Background
Ventilator-associated pneumonia (VAP) is a frequent complication in patients requiring mechanical ventilation and the associated mortality ranges from 20 to 50% [
1,
2].
Pseudomonas aeruginosa is one of the most common pathogens causing VAP and is independently associated with increased mortality; in China, it has been staying in the top three pathogens [
3‐
7]. Antibiotic treatment is the primary method for managing
P. aeruginosa VAP; however, it constitutes a risk factor for the development of multi-drug resistant
P. aeruginosa [
2]. Increasing drug resistance, especially in intensive care units (ICUs), could result in
P. aeruginosa VAP becoming uncontrollable [
8,
9].
Recent findings suggested that the lower respiratory tract (LRT) is inhabited by niche-specific microbiota and VAP occurs mainly when the micro-ecology balance is damaged [
10‐
12]. Probiotics play a preventive role on VAP occurrence, especially the VAP induced by
P. aeruginosa [
13‐
15]. Probiotics pre-treated patients obtained a decreased risk of LRT colonization with
P. aeruginosa [
13,
15]. In addition, Khailova et al. found that
Lactobacillus rhamnosus can decrease lung
P. aeruginosa load and increase the survival rate of
P. aeruginosa pneumoniae mice [
16]. It indicated that microbiota regulation could have significant effect on
P. aeruginosa infection. However, we know less about the characteristics of LRT microbial composition in
P. aeruginosa VAP patients.
In this study, our aim was to examine LRT microbiota characteristics in P. aeruginosa VAP first and then analyze the relationship between LRT microbial characteristics and patient prognosis.
Methods
Subjects
This study was a prospective study conducted at intensive care unit (ICU) of Ruijin Hospital, China. Inclusion criteria of
P. aeruginosa VAP patients included [
1]: (1) mechanical ventilation > 48 h; (2) satisfied two of the following: body temperature > 38 °C or < 36 °C, leukopenia or leukocytosis, or purulent secretions; (3) new or progressive chest infiltrates, for patients with underlying pulmonary or cardiac disease, two serial chest radiographs were required for assessment; (4) endotracheal aspiration cultured
P. aeruginosa at least + 2 growth using semi-quantitative measurements. Exclusion criteria included: (1) age < 18 years; (2) pregnant woman; (3) sputum cultured
P. aeruginosa prior to intubation. Clinical data collection was performed at the hospital upon admission and terminated following study withdrawal, discharge, or death. Sequential Organ Failure Assessment (SOFA) score was evaluated at ICU admission and the initial sample collection day. The severity of pulmonary infection in
P. aeruginosa VAP patients was assessed using the Clinical Pulmonary Infection Score (CPIS). The criteria of CPIS were performed as previously described [
17]. CPIS could be used to assess the clinical outcome of pulmonary infection in patients with VAP [
18]. We took the CPIS of the patient at the time of diagnosis as the baseline parameter and reassessed it within 7 days to 14 days after antibiotic treatment; the CPIS score was reduced to within 6 points to be recognized as clinical improvement of pulmonary infection [
17]. Control subjects were selected from selective operation patients without acute or chronic respiratory disease, any infection, and antibiotic use for three months prior. Written informed consents were obtained from all study subjects or their lineal consanguinities prior to enrollment. The protocol of this study was approved by the Ruijin Hospital Ethics Committee Shanghai Jiao Tong University School of Medicine.
Sample collection
Initial samples were collected within 24 h post P. aeruginosa VAP diagnosis. Sequential sample collection was performed at day 7 and day 14 post initial sample collection. Endotracheal aspiration samples were collected using an endotracheal tube. Collected samples were stored in sterile 15 mL centrifuge tubes at − 80 °C.
Freeze-drying
We set up negative control using sterile deionized water. Prior to freeze-drying, the sample tubes were transferred into a portable liquid nitrogen container and the sample tubes caps were substituted with parafilm; five small holes were made using sterile yellow tips in a bio-safety cabinet while the sample tubes remained in liquid nitrogen. The sample shelf of the freeze-dryer (FreeZone 6 Liter Console Freeze Dry Systems, Labconco, USA) was pre-cooled for at least 2 h in advance by setting the trap-temperature to − 86 °C. The pre-treated frozen samples were then transferred into the sample shelf in the freeze-dryer (chamber space = 2.82 × 107 mm3, trap-temperature = − 86 °C, and vacuum pressure = 0.165 Torr) for drying. Samples were incubated for 60 h. The samples were hermetically sealed immediately following vacuum release and then stored at − 80 °C.
