Distinct relative abundances in pathogens detected in mechanically ventilated patients with suspected pneumonia in the intensive care unit at King Abdulaziz University Hospital

Abstract

In this study, we present for the first time the landscape of the lung microbiota in patients with ventilator-associated pneumonia in Intensive Care Units in Saudi Arabia. DNA from 83 deep endotracheal aspirate lung samples was subjected to PacBio sequencing to identify pathogens in comparison with conventional diagnostic techniques. Patients on ventilation with pneumonia presented with similar lung flora to those of patients on ventilation without pneumonia. Proteobacteria, Firmicutes, and Bacteroidetes were detected in the majority of the samples. Samples treated with different antibiotics exhibited similar abundances of phyla and families. In order, the ten most common species detected in 16 clusters were Klebsiella pneumoniae, Stenotrophomonas maltophilia, Haemophilus influenzae, Pseudomonas aeruginosa, Metamycoplasma salivarium, Elizabethkingia anophilis, Staphylococcus aureus, Acinetobacter baumannii, Prevotella oris and Klebsiella africana. Of 51 on ventilation with pneumonia, the pathogens identified through sequencing corresponded with the findings from culture-dependent tests in 26 patients (50.98%), whereas the results differed in 30 patients (58.82%). Of 32 patients on ventilation without pneumonia, the pathogens identified through sequencing matched the conventional diagnostics results in only two patients (6.25%) but differed in 25 patients (78.13%). In summary, patients on mechanical ventilation with pneumonia did not display notable phenotypic traits. K. pneumoniae and S. maltophilia were the most common taxa detected in the samples, although some variations in microbial composition were observed. We conclude that Intensive Care Units exhibit distinct patterns of microbial colonization.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-88087-0.

Introduction

Active surveillance plays a crucial role in controlling and decreasing the incidence of Ventilator-Associated Pneumonia (VAP) and evaluating preventive measures. A collaborative group of multiple research institutions conducted a two-year study to determine the incidence rates of VAP in hospitals under the auspices of the Saudi Ministry of Health (MOH). The standardized infection ratio for VAP across all types of Intensive Care Units (ICUs) in MOH hospitals was 191% greater than that of hospitals in the U.S. National Healthcare Safety Network and 29% lower than that of Gulf Cooperation Council (GCC) hospitals¹. Furthermore, multiple antibiotic therapy, hospital-acquired infection, and Mechanical Ventilation (MV) were identified as independent risk factors for infection². The antibiotic resistance and virulence of the pathogenic bacteria influence the prognosis of VAP patients³.

The occurrence of multiantibiotic-resistant infections has become highly prevalent in the ICU. The most prevalent carbapenem-resistant bacteria are Acinetobacter baumannii, Klebsiella pneumoniae, and Pseudomonas aeruginosa. The ability of these bacteria to hydrolyze carbapenems determines their prevalence. Studies have demonstrated that patients infected with K. pneumoniae strains with Carbapenem Resistance (CRE) have a higher death rate than those infected with regular K. pneumoniae⁴. In K. pneumoniae strains detected in Saudi patients, the resistance rate to imipenem increased from 6.6% in 2011 to 59.9% in 2021⁵. blaNDM-1 and blaOXA-48-like resistance genes have been detected in Saudi patients with various illnesses^6,7. Reports of P. aeruginosa isolates with GES-type extended-spectrum β-lactamases are increasing in frequency. Acquired carbapenemases, specifically those in the VIM and IMP families, have also been detected in some isolates^8,9. Stenotrophomonas maltophilia, which produces ß-lactamase enzymes that decrease the effectiveness of β-lactam antibiotics, is becoming less sensitive to tigecycline and ticarcillin-clavulanic acid¹⁰. To treat S. maltophilia infections, newer fluoroquinolones and trimethoprim-sulfamethoxazole should be utilized to ensure that elderly patients are hospitalized for the shortest possible duration¹¹. A. baumannii is the predominant bacterial species identified in ICUs; this bacterium is resistant to carbapenems, which is mostly caused by three genes: blaOXA-23, blaOXA-66, and blaOXA-51¹².

