Exploration of key mechanisms underlying the therapeutic effects of AMD3100 on attenuating lipopolysaccharide-induced acute lung injury in mice

Animals

Male ICR mice, aged 7–9 weeks and weighing 23–28 g, were procured from Shanghai Ji hui Laboratory Animal Care Co., Ltd. They were housed in groups of three in a specific pathogen-free environment at the Shanghai Xinhua Hospital animal laboratory, with a humidity range of 40–60% and a constant temperature of 24 ± 2 °C. The study comprised four experimental groups, where animals were initially weighed and subsequently numbered by weight before being randomly allocated to either the experimental groups or the control group (n = 7) using a randomization tool. No specific criteria were set for the inclusion or exclusion of animals in the study; all groups and animals were included in the analysis. The study was ethically approved by the Xinhua Hospital ethics committee (Approval ID: XHEC-F-2020-026, April 12th, 2020). The researchers adhered strictly to the principles of “Replacement, Reduction, and Refinement” for the welfare of experimental animals, implementing effective care measures to safeguard animal welfare and prevent unnecessary harm. The researchers possessed the requisite ability and qualifications for conducting the project, backed by clinical research experience and infrastructure. The approval letter is provided as Supplemental Material.

Drug treatment

The animals were randomly assigned to four groups: control, LPS, AMD3100, and LPS plus AMD3100 (LPS + AMD3100), each comprising seven mice. Mice in the LPS group were intraperitoneally injected with LPS (5 mg/kg; Sigma-Aldrich, St. Louis, MO, USA). AMD3100 was administered intraperitoneally at a dosage of 10 mg/kg, one hour before LPS injection. The control group mice received an equivalent volume of saline. After a 24-hour period post-treatment, the mice were euthanized. Anesthesia was induced using 5% isoflurane in oxygen within a plexiglass cage, following which the mice were sacrificed, and lung tissues were harvested.

Lung tissue histological evaluation

The lung tissues underwent fixation with 4% paraformaldehyde, followed by dehydration utilizing various concentrations of ethanol. These dehydrated tissues were then embedded in paraffin and sectioned into slices measuring 5 μm in thickness. Hematoxylin and eosin staining were performed on these sections. Slices are scanned and imaged under the microscope lens of a slice scanner (Pannoramic MIDI, Pannoramic 250 FLASH, Pannoramic DESK; 3Dhistech, Budapest, Hungary), and seamlessly pieced together into a full field digital slice through a control software system (CaseViewer, C.V.2.4). The total surface of the slides was scored by two blinded pathologists with expertise in lung pathology. The criteria for scoring lung inflammation was set up as previously described with some modifications (Wu et al., 2013): 0, normal tissue; (1) minimal inflammatory change; (2) mild to moderate inflammatory changes (no obvious damage to the lung architecture); (3) moderate inflammatory injury (thickening of the alveolar septae); (4) moderate to severe inflammatory injury (formation of nodules or areas of pneumonitis that distorted the normal architecture); (5) severe inflammatory injury with total obliteration of the field.

Enzyme-linked immunosorbent assay

ELISA was used to quantify the secretion of pro-inflammatory cytokines included IL-6, IL-1β in the BALF and Cxcl12, Cxcl10, Cxcl1 and Cxcl9 in the lung tissue. The lung tissue and BALF were collected after sacrifice of mice, stored at −80 °C and tested using ELISA kits.

Transcriptome sequencing

The lung tissues from the four groups underwent transcriptome sequencing, and four samples were selected from each group. Total RNA was extracted from the lung tissues using the mirVana miRNA Isolation Kit (Ambion, Austin, TX, USA). After detecting RNA integrity, the libraries were established using the TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) and then sequenced on the HiSeq 2500 sequencing platform (Illumina, San Diego, CA, USA), resulting in the generation of paired-end reads with lengths of 125 bp/150 bp.

Data filtering and mapping

The raw reads generated were processed using Trimmomatic (Bolger, Lohse & Usadel, 2014). After filtering out the reads containing poly-N and low-quality reads, clean reads were obtained and aligned to the reference genome using HISAT2 (Kim, Langmead & Salzberg, 2015). Cufflinks (Trapnell et al., 2010) were used to calculate the FPKM value, and htseq-count (Anders, Pyl & Huber, 2015) was applied to evaluate the read counts.

