K. pneumonia, P. aeruginosa and S. maltophilia), were attributed by the alignment method. However, they were only attributed to the A. baumannii by the assembly method. Similarly, the other genes including MEX and OPM were only attributed to the P. aeruginosa, were aligned to multiple bacteria (Table S5). In addition, there were 34 ARGs detected only by assembly method, including C. striatum (carA), Neisseria sicca (farB, mtrC, mtrD), Streptococcus mitis (patB, pmrA, RlmA (II)) (Table S5, blue font), which may play a role in improving the reference antibiotic resistance database. The results revealed that the interactions potentially could contribute to determining the ARG-bacteria attribution and optimizing the antibiotic resistance database, via the introduction of mNGS assembly method.

In line with our hypothesis, improved the determining of ARG-bacteria attribution was obtained by combining mNGS alignment method and assembly method, suggesting an additive detection value for the combination of those two approaches. Based on the common detected numbers of ARGs, we explore the most prevalent ARGs and drug class in its corresponding ARB. As follows: A. baumannii (ADE, multidrug), P. aeruginosa (Mex, multidrug), K. pneumonia (MDT, aminocoumarin; EMR, fluoroquinolone), S. maltophilia (SME, multidrug) and C. striatum (carA, MLSB) (Table S6, red/blue font). The ARB had high drug resistance to antibiotics and multidrug resistance. Multidrug resistance of ARB to antibiotics was also more severe in BALF and sputum samples (Fig. 3, bottom right plot).

Identification of the predominant ARB in pulmonary infection patient samples

Since combination with two approaches were more effective and accurate for ARB detection, we next aimed to identify the predominant ARB of the pulmonary infection patient samples. In this study, 48 predominant ARB were identified by both methods and only by the assembly method. The ARB and its corresponding drug resistance are also shown in Fig. 4. To identify a predominant ARB in the pulmonary infection patient samples, we choice features according to their ranking (detection frequency based the number of samples). First, pathogens detected high frequency by both method, that is, A. baumannii (75.0%, 48.0%), P. aeruginosa (65.4%, 11.5%), S. maltophilia (71.2%, 11.5%), K. pneumoniae (63.5%, 11.5%) were among the top four most predominant ARB.

Additionally, pathogens detected high frequency (> 50%) by the alignment method, including Pseudomonas fluorescens (59.6%), Staphylococcus aureus (57.7%), Enterococcus faecium (51.9%), and Staphylococcus epidermidis (51.9%) (Fig. 4, blue font) also contributed to the predictability of the ARB. Meanwhile, pathogens detected high frequency by the assembly method, such as C. striatum (38.5%), Streptococcus oralis (9.6%), N. sicca (9.6%), S. mitis (9.6%), were also identified as predominant ARB (Fig. 4, red font). In addition, we also identified the main antibiotic drug class according to CARD. Multidrug resistances were dominant among most common bacteria. In A. baumannii, P. aeruginosa and S. maltophilia, most ARGs were resistance-nodulation-division (RND)-type efflux pump genes, which conferred resistance to multiple drugs. Antimicrobial agents should be used rationally to decrease multiple antibiotic resistant strains. Altogether, this comprehensive analysis validated the popular ARB and its drug class in pulmonary infection patient samples from this study across a combination of the mNGS alignment and assembly methods.

Conditions of mNGS alignment and assembly methods in application for the clinical antibiotic resistant detection.

We compared the ARG positive rates of 151 clinical samples based on the alignment and assembly methods. In this study, the ARG positive rates referred to the ARGs detected in the sample alignment method and assembly method. The ARG positive rates were much higher by the alignment method than by the assembly method (75.5%, 114/151 vs. 34.4%, 52/151) (Fig. 5A).

We also analyzed the ARG positive rates among the samples with varied amounts of effective reads and different filtering conditions based on the two approaches. First, in the BALF and sputum samples, when the effective reads were approximately 10⁷ counts, the positive rate was 100% both the alignment and assembly methods. The ARG positive rate was higher with the alignment than in the assembly method, with approximately 10⁶ counts (BALF: 95.5% > 49.3%; sputum: 100% > 50%) and approximately 10⁵ counts (BALF: 71.4% > 7.1%; sputum: 66.7% > 33.3%), respectively (Fig. 5B). Herein, setting a cutoff over 10⁷ effective reads may have been more suitable for both methods to detect ARGs in BALF and sputum samples. Generally, the ARG positive rates relied more on sufficient effective reads (over 10⁷ counts) in these sample types.