Recent research on the microbiota involved in VAP occurring in the ICU¹³ revealed unique microbiomes associated with diverse chronic respiratory disorders and reported that dysbiosis levels correlate with clinical state. In critically ill patients, especially those receiving MV, microbiota diversity declines over time, with specific bacterial species becoming predominant, and is influenced by gut bacteria. The authors highlight the need for more research to fully understand these complex microbial dynamics. Additionally, the lung microbiomes of Streptococcus, Haemophilus, and Moraxella are less diverse in patients with bacterial pneumonia, and viral pneumonia patients have increased abundances of Prevotella, Veillonella, and Fusobacterium¹⁴. In the diagnosis of hospital-acquired pneumonia (HAP), metagenomics has numerous advantages over conventional methods. Metagenomics can also be used to diagnose polymicrobial infections and improve the understanding of the microbial community¹⁵. Statistical analysis revealed no correlation between VAP prevalence and death. A. baumannii was the causative agent in most patients with both early- and late-onset VAP, which were also associated with the bacteria K. pneumoniae and P. aeruginosa. An association between death and the presence of S. maltophilia and Escherichia coli was detected¹⁶.

Recently, preexisting conditions were found to affect death rates more than VAP. S. maltophilia, K. pneumoniae and Serratia marcescens may also be present. VAP patients presented lower abundances of Pseudomonadales and greater abundances of Staphylococcus aureus and Burkholderia¹⁸. These findings indicate that MV, not antibiotics, causes microbial dysbiosis in patients with VAP. Furthermore, polymicrobial infection may increase antimicrobial resistance. Biofilm alliteration allows K. pneumoniae to feed A. baumannii¹⁹, which can be protected against lethal cefotaxime doses by K. pneumoniae. Interestingly, carbapenem-resistant K. pneumoniae decreased the success of meropenem in killing S. aureus²⁰.

In this study, we examined the lower airway microbiota at the start of intubation and after 48 h and compared the microbial features between patients who developed VAP and those who did not. We hypothesized that increased lung dysbiosis and gene dominance would be detected in VAP patients.

Results

Quality control statistics for microbial sequencing

Among 90 Deep EndoTracheal Aspirate (DETA) samples, 83 presented high-quality DNA. Fifty-one patients (61.45%) were positive for pneumonia (VP), whereas 32 patients (38.55%) were negative for pneumonia (VN). The total number of reads following quality filtering was 1,889,193, representing 93.16% of the initial dataset. A total of 4089 Amplicon Sequence Variants (ASVs) were identified. The mean number of ASVs per sample was 115. The total number of reads in 4089 ASVs was 1,212,253, representing 64.17% of all the input reads.

Diversity of microbial communities within and between sample phenotypes

The predominant bacteria found by Conventional Diagnostic (CD) techniques in the VP and VN groups included P. aeruginosa (n = 10), S. aureus (n = 6), E. coli (n = 1), yeast (n = 11), and S. maltophilia (n = 6). Additionally, K. pneumoniae (n = 17) and A. baumannii (n = 6) were exclusively detected in the VP samples. Twenty-one samples had negative culture results for DETA. Nine patients did not receive antibiotic treatment during the aspirate sampling period. K. pneumoniae alone was identified in 19 patients in the VP group. Conversely, two microorganisms were detected in 10 patients, suggesting polymicrobial VAP.

Species identification and the antibiotic sensitivity minimum inhibitory concentration (MIC) before DETA culture revealed that 51 samples did not have sufficient quality for further testing. The poor quality of these samples may be because the isolates were gram-positive and not sensitive to gram-negative drugs and vice versa. Twenty-two samples were resistant, and 13 were sensitive. BioMérieux developed the Vitek II system to identify species and determine antibiotic sensitivity. K. pneumoniae (CRE-OXA 48) was the most frequently observed species in the VP group, with 14 isolates. Furthermore, we discovered that one isolate produced both OXA-48 and NDM. Three VIM-type bacterial isolates were obtained from two wound swabs and one blood culture. KPC-type or IMP-type bacteria were not detected. We identified two isolates with carbapenem-resistant P. aeruginosa (CRPA), five isolates with multidrug-resistant (MDR) P. aeruginosa, and methicillin-resistant S. aureus (MRSA).

