Effects of high-fat diet on the epithelial-mesenchymal transition of respiratory tract through the glyoxylic acid cycle of pulmonary microbes and the intervention of saturated hydrogen

Background High fat diet is extensively studied to be associated with trending metabolic diseases. In addition to type 2 diabetes and hypertension, high fat diet is strongly associated with asthma and other respiratory diseases among children however, the pathogenicity regarding these pulmonary diseases begs for extensive research. This study investigated the mechanism of the epithelial-mesenchymal transition of respiratory tract, induced by changes in lung microecology with the intake of high-fat diet. 80 ve-week-old C57BL6/J male mice were randomly divided into normal control group, normal hydrogen group, high-fat group and high-fat hydrogen group, making 20 mice in each group. The weight of the mice were measured on weekly basis. 6 mice from each group were executed at every second week. Blood sample was collected for lipid testing, lung tissues were collected for 16SrRNA gene sequencing, HE staining, immunouorescence and quantitative real-time PCR (qPCR). Results Compared with the normal diet group, mice on the high-fat diet group showed increased inammatory cell inltration, decreased expression of e-cadherin (E-cad) and increased expression of Twist. There were signicant differences in the composition of bacteria in the lung, and the expression of isocitrate lyase (ICL) gene in Pseudomonas aeruginosa , Staphylococcus aureus and Acinetobacter baumannii , which were signicantly associated with asthma were seen with a signicant increasing trend. After the treatment of saturated hydrogen, the changes in lung microbial population, lung tissue inltration of inammatory cells and the transformation of epithelial stroma caused by high-fat diet were moderately alleviated. Conclusion High fat diet can affect the process of airway epithelial stroma by altering the glyoxylate cycle of pulmonary microbes while the pathological process are alleviated by saturated hydrogen by acting on glyoxylate


Introduction
High fat diet is not only associated with type 2 diabetes (T2DM), hypertension and other metabolic diseases, but is also studied to be correlating signi cantly with asthma and other respiratory diseases.
Based on the previous studies, the incidence of asthma and asthmatic diseases in children with high-fat diet is signi cantly higher than children without high-fat diet [1,2] however, the pathogenicity underpinning high-fat diet with asthma is still in the mystery. Asthma is a chronic airway in ammatory disease, which often occurs with airway remodeling [3,4]. Airway remodeling is based on chronic airway in ammation, which leads to the irreversible changes of airway structure. Subepithelial brosis is one of the core characteristics of airway remodeling, which is associated with the severity of asthma and one of the main causes of death in patients with severe asthma [5,6]. Airway epithelial mesenchymal transition (EMT) is a cell biological process in which epithelial cells are transformed into stromal phenotypes through a speci c process, which plays an important role in airway remodeling of asthma. Airway remodeling can improve the degree of subepithelial brosis of airway through EMT [7]. Early prevention and treatment of airway remodeling and reduction of brosis in asthmatic patients is an intervention that do not only control the level of asthma but also enhance the prognosis.
Modern day science is blessed with 16SrRNA gene sequencing technology that enhanced microbial characterization and ascertained more than 30 bacterial species planted in the lungs [8]. Going by the published studies, the diversity of microbes in the lung is biological imperative in maintaining the normal airway function, and in addition to bacterial, fungi and viruses constitute the intrapulmonary microecology. Studies based on common respiratory diseases, alteration or changes in the diversity of microbes are associated with asthma and pulmonary brosis [9][10][11]. A number of studies meet a consensus that in respiratory diseases such as asthma and chronic obstructive pulmonary diseases (COPD), abnormal increase in abundance of species such as Pseudomonas aeruginosa, Acinetobacter baumannii, Staphylococcus aureus, Klebsiella pneumoniae, and Haemophilus in uenzae was observed [12,13].