In terms of filtering conditions, a gradual decrease in ARG positive rates by the alignment method was observed when the cutoff values of unique reads increased as the key filtering conditions. When unique reads = 1, the ARG positive rates of BALF and sputum samples were 77.6% and 58.8%, respectively. When unique reads = 8, it fells to 57.1% and 50.4%, respectively (Fig. 5C). Regarding effective reads, when unique reads = 1, the positive rates were 100%, 90%, 70% and 0% for the effective reads of 10⁷, 10⁶, 10⁵ and 10⁴ counts, respectively (Fig. 5D). Henceforth, unique reads = 1 could be introduced as the minimum filtering condition for ARG detection in the case of insufficient effective reads (≤ 10⁷).

Moreover, we calculated the CPU time occupied based on assembly and alignment. As shown in Fig. 5E, when the effective reads were approximately 10⁷ counts, the average times for alignment, assembly, and the combination of these two approaches were approximately 152.5 s, 567.8 s and 600 s, respectively. The data running time of the two approaches could still meet the clinical report requirement (48–72 h). Finally, we proposed a combined condition of unique reads ≥ 1 and effective reads ≥ 10⁷, which in most cases met the requirement of ARG detection for BALF and sputum samples.

Identifying and understanding antimicrobial resistance are imperative for clinical practice to treat resistant infections and for public health efforts to limit the spread of resistance. The detection of antibiotic resistance of bacteria by traditional methods is limited due to complexity and low sensitivity, as well as the presence of bacteria that are fastidious or slow growing[20]. mNGS analysis including alignment and assembly methods effectively overcomes the deficiencies of traditional detection methods and expand our abilities to detect and study antibiotic resistance. Presently, there is no consensus on which sequence analysis approach is better, and the choice of analysis mainly depends on the type of sequencing, availability of computational resources and the study objective[11]. Both approaches have trade-offs. The alignment method, enabling identification of antibiotic resistance genes from low-abundance organism, but was likely to be with false-positive predictions [17]. As shown in this study, the ARB detected by mNGS alignment method was significantly higher than that by the assembly method (142 > 48). However, it showed high false-positive predictions (Fig. 2). Additionally, the alignment method relies more on the comprehensive reference databases contain all variants of the reference genes with ARG-bacteria attribution. For example, in our study, some positive ARB, especially the 23 ARB (such as Neisseria sicca), evaluated by the alignment method were mistakenly classified as non-ARB, due to the absence of ARG-bacteria attribution in the reference database(Fig. 2B). Regarding with the assembly method, considering the higher requirements on the effective reads, the positive rate might be lower owing to incomplete or poor assemblies. For example, the P. aeruginosa and A.baumannii were assembled only when the effective reads more than 10⁵ counts (Table S3). However, the assembly method, corroding to the long sequences assembled more sequences to support the assembly results [21], potentially assist the alignment method to judge whether the detection results are valid or not. As the results shown, the ARB detection consistency by the assembly method was relatively higher than that by alignment method (46% >13%). Taken together, these findings suggest an important role for both alignment and assembly combinations for the antibiotic resistance analysis.

Previous studies showed many genes contributing to antibiotic resistance have been identified, and the high concordance of genotype-phenotype for drug resistance have been reflected along with the broader usage of mNGS technologies in evaluating the antibiotic resistance genetic determinants [22]. In our study, through the ARG-ARB network analysis(Fig. 3), it is interesting to note that the introduction of mNGS assembly method could contribute to determining the ARG-bacteria attribution. What’s more, there were many ARGs undetected by the mNGS alignment method for the absence of ARG and ARB relationships in the reference database, examples included C. striatum (carA) and N. sicca (mtrD). We verified the existence of mtrD in N. sicca by RT‒qPCR (Figure S1). Those ARGs were assembled and should be attributed to the corresponding ARB, which was meaningful for complementing the antibiotic resistance database and clinical antibiotic treatment. As detection methods move to mNGS techniques, our knowledge about which type of bacteria carry which resistance gene(s) will become more important to ensure that the whole spectrum of bacteria is included in future surveillance studies. Antibiotic resistance databases update should be a continuous effort, as all downstream analyses depend on the accuracy of reference databases[11]. Establishing best practices for systematically assigning annotations to newly discovered genes and preventing misinterpretations will pay dividends for public health and basic science.