Analysis of the full-length 16 S rRNA sequencing data indicated that each sample was influenced by at least one predominant bacterial species, including K. pneumoniae, Haemophilus influenzae, and P. aeruginosa (Fig. 1). In a comparison of the five most common species identified by PacBio with those detected by CD techniques and DETA in each sample, the latter demonstrated a diagnostic yield of 48.2% (Supplementary Table S1). Eleven samples indicated that the lung microbiome undergoes dynamic changes in response to selection pressure. Four patients (VP16/VN11, (VP24/VP26, (VN26/VN27, and VN29/VN31) underwent CD and PacBio methods on two different days, and one patient was assessed three times. In this patient, no organisms were detected in the initial sample (VN15) using CD methods, but S. maltophilia was identified by PacBio sequencing. The second sample (VP28) tested positive for S. maltophilia on CD methods and P. aeruginosa on PacBio. The third sample (VP30) revealed the presence of P. aeruginosa (MDR) through culture and S. maltophilia by sequencing. These findings may have implications for therapeutic efficacy and the improvement of current clinical methods, indicating that PacBio and CD methods are complementary (Table 1).

Fig. 1.

Ten most abundant species per sample as identified by full-length 16 S rRNA sequencing.

Table 1.

Observations of variability in replicated samples from the same patient.

Sample name	Conventional diagnostic method	PacBio
Positive—VP16	Klebsiella pneumoniae (CRE-OXA 48)	Klebsiella pneumoniae
Negative—VN11	Yeast	Haemophilus influenzae
Positive—VP24	Pseudomonas aeruginosa (MDR)	Pseudomonas aeruginosa
Positive—VP26	Pseudomonas aeruginosa (CRPA) and Klebsiella pneumoniae (CRE-OXA 48)	Klebsiella pneumoniae
Negative—VN26	No growth	Haemophilus influenzae
Negative—VN27	Yeast	Metamycoplasma salivarium
Negative—VN29	Yeast	Klebsiella pneumoniae
Negative—VN31	Yeast	Elizabethkingia anophilis
Negative—VN15	No growth	Stenotrophomonas maltophilia
Positive—VP28	Stenotrophomonas maltophilia	Pseudomonas aeruginosa
Positive—VP30	Pseudomonas aeruginosa (MDR)	Stenotrophomonas maltophilia

There were notable differences in the number of observed taxa and Shannon index values among the sample clusters identified by the 10 most abundant species according to the PacBio sequencing results (Fig. 1, Supplementary Table S1). The predominant species, K. pneumoniae, was identified in 10 VN samples and 19 VP samples. S. maltophilia was the second most abundant bacterium and was found in 7 samples in the VP group and only 2 samples in the VN group (Supplementary Table S1).

Analysis of the alpha diversity metrics among the sample clusters generated through the previously described hierarchical clustering (Fig. 1) revealed notable differences in the microbial communities (Fig. 2). Clusters 3 (K. pneumoniae), 4 (A. baumannii), 6 (S. maltophilia), and 10 (P. aeruginosa) included the lowest number of taxa. Analysis of the Shannon index values revealed that clusters 4 and 6 presented minimal within-group diversity. No difference in alpha diversity was detected between samples when categorized by ICU outcome (deceased, discharged) or antibiotics administered (broad-spectrum, narrow-spectrum, both, or none) (Supplementary Fig. S1).

Fig. 2.

Alpha diversity differed among the sample clusters based on the 10 most abundant species. The number of observed taxa and the Shannon diversity index were evaluated separately. Clusters with fewer than 3 samples were not included in the analysis.