The glyoxylic acid cycle is a metabolic pathway in prokaryotes, lower eukaryotes, and plants, and it is essential for the energy metabolism. Studies have shown that the glyoxylate cycle also exists in many opportunistic pathogens. These opportunistic pathogens use the host's fatty acid degradation products to synthesize substances required for their normal growth and reproduction, split macrophages from the inside, and then expand in abundance in the host [14]. Fatty acids are decomposed into acetyl-CoA by beta-oxidation, which form citric acid with oxaloacetate produced from two acetates through the glyoxylic cycle. Under the action of aconitase, citric acid produces isocitrate, which is broken by isocitrate lyase (ICL) to produce glyoxylic acid and succinic acid. Next, glyoxylic acid and acetyl-CoA produce malic acid under the catalysis of malic acid synthetase and malic acid is dehydrogenated to oxaloacetic acid under the catalysis of malic acid dehydrogenase. The succinic acid produced in this cycle can be used to synthesize sugars and other cellular components, which de ned ICL as the most critical enzyme in the glyoxylic acid cycle [15].
Number of studies have shown that viral infection and high-fat diet can lead to the accumulation of fatty acids in cells. Therefore, we speculate that the high-fat airway microenvironment caused by high-fat diet leads to an imbalance of airway microecology through the promotion of the glyoxylate cycle of microbes. We further hypothesized that the abnormal increase of opportunistic pathogens mediate chronic airway in ammation, thus promoting the occurrence of asthma and airway remodeling.
Hydrogen (H 2 ) has antioxidant, anti-in ammatory, and anti-apoptotic properties and is the most abundant chemical element in the world [16]. In 2007, Ohsawa et al [17] con rmed that inhaling 2% hydrogen could lead to selective scavenging of hydroxyl radicals (OH) and peroxynitrite anions (ONOO − ), which signi cantly enhance cerebral ischemia-reperfusion injury in rats. After long-term drinking of hydrogenrich water by obese mice with T2DM, liver malondialdehyde level, fat content, blood sugar, and blood total glyceride (TG) levels all decreased [18,19], suggesting that hydrogen could improve metabolic disorders of lipids and glucose, as well as the metabolic syndrome. Therefore, we use saturated hydrogen as a treatment factor to explore whether it can improve the effect of lipid metabolism disorder caused by highfat diet on the transformation of respiratory epithelial stroma.

Results
The mice on the high-fat diet show signi cant increase in body weights and dyslipidemia, which was aborogated by saturated hydrogen.
As shown in Figure 1A, the average weight of mice in the HF+H 2 group on the 42 nd day was 34.05 ± 2.15 g, 20% higher than that of mice in the Control group, which was 27.45 ± 1.15 g, meaning that the average weight of mice in the HF+H 2 group meet the weight standard of obese mice. The average weights of mice in the H 2 and HF+H 2 groups decreased after treatment with saturated hydrogen when compared with those in the corresponding Control and HF groups, as shown in Figure 1B, indicating that saturated hydrogen inhibited the increase in body weights of mice.
On day 28, total cholesterol (TC) and low-density lipoprotein (LDL) in the HF group increased signi cantly compared with those in the Control group, and LDL in the HF+H 2 group decreased considerably compared with the HF group (Figure 1 C, E). On day 42, TG, TC, and LDL in the HF group increased markedly compared with those in the Control group, which decreased substantially after saturated hydrogen treatment ( Figure 1C, D, and E). However, high-density lipoprotein (HDL) in the HF group decreased remarkably compared with the Control group, and it increased dramatically in the HF+H 2 group compared with the HF group ( Figure 1F), suggesting that saturated hydrogen reduces TC, TG, and LDL levels and increases the content of HDL in the peripheral blood of obese mice.
In the high-fat diet group, the in ltration of in ammatory cells in the lung tissues increased and EMT increased.
