Olink proteomics analysis revealed that 25 DEPs were identified in the MTB group (16 upregulated and 9 downregulated), 29 DEPs in the RGM group (24 upregulated and 5 downregulated), and 54 DEPs in the SGM group (46 upregulated and 8 downregulated) compared to the normal group (Fig. 2a, Table S2). The differential expression profiles of DEPs between the comparison groups are shown in Fig. 2b-d.
A Venn diagram showed that 16 proteins exhibited a consistent trend of differential expression in the MTB, RGM, and SGM groups compared to the control group. This suggests that these proteins represent alterations in the host immune landscape that are shared among all three Mycobacteria (Fig. 2e, f). These 16 proteins consist of CCL3, CD83, CXCL10, CXCL11, CXCL13, CXCL9, Gal-9, HGF, IFNG (IFN-gamma), IL12, MMP12, MUC-16, PD-L1, TNF, TRAIL, and VEGFR-2.
PPI network analysis revealed that these DEPs were enriched in Gene Ontology (GO) biological process terms such as positive regulation of CD4-positive, CD25-positive, alpha-beta regulatory T cell differentiation involved in immune response, T cell chemotaxis, and chronic inflammatory response (Fig. 2i, Table S3). Furthermore, these DEPs included four out of the five proteins in the GO molecular function term CXCR3 chemokine receptor binding, namely CXCL13, CXCL10, CXCL11, and CXCL9 (Table S4).
C-X-C motif chemokine ligands (CXCL9, CXCL10, CXCL11, and CXCL13) belong to the CXC subfamily of chemokines. They are ligands that target CXCR3, inducing chemotaxis and promoting the differentiation of immune cells [18]. It has been demonstrated that CXCL9, CXCL10, CXCL11, as well as CXCR3, are significantly elevated in patients with active tuberculosis compared to patients with other types of lung disease and healthy controls [19,20,21,22]. The immune response occurs through the CXCL9, CXCL10, and CXCL11/CXCR3axis by recruiting immune cells such as cytotoxic lymphocytes, natural killer cells, NKT cells, and macrophages. The ligands are typically expressed at low levels under steady-state conditions but are upregulated in response to cytokine stimulation. CXCL9, CXCL10, and CXCL11 are primarily secreted by monocytes, endothelial cells, fibroblasts, and carcinoma cells in response to IFN-gamma [22]. Correlation analysis of our results showed that IFN-gamma was positively correlated with CXCL9 (r = 0.77, p-value < 0.001, Fig. 2g), CXCL10 (r = 0.74, p-value < 0.001), and CXCL11(r = 0.64, p-value < 0.001). Therefore, it can be assumed that IFN-gamma induces activation of the CXCL9, CXCL10, and CXCL11/CXCR3 axis in all three infection groups.
The results of the ROC analysis showed that many immune-related proteins are promising diagnostic markers for Mycobacterium (Fig. 2h). IL8 was demonstrated to be an effective diagnostic marker for MTB infection, with an area under the curve (AUC) of 0.9186. It is interesting to note that CXCL9 showed robust diagnostic potential for both RGM and SGM.
Fig. 2Olink proteomic analysis comparing the MTB, RGM, and SGM groups with the normal group. a Histogram of the numbers of DEPs among the four groups. b Volcano plot depicting the DEPs between the MTB group and the normal group. Red dots represent upregulated DEPs, while blue dots represent downregulated DEPs. c Volcano plot depicting the DEPs between the RGM group and the normal group. d Volcano plot depicting the DEPs between the SGM group and the normal group. e Venn diagram of DEPs between the MTB, RGM, and SGM groups and the normal group. Proteins in the red boxes are DEPs that are co-upregulated in the three infection groups, and proteins in the blue boxes are DEPs that are co-downregulated. f Boxplots of co-upregulated or co-downregulated DEPs. g Correlation heatmap of co-upregulated or co-downregulated DEPs. The color and area of the pie chart represent the correlation coefficient, with colors ranging from dark blue to dark red indicating correlation coefficients from − 1 to 1. h ROC curves for the top 5 DEPs with the highest AUC values. i PPI network analysis of co-upregulated or co-downregulated DEPs. Proteins highlighted in red are involved in the GO term positive regulation of CD4-positive, CD25-positive, alpha-beta regulatory T cell differentiation involved in immune response. Proteins highlighted in blue are associated with the GO term T cell chemotaxis. Proteins in green are associated with the GO term chronic inflammatory response
Immune protein differences among the three mycobacterium groupsCompared with the MTB group, a total of 12 upregulated proteins were identified in the RGB group, and 22 upregulated proteins and 1 downregulated protein were identified in the SGM group (Fig. 3a-c). The identified DEPs are listed in Table S1.
