3.1 DLX gene alterations and mRNA expression in COAD
The cBioPortal online tool was used to analyze the gene expression of DLX genes in COAD patients. Alterations in the DLX genes in COAD ranged from 0.7–3% (Fig. 1). The mutation data, CNA (copy number alteration) data, and deep deletion from the 2 studies are depicted in Fig. 2. The analysis of DLX gene expression was performed based on 41 COAD tumor tissue samples and 41 paired samples of normal colon tissues (Fig. 3). The results showed that the expression level of DLX2 in COAD tumor samples was significantly higher than that of DLX2 in normal colon tissues (0.317 ± 0.448 vs. 0.133 ± 0.101, P = 0.015), the expression level of DLX3 in COAD tumor samples was significantly higher than that of DLX3 in normal colon tissues (0.928 ± 1.246 vs. 0.288 ± 0.165, P = 0.003), the expression level of DLX4 in COAD tumor samples was significantly higher than that of DLX4 in normal colon tissues (0.726 ± 0.525 vs. 0.166 ± 0.083, P < 0.001), the expression level of DLX5 in COAD tumor samples was significantly higher than that of DLX5 in normal colon tissues (0.413 ± 0.415 vs. 0.211 ± 0.216, P = 0.005), the expression level of DLX6 in COAD tumor samples was significantly higher than that of DLX6 in normal colon tissues (0.511 ± 0.693 vs. 0.071 ± 0.076, P < 0.001). There was no significantly. However, there was no significant difference between DLX6 expression in COAD tumor samples and in normal colon tissue (0.392 ± 0.643 vs. 0.337 ± 0.263, P = 0.641). We examined the correlation between DLX genes using Pearson correlation analysis. As shown in Fig. 4, there was no significant correlation between DLX1 and DLX3, DLX1 and DLX6, and there was a significant positive correlation between other DLX genes.
3.2 Relationship between DLX gene expression and clinical characteristics and prognosis of COAD patients
Clinical characteristics data and gene expression data for 478 COAD tumor samples were downloaded from the TCGA database (Table S1). DLX2 expression was associated with M stage (P = 0.005), pathologic stage (P = 0.014), primary therapy outcome (P = 0.036), residual tumor (P = 0.002), and lymphatic invasion (P = 0.013) in COAD patients. DLX3 expression was associated with N stage (P < 0.001), M stage (P < 0.001), pathologic stage (P < 0.001), height (P = 0.045), and residual tumor (P < 0.001) in COAD patients. DLX5 expression was associated with T stage (P < 0.001), N stage (P < 0.001), M stage (P = 0.005), pathologic stage (P < 0.001), primary therapy outcome (P = 0.005), age (P < 0.001), perineural invasion (P = 0.023), lymphatic invasion (P < 0.001), and history of colon polyps (P = 0.009). However, the expression of DLX1, DLX4 and DLX6 did not significantly correlate with the clinical characteristics of COAD patients.
As shown in Fig. 5 and Figure S1, high DLX1 expression was associated with the PFI (HR: 1.50; 95% CI: 0.47–0.91; P = 0.024) of COAD; high DLX2 expression was associated with the OS (HR: 1.74; 95% CI: 1.17–2.59; P = 0.006), PFI (HR: 1.66; 95% CI: 1.16–2.37; P = 0.005), and DSS (HR: 2.02; 95% CI: 1.20–3.38; P = 0.008) of COAD; high DLX3 expression was associated with the OS (HR: 1.58; 95% CI: 1.07–2.35; P = 0.022), PFI (HR: 1.64; 95% CI: 1.15–2.34; P = 0.006), and DSS (HR: 1.98; 95% CI: 1.19–3.30; P = 0.009) of COAD; high DLX2 expression was associated with the OS (HR: 1.74; 95% CI: 1.17–2.59; P = 0.006), PFI (HR: 1.66; 95% CI: 1.16–2.37; P = 0.005), and DSS (HR: 2.02; 95% CI: 1.20–3.38; P = 0.008) of COAD; high DLX4 expression was associated with the PFI (HR: 1.54; 95% CI: 1.09–2.19; P = 0.015) of COAD; high DLX5 expression was associated with the PFI (HR: 1.69; 95% CI: 1.18–2.41; P = 0.004) and DSS (HR: 1.86; 95% CI: 1.11–3.11; P = 0.019) of COAD. However, high DLX6 expression was not significantly associated with the prognosis of COAD.