It can be seen from the HE staining results of the lung histopathological sections ( Figure 2A) that after day 42 on high-fat diet, the in ltration degree of in ammatory cells around the bronchus in HF group increased compared with that of the Control group, with a small amount of in ammatory cells in ltrating in the lumens and a little shedding of bronchial cilia. Compared with the HF group, the pathological changes were decreased in the HF+H 2 group. The results of qPCR showed that the expression of Ecadherin was signi cantly decreased and twist was signi cantly increased in the HF group compared with the Control group. Compared with the HF group, the expression of E-cadherin gene in the HF+H 2 group increased, and twist decreased, but there was no signi cant statistical difference ( Figure 2B). The results of immuno uorescence showed that the expression of E-cadherin protein decreased signi cantly at day 42, while the expression of twist protein increased signi cantly in the HF group compared with the Control group, which was in consistent with the results of qPCR (Figure 2 C, D). These results have shown that high-fat diet can lead to in ammation of lung tissue and accelerate the process of EMT of respiratory tract.
The normal diet group and the high-fat diet group show signi cant differences in the number and composition of microbes in the lungs.
After the question sequence is eliminated, the total sequencing quantity is 559453. The length distribution of the sequence contained in all samples was statistically analyzed in R software, and the sequence length was mainly distributed between 250-450bp (Supplementary gure 1). Operational taxonomic unit (OTU) whose abundance value was less than 0.001% (1/100,000) of the total sequencing amount of all samples was removed, and the distribution of the remaining OTU in each group is shown as ( Figure 3A, Supplementary gure 2). The results showed that the number of OTU on the HF group was signi cantly higher than that of the Control group at the level of phylum, class, order, genus and species. The analysis of the same OTU between the groups showed that the number of unique OTU in the HF group increased signi cantly compared with that of the Control group, and the number of unique OTU in the Control and HF groups all decreased after hydrogen administration ( Figure 3B). According to the OTU classi cation and classi cation status, the speci c composition of each sample at the level of phylum, class, order, family, genus and species classi cation was obtained. The results showed that the number of bacteria in the lung of mice in the HF group was signi cantly higher than those in the Control group, and there was no signi cant difference between the hydrogen group and the non-hydrogen group ( Figure   3C, D).
The diversity and difference of micro ora in the lung of each group.
The rarefaction sparse curve (Chao1), speccaccum species accumulation curve and abundance level curve in alpha diversity analysis tend to be gentle, indicating that the sequencing depth is enough to re ect the diversity and richness of the microbial community contained in the community sample, and the total number of OTUs in the community will no longer increase by signi cant. With the addition of new samples, and the abundance difference among OTUs in the community, the difference is small and the community composition is high ( Figure 4A). According to the alpha diversity index (Chao1, ACE, Shannon, Simpson) ( Figure 4B), the diversity of bacteria in the lung of mice in the high fat diet group was signi cantly higher than that of the normal diet group, and the diversity of mice in the normal hydrogen feeding group was lower than that of the normal control group.
Discriminant analysis by PLS-DA partial least squares and beta diversity based on weighted UniFrac NMDS UPGMA cluster analysis chart of UniFrac distance matrix ( Figure 4C) shows that there are obvious differences in the structure of microbial community in the lung of the HF group compared with the Control group, and also some differences in the hydrogen feeding group compared with the non-hydrogen feeding group, and the difference of each mouse within the same group is small. The results show that the animal classi cation model is effective.
Identi cation of key microbes with signi cant difference in pulmonary micro ora of mice in each group.
Using R software, cluster analysis and heat map ( Figure 5A) were made for the rst 50 genera of abundance, which showed that there were signi cant differences between the high-fat diet group and the normal diet group, and there were also some differences between the hydrogen group and the nonhydrogen group.
Compared with the Control group, the abundance of the common opportunistic pathogens Acinetobacter, Pseudomonas, Corynebacterium, Strepcoccus, Clostridium, Haemophilus and Porphyromonas increased signi cantly in the HF group. After hydrogen administration, the abundance of Acinetobacter, Clostridium and Porphyromonas decreased signi cantly. There was a signi cant decrease in Bi dobacterium in the HF group compared with that of the Control group ( Figure 5B, C). These results showed that high-fat diet signi cantly alters the diversity of microbes in the lung of mice, opportunistic pathogens increased while probiotic microbes decreased, thus increased susceptibility of respiratory in ammation.