An exponential increase of antibiotic drug resistant microbes, especially the multi-drug resistance (MDR) pathogens, is a serious challenge that encompass antibiotic therapy[23]. In this study, A. baumannii, P. aeruginosa, K. pneumoniae, were the predominant multidrug-resistant bacteria in the pulmonary infection samples (Fig. 4). The results mostly consistent with previous reports that MDR pathogens including E. faecium, S. aureus, K. pneumonia, P. aeruginosa, A. baumannii, Enterobacter species have been classified as ESKAPE and are the most common microbial causes of acquired infection including the pulmonary infection, and antibiotics resistance patterns are highly endemic[23, 24]. Thus, our ARB ranking analysis highlighted the need to combine methods from mNGS alignment and assembly methods for maximized predictive value. What’s more, ARB detected as predominant by the assembly method including C. striatum, Neisseria (N. sicca, N. meningitidis), Streptococcus (S. mitis, S. oralis, S. anginosus), etc, should be attracted attention (Fig. 4). Especially, emerging evidence indicates that C. striatum-infected individuals are at increased risk for serious infections in immunocompetent hosts and has been reported as an emerging multidrug-resistant nosocomial pathogen in recent studies [25, 26]. The mNGS alignment method only detected the cmx gene (phenicol resistance) in C. striatum and has been reported in several studies [27, 28], which was failed to detect carA and mtrA of C. striatum for the absence of ARG-bacteria attribution in the incomplete antibiotic resistance database. Therefore, our study innovatively introduced the mNGS assembly method combined with the alignment method, may represent a more effective strategy for the rapid and precise detection of ARB and ARGs in clinical samples. Multidrug resistance of these bacteria to antibiotics is becoming a serious problem. Targeted therapies for combatting multidrug-resistant bacteria in pulmonary infection patients should be promptly addressed. Our findings provide potentially antibiotic treatment targets for the pulmonary infection.

In addition, we also found that there is a large relationship between ARG detection efficiency and effective reads. In this study, effective reads more than 10⁷ counts maybe a good recommendation. However, there is need more enough samples including different types to verify the recommended effective reads range for ARG validity analysis in future studies. Future research should pay more attention to obtain more effective information (e.g., the effective ARG detection results, ARG-bacteria attribution) to promote the rapid ARGs detection in clinical samples. Additionally, The ARG detection efficiency is also greatly affected by the filtering threshold (unique reads) [29]. For example, the positive rate of the alignment method was 70.5% (unique reads = 1) and 10% (unique reads = 65) (Fig. 5) (Fig. 5). Therefore, there should be more research on the alignment method for ARG verification. It is also a problem that needs our further research to explore the threshold setting with a relatively fixed number of effective reads.

There were several limitations in this study. First, the retrospective study design did not allow for the use of PCR, complement-fixation testing, or micro-immunofluorescence to confirm all the mNGS results. Further study including drug resistance experiments, authoritative guidance about the application conditions for the alignment and assembly methods in interpreting the resistance results is required. Second, for limitations of sample volume, we still need more data to reflect the spectrum of antibiotic resistant determinants in the samples from patients with pulmonary infections. However, despite this, our study still provided a new perspective on the applicability of mNGS in antibiotic resistance detection.

The mNGS alignment and assembly methods are an efficient and useful diagnostic technology for antibiotic resistance in identifying bacteria with ARGs. Although many challenges remain necessary to overcome for its use in clinical applications, in the case of sufficient effective data, mNGS will be a revolutionary technology for clinical antibiotic resistance diagnostics including reducing the false-positive ARB prediction, making the clear ARG-bacteria attribution, and promoting the antibiotic resistance database.

mNGS metagenomic next generation sequencing

ARGs antibiotic resistance genes

ARB bacteria with antibiotic resistance genes

DST drug susceptibility tests

MLSB macrolide-lincosamide-streptogramin B

CARD Comprehensive Antibiotic Resistance Database

NCBI National Center for Biotechnology Information

RT-qPCR Real-time Quantitative polymerase chain reaction

Ethics approval and consent to participate.

