Dynamic alterations in the lung microbiota in a rat model of lipopolysaccharide-induced acute lung injury

The lung microbiota have been found to be substantially altered in numerous pulmonary disorders, and crosstalk between the host pathophysiology and lung microbiota plays critical roles in the regulation of disease states. The aim of this study was to investigate dynamic changes in the lung microbiota during different stages of acute lung injury and acute respiratory distress syndrome (ALI/ARDS). Rats receiving an intraperitoneal administration of lipopolysaccharide (LPS) were sacrificed at 12 and 48 h after injection, and the hematological parameters, serum cytokine levels, and histological characteristics of the lung tissue and lung microbiota were assessed. After LPS injection, along with fluctuations of systemic cytokine levels and the onset and regression of pulmonary edema, the diversity, components, and functionalities of the pulmonary microbiota underwent significant dynamic changes. The volatility of the α-diversity indices narrowed after LPS injection, and the indices significantly decreased 48 h later. The abundance of 18 genera and functionality of adenosine triphosphate–binding cassette (ABC) transporters, pentose phosphate, and bacterial chemotaxis pathways were found to significantly differ between specified time points. Several significant correlations between the components and functionalities of the lung microbiota and indicative symptoms of ALI/ARDS were also observed. Brevibacterium was correlated with cytokines tumor necrosis factor (TNF)-α, interleukin (IL)-10, and IL-6 and with hematological percentage of neutrophils (NEU%); Wnt, Notch, and chronic myeloid leukemia signaling pathways were correlated with IL-1β; mitogen-activated protein kinase (MAPK) signaling pathway–yeast was correlated with IL-10; and the pathways of ascorbate and aldarate metabolism and basal transcription factors were correlated with platelet-related indicators. The correlations between the lung microbiota and indicative symptoms of ALI/ARDS identified in this study support further investigation into the underlying mechanism of host–microbiota interactions during lung injury and repair.

In recent years, with the widespread application of culture-independent, second-generation sequencing methods in microbiota research, the role of the microbiota in various human habitats in maintaining normal homeostasis and regulating disease states has attracted increasing attention 1 . The lungs, previously considered sterile, have been shown to have a diverse and dynamic bacterial community 2 . Although a relative paucity of data on the physiological implications of the respiratory microbiome exists in comparison to research involving the gut microbiome, the lung microbiota have been found to be substantially altered in the context of numerous respiratory disorders 3 . Studies have reported that the lung microbiota are strongly correlated with alterations of both systemic and alveolar immunity 4,5 . On one hand, the host inflammatory response can significantly change the lung microbial community by altering the lung microenvironment, which is conducive to the proliferation of certain microbial species 5 ; on the other hand, such changes can result in greater susceptibility to several lung diseases, such as infections by various pathogens, lung injury, and asthma [5][6][7] . Growing evidence for the correlation between variations in the lung microbiota and the physiological severity of pulmonary disease and response to www.nature.com/scientificreports/ therapy [8][9][10] indicates that the lung microbiome contains unexplored therapeutic targets for the prevention and treatment of lung disease. Application of lipopolysaccharide (LPS) can stably trigger a systemic immune response in experimental animals and cause lung histological changes, such as severe epithelial and endothelial damage, inflammatory cell infiltration into the lung, and accumulation of protein-rich edema fluid in alveolar spaces [11][12][13] . Therefore, LPS-induced acute lung injury and acute respiratory distress syndrome (ALI/ARDS) has become one of the most widely used in vivo models for studying the molecular mechanisms of inflammation-associated lung injury and researching pharmacological treatments thereof. However, the dynamic impact of the changes of inflammatory cells and immune factors on the pulmonary microbiota during the occurrence and development of LPS-induced ALI/ARDS is still lacking.
In this study, we present for the first time the dynamic effects of single-dose intraperitoneal (i.p.) administration of LPS on the lung microbiota. Herein, we also analyzed the association between the lung microbiota and other quantitative ALI/ARDS symptom indicators, including blood cell count and serum cytokine levels.