Functional prediction of bacterial metabolism
The 16S rRNA gene sequence was predicted in KEGG, COG and rfam3 functional spectrum databases by PICRUSt functional prediction analysis. The predicted functional spectrum data are clustered according to the abundance distribution of functional groups or the similarity between samples, and the functional groups and samples are sorted according to the clustering results. Using R software, the functional groups in the top 50 of the abundance were clustered and analyzed, and a heat map ( Figure 6A) was drawn, showing that there were signi cant differences between the high-fat diet group and the normal diet group. R software was used to calculate the number of common functional groups of each sample, and the proportion of common and unique functional groups of each sample was visualized through Venn diagram. The results showed that the number of functional groups in the HF group was signi cantly increased compared with the Control group, and that in the hydrogen group was decreased compared with the non-hydrogen group ( Figure 6B).
Violin diagram was drawn to show the abundance distribution of the predicted functional groups in each sample ( Figure 6C). The results showed that the ora abundance of the HF group involved in lipid metabolism, energy metabolism and amino acid metabolism was signi cantly increased compared with the Control group, while the ora abundance of the HF group involved in glycan biosynthesis, metabolism and glucose metabolism was decreased. The abundance of bacteria involved in cell communication, cell migration, cell growth and death increased signi cantly. The number of bacteria involved in the transcription of genetic information increased and the abundance of bacteria involved in translation decreased signi cantly. The abundance of bacteria involved in neurodegenerative diseases and cardiovascular diseases increased signi cantly. The abundance of bacteria involved in the body's immune system decreased signi cantly and the abundance of bacteria involved in the endocrine system and circulatory system increased signi cantly.
The expression of ICL gene in P. aeruginosa, S. aureus, A. baumannii, K.pneumoniae, C. albicans and X. maltophilia To investigate whether changes in microbes in the lungs of mice on a high-fat diet were due to changes in glyoxylic acid circulation in bacteria caused by a high-fat environment. We designed primers using the ICL sequence, a key enzyme in the glyoxylic acid cycle, to detect the expression of ICL genes in several clinical pathogens signi cantly associated with the development of asthma. qPCR results (Figure 7) showed that the ICL gene expression in P. aeruginosa, S. aureus and A. baumandii in the HF group was signi cantly increased compared with that of the Control group, and P. aeruginosa was signi cantly decreased after hydrogen supplementation. However, K. pneumoniae, C. albicans and X. maltophilia showed no signi cant changes.

Discussion
The diversity of microbes found in the lung is imperative to maintaining the normal airway function. Our results show that high-fat diet can change the number and microbial species in the lung, some opportunistic pathogens that are signi cantly associated with the occurrence of respiratory diseases are increasingly abundant, while the abundance of Bi dobacterium is reduced. These increased opportunistic pathogens are believed to mediate the occurrence of chronic airway in ammation, airway hyperresponsiveness and other pathological changes. The results obtain from qPCR and immuno uorescence also showed a signi cant reduction in the expression of E-cad, a gene related to the epithelial stromal transformation, while the expression of twist was signi cantly increased in the HF group when compared with the Control group. This indicates that high-fat diet accelerated the process of the epithelial stromal transformation of the respiratory tract, which may affect the occurrence of airway remodeling.
Although we could not establish a correlating result from a similar study, there are couple of studies on obesity related high-fat diet and one such concluded that TH2 immune response and immunopathologic characteristics in murine model of eosinophilic oesophagitis was worsen severely due to high-fat diet [21]. Furthermore, maternal diet enriched with high-fat is studied to be associated with impaired fetal lung development [22].