Bacterial DNA amplification and sequencing
LRT bacterial DNA extraction from freeze-dried powder was performed as previously described [
19], including negative controls which set up in the freeze-drying procedure. The V3-V4 region of 16S ribosomal RNA (16S rRNA) gene from each DNA sample was amplified with primers F1 and R2 (5′- CCTACGGGNGGCWGCAG-3′ and 5′-GACTACHVGGGTATCTAATCC-3′) corresponding to positions 341 to 805 of the
Escherichia coli 16S rRNA gene using an EasyCycler 96 PCR system (Analytik Jena Corp, AG) with the following program: 3 min at 95 °C (denaturation); 21 cycles of 30 s at 94 °C (denaturation), 30 s at 58 °C (annealing), and 30 s at 72 °C (elongation); 5 min at 72 °C (final extension). The products from different samples were indexed and mixed at equal ratios for sequencing using the Miseq platform (Illumina Inc., USA) according to the manufacturer’s instructions.
Data processing
In this study, negative control samples were not detected any bands when the DNA amplification process was completed. We also sequenced those negative control samples, and a total of 90 sequences were obtained from negative control samples, which can be classified as 31 genera. Within the 31 genera, only two (
Escherichia and
Streptococcus) belong to the contaminant genera detected in negative controls by previous study [
20]. The
Escherichia was not found in any particular sample in our study;
Streptococcus was the commensal of LRT, and the average relative abundance was 9.7% in the samples from controls and 0.5% in the samples from
P. aeruginosa VAP patients.
Raw FASTQ files were demultiplexed and quality-filtered using USEARCH 8.0 with the following criteria: (1) exact index matching, (2) only sequences with > 50 bp overlaps were assembled according to their overlap sequence, (3) merged sequences > 400 bp, and (4) a maximum mismatch in overlap area <0.1. Reads that could not be assembled were discarded. Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff using UPARSE (version 7.1) after chimeric sequences removed. The phylogenetic affiliation of each 16S rRNA gene sequence was analyzed using RDP Classifier against the SILVA (SSU123) 16S rRNA database using a confidence threshold of 70%. The raw sequencing data have been uploaded in the NCBI GenBank Sequence Read Archive database (accession number SRP112361).
Statistical analysis
To access the sequencing depth of the LRT samples, a rarefaction curve was generated as previous study [
21] and it indicated that the sequencing depth of samples was reasonable (Additional file
1: Figure S1). Shannon’s diversity was calculated by Mothur. The significant difference in Shannon index between groups was calculated using the Wilcoxon nonparametric test. The unweighted UniFrac distance and weighted UniFrac distance were calculated using Quantitative Insights into Microbial Ecology (QIIME) to assess compositional dissimilarity between samples and finally showed as principal co-ordinates analysis (PCoA) plots conducted in R version 3.2.1 [
22]. To find factors that related to the LRT microbial composition in
P. aeruginosa VAP patients, patient samples from the initial collection day were subjected to a similarity-based, unsupervised hierarchical clustering [
23,
24]. Clustering analysis was performed by unweighted pair-group method with arithmetic means (UPGMA) with Bray Curtis distance [
23]. Spearman’s correlation was used to test the relationship between the relative abundance of each genus found in
P. aeruginosa VAP patients and CPIS score, SOFA score. The statistic comparison of relative abundance of all taxa between the groups was analyzed using the Wilcoxon rank sum test and the
p-values were calculated for the false discovery rate (FDR) (q value) [
25]. In the dynamic analysis, the analysis of significant differences in Shannon diversity at different time points using paired nonparametric tests.
Clinical data were analyzed using SPSS version 23 (Armonk, New York, USA). Differences between groups were tested using a two-tailed t-test, Mann-Whitney U test, or chi-squared test as appropriate. Continuous variables were presented as the mean ± standard deviation for normally distributed data and median [interquartile range, IQR] for non-normally distributed data. Categorical variables were presented as number of subjects and percentages. Figures were created using GraphPad Prism version 6.0.
Discussion
In this study, we described the dynamic characteristics of the LRT microbiota in patients with P. aeruginosa VAP and their association with clinical characteristics and severity of pneumonia. Microbial composition varied among P. aeruginosa VAP patients, forming two clusters. The microbial clusters related to the primary diseases of P. aeruginosa VAP patients. We also analyzed the correlation between microbial composition and the severity of pneumonia in P. aeruginosa VAP patients. We found that abundance of Pseudomonas was positively correlated with the severity of pneumonia, while the abundance of Lactobacillus was negatively correlated with the severity of pneumonia. .
We found
P. aeruginosa VAP patient samples were depleted for LRT commensal bacteria (
Streptococcus and
Veillonella) and the predominant taxa differed by cluster. Fir-Bac cluster was enriched for
Lachnospiraceae_incertae_sedis, Bacteroides,
Blautia, and
Alloprevotella, and Pro cluster was enriched for
Proteobacteria including
Enterobacteriaceae and
Pseudomonas.
Lachnospiraceae_incertae_sedis, Bacteroides, and
Blautia are the predominant (and harmless) taxa in normal healthy intestines [
26].