After calculating the beta diversity in the samples, we employed the Aitchison distance for clustering purposes to uncover potentially noteworthy groups (Fig. 3). Interestingly, although samples could not be grouped according to variables such as ICU outcome, registered antibiotics, sample sex, and antibiotic combination, a substantial number of cluster 3 samples (Table 2) were grouped together (Fig. 3).

Fig. 3.

Beta diversity-based Aitchison distance clustering reveals a solid group of samples infected with Klebsiella pneumoniae.

Table 2.

Sample clusters identified by examination of the 10 most abundant bacterial species.

Cluster	Sample	Species
1	VN1, VN21, VN26, VN4, VN5, VN6, VP11	Haemophilus influenzae
2	VN10, VN16, VN2, VN27, VN32, VP32	Metamycoplasma salivarium
3	VN11, VN14, VN17, VN19, VN29, VN7, VP16, VP2, VP20, VP22, VP26, VP27, VP36, VP38, VP39, VP4, VP41, VP42, VP44, VP45, VP46, VP49, VP50, VP51	Klebsiella pneumoniae
4	VN12, VN23, VP17, VP25, VP40	Acinetobacter baumannii
5	VN13, VN33	Metamycoplasma salivarium and Klebsiella pneumoniae
6	VN15, VP10, VP23, VP30, VP33, VP34, VP43, VP5	Stenotrophomonas maltophilia
7	VN18, VN20, VN3, VN8, VP6	Stenotrophomonas maltophilia and Klebsiella pneumoniae
8	VN22, VN24, VP9	Staphylococcus aureus
9	VN25, VN30, VN31, VP48	Elizabethkingia anophilis
10	VN9, VP1, VP24, VP28, VP31, VP35, VP37	Pseudomonas aeruginosa
11	VP12	Klebsiella pneumoniae, Haemophilus influenzae, Acinedobacter baumannii and Metamycoplasma salivarium
12	VP13	Klebsiella pneumoniae and Acinedobacter baumannii
13	VP14, VP19	Klebsiella pneumoniae and Prevotella oris
14	VP15, VP21, VP29, VP47, VP7, VP8	Prevotella oris
15	VP18	Klebsiella africana
16	VP3	Haemophilus influenzae and Staphylococcus aureus

These results suggest that K. pneumoniae infection has a unique impact on microbial diversity compared with the other most abundant bacterial species. Permutational multivariate analysis of variance (PERMANOVA) revealed differences in beta diversity among the sample groups based on the top microbial species (R² = 0.197; p = 0.001). The underlying variance heterogeneity identified by ANOVA and permutation tests at least partially explains the PERMANOVA results. Finally, PERMANOVA confirmed the earlier findings that no differences in microbe types were found between sample groups with different ICU outcomes or different antibiotic treatments (Supplementary Table S2).

We subsequently assessed the relative abundance of organisms in each sample utilizing gene copy number (GCN)-corrected abundance values. Figure 4 shows that Proteobacteria, Firmicutes, and Bacteroidetes were the predominant phyla in most of the analyzed samples. Statistical analysis of relative abundance at the phylum level revealed multiple differences among sample clusters, except the phylum Spirochaetes (Fig. 5). Cluster 2 presented greater abundances of Actinobacteria, Bacteroidetes, Campylobacterota, and Firmicutes but a lower abundance of Proteobacteria than those in the other clusters. Clusters 8 and 14 exhibited a comparable pattern, characterized by a predominance of Bacteroidetes, Campylobacterota, Firmicutes, and Fusobacteria and a relative decrease of Proteobacteria in comparison to other sample clusters. No differences in relative abundance were observed between the sample groups defined by antibiotic regimens. The expansion of our search to the family level revealed no significant differences.

Fig. 4.

Per-sample phyla relative abundance calculated after GCN correction. (A) Relative abundance of phyla per sample. (B) Relative abundance of phyla per cluster based on the 10 most abundant species. For (B), only clusters with at least 3 samples were considered.

Fig. 5.

Differences in the relative abundance of phyla in the sample clusters defined by the top 10 species. Clusters with fewer than 3 samples were not included in the analysis.