Both bacteria and fungi can use glyoxylic acid cycle to synthesize their own pathogenic substances, which can survive in macrophages for a long period of time and expand in the host. It has been shown that many pathogenic and opportunistic microbes express isocitrate lyase and start glyoxylate cycle. In the process of infection, P. aeruginosa can activate glyoxylate cycle as an energy metabolism pathway, make it survive in macrophages and spread in large quantities [23]. Glyoxylate cycle is also the main pathway of material and energy metabolism of A. baumannii in macrophages. A. baumannii enters the host via respiratory tract, connects laminin and bronectin via sialic acid, and enters macrophages via multiple auxiliary receptors on the surface of the host cell [24,25]. In macrophages, A. baumannii can start the glyoxylate cycle as an energy metabolism pathway to synthesize its own pathogenic substances [26]. C. albicans, a human opportunistic pathogenic fungus, is engulfed by macrophages to form phagocytes at the early stage of infection. Through glyoxylic acid cycle, it uses fatty acids as energy source, and synthesizes its own virulent factor to split macrophages from the inside, and expand in the host [27]. Therefore, in order to investigate whether the changes of microbes in the lung during high-fat diet are related to glyoxylate cycle, we detected the expression of ICL gene of several conditional pathogens intensively related to respiratory diseases. The results showed that the ICL gene of P. aeruginosa, S. aureus and A. baumannii in the lungs of mice fed with high-fat diet had abnormal high expression. The high-fat diet changed the glyoxylate cycle of these pathogenic bacteria in the lung, and the speed in utilizing fatty acids to synthesize the substances needed for self-metabolism and pathogenic substances was accelerated, resulting in the increase abundance of these opportunistic pathogens. However, we did not nd any signi cant changes in ICL of fungi, and we speculated that this may be associated with the antagonistic effect of bacteria on fungi.
Our results also showed that saturated hydrogen treatment could inhibit the growth and expansion of opportunistic pathogens in the lung and the increase ICL gene expression in S. aureus. Therefore, we speculate that saturated hydrogen can regulate the bacterial ora in the lung by acting on the glyoxylate cycle of the microbes and improve the epithelial stromal transformation caused by high-fat diet. However, we recommend further in vitro and in vivo studies to validate our ndings.

Conclusion
This study found that the high-fat airway microenvironment caused by high-fat diet can change the abundance of respiratory tract microbes via glyoxylic acid cycle, thus affecting the process of epithelial stromal transformation and the occurrence of respiratory diseases. Saturated hydrogen is highly recommended based on our study as a new substance for improving respiratory diseases. In the annals of prevention and treatment of respiratory infections and diseases, this study provides new ideas and research basis as a new therapeutic approach.

Animal grouping
A litter of ve-week-old C57BL6/J male mice weighing 16-20 g (purchased from Hunan Tianqin Biological Technology Co., Ltd.) were acclimatized for a week during 12hrs of a light/dark cycle at a constant temperature of 25°C and supplied with sterile water (once a day). The mice were randomized into normal control (Control or A), saturated hydrogen (H 2 or B), high-fat (HF or C), and saturated hydrogen + high-fat (HF+H 2 or D) groups (n = 20 each). Control and H 2 group mice were fed with sterile water and low-fat control feed, while HF and H 2 +HF group animals were given sterile water and 60% high-fat model feed (purchased from Nantong Trophy Feed Technology Co., Ltd.). Additionally, mice in the Control and HF groups were given 0.5 ml of saline intragastrically once a day, and those in the H 2 and HF+H 2 groups were given 0.5 ml of saturated hydrogen saline once a day. Saturated hydrogen saline was prepared at the Center of Modern Analysis and Detection of Central South University by dissolving molecular hydrogen into normal saline at high pressure (13.5 Mpa). The formulated saturated hydrogen solution was freshly prepared once every week and stored in aluminum packaging to ensure the hydrogen concentration remained > 0.6 mmol/L. The weights of the mice were taken once in a week. Six mice from each group were sacri ced in the 2 nd , 4 th , and 6 th weeks from the onset of the experiment. Blood, and lung tissues were collected for subsequent analyses. This study was conducted according to the Declaration of Helsinki and approved by the Medical Ethical Committee of the Xiangya School of Medicine. The same experiments were repeated 3 times.
