Qing-Kai-Ling oral liquid was supplied by Guangzhou Baiyunshan Mingxing Pharmaceutical Co., Ltd. (Guangzhou, China). It was extracted as a mixture including Margaritifera Concha (50 g), Gardenia Jasminoide (25 g), Buffalo Horn (25 g), Radix Isatidis (200 g), Baicalin (5 g), Flos Lonicerae (60 g), cholic acid (3.25 g), and hyodeoxycholic acid (3.75 g) in 1,000 mL of sterile water. The above 1,000 mL QKL oral liquid was further concentrated into 500 mL liquids for animal experiments.

Male Sprague-Dawley (SD) rats weighing 170–200 g were supplied by Bes Test Bio-Tech Co., Ltd. (Zhuhai, China). The scheme was authorized by the Laboratory Animal Center of Sun Yat-sen University (ethics No: SYSU-IACUC-2021-000511). After a 7-day acclimation period, 24 SD rats were randomly allocated to the Control group, Model group, QKL group and dexamethasone (Dex) group (n = 6 in each group). Lipopolysaccharide (LPS) was continuously injected intraperitoneally into rats (4 mg/kg/day) for 7 days to induce pneumonia except in the Control group. After model establishment for 7 days, the rats in the QKL and Dex groups were orally treated with QKL (10 mL/kg/day) and Dex (2 mg/kg/day), respectively, for 12 days. While the other two groups received equal volumes of sterile PBS buffer by gavage for 12 days. On the 19th days, rat feces were collected using metabolic cages with ice-immersed 10 mL EP tubes underneath. Rats were sacrificed under anesthesia with ether after overnight fasting. Blood samples (plasma and serum), lung tissue and intestinal contents of rats were collected and preserved at −80°C. The whole animal experiment process was depicted in Figure 1A.

The right upper lung tissue was excised, immersed in 10% formaldehyde, and embedded in paraffin. Then, sliced 5-μm sections were stained using hematoxylin and eosin (H&E). A semi-quantitative score was employed to assess lung tissue lesions, including edema, inflammation of alveolar and interstitial tissues, and hemorrhage of alveolar and interstitial tissues. The grades were as follows: no injury, mild injury, mild to moderate injury, moderate to severe injury and severe injury, representing grades 0, 1, 2, 3, and 4, respectively (Rotta et al., 1999).

The levels of four inflammatory cytokines in plasma, including IL-1β, TNF-α, IL-10, and monocyte chemotactic protein-1 (MCP-1), were measured using the Bio-Plex 200 chip system (Bio-Rad Laboratories, CA, United States). Additionally, two oxidative stress indices in serum, superoxide dismutase (SOD) and nitric oxide (NO), were detected by testing kits (Nanjing Jiancheng Biotechnology, Nanjing, China). The content of myeloperoxidase (MPO) in serum was detected by an enzyme-linked immunosorbent assay kit (Cloud-Clone Corp., Houston, TX, United States). All procedures were operated in accordance with the manuals of the manufacturer.

The microbiome of 18 feces samples (the Control, Model and QKL groups with 6 samples in each group) was for testing. Bacterial DNA was extracted from the samples using the HiPure Stool DNA Kits from Magen, China. Then, the V3-V4 regions of bacterial gene were amplified by the primer pairs: 341F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GGACTACHVGGGTATCTAAT-3′). After the amplified DNA was purified and detected, the obtained amplicons were mixed in equimolar amounts and sequenced using the Illumina 2500 platform at Gene Denovo Bio-Tech Co., Ltd., in Guangzhou, China. Then, the sequenced data treatment was as follows: raw reads obtained from the sequencing were filtered using FASTP v0.18.0 to acquire clean reads. Then paired and clean reads were merged as raw tags using FLASH v1.2.11, and noisy sequences of raw tags were also further filtered. Furthermore, the obtained clean tags were classified into operational taxonomic units (OTUs) of more than 97% similarity. The obtained OTU sequences were annotated using the RDP Classifier. Finally, a PICRUSt analysis was conducted to infer the ecological functions of intestinal microbial communities by predicting gene functions. As a result, the sequencing data yielded between 22,038 and 50,820 sequences per sample, with a minimum length of 201 bp and a maximum length of 469 bp. These data could be employed to analyze the composition and diversity of the intestinal flora and to identify alterations that may be associated with QKL treatments.

Data were presented as the mean ± SEM. Using OriginPro v9.1.0 and GraphPad Prism v9.0 to analyze and visualize the data. Statistical differences among groups were analyzed using one-way analysis of variance (ANOVA) followed by Tukey’s test, and non-normal distribution data were analyzed by non-parametric test.