As shown in Table 1, univariate cox regression analysis for OS showed that DLX2 (HR: 1.738; 95%CI: 1.168–2.586, P = 0.006) and DLX3 (HR: 1.583; 95%CI: 1.068–2.347, P = 0.022) were negative predictors of OS outcome in COAD patients, and DLX1 (HR: 1.500; 95%CI: 1.055–2.131, P = 0.024), DLX2 (HR: 1.661; 95%CI: 1.165–2.369, P = 0.005, DLX3 (HR: 1.641; 95%CI: 1.153–2.335, P = 0.006), DLX4 (HR: 1.544; 95%CI: 1.086–2.194, P = 0.015), and DLX5 (HR: 1.687; 95%CI: 1.180–2.410, P = 0.004) were negative predictors of PFS outcome in COAD patients. DLX2 (HR: 1.541; 95%CI: 1.011–2.348, P = 0.044) was independently correlated with OS of COAD in multivariate analysis, and DLX5 (HR: 1.539; 95%CI: 1.072–2.210, P = 0.019) was independently correlated with PFS of COAD in multivariate analysis.
Table 1
Univariate and multivariate Cox regression analyses with DLX genes and prognosis of COAD patients.
|
|
|
Univariate analysis
|
Multivariate analysis
|
|
Characteristics
|
Total(N)
|
HR (95% CI)
|
P value
|
HR (95% CI)
|
P value
|
Overall
survival
|
DLX1 (Low vs. High)
|
477
|
1.245 (0.844–1.836)
|
0.268
|
|
|
DLX2 (Low vs. High)
|
477
|
1.738 (1.168–2.586)
|
0.006
|
1.541 (1.011–2.348)
|
0.044
|
DLX3 (Low vs. High)
|
477
|
1.583 (1.068–2.347)
|
0.022
|
1.325 (0.870–2.018)
|
0.19
|
DLX4 (Low vs. High)
|
477
|
1.437 (0.973–2.124)
|
0.068
|
1.123 (0.736–1.714)
|
0.592
|
DLX5 (Low vs. High)
|
477
|
1.465 (0.988–2.171)
|
0.058
|
1.333 (0.892–1.990)
|
0.161
|
DLX6 (Low vs. High)
|
477
|
0.951 (0.646–1.402)
|
0.801
|
|
|
Progress
free
interval
|
DLX1 (Low vs. High )
|
477
|
1.500 (1.055–2.131)
|
0.024
|
1.262 (0.867–1.838)
|
0.225
|
DLX2 (Low vs. High )
|
477
|
1.661 (1.165–2.369)
|
0.005
|
1.299 (0.874–1.930)
|
0.196
|
DLX3 (Low vs. High )
|
477
|
1.641 (1.153–2.335)
|
0.006
|
1.321 (0.906–1.927)
|
0.149
|
DLX4 (Low vs. High )
|
477
|
1.544 (1.086–2.194)
|
0.015
|
1.244 (0.855–1.809)
|
0.254
|
DLX5 (Low vs. High )
|
477
|
1.687 (1.180–2.410)
|
0.004
|
1.539 (1.072–2.210)
|
0.019
|
DLX6 (Low vs. High )
|
477
|
0.939 (0.664–1.329)
|
0.723
|
|
|
As shown in Fig. 6, it can be obtained that in predicting normal and tumor outcomes, variable DLX1 has some accuracy in predicting (AUC = 0.893, CI = 0.867–0.920), variable DLX2 has some accuracy in predicting (AUC = 0.731, CI = 0.691–0.771), variable DLX3 has a lower accuracy (AUC = 0.561, CI = 0.512–0.611), variable DLX4 had some accuracy in predictive ability (AUC = 0.834, CI = 0.802–0.867), variable DLX5 had low accuracy in predictive ability (AUC = 0.590, CI = 0.546–0.635), and variable DLX6 had had poor diagnostic efficacy (AUC = 0.486, CI = 0.439–0.534).
