RAW264.7 cells (ATCC® TIB-71, purchased from American Type Culture Collection, Maryland, USA) infected with BP and control RAW264.7 cells were collected at 3 time points (0.5, 8, and 16 h) after infection. The cells were grown in 6-well plates (1.5 × 10⁶ cells per well). A total of 12 samples were collected, including 9 from the infection time points and 3 from the control time points. The cells were washed with phosphate buffered saline (PBS), lysed using TRIzol (15596026, ThermoFisher Scientific, USA), and transferred to RNase-free EP tubes. The samples were frozen in liquid nitrogen and sent for microarray assay (CapitalBio Technology Corporation, Beijing, China). Expression profiling of the mRNA was performed using the Mouse 8*60K expression profile chip from Agilent (Agilent Technology Corporation, USA). The detection protocol followed the specifications for each chip. DNA contaminants in host cells are usually removed during sequencing and analysis.

A t-test was employed to analyze the significant differences between the two groups of samples to identify genes with significant differential expression. Genes with a higher absolute value of the fold change (Absolute Fold change) exhibited more significant differences between the two samples. The criteria for differential gene screening were absolute fold change ≥ 2 and p < 0.05.

GO enrichment and KEGG pathway analyses of the differential genes were performed using DAVID 6.8. The study focused on cellular components (CC), biological processes, and molecular functions (MF), which provided insights into the cellular functions and components associated with the key genes. The KEGG pathway analysis identified the key signaling pathways involved in the activity of the differential genes, with a significance threshold of p < 0.05.

Macrophages contain both M1 and M2 subtypes. Subsequently, the “WGCNA” R package was utilized to analyze gene modules associated with M1 and M2 macrophages. Hierarchical clustering constructed a clustering tree based on dissimilarity coefficients. Genes with similar expression patterns were assigned to the same modules, while genes with low similarity were assigned to different modules. The module containing macrophage-associated genes was identified and visualized.

LASSO logistic analysis is a method for reducing linear regression and performing variable selection and parameter estimation. This study used the “glmnet” and “mice” R packages to model and evaluate macrophage-related genes. A random seed (seed) number of 500 was set to ensure reproducible results. The relationship between the genes and the matrix was analyzed as the dependent variable. A risk score was calculated based on the obtained gene signature.

SVM is a non-linear binary classification model that constructs a hyperplane in a high-dimensional space for classification and regression tasks. This study utilized the “e1071” R package to analyze samples from different subgroups, including gene selection between high-risk and low-risk groups and between experimental and control groups.

Gene expression values were used to calculate gene enrichment scores in the samples. The correlation between genes and macrophages was analyzed using the “ggpmisc” and “MASS” R packages, employing the Pearson correlation coefficient method. Correlation coefficients with absolute values < 0.4 were considered relevant.

The protein levels of FZD4 were measured by western blotting, following an earlier protocol [16]. RAW264.7 cells were pretreated with PBS for 24 h, then infected with BP (MOI = 10) for 8 h. Cells were washed with PBS, lysed in protein lysis buffer containing Triton X-100 (T8787, Thermo Scientific, USA), and protease inhibitors (P7626, Thermo Scientific, USA), incubated on ice, and heated. After centrifugation, the protein concentration was determined using the BCA Protein Assay (23227, Pierce, USA). Lysates were subjected to standard sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and the proteins were transferred to a polyvinylidene difluoride membrane (Millipore, Massachusetts, USA). The membranes were blocked and incubated with primary antibodies (β-actin and FZD4 were purchased from Thermo Scientific, cat Nos. A5441 and SAB1402984). Subsequently, horseradish peroxidase (HRP)-conjugated anti-mouse antibodies were used (AP160P, Thermo Scientific, USA). Protein was visualized using the Supersignal West Dura Duration substrate reagent (34080, Thermo Scientific, USA). A commercial protein marker was used to determine protein sizes. The R v3.6.1 was used for all bioinformatics analyses. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated to assess the predictive power of the biomarkers. A significance level of p < 0.05 was considered statistically significant.