Results
Histopathological characteristics and W/D ratio. The H&E-stained sections of lung tissue of the Control group had normal architecture with no signs of inflammation. However, tissue from the LPS12 group showed interalveolar septum thickening, irregular distribution of air spaces, pulmonary interstitial edema, and inflammatory cell infiltration with a predominance of polymorphonuclear cells. In the LPS48 group, although the alveolar architecture and histological changes were slightly relieved, the lung tissue still showed perivascular and peribronchial inflammatory infiltration with inflammation in the alveolar septa, and we also observed patchy intra-alveolar and interstitial hemorrhage (Fig. 1).
We estimated the volume of lung water in each group by the wet/dry (W/D) ratio. Compared with the Control group, the W/D ratio increased significantly in the LPS12 group, suggesting the occurrence of pulmonary Control group presented normal architecture with no damage to the lung tissue. In the LPS12 group, the tissue showed interalveolar septum thickening, irregular distribution of air spaces, pulmonary interstitial edema, and inflammatory cell infiltration. In the LPS48 group, the tissue still showed perivascular and peribronchial inflammatory infiltration, but the alveolar architecture and histological changes were slightly relieved. www.nature.com/scientificreports/ edema. The ratio was reduced significantly in the LPS48 group as compared with the LPS12 group, but it was still significantly higher than that of the Control group (Fig. 2a). Quantification of histological changes revealed similar results. The proportion of the tissue area (red signal) of the LPS12 group was significantly higher than that of the Control group and was decreased significantly in the LPS48 group (Fig. 2b). In addition, the alveolar count ( Fig. 2c) and circumferential measurement (Fig. 2d) showed that as compared with the Control group, the number of alveoli used for effective gas exchange and the area of gas exchange were significantly reduced after LPS injection. Furthermore, as time progressed, the LPS48 group showed signs that these symptom indicators were relieved, compared with the LPS12 group.
Routine blood examination and serum cytokine measurement. Altogether, we captured 24 hematological parameters in this study. Of these, six were found to have significantly changed after LPS challenge. As shown in Fig. 3, the percentage of neutrophils (NEU%) significantly increased 12 h after LPS injection and returned to a level close to that of the Control group 48 h after LPS injection. Although lymphocyte percentage (LYM%) showed the opposite trend, it was significantly decreased at the 12-h time point, and it returned to normal at the 48-h time point. Mean platelet volume (MPV) and platelet-large-cell ratio (PLCR) gradually increased, while platelet count (PLT) and platelet packed volume (PCT) significantly decreased, as time progressed after LPS injection.
We also analyzed levels of interleukin (IL)-1β, IL-6, IL-10, and tumor necrosis factor (TNF)-α in serum samples. The levels of all four cytokines increased at 12 h after LPS injection and were normalized at 48 h ( Fig. 3g-j).
Overall assessment of lung microbiota. We ultimately generated a total of 24,317 zero-radius operational taxonomic units (ZOTUs) from 864,727 high-quality reads. A total of 23,459 ZOTUs were successfully annotated by the Ribosomal Database Project (RDP) classifier, comprising 35 phyla, 79 classes, 137 orders, 270 families, and 582 genera (Fig. 4a,b).
Comparison of lung microbiota diversity. We compared the three α-diversity indices of PD whole tree, Chao1, and Shannon between the groups. After LPS injection, the variation ranges of all three indices narrowed ( Fig. 4c-e). A non-parametric test of differences between groups found that these indices did not significantly differ between the Control and LPS12 groups; however, the LPS48 group showed different degrees of decline in all three indices as compared with the other two groups. In the LPS48 group, the PD whole tree index was significantly lower than in the Control group, the Chao1 index was significantly lower than in both the Control and LPS12 groups, and the Shannon index was significantly lower than in the LPS12 group.
We created a Bray-Curtis distance metric from the subsampled ZOTU table to explore structural differences in bacterial communities among the three groups. Principal coordinate analysis (PCoA) showed partial separation of microbiota communities by group. Adonis tests and distance-based redundancy analysis (db-RDA) both confirmed a significant association between the different treatment groups and bacterial-community structures (Fig. 4f).