3.3 The function of genes associated with DLX genes
The top 10 significantly associated genes for each DLX gene are shown in the single gene co-expression heat map (Fig. 7). Genes significantly associated with DLX1 include DLX2, KLF14, CHRND, KCNN1, IGDCC3, ARHGAP36, NCAN, TFAP2B, CNPY1, and CACNG7. Genes significantly associated with DLX2 include DLX1, CNPY1, CHRND, NEUROD1, IGDCC3, TNFRSF19, KLF14, NELL2, HS3ST4, and SLC38A8. Genes significantly associated with DLX3 include NOTUM, NKD1, APCDD1, ADAMTSL2, MYH7B, PRR9, LRRC43, CAB39L, ABCC2, and DLX4. Genes significantly associated with DLX4 include DLX3, TTLL4, DNMT3B, CDK5R1, IGF2BP1, STK36, UNK, AMER3, PHF12, and WNT3. Genes significantly associated with DLX5 include DYNC1I1, DLX6, RASL11B, ID4, SP7, AMBN, KRT31, MYL3, VENTX, and ISM1. Genes significantly associated with DLX6 include DLX5, TRIM71, SH3GL2, SLC46A1, DYNC1I1, PGBD5, GAL, COCH, AXIN2, and CKB. The top 30 genes significantly associated with each DLX gene (147 in total) were analyzed for GO and KEGG enrichment (Table S2). The top five biological processes, including pattern specification process, regionalization, skeletal system morphogenesis, endochondral bone morphogenesis, and hippocampus development; the top five cytological components, including integral component of synaptic membrane, intrinsic component of synaptic membrane, hippocampal mossy fiber to CA3 synapse, integral component of postsynaptic specialization membrane, and intrinsic component of postsynaptic specialization membrane; the significantly related molecular function including motor activity, microtubule binding, cadherin binding, and cell adhesion molecule binding, are shown in Fig. 8 and Table S3. The significantly related pathways, including breast cancer, gastric cancer, Hippo signaling pathway, Wnt signaling pathway, signaling pathways regulating pluripotency of stem cells, basal cell carcinoma, melanoma, and staphylococcus aureus infection, were shown in Fig. 9 and Table S3.
3.4 Correlation of DLX gene expression and immune cells in COAD
As shown in Fig. 10, there was a correlation between DLX gene expression and immune cells in COAD. DLX1 gene expression was positively correlated with some TIICs, including aDC, Cytotoxic cells, DC, Eosinophils, iDC, Macrophages, Mast cells, Neutrophils, NK CD56dim cells, NK cells, Tem, TFH, Tgd, Th1 cells, and TReg, and negatively correlated with Th17 cells. DLX2 gene expression was positively correlated with Mast cells and TFH, and negatively correlated with pDC and Th17 cells. DLX3 gene expression was negatively correlated with some TIICs, including aDC, CD8 T cells, Cytotoxic cells, DC, Macrophages, Neutrophils, T cells, T helper cells, Th1 cells, Th2 cells, and TReg. DLX4 gene expression was positively correlated with NK cells, and negatively correlated with some TIICs, including Cytotoxic cells, DC, Macrophages, pDC, Th1 cells, and Th2 cells. DLX5 gene expression was positively correlated with some TIICs, including B cells, CD8 T cells, DC, iDC, Macrophages, Mast cells, Neutrophils, NK cells, pDC, Tem, TFH, Tgd, and TReg, and negatively correlated with Th17 cells and Th2 cells. DLX6 gene expression was negatively correlated with some TIICs, including aDC, Cytotoxic cells, DC, Macrophages, Neutrophils, NK CD56dim cells, T cells, Tem, and Th1 cells.