Discussion

The lung microbiome presents low bacterial abundance and few microbial signatures that are associated with diagnosis. In addition, the variety of findings complicate the characterization of lung dysbiosis within the same individual. Consequently, delineating microbial resilience within a dynamic context is challenging. Understanding these concepts is crucial for developing innovative approaches to diagnose infections involving microbes and humans²¹, particularly as multiomics methods gain popularity. In another study, which is the only study focused on VAP and conducted with PacBio sequencing²², full-length sequencing results helped doctors diagnose VAP lung bacterial infection patterns more accurately than the results of CD methods did. Limitations of the CD methods and their associated rates of false-negative/false-positive results were also addressed. A standard CD method can identify a major pathogen, and quorum sensing can identify how harmful the dominant pathogen is based on the common, non-pathogenic bacterial community. In another study of BronchoAlveolar Lavage (BAL) fluid samples from VAP patients, sequencing had greater yields and detection rates for K. pneumoniae, A. baumannii, and Corynebacterium striatum than the CD method did. An increase in the number of unique reads and the relative abundance of detected bacteria correlates with a greater likelihood of these microorganisms being classified as causative agents²³.

The primary phyla in the lung microbiome include Bacteroidetes, Firmicutes, and Proteobacteria, while Actinobacteria are present to a lesser extent²⁴. Furthermore, Firmicutes/Bacteroidetes have been previously reported as potential biomarkers for VAP²⁵. An extended duration of MV resulted in increased variability in the pulmonary microbiota. Staphylococcus, Pseudomonas, and Prevotella predominated as the diversity decreased. A limited number of taxa subsequently predominated in the tracheal specimens²⁶. Our research revealed these relationships across 16 bacterial clusters. Cluster 1 primarily comprised H. influenzae in the VN group. Patients may demonstrate increased baseline levels of H. influenzae colonization despite the absence of pneumonia symptoms²⁷.

Cluster 3, containing K. pneumoniae, was prevalent in 19 VP and 10 VN samples. This result is consistent with a Saudi study that revealed a distinct correlation between the acquisition of multidrug-resistant species and ICU hospitalization, suggesting that the use of antibiotics in ICUs could increase resistance rates. In addition, alternative drugs and cross-transmission may increase the circulation of these organisms within hospital wards²⁸. When K. pneumoniae is present, the metabolome of the lung changes, suggesting that infection with this bacterium alters both the variety of microbes and the biochemical environment of the lungs²⁹. There was a consistent increase in the Klebsiella population, while the levels of Streptococcus in the lung decreased throughout the course of the infection³⁰. A study of critical care patients revealed that 50% of K. pneumoniae infections originated from the resident strains in the patient’s gastrointestinal or pharyngeal tracts; these strains can spread from the gut to other body parts through ventilators³¹.

The lung samples, which were dominated by P. aeruginosa, presented the lowest degree of species diversity. This observation is consistent with other studies, as P. aeruginosa outcompetes other species and is associated with the frequency of exacerbations and mortality³². Furthermore, P. aeruginosa is more likely to develop antibiotic resistance due to exposure to subinhibitory doses of antibiotics during infection³³. A greater number of microbes in the lungs, primarily Staphylococcus and Pseudomonas species, is linked to worse outcomes and higher death rates after invasive MV³⁴. In addition, Pseudomonas can migrate from the intestine to the lungs, potentially leading to dysbiosis. One study suggested that hospitalized patients may benefit from removing antimicrobial-resistant bacteria such as P. aeruginosa from their gut microbiome, even if their infections are not severe or life-threatening³⁵.