Comparison of lung microbiota components. Comparison of the Control and LPS12 groups by linear discriminant effect size (LEfSe) showed the presence of 11 taxa (including 7 genera) with significant differences (Fig. 5a). At the genus level, Brevibacterium, Motilimonas, Enterococcus, Faecalibacterium, and Ligilactobacillus were significantly increased in the LPS12 group versus the Control group, whereas Lactiplantibacillus and Paracoccus were significantly decreased.
In our comparison of the LPS12 and LPS48 groups, the abundances of a total of 12 taxa (related to 5 genera) were found to be significantly changed (Fig. 5b). At the genus level, Pseudarthrobacter was significantly increased www.nature.com/scientificreports/ in the LPS48 group as compared with the LPS12 group, whereas Olsenella, Escherichia Shigella, Acinetobacter, and Phascolarctobacterium were significantly decreased. Finally, in our comparison of the Control and LPS48 groups, the abundances of a total of 28 taxa (including 8 genera) were found to be significantly changed (Fig. 5c). At the genus level, Lactobacillus was significantly increased in the LPS48 group as compared with the Control group, whereas Paramuribaculum, Escherichia Shigella, Enterobacter, Gp2, Brachybacterium, Lachnospira, and Olsenella were significantly decreased.
Comparison of lung microbiota functionalities. We predicted a total of 6616 Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs (KOs) and 284 pathways from the datasets. LEfSe analysis in our comparison of between-group functionalities showed that the adenosine triphosphate-binding cassette (ABC) transporters pathway (ko02010) had significantly lower inferred abundances in the LPS12 group than in the Control group (Fig. 5d), the pentose phosphate pathway (ko00030) had significantly lower inferred abundances in the LPS48 group than in the Control group (Fig. 5e), and the bacterial chemotaxis pathway (ko02030) had significantly higher inferred abundances in the LPS48 group than in the LPS12 group (Fig. 5f). www.nature.com/scientificreports/

Analysis of correlation between lung microbiota and ALI/ARDS symptom indicators.
To further assess potential correlations between LPS-induced ALI/ARDS symptoms and certain lung microbiota components or functionalities, we conducted Spearman's correlation tests on genus-level and KEGG pathway relative abundance data using the psych package in R. The genus Brevibacterium and six functionality pathways were found to have significantly strong correlations (P < 0.05; correlation coefficient > 0.7 or < − 0.7) with LPS-induced ALI/ARDS symptoms (Fig. 6a). Brevibacterium was positively correlated with serum TNF-α, IL-10, and IL-6 levels and with hematological NEU%. In terms of microbiota functionalities, three signaling pathways-Wnt (ko04310), Notch (ko04330), and chronic myeloid leukemia (ko05220)-had negative correlations with IL-1β levels, while the mitogen-activated protein kinase (MAPK) signaling pathway-yeast (ko04011) was negatively correlated with IL-10 levels. Ascorbate and aldarate metabolism (ko00053) was positively correlated with PCT and negatively correlated with PLT. Finally, basal transcription factors (ko03022) were negatively correlated with PLT and PCT and positively correlated with PLCR and MPV (Fig. 6b).

Discussion
Acute lung injury (ALI) is a common pulmonary response to a broad range of injuries or inflammation in the lungs or at other sites in the body. It is characterized by damage to the alveolar endothelial and epithelial barriers, recruitment of inflammatory cells, and onset of pulmonary edema 14 . ALI often leads to acute respiratory distress syndrome (ARDS), for which few effective therapeutic approaches exist to date 15 . Therefore, studies aimed at better understanding the pathogenic processes that develop in the injured lung are essential and might lead to the development of new methods to prevent and treat this disease.