A Venn diagram showed that 7 proteins were co-upregulated in both the RGM and SGM groups compared to the MTB group. These proteins include adenosine deaminase (ADA), CCL3, CD40, CXCL5, Gal-1, IL8, and MCP-2 (Fig. 3d, g). PPI analysis showed that these 7 DEPs were involved in GO biological processes such as granulocyte activation, neutrophil chemotaxis, and eosinophil chemotaxis (Fig. 3e, Table S5).
Interestingly, interleukin-8 (IL-8), a member of the CXC chemokine family and a major mediator of the inflammatory response, was most significantly increased in both the RGM group (FC = 1.56, p-value = 0.01) and the SGM group (FC = 1.31, p-value = 0.03) compared to the MTB group. Therefore, it can be speculated that NTM, especially RGM, induced the host to produce more IL-8, enhanced the IL-8 pro-inflammatory signaling cascade, and augmented the host inflammatory response.
Compared with the SGM group, 7 downregulated DEPs were identified in the RGB group, including CCL17, CXCL12, IL7, LAP TGF-beta-1, CD27, PGF, and TWEAK. PPI analysis showed that these 7 DEPs were involved in GO biological processes such as induction of positive chemotaxis, positive regulation of B cell differentiation, and extrinsic apoptotic signaling pathway (Fig. 3f, Table S6). Furthermore, CCL17 exhibits chemotactic activity for T lymphocytes, but not for monocytes or granulocytes [23]. Therefore, it can be assumed that the differences between RGM and SGM mainly focus on the chemotactic activity of T and B cells and extrinsic apoptosis.
In addition, the results of the ROC analysis revealed several promising Mycobacterium differential diagnostic markers (Fig. 3i). It was demonstrated that MCP-2 was a potent differential diagnostic marker for RGM and MTB with an AUC of 0.8051, whereas MCP-4 was a potent differential diagnostic marker for SGM and MTB. CCL17 was demonstrated to be a promising differential diagnostic marker for RGM and SGM with, an AUC of 0.7587.
Fig. 3Olink proteomic analysis among the MTB, RGM, and SGM groups. a Volcano plot of DEPs between the RGM group and the MTB group. Red dots represent upregulated DEPs, while blue dots represent downregulated DEPs. b Volcano plot of DEPs between the SGM group and the MTB group. c Volcano plot of DEPs between the RGM group and the SGM group. d Venn diagram of DEPs among the MTB, RGM, and SGM groups. Proteins in the red boxes are DEPs that are co-upregulated in the three infection groups. e PPI analysis of co-upregulated DEPs in RGM/MTB and SGM/MTB. Proteins highlighted in red represent the proteins in the GO term granulocyte activation, those in blue are associated with neutrophil chemotaxis, and the ones in green are related to eosinophil chemotaxis. f PPI analysis of DEPs between the RGM group and the SGM group. Proteins in red represent the proteins in the GO term induction of positive chemotaxis, proteins in blue represent the proteins in the positive regulation of B cell differentiation, and proteins in green represent proteins in the extrinsic apoptotic signaling pathway. g Boxplots of co-upregulated DEPs in RGM/MTB and SGM/MTB. h Boxplots of DEPs between the RGM group and the SGM group. i ROC curves for the top 5 DEPs with the highest AUC values
Serum lipid variations caused by mycobacterium infectionLipidomics analysis showed that 61 DEMs were identified in the MTB group (30 upregulated and 31 downregulated), 130 DEMs in the RGM group (79 upregulated and 51 downregulated), and 145 DEMs in the SGM group (127 upregulated and 18 downregulated) compared to the normal group. The differential expression profiles of DEMs between the comparison groups are shown in Fig. 4a-c. The identified DEMs are provided in Table S7.
According to the Venn diagram, 10 DEMs (PI 40:6, CE 22:6, SM 17:0, SMGDG O-11:0, PC 40:7, PC 17:0, LNAPE 18:1, LPC 40:8-SN2, PC O-39:7, and LPC O-24:1) were co-upregulated, and 6 DEMs (FAHFA 22:3, FAHFA 26:4, FAHFA 24:4, FAHFA 20:5, FAHFA 18:2, and CAR18:2) were co-downregulated in the MTB, RGM, and SGM groups compared with the control group (Fig. 4d).