This study was approved by the Ethics Committee of Huayin Medical Laboratory Center (2022-004-02).

Consent for publication

All the patients included provided written informed consent for the use of their samples in research.

Availability of data and materials

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at NCBI, PRJNA899589.

Competing interests

The authors declared that they had no competing financial interests.

Funding

This work was supported by the Key Research and Development Program of Jiangsu Province of China (BE2021707).

Authors' contributions

TD contributed to data interpretation and drafting of the article; YW, CQ, and WF collected the data; JX and HC performed the data analysis; HZ was responsible for reviewing the manuscript; XH and XW were responsible for the conception, design, and critical revisions of the article; and all authors contributed to the article and approved the submitted version.

Acknowledgements

The authors thank all patients and their families who participated in this study. They also thank Dr. Yuhua Ye for his helpful suggestions.

Qian YY, Wang HY, Zhou Y, Zhang HC, Zhu YM, Zhou X, Ying Y, Cui P, Wu HL, Zhang WH et al: Improving Pulmonary Infection Diagnosis with Metagenomic Next Generation Sequencing. Front Cell Infect Microbiol 2020, 10:567615.
Zheng Y, Qiu X, Wang T, Zhang J: The Diagnostic Value of Metagenomic Next-Generation Sequencing in Lower Respiratory Tract Infection. Front Cell Infect Microbiol 2021, 11:694756.
Iregui M, Ward S, Sherman G, Fraser VJ, Kollef MH: Clinical importance of delays in the initiation of appropriate antibiotic treatment for ventilator-associated pneumonia. Chest 2002, 122(1):262-268.
Ibrahim EH, Sherman G, Ward S, Fraser VJ, Kollef MH: The influence of inadequate antimicrobial treatment of bloodstream infections on patient outcomes in the ICU setting. Chest 2000, 118(1):146-155.
Zhang P, Chen Y, Li S, Li C, Zhang S, Zheng W, Chen Y, Ma J, Zhang X, Huang Y et al: Metagenomic next-generation sequencing for the clinical diagnosis and prognosis of acute respiratory distress syndrome caused by severe pneumonia: a retrospective study. PeerJ 2020, 8:e9623.
Ruppe E, Baud D, Schicklin S, Guigon G, Schrenzel J: Clinical metagenomics for the management of hospital- and healthcare-acquired pneumonia. Future Microbiol 2016, 11(3):427-439.
Li ZJ, Zhang HY, Ren LL, Lu QB, Ren X, Zhang CH, Wang YF, Lin SH, Zhang XA, Li J et al: Etiological and epidemiological features of acute respiratory infections in China. Nat Commun 2021, 12(1):5026.
Li N, Cai Q, Miao Q, Song Z, Fang Y, Hu B: High-Throughput Metagenomics for Identification of Pathogens in the Clinical Settings. Small Methods 2021, 5(1):2000792.
de Abreu VAC, Perdigao J, Almeida S: Metagenomic Approaches to Analyze Antimicrobial Resistance: An Overview. Front Genet 2020, 11:575592.
Sun T, Wu X, Cai Y, Zhai T, Huang L, Zhang Y, Zhan Q: Metagenomic Next-Generation Sequencing for Pathogenic Diagnosis and Antibiotic Management of Severe Community-Acquired Pneumonia in Immunocompromised Adults. Front Cell Infect Microbiol 2021, 11:661589.
Boolchandani M, D'Souza AW, Dantas G: Sequencing-based methods and resources to study antimicrobial resistance. Nat Rev Genet 2019, 20(6):356-370.
Flicek P, Birney E: Sense from sequence reads: methods for alignment and assembly. Nat Methods 2009, 6(11 Suppl):S6-S12.
Chandrakumar I, Gauthier NPG, Nelson C, Bonsall MB, Locher K, Charles M, MacDonald C, Krajden M, Manges AR, Chorlton SD: BugSplit enables genome-resolved metagenomics through highly accurate taxonomic binning of metagenomic assemblies. Commun Biol 2022, 5(1):151.