Monitoring body weight and TC, TG, LDL, and HDL concentrations in peripheral blood Mice were weighed on days 0, 7, 14, 21, 28, 35, and 42, and the gures in kilogram were recorded appropriately. Blood samples were collected from veins using the eyeball extraction method; 35µl of blood was collected from each mouse for blood lipid tests according to the speci c operation steps of blood lipid tester and blood lipid test card (Aikang Biotechnology Co., Ltd., Hangzhou, China).

Hematoxylin-eosin (HE) staining
After anesthetizing mice with ether, lung tissues were excised and rinsed with a 0.9% sodium chloride solution. The resected lung tissues were xed in a 10% formaldehyde solution for 48hrs, embedded in para n, and cut into 4µm-thick sections. The tissue sections were then heated to 60°C for 2hrs, dewaxed with xylene (twice, 15 min each), and dehydrated using an increasing alcohol gradient (75%, 95%, and 100%) for 5 min each. HE staining was performed, and the specimens were observed for pathological morphology under an optical microscope.
Determining the expression of E-cad, N-cad, Slug, Snail, Twist and the ICL gene of S. aureus, P. aeruginosa, K. pneumoniae, A. baumannii, X. maltophilia, C. albicans in the lung tissues using real-time quantitative PCR (qPCR) Primers were designed using the Primer Premier 5.0 program (Supplementary table 1) and synthesized by Dingguochangsheng Biotechnology Co., Ltd., Beijing, China. Lung tissues were collected from each group of mice, and total RNA was extracted with TRizol. RNA purity was analyzed using a nucleic acid analyzer. cDNA was synthesized by reverse transcription and ampli ed using a qPCR kit (TansGen Biotech Technology Co., Ltd., Beijing, China) according to the manufacturer's instructions. The reaction mixture consist of 1.5µl cDNA, 12.5µl SYBR Green qPCR Master Mix, 1.5µl each of the forward and reverse primers, and diethylpyrocarbonate (DEPC) water, making a total volume of 25µl. qPCR cycling conditions were: 94°C for 3 min, and 35 cycles of [94°C for 45s, 51°C for 45s, and 72°C for 45s], and the relative mRNA expression was calculated using the 2 △△Ct method [20].

Immuno uorescence
Excised lung tissue sections were processed as described above and boiled with 0.01M citrate buffer for antigen retrieval. The tissues were then washed twice with PBS (3 min each), xed with 4% paraformaldehyde for 5 min, washed with phosphate buffered solution (PBS) again (same condition as the rst wash), blocked with normal goat serum for 20 min, and incubated overnight with 5 µl of E-cad (bs-10009R, BEIJING Bioss Biological Technology Co., Ltd.) or Twist (bs-2441R, BEIJING Bioss Biological Technology Co., Ltd.) rabbit anti-mouse primary antibody in a wet box at 4°C. On the following day, the tissue sections were washed thrice with PBS, incubated with a CY3-conjugated goat anti-rabbit secondary antibody (GB21303, Wuhan Seville Biotechnology Co., Ltd.) at room temperature for 1 h, washed three times with PBS, and counterstained with 4'6-diamidino-2-phenylindole (DAPI) (100 ng/ml) for 10 min. After removing moisture from the tissues, the slides were sealed with an anti-uorescence quenching agent.
Changes in lung microbiota were detected by 16SrRNA gene sequencing Technology DNA was extracted from lung tissues of four groups of mice. The extraction quality of DNA was detected by 0.8% agarose gel electrophoresis, and the DNA was quanti ed by ultraviolet spectrophotometer.