Eight Balb/C mice (aged 4–6 weeks old), age- and sex-matched, were selected and used in all experiments. The mice were exposed to inhalation of 3 × 10⁵ colony forming units (CFU) of BP or an equivalent volume of PBS (control). The mice were monitored daily for 10 days during the experiment, and the number of deaths was recorded. After 10 days of observation, the surviving mice were sacrificed, and lung tissues were harvested for immunohistochemistry analysis. All BP infection experiments were performed in biosafety level 2+ laboratory spaces. All animal care and use protocols in this study were performed following the Regulations for the Administration of Affairs Concerning Experimental Animals approved by the State Council of People’s Republic of China. All animal experiments in this study were approved by the Animal Ethical and Experimental Committee of the General Hospital of Center Theater of PLA (Wuhan, Hubei Province. Permit No. 2011-04) under their rules and regulations. All surgery was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize suffering.

The lung tissues were subjected to fixation, embedding, sectioning, antigen retrieval, blocking, primary antibody incubation (rabbit anti-mouse FZD4 antibody, SAB1402984, Thermo Scientific, USA), secondary antibody incubation (HRP-conjugated anti-rabbit antibodies, A0208, Beyotime Biotechnology, Shanghai, China), enzyme-substrate reaction, and mounting for observation. Reference immunohistochemical staining procedure: mouse lung tissue paraffin sections were routinely dewaxed to water, washed for 5 min x three times in PBS, repaired by autoclaving in 10 mmol/L (pH 6.0) citrate buffer for 3 min, and cooled naturally for 30 min. This was washed thrice for 5 min in PBST, incubated for 10 min in 30 mL/L H 20 2/PBS solution, washed thrice for 5 min in PBS, and incubated overnight at 4°C in a wet box. This tissue was washed with PBS for 5 min thrice, and 50 mL of Immunohistochemistry Kit A was added and incubated at 20°C for 10 min and washed thrice with PBS for 5 min. To this, 50 mL of Immunohistochemistry Kit B reagent was added, incubated at 20°C for 10 min, and washed thrice with PBS for 5 min. Finally, 50 mL of Immunohistochemistry Kit B was added, incubated at 20°C for 10 min, and washed thrice with PBS for 5 min, and 3,3-diaminobenzidine (1:50) was used to develop the color at room temperature.

Statistical analysis of the obtained data from GEO was performed using the R software. The t-test was employed throughout the research process to compare the two independent samples, and a significance level of p < 0.05 was used to determine statistical significance.

The gene expressions were comprehensively analyzed using Agilent DNA microarrays, and 33,623 annotated and unannotated genes were identified. The experimental groups were primarily compared to their respective control groups using a single-channel hybridization approach. Specifically, the following comparisons were made: 30-min infection vs. 30-min negative control, 8-h infection vs. 8-h negative control, 24-h infection vs. 24-h negative control, 8-h infection vs. 30-min infection, and 24-h infection vs. 30-min infection. Scanning pseudo-color plots of the microarrays were generated to ensure the reliability of the data. The samples satisfied the quality control criteria for each assay parameter (Fig. 1). These results validated the suitability of the data for subsequent analyses, ensuring the robustness and accuracy of the findings.

In this study, we integrated DEGs by filtering the raw data using a significance threshold of p < 0.05 and a fold change of ≥2.0. We obtained 12,813 DEGs. To assess the relationship between these genes, we calculated the Pearson coefficient (r) between pairs of DEGs. Gene pairs with an absolute value of r > 0.9999 were considered related gene pairs. This analysis resulted in 5,520 corresponding gene pairs involving 898 unique genes. Further analysis was conducted on these 898 genes to explore their functional enrichment. These 898 genes were further analyzed by GO to explore the relationship between the structure and biological function of the genome, the possible regulatory mechanisms in the genome, and its functional enrichment. These genes were primarily associated with biological processes such as the G protein-coupled receptor signaling pathway, signal transduction, and the development of multicellular organisms (Fig. 2A). Considering CC, the analysis indicated that these genes predominantly participated in membrane components, including plasma membranes, plasma membrane components, synaptic and posterior membranes, as well as extracellular regions and cell surfaces. Regarding MF, the identified genes were mainly involved in G protein-coupled receptor activity, transmembrane signaling receptor activity, ion channel activity, and signaling sensor activity. KEGG enrichment analysis of DEGs in the infection group and the control group were compared (Fig. 2B). The top six significant pathways were selected for visualization; the size of the dots in the graph represents the number of genes, and the color represents p.adjust. The infection group was enriched in cytokine-cytokine receptor interactions, nucleotide oligomerization domain-like receptor signaling pathway, tumor necrosis signaling pathway, interleukin (IL)-17 signaling pathway, etc. The clustering heatmap showing differential genes and clustering between samples is presented in Fig. 2C. Horizontal coordinates show sample information and hierarchical clustering results. In contrast, vertical coordinates show differential genes and hierarchical clustering results.