In order to investigate the components of QKL, ultra performance liquid chromatography-quadrupole time-of-flight-mass spectrometry (UPLC-Q-TOF-MS) was employed. And total ion chromatography (TIC) of QKL in ESI+ and ESI– modes was depicted in Supplementary Figure 1. Additionally, totally 46 compounds of QKL identified were shown in Supplementary Table 1, which included 8 bile acids, 5 flavonoids, 10 iridoids, 9 organic acids, 9 amino acids, and 5 other components.

According to the results of histopathological examination of the lung, we found that the Control group displayed a normal structure of lung tissue. LPS exposure in rats dramatically exacerbated lung lesions, including: infiltration of the lung interstitium by a rising number of inflammatory cells, widening of the distance between the alveolar septum, alveolar disarray and alveolar hemorrhage (Figure 1B). However, the QKL group showed significant improvement of lung damage, which caused a significant decline in the histopathologic score (Figure 1C).

The level of inflammatory cytokines is generally upregulated during the development of pneumonia. Compared with the Control group, IL-1β (P = 0.0079), TNF-α (P < 0.0001), and MCP-1 (P < 0.0001) were significantly higher in the Model group. Additionally, compared with the Model group, QKL-treated rats exhibited a marked decrease in TNF-α (P = 0.0437) and a declining tendency in IL-1β, and MCP-1 but without significance (P > 0.05) (Figures 1D–G). Simultaneously, the anti-inflammatory cytokine IL-10 was increased in the QKL and Dex groups. Taken together, these findings revealed that QKL acted as an anti-inflammatory agent, effectively suppressing the levels of inflammatory cytokines. LPS-stimulated pneumonia may increase oxidative stress in rats. Therefore, we examined the contents of NO, SOD and MPO. Compared with the Control group, the NO level (P = 0.0040) and MPO concentration were increased, while SOD (P = 0.0020) was significantly depleted in the Model group. However, QKL treatment not only significantly increased SOD activity (P = 0.0036), but also reduced the generation of NO (P = 0.0016) and MPO (P > 0.05) (Figures 1H–J). In summary, QKL treatment effectively attenuated LPS-induced pneumonia in rats.

After 16S rDNA sequencing, a total of 1,105,686 raw reads were obtained. Totally 661,452 effective sequences were obtained after removing whole chimeric tags by USEARCH, which remained for further analysis. No apparent changes were observed in bacterial richness among the Control, Model and QKL groups (Figure 2A). However, the Shannon index, reflecting both richness and evenness of intestinal flora, was the highest in the QKL group. Meanwhile, as depicted in the rank abundance curve, the intestinal flora αlpha diversity in the QKL group was between that of the other two groups. Moreover, according to the unweighted UniFrac distance of principal coordinates analysis (PCoA) analysis and hierarchical clustering results, excellent isolation was observed among the three groups, representing an overall structural difference of intestinal flora in rats (Figure 2B). In summary, by contrast with the Model group, the αlpha and beta diversity of the QKL group were significantly altered. Then, we investigated the taxonomic profile of intestinal flora. Two predominant phyla, consisting of Bacteroidetes and Firmicutes, occupied higher than 90% of the cumulative relative abundance (Figure 2C). An increased relative abundance of Bacteroidetes but a lower proportion of Firmicutes, and a smaller ratio of Firmicutes/Bacteroidetes (F/B) (Supplementary Figure 2) were observed in the QKL group as compared to the Model group. Meanwhile, Bacteroidetes, Verrucomicrobia and Proteobacteria were accumulated, while Firmicutes, Cyanobacteria and Deferribacteres were depleted after QKL intervention. At the family level, the differences were assessed by the Wilcoxon test. Compared with the Model group, QKL treatment markedly upregulated the relative abundance of Lachnospiraceae, Bifidobacteriaceae and Akkermansiaceae. In contrast, it significantly reduced the abundance of Lactobacillaceae, Corynebacteriaceae, Staphylococcaceae, Deferribacteraceae and Moraxellaceae (Figure 2C). The linear discriminant analysis effect size (LEfSe) was conducive to further identifying abundance variations of the specific intestinal microbiota and screening the potential microbial biomarkers (Figure 2D). In addition, compared with the Model group, the relative abundance of 16 genera was significantly altered following treatment with QKL, including 6 increasing genera (Bifidobacterium, Dorea, Ruminococcus_torques_group, Globicatella, etc.) and 10 decreasing genera (Mucispirillum, Corynebacterium_1, Staphylococcus, Facklamia, etc.). Moreover, the 16 genera were identified as the key characteristics (Figure 2E). Namely, the 16 genera were conducive to distinguishing the microbial communities between the Model and QKL groups. Collectively, QKL treatment induced overwhelming changes in community structures of intestinal flora in rats with pneumonia, which might affect metabolism and host health.