In cluster 7, S. maltophilia and K. pneumoniae were identified in four VP samples and one VN sample, and in cluster 6, one VP sample and seven VN samples exclusively contained S. maltophilia. The reasons why the bacteria in the VN samples resemble those in the VP samples remain unclear. Factors that may play a role include antibiotic treatment, which can result in bacteria becoming nonviable and unculturable but still allowing for the detection of DNA or the presence of various strains within the same species. Evaluating the pathogenicity of S. maltophilia in polymicrobial infections presents challenges³⁶. Compared with polymicrobial bloodstream infections, monomicrobial S. maltophilia in bacteremia patients was associated with a higher mortality rate. Therapeutic interventions for S. maltophilia infections have led to a notable decrease in mortality rates¹⁴. Additionally, Metamycoplasma salivarium was detected for the first time in Saudi ICUs; this bacterium is challenging to culture because of its fastidious nutritional requirements. One study linked poor clinical outcomes to lower airway enrichment, which is a marker for MV patients³⁷. Elizabethkingia is the other non-frequent cause of lung infection that is clinically significant, particularly when caused by Elizabethkingia anophilis. In the ICU, a patient developed nosocomial septicemia caused by polymicrobial infection with Elizabethkingia meningoseptica, K. pneumoniae, and A. baumannii while being treated for COVID-19³⁸. Additionally, infections caused by Elizabethkingia species have been reported in VAP patients in Saudi Arabia³⁹.

In conclusion, the lungs contain a multitude of bacterial species; however, predominant genera such as Klebsiella, Stenotrophomonas and Pseudomonas frequently prevail over others in the competition for resources, resulting in diminished diversity within the microbial community associated with pneumonia. Although VAP is reliably linked to extended periods of MV and prolonged stays in the ICU, the direct influence of VAP on mortality is subject to ongoing debate. Mortality seems to be influenced by multiple factors, including VAP as well as preexisting conditions. We observed notable discrepancies in the DETA microbiota profiles of VAP patients. Our findings revealed that both VP patients and VN patients exhibited a significantly increased rate of infection with K. pneumoniae and Stenotrophomonas, and statistical analysis revealed no significant differences between the groups. This result suggests that PacBio sequencing data should be interpreted in conjunction with CD and clinical data.

This study was limited by the setting of the ICU environment and the lack of healthy control samples. The complexity of antibiotic regimens and sensitivity, as well as the small size or individual nature of some microbiomics-defined clusters (e.g., 11, 12, 13, 16), limits the ability to derive reliable results. For this reason, these factors were excluded from analyses such as alpha diversity or relative abundance analyses. Future prospective studies should collect specimens from healthy individuals, as its essential for comparison and provide a baseline for understanding the alterations that occur during critical illness.

Methods

Statement of ethics

This study was executed in compliance with the pertinent rules of the Saudi Committee. Written informed consent was obtained from each admitted patient with suspected pneumonia prior to the collection of DETA samples. The Unit of Biomedical Ethics, Research Ethics Committee (REC), approved this investigation (NCBE Registration Number: HA-02-J-008; Monday, February 27, 2023).

Study design

The research team collected DETA samples from patients hospitalized in the ICU at King Abdulaziz University Hospital (KAUH) between March 2023 and February 2024. Of the 83 DETA samples, 51 patients developed pneumonia (VP), 48 with late onset and 3 with early onset, and the remaining 32 patients did not develop pneumonia (VN). Patients with VP were defined according to the most recent guidelines provided by the Infectious Diseases Society of America/American Thoracic Society. The inclusion criteria were as follows: intubation and MV for a duration exceeding 48 h; clinical suspicion of pneumonia; pus-filled secretions in the trachea and other signs of illness, such as a body temperature above or below 36.5 °C, white blood cell count below or above 4000/mm³, and decreased oxygenation; symptoms of suspected pneumonia including tachypnea, increased or purulent secretions, hemoptysis, rhonchi, crackles, reduced breath sounds, and bronchospasm; MV difficulties, reduced tidal volume, or increased inspiratory pressures; laboratory findings including worsening hypoxemia and leukocytosis; and the presence of a new or progressive infiltrate on a chest radiograph or computed tomography (CT) scan. We excluded patients if we deemed them medically unfit for lung sample collection or if their medical proxy declined their participation in the trial. Table 3 presents the clinical characteristics of the patients. We obtained written informed consent from all participants.

