Fig 2: CD8+ T cells inversely correlate with platelets and predict metastasis-free survival.a CXCL4 is selectively expressed in platelets (CD61+) and not by cultured 67NR and 4T1 cells, or the 67NR or 4T1 tumour cells isolated from BALB/c mouse harbouring these tumours, or by CD45+ fractions from these mice (n = 3). b Representative images of 67NR and 4T1 tumours grown in wild-type BALB/c mice stained with CD8a and CD61 (platelets). Scale bars: 200 µm. c Quantification of immunohistochemistry of 67NR and 4T1 tumours grown in wild-type BALB/c mice stained with CD8a and CD61(n = 20, analysed five areas per tumour, a total of four tumours). d Representative images of human triple-negative breast cancer tissues stained with CD8a and CD61 (platelets), scale bars, 200 µm. e Quantification of immunohistochemistry of human triple-negative breast cancer tissues stained with CD8a and CD61 (n = 68). f Negative correlation between platelets and CD8+ T cells in breast cancer using TCGA data. g Metastasis-free survival probability of patients with a low frequency of CD8+ T cells and high frequency of platelets (CD61+ cells) relative to breast cancer patients with a high frequency of CD8+ T cells and low frequency of CD61+ cells (n = 68) in breast tissue TMA. h Overall survival probability of breast cancer patients with high levels of CD8+ T cells and low levels of platelets compared to patients with low levels of CD8+ T cells and high levels of platelets (data obtained from TCGA, n = 1093). Quantitative data are mean ± SEM. Statistical analyses were performed by unpaired two-tailed Student’s t test for panels a and c or by log-rank test for panel e; ns indicates not significant, ∗P ≤ 0.05, ∗∗P ≤ 0.01, ∗∗∗P ≤ 0.001, ∗∗∗∗P ≤ 0.0001.

Fig 3: HOXB5 promotes CRC metastasis through upregulating CXCR4 and ITGB3 expression. (A) Western blotting analysis of CXCR4 and ITGB3 expression in the indicated CRC cell lines after manipulation of HOXB5 expression via lentivirus transfection. (B) Real-time PCR analysis of CXCR4 and ITGB3 mRNA expression in the indicated CRC cell lines. (C) HOXB5 transactivates CXCR4 and ITGB3 promoters. Caco-2 cells were co-transfected with pCMV-HOXB5 and CXCR4 or ITGB3 promoter luciferase constructs, then promoter activities were examined by a luciferase reporter assay. (D-E) Deletion and selective mutation analysis identified HOXB5-responsive regions in the CXCR4 (D) or ITGB3 (E) promoter. PGL3-luciferase reporter plasmids containing serially truncated or mutated CXCR4 or ITGB3 promoter constructs were co-transfected with pCMV-HOXB5 into Caco-2 cells, and relative luciferase activities were detected. (F-G) A CHIP assay indicated that the direct binding of HOXB5 to the CXCR4 promoter (F) and ITGB3 promoter (G) in both Caco-2-HOXB5 cells and human primary CRC tissues. Cells were cross-linked and the chromatin was immunoprecipitated by anti-HOXB5 or control antibody, respectively. Real-time PCR was then carried out to quantitate the immunoprecipitated products by using primers within the CXCR4 or ITGB3 promoter. (H) Western blotting analysis of CXCR4 and ITGB3 expression in the Caco-2 and SW620 cells after lentivirus transfection. (I) Transwell assay analysis demonstrated that downregulation of CXCR4 and ITGB3 inhibited the migration and invasion capacities of Caco-2-HOXB5 cells, and upregulation of CXCR4 or ITGB3 promoted the migratory and invasive potentials of SW620-shHOXB5 cells. (J-O) In vivo lung metastatic assays. Indicated cell lines were injected into the tail veins of nude mice (n = 10 mice per group). (J) Representative bioluminescent images of the different groups at 9 weeks after implantation were shown. (K) The intensity of recorded bioluminescence signals for 9 consecutive weeks of the different groups was presented. (L) Incidence of lung metastasis in the different groups. (M) The number of lung metastatic nodules in the different groups. (N) Overall survival time of nude mice in the different groups. (O) Representative images of H&E staining of lung tissues from the different groups were shown. The scale bars represent 1 mm (low magnification) and 100 µm (high magnification). (P-U) In vivo liver metastatic assays. Indicated cell lines were injected into the spleens of nude mice (n = 10 mice per group). (P) Representative bioluminescent images of the different groups at 9 weeks after implantation were shown. (Q) The bioluminescence signals of the different groups were recorded for 9 consecutive weeks after cell implantation. (R) Incidence of liver metastasis in the different groups. (S) The number of metastatic nodules in the liver was calculated. (T) Overall survival time of nude mice in the different groups. (U) Representative images of H&E staining of liver tissues from the different groups were shown. The scale bars represent 500 µm (low magnification) and 100 µm (high magnification). All the data are shown as the mean ± s.d. *P < 0.05 **P ˂ 0.01.