The results of this study demonstrated that at different time points after LPS injection, along with fluctuations in systemic cytokine levels and the onset and regression of pulmonary edema, the diversity, components, and functionalities of the pulmonary microbiota in our rat model underwent significant dynamic changes. We also observed that several components and functionalities of these microbiota correlated with symptoms indicative of LPS-induced ALI.
Bacteria suspended in or attached to particles in the air are considered to be the main source of the lung microbiota. After the bacteria disperse into the lung, the microbiota is delicately balanced by bacterial reproduction rates and bacterial immigration and elimination 16 . Under normal circumstances, the rate of bacterial www.nature.com/scientificreports/ reproduction is low, while those of immigration and elimination are high 17 . We can therefore infer that under healthy conditions, the lung microbiota are highly diverse and affected by the surrounding environment. Indeed, this speculation is supported by a study by Dickson et al. on the lung microbiota of healthy mice, which demonstrated that this bacterial community is highly variable depending on cage, transportation, and supplier of mice and can better reflect the natural immune status of the host lung 18 . In lung disease, the balance of immigration and elimination is disturbed, resulting in bacteria that display competitive advantages to predominate, and thus, a decrease in diversity 19,20 .
In the present study, we observed that after LPS injection, the variation ranges of all three α-diversity indices were narrowed. These indices had all decreased significantly 48 h after LPS injection. These results, which were in line with those of the above-mentioned studies noting disturbance of the immigration/elimination balance and theorizing on the decrease in diversity, demonstrated the dynamic characteristics of changes to the microbiota diversity during systemic inflammation. In the early stage of inflammation (12 h after LPS injection), with the appearance of severe lung injury and edema, the host environment undergoes significant changes, such as increased levels of cytokine 21 and endogenous antigenic peptides 22 . The influence of host factors on the lung microbiota was strengthened, overwhelming the influence of external environmental factors; this was reflected www.nature.com/scientificreports/ by the narrowing of the microbiota diversity index ranges within groups. In the later stage of inflammation (48 h after LPS injection), after the lung microenvironment had changed and competitive bacteria became dominant, α-diversity significantly decreased. This decrease in lung bacterial diversity in the late stage of acute inflammation has been reported in other studies. For example, Poroyko et al. reported that the lung microbiota diversity was significantly reduced in a mouse model 72 h after LPS-induced ALI. They also demonstrated that the change in the lung microbiota could increase susceptibility to lung injury in model mice 5 . The association between lung microbiota diversity and lung diseases has also been reported in studies of other lung diseases, such as pneumonia, chronic obstructive pulmonary disease, cystic fibrosis (CF), and idiopathic pulmonary fibrosis, indicating that the diversity of the lung microbiota might play a critical role in bacteria-host interactions in the lung 6,23-26 . However, until now, the compositional and functional basis of the lung microbiota diversity change remained to be clarified. In our comparisons of lung microbiota components and functionalities between different groups, we found that the component abundance varied drastically between groups, while changes in the functional abundance were relatively mild. A total of 51 taxa were found in our pairwise comparisons of species abundances in the three groups, but only three differential pathways were found in functionality abundance. The very high richness and diversity of the lung microbiota could be the cause of this phenomenon. A study analyzing the microbiota in different bodily sites of mice found a total of 226 genera in the lungs, but only 24 in the gut microbiota 27 . Similarly, using a more stringent method of selecting representative sequences (aligning reads to OTUs at a 100% rather than a 97% identity threshold) in the present study, we found 582 genera in the lungs as compared with only 38 genera in the rat gut microbiota in our previous study 28 . These findings all indicated that the lung microbiota contains many more species than other bodily habitats. In contrast to its components, the functionalities of the lung microbiota seem to be less variable and more stable. A total of 284 KEGG pathways were enriched from the lung microbiota in the present study-a number comparable to the 258 KEGG pathways enriched from the gut microbiota data in another study 29 .