Table 3.

Clinical characteristics (count or mean ± standard error of the mean).

Variables	VAP-positive (n = 51)	VAP-negative (n = 32)
Age	56 ± 3	53 ± 3
Sex
Female	25	12
Male	26	20
Temperature (°C)
36.5–38.4 °C	37 ± 0.26	38 ± 0.13
38.5–38.9 °C	39 ± 0.00	38.5
≥ 39 °C	39 ± 0.18	39.4
Leukocytes in blood (cells/mm3)
4000–11,000/mm3	8 ± 0.34	8 ± 0.69
> 11,000/mm3-38,000/mm3	18 ± 1.20	17 ± 1.00
Radiographic findings
Infiltrate	23	17
Combined treatment	24	15
Broad spectrum (meropenem and piperacillin-tazobactam)	15	11
Narrow spectrum (vancomycin and colistin)	8	1
Comorbidities
Diabetes mellitus
Positive	22	16
Negative	29	16
Hypertension
Positive	26	17
Negative	25	15
Ischemic heart disease
Positive	5	9
Negative	46	23
Stroke
Positive	9	3
Negative	42	29
Days of MV
VAP onset	28 ± 2.87	24 ± 2.97
Early onset	3	0
Late onset	48	0
Tracheostomy	3	6
Length of stay in ICU (days)
Outcome	11 ± 1.60	9 ± 1.88
Death	28	12
Recovery	23	20

Sampling procedures and antimicrobial susceptibility testing

The DETA samples were obtained by inserting a catheter into the endotracheal tube until it encountered resistance, after which the catheter was withdrawn approximately 2 cm. After the vacuum was released, the probe was carefully extracted utilizing rotational motions, and the secretion was immediately aspirated into a sterile specimen trap for subsequent microbiological investigation. The use of saline solution to liquefy secretions was avoided, and we strictly adhered to aseptic standards⁴⁰. The DETA samples were collected before the patients started or changed antibiotic treatment. Within 1–2 h, we transported the DETA samples to the clinical and molecular microbiology laboratory at KAUH, where gram staining and bacteriology culture were performed. We used the most infected or blood-stained parts for gram staining, microscopic analysis⁴¹, and bacterial cultures. The DETA bacterial CD sample was inoculated on 5% sheep medium. We incubated blood on chocolate agar from Saudi Prepared Media Laboratories aerobically at 35–37 °C with 5% CO₂ for 24–48 h. Additionally, a plate of MacConkey agar from Saudi Prepared Media Laboratories was used and incubated in a regular incubator at 35–37 °C for 24–48 h. The presence of moderate to heavy growth of an isolate in the second, third, or fourth quadrants of each plate following incubation indicated significant bacterial growth. Organisms were identified using VITEK MS, a mass spectrometry instrument specifically designed for microbial identification and manufactured by BioMérieux, Durham, NC, USA.

The VITEK MS system employs matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) technology. The Vitek 2 identification card and susceptibility were assessed using Vitek 2 antibiotic susceptibility testing (AST) cards (BioMérieux, Marcy l’Etoile, France). The tests followed the normal operating procedures used by the clinical and molecular microbiology laboratory at KAUH. The laboratory derived the criteria for evaluating resistance, sensitivity, and intermediate resistance in antibiotic susceptibility reporting from the performance standards for antimicrobial susceptibility testing from CLSI M100, specifically Editions 33 and 34. In addition, the selected antibiotic may be appropriate, but the dosage or frequency may be too high. It is imperative to focus on not only treating resistant organisms but also preventing the development of resistance through eradication activities. Antibiotic effectiveness can be predicted in three ways: the highest concentration reached, the amount of time the antibiotic stays above the MIC, and how much greater the area under the concentration‒time curve is than the MIC. By integrating our understanding of bacterial eradication strategies with pharmacokinetic/pharmacodynamic (PK/PD) parameters, we can effectively reduce mortality and morbidity.