Langmead B, Trapnell C, Pop M, Salzberg SL: Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 2009, 10(3):R25.
Chen H, Bai X, Gao Y, Liu W, Yao X, Wang J: Profile of Bacteria with ARGs Among Real-World Samples from ICU Admission Patients with Pulmonary Infection Revealed by Metagenomic NGS. Infect Drug Resist 2021, 14:4993-5004.
Andreas Bremges ACM: Critical Assessment of Metagenome Interpretation Enters the Second Round. mSystems 2018 Jul 10, 3(4):e00103-18.
Carr R, Borenstein E: Comparative analysis of functional metagenomic annotation and the mappability of short reads. PLoS One 2014, 9(8):e105776.
Liang Y, Dong T, Li M, Zhang P, Wei X, Chen H, Wang Y, Gao X: Clinical diagnosis and etiology of patients with Chlamydia psittaci pneumonia based on metagenomic next-generation sequencing. Front Cell Infect Microbiol 2022, 12:1006117.
Langmead B, Salzberg SL: Fast gapped-read alignment with Bowtie 2. Nat Methods 2012, 9(4):357-359.
Jie Yuan, 2,* Weiqi Li,3,* Enming Qiu,1 Shuai Han,1 and Zhou Li1: Metagenomic NGS optimizes the use of antibiotics in appendicitis patients: bacterial culture is not suitable as the only guidance. Am J Transl Res 2021, v.13(4); 2021
Jay S Ghurye VC-E, Mihai Pop: Metagenomic Assembly: Overview, Challenges and Applications. Yale J Biol Med 2016 Sep 30, 89(3):353-362.
Han D, Li R, Shi J, Tan P, Zhang R, Li J: Liquid biopsy for infectious diseases: a focus on microbial cell-free DNA sequencing. Theranostics 2020, 10(12):5501-5513.
Moore LS, Freeman R, Gilchrist MJ, Gharbi M, Brannigan ET, Donaldson H, Livermore DM, Holmes AH: Homogeneity of antimicrobial policy, yet heterogeneity of antimicrobial resistance: antimicrobial non-susceptibility among 108,717 clinical isolates from primary, secondary and tertiary care patients in London. J Antimicrob Chemother 2014, 69(12):3409-3422.
Boucher HW, Talbot GH, Bradley JS, Edwards JE, Gilbert D, Rice LB, Scheld M, Spellberg B, Bartlett J: Bad bugs, no drugs: no ESKAPE! An update from the Infectious Diseases Society of America. Clin Infect Dis 2009, 48(1):1-12.
Wang X, Zhou H, Chen D, Du P, Lan R, Qiu X, Hou X, Liu Z, Sun L, Xu S et al: Whole-Genome Sequencing Reveals a Prolonged and Persistent Intrahospital Transmission of Corynebacterium striatum, an Emerging Multidrug-Resistant Pathogen. J Clin Microbiol 2019, 57(9).
Asgin N, Otlu B: Antimicrobial Resistance and Molecular Epidemiology of Corynebacterium striatum Isolated in a Tertiary Hospital in Turkey. Pathogens 2020, 9(2).
Shariff M, Aditi A, Beri K: Corynebacterium striatum: an emerging respiratory pathogen. J Infect Dev Ctries 2018, 12(7):581-586.
Ramos JN, Souza C, Faria YV, da Silva EC, Veras JFC, Baio PVP, Seabra SH, de Oliveira Moreira L, Hirata Junior R, Mattos-Guaraldi AL et al: Bloodstream and catheter-related infections due to different clones of multidrug-resistant and biofilm producer Corynebacterium striatum. BMC Infect Dis 2019, 19(1):672.
Hunt M, Mather AE, Sánchez-Busó L, Page AJ, Parkhill J, Keane JA, Harris SR: ARIBA: rapid antimicrobial resistance genotyping directly from sequencing reads. Microbial Genomics 2017, 3(10).

No competing interests reported.

supplementarymaterials.zip

Download PDF

Version 1

posted

You are reading this latest preprint version

Sequencing methods to study the microbiome with antibiotic resistance genes in patients with pulmonary infections

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Introduction

Methods

Sample preparation

mNGS sequencing

Uratio: unique read ratio (how many aligned reads are unique reads in X species)

Results

Identification of the predominant ARB in pulmonary infection patient samples

Discussion

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1