In our functionality comparison between the Control and LPS12 groups, we found the ABC transporter pathway in lung bacteria to be significantly decreased in the LPS12 group. The ABC transporters are a family of transmembrane proteins that can transport a wide variety of substrates across biological membranes 30 . The ABC transporter protein is reported to be involved in the process of bacterial-iron acquisition and plays an important role in bacterial-iron competition 31 . Under physiological conditions, iron metabolism is tightly regulated to prevent iron toxicity. During infection, the host's iron-scavenging activity level increases, and the response to inflammatory mediators leads to an increase in hepcidin, which can lead to hypoferremia 32 . Bacteria, like other cells, require iron for many basic cellular functions and metabolic pathways 33 , and therefore iron acquisition is essential for bacterial colonization in the competitive environment of the inflamed lungs. Studies have reported that in CF, lung bacteria adapt to and deploy specific systems based on the availability of iron, bioavailability of iron pools, stage of infection, and presence of competing bacteria 31 . Therefore, the change in the abundance of the ABC transporter pathway 12 h after LPS injection might reflect the adaptation of the lung bacteria to inflammatory conditions. In our comparison of the LPS12 and LPS48 groups, we found that the bacterial-chemotaxis pathway in lung bacteria was significantly increased. This pathway is reportedly correlated with bacterial colonization on epithelial surfaces 34 . Tamar et al. showed that an acquired mutation in bacterial chemotaxis can change the inherent swimming behavior of bacteria that could barely reach the epithelial surface beyond the hydrogel mucosal layer, leading to a key enhancement in surface colonization 35 . Upregulation of the bacterial-chemotaxis function of the lung microbiota after 48 h of LPS infection might therefore have resulted from bacteria adapting to changes in an environment of edema.
Furthermore, in our comparison of the Control and LPS48 groups, we found the pentose phosphate pathway of the lung microbiota to be significantly decreased. The products of this pathway are known to be nicotinamide adenine dinucleotide phosphate (NADPH) and pentose phosphate. NADPH can provide hydrogen donors that are involved in a variety of metabolic reactions, preventing oxidation of various proteins and enzymes 36,37 , while pentose phosphate is necessary for nucleic acid biosynthesis. Studies have shown that nucleosides can increase the flux of the pentose phosphate pathway, thereby improving the effectiveness of DNA damage repair and improving cell survival 38 . Based on this evidence, we presumed that the reduction of the pentose phosphate pathway might reflect microbiota with lower antioxidant capacities and less efficient DNA damage repair ability after lung injury.
In this study, we also explored the correlation between physiological indicators that changed significantly during the development of ALI, as well as lung microbiota components and functionalities. We obtained a number of indicative results. In terms of microbiota composition, the abundance of the genus Brevibacterium was found to have a strong positive correlation with parameters, such as TNF, IL-6, IL-10, and NEU%. This genus is composed of organisms commonly found in the CF lung microbiome, as demonstrated in another study 39 . Brevibacterium has also been reported to be a causative agent of allergic alveolitis 40 . Our results were consistent with the above findings, suggesting that Brevibacterium might play a role in the progression of inflammation in the lungs.
In terms of lung microbiota functionality, the abundances of four signaling pathways-Wnt (ko04310), Notch (ko04330), chronic myeloid leukemia (ko05220), and MAPK-yeast (ko04011)-were found to have strong negative correlations with serum IL-1β and IL-10 levels, indicating that these pathways might play roles in the interaction between the lung microbiota and host immunity and should be addressed in more detail in follow-up studies. Finally, the abundances of the ascorbate and aldarate metabolism (ko00053) and basal-transcription factor (ko03022) pathways of the lung microbiota were found to be strongly correlated with four platelet-related physiological indicators: PLT, PCT, PLCR, and MPV. Current studies have found that platelets are important effectors of experimental ALI and clinical ARDS, playing diverse and complex roles in vascular-barrier integrity, organ repair, inflammation, activities across the immune continuum, and antimicrobial host defense 41  www.nature.com/scientificreports/ study 42 . The strong correlation we found between the two above-mentioned bacterial-function pathways and platelet-related indicators can provide clues for further elucidating the complex role of platelets in the occurrence and development of ALI.