Enterobacteriaceae and
Pseudomonas are common opportunistic pathogens causing lung infections [
2]. Microbial clusters of the LRT microbiota in patients with
P. aeruginosa VAP appear to be related to the primary disease. The primary disease in the Pro group was mostly digestive diseases, whereas in the Fir-Bac cluster the patients were mainly respiratory diseases. Previous studies have shown that
Lachnospiraceae and
Bacteroides are the predominant taxa in fecal samples of patients with respiratory disease such as cystic fibrosis [
27], while
Enterobacteriaceae increases more frequently in the gut of patients with gastrointestinal diseases [
28‐
30]. Intestinal microbiota can translocate into the lung and that increased intestine permeability is the most probable mechanism underlying microbiota translocation [
31]. Few studies have focused on the relationship between LRT microbiota and different primary diseases in patients with hospital infection. There was one study suggested that the predominant taxa in patients with pulmonary sepsis were different from that of the abdominal sepsis [
29]. Thus, our findings suggested that the variation in LRT microbiota may be related to the intestinal microbiota of
P. aeruginosa VAP patients. However, further analysis of the intestinal microbiota of
P. aeruginosa VAP patients is needed.
The CPIS is commonly used to assess the severity of lung infections in patients and the efficacy of antibiotic therapy [
17,
18]. In this study, we found that
Burkholderia,
Pseudomonas, and
Enterobacter were positively correlated with CPIS, while
Lactobacillus was negatively correlated with CPIS. These results suggested that the LRT microbiota in patients with
P. aeruginosa VAP was associated with the severity of pneumonia. We also found that
Pseudomonas showed a decreasing trend in the patients that acquired clinical improvement, and
Lactobacillus and
Bifidobacterium showed an upward trend in those patients.
Burkholderia,
Pseudomonas, and
Enterobacter were common pathogens of hospital-acquired pneumonia [
2]. Although they are also present in normal respiratory tract, their abundance is extremely low, and their abundance increases are likely to exacerbate the severity of pneumonia [
10,
32].
Lactobacillus and
Bifidobacterium could inhibit pathogen proliferation, affect the behavior of other taxa, and regulate the host immune response [
33]. Several studies have shown that
Lactobacillus and
Bifidobacterium have an inhibitory effect on
P. aeruginosa, Escherichia coli, Klebsiella pneumonia, and
Enterococcus faecalis in vitro [
33‐
36]. Cotar et al. found that
Lactobacillus decreases
P. aeruginosa virulence by decreasing the expression of
lasI,
lasR,
rhlI, and
rhlR involved in quorum sensing and inhibition of biofilm formation [
37]. It has also been demonstrated that
Bifidobacterium has the ability to inhibit the adhesion of
P. aeruginosa to epithelial cells [
34]. Therefore, their decline was likely to exacerbate the severity of pneumonia in
P. aeruginosa VAP patients. Microbiota regulation may become a potential therapy for
P. aeruginosa VAP, however, the issues of delivery and probiotic selection require careful consideration. Although we found LRT microbiota was associated with the severity of pneumonia and early treatment effect, we did not observed difference between survival group and non-survival group. The sampling collection time was far from the death time, ranged from one month to 12 months, which may be one of the reasons that resulted in no significant difference in the composition of the LRT microbial composition between the survivors and non-survivors. On the other hand, we found the SOFA score which was associated with patients’ survival showed little correlation with LRT microbiota in
P. aeruginosa VAP patients. It seemed that the LRT microbiota had greater effect on lung than the whole organ system.
Though antibiotic species and numbers were individual in each
P. aeruginosa VAP patient, patients of each cluster showed similar microbial composition. Our findings were similar to previous lung-infection-disease studies which showed the lung dominant taxa were stable following the antibiotic treatment [
38,
39]. Rogers et al. found that long-term antibiotic therapy did not changed the lung microbial composition in non-cystic fibrosis bronchiectasis patients with baseline airway dominated by
P. aeruginosa but changed in those not dominated by
P. aeruginosa [
40]. The mechanisms were complicated: competitions between bacterial species and antibiotic-bacterial interactions may involve in it.
Pseudomonas, Enterobacteriaceae and
Acinetobacter were common antibiotic-tolerant pathogens; it was not surprising that they could exist consistently with antibiotic exposure. The consistent microbial dysbiosis had significant negative impact on patient health, so identifying community member interactions and how to break up the network should be pay attention in future studies.
This study had several limitations. First, the sample size was limited. We could not exclude other microbial types that may exist in P. aeruginosa VAP. Second, we characterized the LRT microbiota using the culture-independent technology, which could identify low-proportion or uncultured taxa; however, we were unable to determine the function of specific species accurately. Further analysis regarding probiotic species and associated mechanisms in P. aeruginosa VAP is needed. Third, in this study, the sample collection time was limited. If we could increase the collecting time points in future, it may be better to reflect the changes of the LRT microbiota of patients with P. aeruginosa VAP.