It is interesting to note that all FAHFAs identified were significantly downregulated in all three Mycobacterium infection groups (Fig. 4e). In addition, correlation analysis showed a highly positive correlation between these FAHFAs (Fig. 4f). FC values and p-values for all FAHFAs are presented in Table S7.
ROC analysis showed that FAHFAs are promising diagnostic markers for Mycobacterium (Fig. 4g). FAHFA 18:2 is a promising diagnostic marker for both MTB and RGM, with AUC values of 0.8708 and 0.944 for diagnosing MTB and RGM, respectively. Meanwhile, FAHFA 24:4 was identified as a promising diagnostic marker for SGM, with an AUC value of 0.8629.
Fig. 4Lipidomics analysis comparing the MTB, RGM, and SGM groups with the normal group. a Volcano plot illustrating the DEMs between the MTB group and the normal group. Red dots represent upregulated DEPs, while blue dots represent downregulated DEPs. b Volcano plot illustrating the DEMs between the RGM group and the normal group. c Volcano plot illustrating the DEMs between the SGM group and the normal group. d Venn diagram of DEMs between the MTB, RGM, and SGM groups and the normal group. e Boxplots of co-upregulated or co-downregulated DEMs. f Correlation heatmap of co-upregulated or co-downregulated DEMs. The color and area of the pie chart represent the correlation coefficient, with colors ranging from dark blue to dark red indicating correlation coefficients from − 1 to 1. g ROC curves for the top 5 DEMs with the highest AUC values
Differences in lipid profiles among the three mycobacterium infection groupsCompared with the MTB group, a total of 61 upregulated and 38 downregulated DEMs were identified in the RGB group, while 68 upregulated proteins and 5 downregulated DEMs were identified in the SGM group (Fig. 5a, b).
Venn diagrams showed that 29 DEMs were co-upregulated in the RGM and SGM groups compared to the MTB group, revealing differences in the lipid profiles of individuals infected with NTM and MTB (Fig. 5d). The most abundant types of these 29 DEMs were phosphatidylcholines (PCs) and lysophosphatidylcholines (LPCs) with 65.5% (19/29), followed by phosphatidylethanolamines (PEs) and lysophosphatidylethanolamines (LPEs) with 20.7% (6/29, Fig. 5e). In addition, digalactosyl diacylglycerol (DGDG) O-19:0 was up-regulated the most significantly in both the RGM group (FC = 2.51, p-value = 0.020) and the SGM group (FC = 2.89, p-value = 0.02) compared to the MTB group. Remarkably, the correlation heatmap revealed a strong positive correlation between LPCs, LPEs, and DGDGs (Fig. 5f). Conclusively, the differences in serum lipid profiling between NTM patients and MTB patients mainly revolved around LPCs, LPEs, DGDGs, PCs, and PEs.
Compared with the SGM group, 8 upregulated and 45 downregulated DEMs were identified in the RGB group (Fig. 5c). LPC 28:7-SN2 was the most significant DEM with an FC = 0.55 and a p-value < 0.001.
ROC analysis revealed that Cer 20:0 was a reliable marker for distinguishing RGM from MTB, with an AUC of 0.8747. It was demonstrated that both LPE O-18:1 and LPE O-18:2 were promising differential diagnostic markers for distinguishing SGM from MTB, with AUC values of 0.7657 and 0.7467, respectively. Moreover, LPC 28:7-SN2 is a potential marker for the differential diagnosis of RGM and SGM with an AUC value of 0.7891.
Fig. 5Lipidomics analysis among the MTB, RGM, and SGM groups. a Volcano plot of DEMs between the RGM group and the MTB group. Red dots represent upregulated DEMs, while blue dots represent downregulated DEMs. b Volcano plot of DEMs between the SGM group and the MTB group. c Volcano plot of DEMs between the RGM group and the SGM group. d Venn diagram of DEMs among the MTB, RGM, and SGM groups. e Boxplots of co-upregulated DEMs in RGM/MTB and SGM/MTB. f Correlation heatmap of co-upregulated DEMs. g ROC curves for the top 5 DEMs with the highest AUC values
Correlation analysis of proteomics data and lipidomics dataSubsequently, Mantel’s test [24] was performed to analyze the correlation between the proteomics and lipidomics data. Figure 6a illustrates the results of the correlation analysis of the 16 DEPs that were co-upregulated or downregulated in the MTB, RGM, and SGM groups with 5 FAHFAs. FAHFA 20:5 showed a positive association with CCL3, IFN-gamma, MUC-16, and TNF. The Mantel’s p-value was less than 0.05, indicating a statistically significant correlation, and the Mantel’s r (correlation coefficient) ranged from 0.1 to 0.3, suggesting a moderate correlation. In addition, FAHFA 18:2 was positively associated with CD83 and MMP12. FAHFA 22:3 was positively correlated with CXCL13. And FAHFA 24:4 was positively correlated with TRAIL. The FAHFAs themselves are modified forms of fatty acids that may have specific roles in the host immune response against Mycobacteria.