Primers were designed with microbial ribosomal RNA as the target and according to the conserved region in the sequence. Sample speci c Barcode sequences were added to further PCR ampli cation of rRNA gene variable regions (single or consecutive multiple) or speci c gene fragments. PCR ampli cation was performed by NEB's Q5 high-delity DNA polymerase, and the number of cycles was strictly controlled to keep the number of cycles as low as possible while ensuring the same ampli cation conditions for the same batch of samples. PCR ampli cation products were detected by 2% agarose gel electrophoresis, and the target fragments were gelled and recovered using AXYGEN gel recovery kit. Based on the preliminary quantitative results of electrophoresis, uorescence quanti cation was performed for the PCR ampli cation of the recovered products. The uorescent reagent was the quantit PicoGreen dsDNA Assay Kit and the quantitative instrument was the Microplate reader (BioTek, FLx800). According to the uorescence quantitative results and the sequencing quantity demand of each sample, the samples were mixed according to the corresponding proportion. Illumina MiSeq sequencing was used as an example to prepare a sequencing Library using Illumina's TruSeq Nano DNA LT Library Prep Kit, and then highthroughput sequencing was performed on the machine. (Detected in Shanghai paisenno Biotechnology Co., Ltd)

Statistical analysis
The SPSS21.0 software was used for statistical analysis, and the data were expressed as mean ± standard deviation. The t-test was used for comparisons between two groups, variance analysis was used for comparisons among multiple groups, and the least signi cant difference (LSD) was used for comparisons between intra-group differences. P-value < 0.05 was considered statistically signi cant.  Control; + p < 0.05, ++ p < 0.01 and +++ p < 0.001 vs. Control; # p < 0.05 and ## p < 0.01 vs. HF.)  The difference analysis of micro ora in the lung of each group and the identi cation of key microbes. A, at the genus level, cluster analysis of the top 50 bacteria in each sample. Samples were rst clustered based on their similarity, and then arranged horizontally according to the clustering results. The taxa are also clustered according to the similarity of their distribution in different samples and arranged vertically according to the clustering results. In gure A, red represents the genus with higher abundance in the corresponding sample, while green represents the genus with lower abundance. B, Cluster analysis diagram of taxa with signi cant differences among groups. From the inner circle to the outer circle are phylum, class, order, family, and genus. The fan area represents the abundance of bacteria. C, The abundance of the rst 20 taxa with the most signi cant differences in each group. In the gure, the abscissa is the rst 20 taxon with the most signi cant difference, and the ordinate is the sequence quantity of each taxon in various groups. The "fat and thin" of "violin" re ects the density of sample data distribution, and the wider the width, the more samples corresponding to this sequence quantity. The border of the boxplot represents the Interquartile range (IQR), the horizontal line represents the median value, the upper and lower tentacles represent the IQR range of 1.5 times beyond the upper and lower quartile, and the symbol "•" represents the extreme value beyond the range. Prediction of microbial metabolism. A, The heat map of clustering analysis of functional groups in the rst 50 abundances. The samples were rst clustered according to the similarity of the abundance distribution of functional groups, and then arranged horizontally based on the clustering results.

Abbreviations
Functional groups are also clustered according to the similarity of their distribution in different samples, and arranged vertically according to the clustering results. In the gure, red represents the functional group with higher abundance in the corresponding sample, while green represents the functional group with lower abundance. B, Venn diagram. Each ellipse represents a set of samples. The overlapping regions between the ellipses indicate the common functional groups among each group, and the number of each block indicates the number of common or unique functional groups contained in the block. C, Analysis diagram of KEGG database for prediction of microbial metabolism function. In the gure, the abscissa is the second functional group of KEGG, and the ordinate is the relative abundance of each functional group in each group. The "fat and thin" of the "violin" re ects the density of the distribution of sample data. The wider the width, the more samples there are under this abundance. The border of the boxplot represents the Interquartile range (IQR), the horizontal line represents the median value, the upper and lower tentacles represent the IQR range of 1.5 times beyond the upper and lower quartile, and the symbol "•" represents the extreme value beyond the range. Changes in the expression of ICL genes in several bacteria or fungi signi cantly associated with respiratory disease (n = 6). A-F was the ICL expression level of P. aeruginosa, S. aureus, A. baumannii, K. pneumoniae, C. albicans and X. maltophilia in the lung tissues of each group detected by qPCR. (* p < 0.05 and ** p < 0.01 vs. Control; # p < 0.05 vs. HF.)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.