Differential expression analysis

The FPKM values were converted into log2 (FPKM+1). The differentially expressed genes between the LPS and control groups, and between the LPS + AMD3100 and LPS groups, were identified using the limma package (version 3.10.3) (Smyth, 2005). The cutoff values were set at log fold change >2 and adjusted p-value <0.05. Genes with opposite expression trends in the LPS vs. control and LPS + AMD3100 vs. LPS groups were used as target genes for subsequent analyses.

Functional enrichment analysis

To elucidate the function of the target genes, Gene Ontology (GO) (Ashburner et al., 2000) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) (Kanehisa & Goto, 2000) pathway enrichment analyses were performed using cluster profile (version 3.18.1) (Yu et al., 2012). The significant enrichment results were obtained with threshold values of count ≥2 and adjusted p-value < 0.05.

Construction of protein-protein interaction network

The PPI relationships between proteins encoded by the target genes were predicted with the STRING (version 10.0) database (Szklarczyk et al., 2015). The parameters were set as follows: mouse as the organism, and a PPI score of 0.4. A PPI network was constructed with Cytoscape (version 3.2.0). The topological properties of each node were analyzed using the CytoNCA plugin (version 2.1.6) (Tang et al., 2015). The hub nodes in the PPI network were screened by ranking the scores of each node. Additionally, significant modules in the network were selected using the MCODE plugin (version 1.4.2) (Bandettini et al., 2012). The threshold value was greater than five. Functional enrichment analyses of the module genes were also conducted.

Prediction of transcription factor

Transcription factors (TFs) potentially targeting module genes were anticipated through the utilization of the Web Gestalt GAST tool (Zhang, Kirov & Snoddy, 2005). Key parameters included the selection of Mus musculus as the Organism of Interest, the application of Overrepresentation Enrichment Analysis as the Method of Interest, a significance level set at p < 0.05, and the focus on the Top 10 outcomes.

Quantitative RT-PCR for the validations of mRNA

To validate the RNA sequencing data, quantitative real-time polymerase chain reaction (qRT-PCR) was employed, revealing significant differences in four mRNAs across the four groups (n = 7 per group). Lung tissue samples were utilized for sequencing, with all gene primers detailed in Table 1. Initially, total RNA was extracted from the lung using the RNA iso Plus Kit (TaKaRa, Shiga, Japan). Subsequently, the PrimeScript™ RT reagent Kit (TaKaRa, Shiga, Japan) facilitated the reverse transcription of RNA into cDNA. qRT-PCR assays were performed using the SYBR^® Premix Ex Taq™ kit (TaKaRa, Shiga, Japan) on the biosystems QuantStudio 5 Flex (Thermo Fisher Scientific, Waltham, MA, USA) as per the manufacturer’s instructions. The relative expression levels of mRNAs in the lung were normalized to 18S, while the relative mRNA expression was normalized to U6. The 2^−ΔΔCt method was utilized for calculating the qRT-PCR results, with each qRT-PCR sample undergoing three replications.

Statistical analysis

All data were analyzed with the statistical program SPSS 25.0 (Chicago, IL, USA). Data are expressed as means ± SEM. Statistical comparisons between experimental groups and control group were performed using two sample t-test (T-T test) with an additional Bonferroni post hoc test. Multiple groups were compared using one-way ANOVA test followed by Bonferroni correction. Statistical significance was defined as p-value < 0.05.

AMD3100 alleviated LPS-induced ALI in mice

The protective effect of AMD3100 was assessed using Hematoxylin and eosin staining. Normal pulmonary structures were observed in the lung tissues in control group. After LPS treatment, the lung tissues showed evident histopathological alterations, such as alveolar wall edema and infiltration of inflammatory cell. However, after further treatment with AMD3100, the alveoli in the LPS + AMD3100 group exhibited remarkable restoration in morphology, accompanied by minor pathological changes (Fig. 1A). We evaluated the histopathological damage scores of the different groups. The histopathological damage score exhibited a marked increase in the LPS group compared with the control group (p < 0.01). Furthermore, compared with the LPS group, the histopathological damage score had remarkably decreased in the LPS + AMD3100 group (p < 0.01) (Fig. 1B). We also measured the lung wet/dry ratio in different groups, which indicated the damage level of lung. The LPS group had an increased ratio compared with the control group. However, the lung wet/dry ratio significantly decreased in the LPS + AMD3100 group (p < 0.01) (Fig. 1C). The concentration of IL-6 and IL-1β in BALF indicated that the LPS group had an increased level of inflammation, which showed a decreased tend in the LPS + AMD3100 group (Fig. 1D). These findings indicate that AMD3100 alleviated LPS-induced acute lung injury of mice.