Pathway analysisPathway analysis was conducted at both the protein and metabolite levels for the four comparison groups, with annotation provided by KEGG. The shared pathways that are regulated at the protein and metabolite levels among the four comparison groups are illustrated in Fig. 6b. After excluding the irrelevant KEGG pathways, only three shared pathways that were regulated at the protein and metabolite levels remained: tuberculosis, sphingolipid signaling pathway, and adipocytokine signaling pathway.
Subsequently, protein-lipid metabolite interaction maps (Fig. 6c) demonstrated the regulated DEPs and DEMs and shared pathways. Tuberculosis, sphingolipid signaling pathway, and adipocytokine signaling pathway were all enriched in the three Mycobacterium-infected groups compared to the normal group. As illustrated in Fig. 6c, the composition of tuberculosis-related DEPs and DEMs was highly similar across the three comparison groups. All three comparison groups consist of TNF, IFN-gamma, and 1-Phosphatidyl-D-myo-inositol, suggesting that these DEPs and DEMs might be regulated in all patients infected with Mycobacterium. In addition, LAP TGF-beta-1 was regulated in both MTB and SGM patients, while CASP-8 was regulated only in SGM patients. The pathways we enriched, and the proteins and lipids in these pathways, are associated with disease pathogenesis and host-microbe-related effects.
Fig. 6Proteomics and lipidomics correlation analysis and pathway analysis. a Correlation analysis of the 16 DEPs co-upregulated or downregulated in the MTB, RGM, and SGM groups with 5 FAHFAs. b Bubble map of KEGG pathways shared by proteomics and lipidomics data. The size of the bubbles indicates the enrichment factor of the KEGG pathway, while the color of the bubbles represents the p-value. c Protein-lipid interaction diagram. The red bubbles represent the KEGG pathway. Other bubbles pointing to the red bubble represent proteins and metabolites in the pathway
Establishment of protein-lipid diagnostic panelsIt has been demonstrated that many DEPs and DEMs serve as effective diagnostic and differential diagnostic markers for MTB, RGM, and SGM, as mentioned earlier. However, the differential diagnostic efficacy of DEPs and DEMs for different groups of Mycobacterium infections needs to be improved. Therefore, we furthermore conducted additional research to combine proteomic and lipidomic data to create predictive panels with increased specificity and sensitivity. The DEPs and DEMs with the highest AUC values were used to develop predictive models.
As shown in Fig. 7, the MTB diagnostic model utilizing FAHFA 18:2 and IL8 demonstrated an AUC value of 0.9754, with a specificity and sensitivity of 96% and 92.3%, respectively. This performance was significantly superior to that of a single marker. Meanwhile, diagnostic panels for RGM and SGM constructed using protein and lipid markers had AUC values of 0.9867 and 0.9562, respectively (Fig. 7b, c).
Differential diagnostic panels constructed from proteins and lipids significantly enhance diagnostic efficacy among patients with different Mycobacteria. The MTB-RGM differential diagnostic model using Cer 20:0 and MCP-2 showed an AUC value of 0.92, with a specificity of 96% and a sensitivity of 86.7%. The MTB-SGM differential diagnostic model using LPE O-18:1 and DCN showed an AUC value of 0.80, with a specificity of 72% and a sensitivity of 85.7%. Moreover, the RGM-SGM differential diagnostic model using CCL17 and LPC 28:7-SN2 showed an AUC value of 0.86, with a specificity of 92.9% and a sensitivity of 81%. Overall, the panels constructed from proteins and lipids were satisfactory for the diagnostic and differential diagnostic efficacy of Mycobacterium infections, except for the relatively low specificity (< 80%) of the MTB-SGM differential diagnostic model.
Fig. 7Establishment of protein-lipid diagnostic panels. a Establishment of the diagnostic panel for MTB. The upper panel is the ROC curve for single diagnostic markers and the lower panel is the ROC curve for a protein-lipid diagnostic panel. b Establishment of the diagnostic panel for RGM. c Establishment of the diagnostic panel for SGM. d Establishment of MTB-RGM differential diagnostic model. e Establishment of MTB-SGM differential diagnostic model. f Establishment of RGM-SGM differential diagnostic model
Comments (0)