To understand the intrinsic community interaction of 16 distinguishing genera, we employed a Spearman correlation analysis (Figure 3A) and the exact P-values were shown in Supplementary Table 2. Subsequently, a Spearman correlation analysis focused on 7 biological parameters and 16 genera was performed to determine the correlation between pneumonia and intestinal flora. Our results showed that Facklamia was positively correlated with IL-1β and MCP-1, and inversely correlated with SOD. Staphylococcus and Jeotgalicoccus were positively correlated with MCP-1, MPO and TNF-α, and inversely correlated with SOD. Corynebacterium_1 was positively correlated with NO and MPO, and inversely correlated with SOD. Meanwhile, Ruminiclostridium_1, Paenalcaligenes and Mucispirillum were positively correlated with NO, and inversely correlated with IL-10 and SOD. Additionally, Globicatella was inversely correlated with NO and MPO. Interestingly, the majority of the genera abundant in the Model group were positively and inversely correlated with pro-inflammatory and anti-inflammatory factors, respectively (Figure 3B), and the P-values were exhibited in Supplementary Table 3. Moreover, PICRUSt analysis is conducive to understanding microbial KEGG functional prediction in fecal samples. Eight pathways were associated with metabolism, and two pathways were associated with genetic information processing. Among the top 10 pathways, in comparison to the Control group, the Model group’s lipid metabolism was significantly higher, and QKL downregulated lipid metabolism (Figure 3C).

A satisfying chromatographic peak shape occurred in ESI+ mode and ESI− mode (Figures 4A, B). Meanwhile, SIMCA v14.1 was used to perform multivariate statistical analysis to better understand the metabolite fingerprints of feces. The OPLS-DA score plot displayed great separation among distinct groups and desirable clustering of each group (Figures 4C, D), which indicated marked metabolic profile alterations in fecal samples under ESI+ mode (R²Y = 0.989, Q² = 0.795) and ESI− mode (R²Y = 0.994, Q² = 0.782). Furthermore, permutation tests (n = 200) were employed to avoid overfitting the model, and validation was performed in both ESI+ mode (R²Y = 0.836, Q² = −0.685) and ESI– mode (R²Y = 0.767, Q² = −0.599). The results showed that the OPLS-DA models were effective and did not overfit. In the OPLS-DA model, 1,861 (ESI−) and 1,642 (ESI+) variables between the Control and Model groups, and 1,511 (ESI−) and 1,095 (ESI+) variables between the Model and QKL groups were screened based on the S-plot (Figures 4E, F). Furthermore, according to the selection criteria (VIP > 1 and P < 0.05), 88 metabolites and 76 molecules were identified as potential metabolic biomarkers between the Control and Model groups (Supplementary Tables 4, 6), and the Model and QKL groups (Supplementary Tables 5, 6), respectively. Moreover, the differential metabolites were most abundant in carboxylic acids and derivatives, fatty acyls, indoles and derivatives, and other small molecules (Supplementary Tables 4–6), suggesting a large-scale metabolite dysregulation. To obtain the essential metabolic pathways, 88 and 76 differentially metabolites were imported into MetaboAnalyst V5.0. Consequently, our findings suggested that QKL treatment has a protective effect against LPS-induced inflammation in rats. This effect may be mediated, at least in part, by the regulation of various metabolic pathways, including arachidonic acid metabolism, glycerophospholipid metabolism, primary bile acid synthesis, steroid hormone biosynthesis and tryptophan metabolism (Figures 5A, B). Furthermore, 24 common metabolites, including 11 steroids and steroid derivatives, 7 fatty acyls, 3 glycerophospholipids, and 3 others were observed among the three groups (Supplementary Table 6), and their peak areas were compared to evaluate the protective effect after QKL therapy. As a result, the study demonstrated that compared with the Control group, the 24 metabolic compounds were altered by LPS administration and subsequently improved following treatment with QKL (Figures 5C–E). The 24 metabolites were involved in arachidonic acid metabolism, i.e., 8,11,14-eicosatrienoic acid (known as dihomo-γ-linolenic acid, DGLA) and arachidonic acid (AA); glycerophospholipid metabolism, i.e., LysoPC (20:0/0:0) and LysoPA (18:0e/0:0); and primary bile acid synthesis pathway, i.e., cholic acid. Collectively, these altered bioactive molecules may work together to reveal the therapeutic effects of QKL treatment.