Transcriptome sequencing data

In total, 113.37 G of clean data were obtained from the sequencing of 16 samples. The effective data size for each sample ranged from 6.75 to 7.38 Gb, with Q30 base distribution ranging from 95.4% to 95.71% and an average GC content of 50.55% (Table 2). The alignment of clean reads to the reference genome had alignment rates ranging from 96.7% to 97.88% (Table 3).

Identification of target genes

Using the limma package, 541 differentially expressed genes (291 upregulated and 250 downregulated) were identified in the LPS vs. control group (Fig. 2A), and 307 genes (131 upregulated and 276 downregulated) were identified in the LPS + AMD3100 vs. LPS group (Fig. 2B). Venn analysis yielded 172 intersecting genes from the upregulated genes in the LPS vs. control group and downregulated genes in the LPS + AMD3100 vs. LPS group, alongside 122 intersecting genes from the downregulated genes in the LPS vs. control group and upregulated genes in the LPS + AMD3100 vs. LPS group (Fig. 3). These 294 intersecting genes were selected as target genes for subsequent analyses.

Functional enrichment analysis for target genes

To elucidate the functions of the target genes, we analyzed the enriched GO terms and KEGG pathways. These target genes were enriched in 551 GO biological process functions, such as the cytokine-mediated signaling pathway, response to the virus, and response to interferon-beta (Fig. 4A). Moreover, nine pathways were significantly enriched; these included cytokine-cytokine receptor interaction, nuclear factor-kappa B (NF-κB) signaling pathway, TNF signaling pathway, and IL-17 signaling pathway (Fig. 4B).

PPI network analysis and hub gene identification

A PPI network was constructed with the target genes, which included 219 nodes and 1,023 interactions (Fig. 5). Hub nodes were identified based on topology scores. The top five nodes with the highest degrees were the C-X-C motif chemokine ligand 10 (Cxcl10, degree = 57), interferon regulatory factor 7 (Irf7, degree = 49), Cxcl9 (degree = 48), Irf1 (degree = 45), and cluster of differentiation 274 (Cd274, degree = 41).

Module analysis and functional enrichment analysis

Using the MCODE plugin (score > 5), two modules were identified in the PPI network (score > 5; Fig. 6A). Module 1 (score = 28.207) contained 30 nodes (such as Cxcl10 and Cxcl9) and 409 interaction pairs. Module 2 (score = 6.667) contained 10 nodes (such as Cxcl12 and Cxcl1) and 30 interaction pairs.

Moreover, the genes in module 1 were enriched in 133 GP-BP terms, such as response to the virus, and 11 KEGG pathways, such as the Toll-like receptor (TLR) and TNF signaling pathways (Fig. 6B). Notably, Cxcl10 and Cxcl9 are involved in the TLR signaling pathway. The genes in module 2 were enriched in 188 GP-BP terms, such as cell chemotaxis, and eight KEGG pathways, such as cytokine-cytokine receptor interaction and NF-κB signaling pathway (Fig. 6C). Remarkably, Cxcl12 and Cxcl1 were involved in cytokine-cytokine receptor interaction and the NF-κB signaling pathway.

TF-target regulatory network analysis

Using the Web Gestalt GAST tool, the top ten TFs that could target module genes were obtained. A TF-target regulatory network was constructed, which included 10 TFs, 15 module genes, and 49 TF-target relationships (Fig. 7). Modular genes in this network were downregulated. Moreover, Cxcl10, Cxcl9, and Cxcl1 were targeted by NF-κB (NFKAPPAB_01 and NFKB_Q6).

Validations of mRNA and protein

In order to verify the DEGs, we selected four significantly changed mRNAs. We performed qRT-PCR from the lung tissues for the validation of mRNA, which was used for mRNA sequencing. The expressions of Cxcl12, Cxcl10, Cxcl9, and Cxcl1 were increased in the LPS group compared with that in the control group and decreased in the LPS + AMD group using 18s for endogenous control (Fig. 8A). We also measured the concentration of Cxcl12, Cxcl10, Cxcl9, and Cxcl1 from the lung tissues with ELISA. The expressions of Cxcl12, Cxcl10, Cxcl9, and Cxcl1 were increased in the LPS group compared with that in the control group and decreased in the LPS + AMD group (Fig. 8B). The results of qRT-PCR and ELISA are significant support to demonstrate our research.