This study was based on mRNA expression data of 898 differential genes in 12 samples for co-expression weighted network construction. The clustering distribution showed no abnormal samples (Fig. 3A). Two gene modules were identified under the parameter settings of minModuleSize = 30 and mergeCutHeight = 0.25 (Figs. 3B and 3C). The two gene modules are represented as the blue and turquoise modules. Subsequently, the correlation between the two modules and the immunological profile was calculated, and the results are shown in Fig. 3D. The blue module shows a significant correlation with macrophages and a total of 416 genes within the module.

This study used tenfold cross-validation to further analyze the effector genes among 416 immune genes that effectively affect macrophages. Using LASSO logistic analysis, the number of seeds was set to 500 and 416 immune gene features and their correlation coefficients were obtained (Fig. 4A). Ten gene features with independent predictive ability were finally selected based on the appropriate lambda (Fig. 4B). The seed number (seed) was set equal to 500; as a result, 10 gene signatures and correlation coefficients with independent predictive power were obtained (PDE5A, TRIM67, TRP73, RAP2C, FZD4, LST1, SLC15A1, FOXD3, ALPI, ACSM2) (Figs. 4A and 4B). Subsequently, the samples’ risk score was calculated. The risk score was calculated as: risk score = (−8.93626) * PDE5A + (6.29332) * TRIM67 + (−5.86773) * TRP73 + (−1.72814) * RAP2C + (−0.38555) * FZD4 + (−0.17740) * LST1 + (0.00206) * SLC15A1 + 0.51196 * FOXD3 + 0.92781 * ALPI + 5.9582 * ACSM2. Samples can be categorised into high and low-risk groups by immunogenetically related risk scoring models.

SVM analysis was performed on each of the 10 gene signatures in this study. The results were obtained based on the gene expression analysis by selecting four genes (FZD4, TRIM67, TRP73, SLC15A1) in the high and low score groups (Fig. 5A). Minimal error rate can be obtained when the above four immunity genes were selected for prediction. Five genes (RAP2C, PDE5A, FZD4, ALPI, ACSM2) were selected in the experimental and control groups (Fig. 5B). Thus, the error rate of the curve change after 10-fold cross-validation was minimal when the features were the five genes mentioned above. The intersecting gene, FZD4, was identified based on the intersection of the different classification results from the two groups.

Expression levels of the seven genes in macrophages in the experiment group compared with the control group were significantly different. The correlation between genes and macrophages was analyzed using the Pearson correlation coefficient method, which refers to the presence or absence of a correlation between seven genes and macrophages. Correlation coefficients with an absolute value <0.4 were considered to be correlated. The crossover gene of interest, FZD4, showed a significant negative correlation with macrophages (cor = −0.66, p = 0.022). The results showed that FZD4 had a significant negative correlation with macrophages (cor = −0.66, p = 0.022) (Figs. 6A and 6B). This result suggests that there may be a strong interaction relationship between FZD4 and macrophages. ROC curve, also known as the “subject operating characteristics curve”, is used to determine whether a specific factor has a predictive value for the disease; an AUC value of 0.704 for the ROC curve indicates a good predictive effect. The ROC curve showed that FZD4 had good predictive efficacy (AUC = 0.704) (Fig. 6B). It suggests that FZD4 is accurate for BP infection.

Based on the above analysis of biomarkers with macrophages, we speculated whether BP induces FZD4 expression in macrophages. Western blotting analysis revealed significant upregulation of FZD4 from 0.5 h post-infection (hpi) to 24 hpi (Fig. 7A). FZD4 expression was significantly upregulated, suggesting that BP induces an increase in the expression of FZD4. Results shown in Fig. 7A are representative of at least three independent experiments. The ratios of FZD4 to beta-actin in macrophages were calculated based on the density of target bands (Fig. 7B) and showed a gradual increase with the time of infection. In addition, the ratios at different time intervals differed significantly. The above results show that BP induces FZD4 expression in macrophages time-dependently.

As Fig. 8A shows, an inhalation infection model with BP was established via nasal infection under the dose of 3 × 10⁵ CFU of BP. The mice were observed daily until 10 days of infection, the survivors were sacrificed, and the lung tissues were harvested (Fig. 8B). The immunohistochemical analysis revealed a high expression of FZD4 in the cytoplasm of pulmonary epithelial cells in BP-infected lungs of mice, compared with the PBS control group (Figs. 8C and 8D).