To better determine the interaction of inflammatory parameters, 16 gut genera and 24 related metabolites, a Spearman correlation analysis was employed (Figures 6A, B), and Supplementary Table 7 displayed the P-values. Nutriacholic acid, cholic acid and 12-ketodeoxycholic acid were positively correlated with Globicatella, Ruminococcus_torques_group and Psychrobacter, and markedly inversely correlated with Mucispirillum, Corynebacterium_1, Jeotgalicoccus, Staphylococcus, Facklamia, and Aerococcus. In addition, the three metabolites were positively associated with SOD, and inversely associated with NO and MPO. AA, DGLA and αlpha-CEHC were significantly inversely correlated with Bifidobacterium, and positively associated with NO except AA. LysoPC (20:0/0:0) was markedly inversely correlated with Bifidobacterium and Dorea, and positively correlated with Staphylococcus, Corynebacterium_1, Jeotgalicoccus and Aerococcus, MG [0:0/22:2(13Z,16Z)/0:0] showed a similar correlation with LysoPC (20:0/0:0). In addition, LysoPC (20:0/0:0) was significantly positively associated with NO and MPO. In summary, the results demonstrated an evident and robust association between mostly gut genera and fecal metabolites, indicating that the modulation of specific metabolites was associated with enormous alterations in intestinal flora composition.

The results of transcriptome analysis revealed that QKL treatment significantly altered the gene expression profile in rats with LPS-induced pneumonia. The Q20 and Q30 were more than 97.23 and 92.40%, respectively, standing for the reliability of the transcriptome data. The PCA plot and hierarchical clustering demonstrated a significant cluster among the three groups. In addition, the PCA axis 1 and PCA axis 2 together explained 88.1% of the variance (Figure 7A). Then, 923 DEGs were discovered from the comparisons of Control vs. Model, and Model vs. QKL, with the cutoff of log2 (| fold change |) ≥ 2 and P < 0.05 (Figure 7B), using the Deseq2 package to analyze the DEGs. Among these, Control vs. Model and Model vs. QKL were screened, yielding 764 and 319 DEGs, respectively. Compared with the Control group, 61 and 703 genes were down-regulated and up-regulated in the Model group, respectively. In contrast, the QKL group generated 49 up-regulated and 270 down-regulated genes when compared with the Model group (Figures 7C, D). These gene alterations also indicated that a distinct gene expression pattern was induced after QKL treatment. In addition, the discriminated metabolic pathways of the 923 DEGs were found to be involved in 29 pathways (P < 0.05), including pathogenic Escherichia coli infection, PI3K-Akt signaling pathway, NF-κB signaling pathway, cytokine-cytokine receptor interaction, etc. (Figure 7E). Several pathways are known to be crucial in the pathogenesis of inflammation and pneumonia, and their modulation by QKL treatment may contribute to therapeutic effects.

PPI network analysis was used to understand the correlation between DEGs. We gave priority attention to genes associated with 29 metabolic pathways. In total, 127 DEGs were obtained and imported to STRING,⁴ where they were visualized using Cytoscape V3.9.1. In addition, we deleted 18 genes that did not correlate with the 109 DEGs. PPI network analysis revealed that the top 20 core genes were highly associated with the DEGs identified in our study (Figure 8A). The maximal clique centrality (MCC) algorithm in the cytoHubba module was employed to further find the core genes. Consequently, the top 20 genes appeared: Cxcl13, Ccl19, Ccl21, Cxcr2, Cr2, Il7, Xcl1, Pla2g2a, Pla2g5, Alox12e, Plb1, Cyp4a1, Cyp4a8, Cyp26b1, Aox2, Cyp1a1, Ckm, Hrc, Trdn, and Pgam2 (Figure 8B), and a heat map of 20 genes was depicted (Supplementary Figure 3). Moreover, the 20 genes were involved in various metabolic pathways, including arachidonic acid metabolism and cytokine-cytokine receptor interaction, etc. Interestingly, some of these pathways were also enriched in the microbiome and metabolome analyses, suggesting a potential microbiota-metabolites-host interaction. Specifically, Il7 was associated with the PI3K-Akt signaling pathway. Il7, Cr2, Xcl1 and chemokine-related genes (e.g., Cxcl13, Ccl19, Ccl21, and Cxcr2), all of which were involved in cytokine-cytokine receptor interaction. Ccl19 and Ccl21 were also related to the NF-κB signaling pathway. Pla2g2a, Pla2g5, Alox12e, Plb1, Cyp4a1, and Cyp4a8 were related to arachidonic acid metabolism. Cyp26b1, Aox2, and Cyp1a1 were involved in retinol metabolism. In addition, Aox2 and Cyp1a1 were associated with tryptophan metabolism.