DNA isolation and sequencing

A total of 83 DETA samples were stored at − 80 °C for subsequent DNA extraction. The PureLink™ Microbiome DNA Purification Kit (Invitrogen™, USA) protocol was used. A DNA amount ≥ 300 ng and a concentration ≥ 10 ng/µL were measured using gel electrophoresis and a NanoDrop-1000 spectrophotometer manufactured by Nanodrop Technologies (Wilmington, DE, USA). PCR amplification was conducted by Novogene Technology Co., Ltd. The PacBio full-length read amplification of 16 S rRNA was conducted with the universal primers 27 F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1514R (5′-GNTACCTTGTTACGACTT-3′). The PCR products were used to generate libraries using SMRTbell™ Template Prep Kit 1.0-SPv3/Pacbio/100-991-900/10 reactions.

Bioinformatics and statistical analyses

The PacBio HiFi 16 S data were analyzed via a Nextflow workflow provided by the manufacturer (https://github.com/PacificBiosciences/HiFi-16 S-workflow) with default parameters. Established tools such as QIIME2 and DADA2 were employed. FASTQ files supplied by NovoGene with adapters and primers trimmed and removed were utilized as inputs. Reads with a minimum quality of Q20 and a length of 1000–1600 bases were retained. Residual reads with fewer than two anticipated errors (max_ee) were denoised into amplicon sequence variations (ASVs). ASVs that were not identified in any samples or represented by fewer than 5 reads across samples were eliminated. ASVs were classified at the species level and subsequently at the genus level using the naïve Bayes classifier function in DADA2. The GTDB r207, Silva v138, and RefSeq + RDP databases were used in decreasing order of importance to obtain the most accurate taxonomic classification of the ASVs. The results were converted into a phyloseq object for subsequent analysis in R. The most abundant species in each sample were identified using absolute abundance values.

Alpha (within-sample) diversity was evaluated by the observed taxa and the fluctuation of the Shannon index, employing Wilcoxon tests for comparisons of two groups or Kruskal–Wallis and Dunn post hoc tests for comparisons of multiple groups^42,43. Across-sample (beta) diversity assessments were conducted to evaluate sample dissimilarity utilizing the Aitchison distance, as previously outlined⁴⁴; this method is a suitable approach that accounts for the compositional characteristics of 16 S rRNA sequencing data. PERMANOVA was employed to evaluate the impact of environmental factors on sample composition.

Differences in taxon relative abundance between sample groups were evaluated after GCN correction using rrnDB data (https://rrndb.umms.med.umich.edu/static/download/rrnDB-5.7_pantaxa_stats_NCBI.tsv.zip). GCN correction was performed at the genus level, as species-level correction led to significant data loss. The Wilcoxon rank-sum test and Kruskal‒Wallis test followed by post hoc Dunn tests were applied to evaluate differences in abundance between two groups and among multiple groups, respectively.

Differential abundance analysis (DAA) was performed at both the genus and species levels on GCN-corrected abundance values using the ALDEx2 R package. Truly differentially abundant taxa were identified using the following criteria:

• |effect size| > 1.

• adjusted p value < = 0.05.

  • Expected Benjamini‒Hochberg corrected P value of Welch’s t test (we.eBH) and/or.
  • Expected Benjamini‒Hochberg corrected P value of the Wilcoxon test (wi.eBH).

The ALDEx2 effect statistic was used instead of raw abundance differences, as this statistic confers stability across datasets and has a stronger relationship with the respective p value. For all the statistical tests performed, the (adjusted) p value threshold of statistical significance was set to 0.05.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Supervision; H.A, Data curation & samples collections; D.H.B, I.A, M.K.B, A.A.A, R.M.A, and S.N.A Microbiology; D.M.A. Methodology; S.A.M & N.R.T. Writing original draft and editing; N.R.T.

Funding

The work presented in this article was self-funded (H.A & N.R.T.).

Data availability

The raw data files containing sequences from each patient were submitted to the NCBI under the Submission: PRJNA1187299.

Declarations

Competing interests

The authors declare no competing interests.