Fig 2: The recruitment factors for MDSCs (CXCL1 and CXCL2) are highly expressed in DCLK1-high colon carcinoma. (A) KEGG analysis shows that genes affected by DCLK1 knockout are enriched in the MAPK–ERK pathway. (B) GO enrichment analysis suggests that down-regulated genes are enriched in processes related to Erk1 and Erk2 cascade. (A and B) Red bars indicate up-regulated genes, while blue bars indicate down-regulated genes. The numbers around the bars are the number of genes in corresponding categories. (C) Expression correlations between DCLK1 and genes in the MAPK–ERK pathway. The X-axis shows Pearson correlation coefficients, and the Y-axis shows –log10 (P values) for the correlations. (D) Co-expression relationships between DCLK1 and the MAPK–ERK pathway in TCGA-COAD project (n = 480). (E) Expression differences among genes in the MAPK–ERK pathway. Levels are normalized by row. (F) Co-expression relationships identified in this study suggest potential crosstalk between DCLK1 and the MAPK–ERK pathway via CXCL1/CXCL2. Nodes represent proteins, and edges represent protein–protein associations, whose colors represent levels of evidence. Relative messenger RNA expression levels of CXCL1 and CXCL2 in (G) MC38-DCLK1WT, MC38-DCLK1-/-, and (I) CT26-DCLK1WT, CT26-DCLK1-/- tumor cells and (K) human CRC cell lines SW480-DCLK1WT, SW480-DCLK1-/-, and (M) HCT116-DCLK1WT, HCT116 DCLK1OE by RT-PCR analysis. ELISA results of cell supernatants of (H and J) CRC cell lines MC38-DCLK1WT, MC38-DCLK1-/-, CT26-DCLK1WT, and CT26-DCLK1-/-, and (L and N) human CRC cell line SW480-DCLK1WT, SW480-DCLK1-/-, HCT116-DCLK1WT, and HCT116 DCLK1OE. (O) High expression of DCLK1 promotes MDSC recruitment in vitro. MC38 or CT26 cells were inoculated subcutaneously into WT C57BL/6J or BALB/c mice. On the seventh day, the tumor tissues were stripped and tumor-infiltrating MDSCs were isolated. MDSCs (2 × 105) were placed into the Transwell upper chamber. DCLK1WT and DCLK1-/- cell-conditioned medium were added into the lower chamber of the Transwell, and the migration of MDSCs was observed 48 hours later. Data are presented as means ± SD. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001, 2-tailed Student t test. Akt, protein kinase B; ECM, Extracellular matrix; EGFR, Epidermal Growth Factor Receptor; NOD, Non Obese Diabetes ; PI3K, Phosphoinositide-3-kinase.

Fig 3: Multiplexed immunofluorescence displays the close association among DCLK, EMT activation and immunosuppression. (a-b). Multiplexed immunofluorescence analysis showing that high DCLK1 expression is often accompanied by increased CD163 and ZEB1 expression levels as well as decreased CD8 and E-cad expression levels. (c) Immunofluorescence quantitative analysis results showing ZEB1 and CD163 expression levels are higher in DCLK1-high group, while CD8 expression level was lower in DCLK1-high group than DCLK1-low group.

Fig 4: DCLK1 correlates with NOTCH pathways in HNSCC tumors. (A) GSEA results of MSigDB Hallmark of cancer gene sets that correlate with DCLK1 mRNA expression in TCGA-HNSC transcriptomic dataset. Blue bars indicate a positive correlation with DCLK1 mRNA expression, whilst red bars indicate negative correlations. Dashed lines indicate Pearson correlation coefficient thresholds for inclusion of the pathways for further analysis. (B) Heatmap demonstrates Hallmark pathways that positively correlate (r > 0.4) with DCLK1 mRNA expression in TCGA-HNSC transcriptomic dataset. (C) GSEA was performed using NOTCH signaling axis genes derived from 4 pathway databases: KEGG, Reactome, PID, and Hallmark. A positive correlation between DCLK1 mRNA expression and estimated NOTCH signaling enrichment using each database is indicated by r-values on the right. (D) iPANDA software suite for analysis of intracellular signaling pathway activation based on transcriptomic data was used to estimate the level of the NOTCH signaling pathway in TCGA-HNSC dataset. NOTCH axis gene sets were derived from 4 pathway databases listed in (C). TCGA transcriptomic data from non-tumorous samples was used as a reference after proper normalization. (E) GSEA querying of the hallmark gene sets using the transcriptomic profiles from TCGA-HNSC and 7 additional independent cohorts of HNSCC patients obtained from either NCBI GEO or ArrayExpress databases was performed. For each pathway, a positive correlations with DCLK1 mRNA expression are indicated in red, whilst negative correlations are shaded in blue.

Fig 5: DCLK1 is a target of miR-223-5p. 6-10B cells were transfected with NC mimics, miR-223-5p mimics, NC inhibitor or miR-223-5p inhibitor. (A) Predicted binding sites between miR-223-5p and DCLK1. (B) Luciferase activity was analyzed in 6-10B cells transfected with NC mimics and miR-223-5p mimics. (C) miR-223-5p expression in 6-10B cells transfected with NC inhibitor or miR-223-5p inhibitor was measured using RT-qPCR. (D) DCLK1 expression in 6-10B cells was measured using RT-qPCR. *P<0.05 vs. NC mimics; #P<0.05 vs. NC inhibitor. DCLK1, doublecortin-like kinase 1; miR, microRNA; RT-qPCR, reverse transcription-quantitative PCR; NC, negative control; WT, wild-type; MUT, mutant.