In conclusion, this study will serve to provide greater insight into the dynamic changes in the diversity, components, and functionalities of the lung microbiota at different time points of lung injury. The representative differential components and functionalities of the lung microbiota at the time points we identified in this study will promote further understanding of the interaction between the host and the airway microbiota during lung injury and repair. Further exploration of the influence of the lung microbiota on host immunity and physiology might allow for better targeted approaches for the prevention and management of ALI and ARDS.

Methods
Animals and ethics statement. Eighteen  Study design. Rats were randomly divided into three groups: Control (Control), LPS12, and LPS48. They were allowed to acclimatize to laboratory conditions for 7 days after arrival in the lab. The LPS12 and LPS48 groups received i.p. injections of LPS (Escherichia coli 055: 10 mg/kg body weight; Sigma-Aldrich, St. Louis, MO, USA), while the Control group received i.p. injections of a 0.9% NaCl solution. Rats in the LPS12 group were anesthetized 12 h after injection, and the Control and LPS48 groups were anesthetized with isoflurane 48 h after injection. Rats were sacrificed by exsanguination after blood was collected from the posterior vena cava. An aliquot of the whole blood from each mouse was centrifuged at 3000 rpm for 10 min, and the serum samples were stored at − 80 °C. The middle lobe of the right lung was dissected using sterile scissors and forceps immediately after opening the chest, and the lobes were placed in 5-ml sterile tubes containing 4 M guanidine thiocyanate solution as a preservative. The tubes were then stored at − 80 °C pending characterization of the lung microbiota. The upper lobe of the right lung was dissected, fixed in 10% formaldehyde solution, and subjected to histopathological examination. Finally, the left lung was dissected for lung W/D weight ratio calculation.
Lung W/D weight ratio. Lung W/D weight ratio was measured to quantify pulmonary edema levels. The wet weight of the left lung was recorded immediately after dissection. The lung was then placed in an incubator at 80 °C for 48 h, and then the dry weight was recorded.
Histopathological examination. Histological tissues were fixed in formalin for 12 h, cut into 5-μm thick sections, and stained with hematoxylin and eosin (H&E). The reagent used for histopathological examination was purchased from Wuhan Servicebio Technology Co., Ltd., Wuhan, China. The slides were de-identified and then observed by a pathologist.
An investigator blinded to group settings screened and analyzed 12-17 adjacent fields (× 100 objective magnification) covering the entire slide of each sample using ImagePro Plus software version 5.0 (Media Cybernetics, Silver Spring, MD, USA). The number of alveoli, sum of alveolar circumferences, and percentage of tissue area (red signal) per field were quantified for all screened images and further adjusted by the ratio of lung tissue area in each image. The tissue (red signal) and alveolar (white signal) areas were each set as hue (0-255) and saturation (0-255), and the intensities of the red and white signal areas were set at 165-255 and 0-164, respectively (Supplemental Fig. 1).
Microbiota sequencing. Total bacterial deoxyribonucleic acid (DNA) was extracted using a Power Soil DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA, USA). After quality and quantity checks of the DNA sample at the ratios of 260/280 nm and 260/230 nm, the V3-V4 hypervariable region of the bacterial 16S ribosomal ribonucleic acid (rRNA) gene was amplified using the common primers 338F (5′-ACT CCT ACG GGA GGC AGC AG-3′) and 806R (5′-GGA CTA CHVGGG TWT CTAAT-3′) combined with adapter and barcode sequences. Second-generation sequencing of purified, pooled polymerase chain reaction (PCR) sample products was performed on a HiSeq 2500 platform (Illumina, Inc., San Diego, CA, USA; 2 × 250 paired ends) at Biomarker Technologies Corporation (Beijing, China).

Bioinformatic analyses.
Raw sequence reads were processed using the UNOISE pipeline of the Usearch v11.0.667linux32 program (www. drive5. com/ usear ch/). High-quality sequences were classified into zero-radius operational taxonomic units (ZOTUs). The Ribosomal Database Project (RDP) classifier was then used to anno-