Fig 4: CXCR4 inhibitor AMD3100 suppresses HOXB5-mediated CRC metastasis. (A) Caco-2 cells were incubated with vehicle or AMD3100 (1 µg/ml, 24 hours) after lentivirus transfection (LV-HOXB5), then Western blotting assays were conducted to measure the protein levels of CXCL12, HOXB5, CXCR4, p-ERK and p-ETS1 in the indicated cell lines. (B) Transwell assays indicated that AMD3100 treatment (1 µg/ml, 24 hours) significantly decreased the migration and invasion of Caco-2-HOXB5 cells. (C-E) In vivo lung metastatic assays showed that AMD3100 administration significantly inhibited HOXB5-mediated CRC metastasis. Tumor cells were injected into the tail veins of nude mice (n = 10 mice per group). (C) AMD3100 or vehicle (3 mg/kg) was administered intraperitoneally to the nude mice every two days from day 8 after implantation to day 63. (D) Representative bioluminescent imaging at 9 weeks after implantation, the intensity of bioluminescence signals for 9 consecutive weeks and the incidence of lung metastasis were shown (E) Overall survival time of nude mice, the number of lung metastatic nodules and representative images of H&E staining of lung tissues from the two groups. The scale bars represent 1 mm (low magnification) and 100 µm (high magnification). (F-H) In vivo liver metastatic assays demonstrated that AMD3100 treatment dramatically reduced HOXB5-mediated CRC metastasis. Tumor cells were injected into the spleens of nude mice (n = 10 mice per group). (F) Nude mice were intraperitoneally administered with AMD3100 or vehicle (3 mg/kg) every two days from day 8 after implantation to day 63. (G) Representative bioluminescent imaging at 9 weeks, the bioluminescence signals of the two groups recorded for 9 consecutive weeks and the incidence of liver metastasis were shown. (H) Overall survival time of the nude mice, the number of metastatic nodules in the liver and representative images of H&E staining of liver tissues from the two groups are presented. The scale bars represent 500 µm (low magnification) and 100 µm (high magnification). All the data are shown as the mean ± s.d. *P < 0.05, **P < 0.01. (I) A schematic diagram of the role of the CXCL12-HOXB5-CXCR4 positive feedback loop in CRC metastasis. The CXCL12-CXCR4 axis upregulates HOXB5 expression through activating the ERK1/2-ETS1 signaling pathway. HOXB5 overexpression promotes CRC invasion and metastasis by transactivating CXCR4 and ITGB3 expression. The administration of AMD3100, a specific CXCR4 inhibitor, effectively disrupts the CXCL12-HOXB5-CXCR4 feedback loop and inhibits HOXB5-mediated CRC invasion and metastasis.

Fig 5: The αVβ3 integrin increases NgR2 expression and is associated with NgR2. (A) Immunoblotting analysis of the expression levels of the αVβ3 integrin and NgR2 (n = 3, left panels), and of the αVβ3 integrin, the NE markers chromogranin A (CHGA) and neuron-specific enolase (NSE; n = 1, middle panels), in C4-2B cells that exogenously express αVβ3 or their mock control cells. Right panel, immunoblotting analysis of the expression levels of the αVβ3 integrin, and NgR2 in PC3 cells in which the αVβ3 integrin expression was downregulated using three different siRNAs against ITGB3 (the gene responsible for the β3 integrin subunit). Non-silencing siRNA (NS) was used as control. Calnexin was used as loading control. (B) PC3 cell lysates were immunoprecipitated with Abs against the αVβ3 integrin, NgR2, the β1 integrin, the αVβ6 integrin or their isotype controls (left panel) and the αVβ3 integrin, NgR2, and the αVβ6 integrin (right panel). Immunoprecipitates were analyzed for NgR2, the β1 integrin, the αVβ3 integrin, and the αVβ6 integrin by immunoblotting (n = 2). Immunoblotting analyses were performed under non-reducing (left panel) and reducing conditions (right panel). NgR2 was detected using two different Abs (left panel).