TANs correlate with LN metastasis and poor prognosis in BCa
To investigate clinical significance of TANs infiltration in LN metastasis of BCa, we performed immunohistochemistry staining for CD66b or MPO-specific markers for neutrophils-in the tumor tissues of 207 patients with BCa. We found that higher level of CD66b+ or MPO+ neutrophil infiltration was positively correlated with high grade and muscle-invasive BCa (MIBC) (Supplementary Fig. 1A, B and Supplementary Table 1, 2). Moreover, patients with LN-metastatic MIBC presented significantly higher level of CD66b+ or MPO+ neutrophil infiltration than those with LN-negative MIBC (Fig. 1A, B and Supplementary Table 1, 2). Additionally, gene‐set enrichment analysis (GSEA) suggested that neutrophil chemotaxis, extravasation and migration pathways were upregulated in LN-metastatic BCa compared to LN-negative BCa (Supplementary Fig. 1C). Furthermore, higher level of CD66b+ or MPO+ neutrophil infiltration was associated with poorer overall survival (OS) and disease-free survival (DFS) in BCa patients (Fig. 1C-F). Moreover, univariate and multivariate Cox regression analyses confirmed that high CD66b+ or MPO+ neutrophil infiltration in BCa tissues was an independent prognostic factor for shorter OS and DFS (Supplementary Table 3-6).
Similarly, the analysis of The Cancer Genome Atlas (TCGA) database through CIBERSORTx (https://cibersortx.stanford.edu) also revealed that patients with higher level of neutrophil infiltration had reduced OS (Fig. 1G). Additionally, the analysis of TCGA database, Lee Bladder Cohort and BLAVERI bladder Cohort from the Oncomine database indicated that higher levels of MPO expression was correlated with shortened OS (Supplementary Fig. 1D-F); this was also corroborated for various types of cancer in TCGA database (Supplementary Fig. 1G-K).
To further validate correlation between TANs infiltration and LN metastasis of BCa, we obtained fresh surgically excised bladder tumor tissues and proceeded flow cytometry analysis of neutrophils. Significantly increased CD11b+CD66b+ (gated from CD45+ cells) neutrophils were observed in tumor tissues of the patients with LN-metastatic BCa compared to those with LN-negative BCa (Fig. 1H). Together, these results suggest that TANs infiltration is positively correlation with LN metastasis and predicts a poor prognosis of BCa.
TANs promote LN metastasis of BCa in mice
To determine whether TANs infiltration increases with tumor progress, we inoculated human bladder cancer UM-UC-3/luc cells into the footpads of nude mice. As shown in Supplementary Fig. 2A, TANs accumulated at footpad tumor over time, suggesting TANs might affects BCa progress in mice. To determine the function of TANs in LN metastasis of BCa, neutrophils were depleted using anti-Ly6G antibody. Treatment was initiated before tumor cells were inoculated and continued until mice developed overt LN metastatic disease (Fig. 1I). Flow cytometry and IHC analysis revealed that neutrophils were efficiently depleted in the peripheral blood and footpad (Supplementary Fig. 2B, C). Interestingly, neutrophil depletion resulted in significant reduction in popliteal LN metastasis, as determined by the in vivo imaging system (IVIS) (Fig. 1J). Moreover, smaller volume of popliteal LN (Mean volume 4.555 mm3 vs. 8.957 mm3) and decreased LN metastasis rate (16.7% vs. 66.6%) were observed in neutrophils-depleted group compared with control group (Fig. 1K, L). Overall, we discover that TANs promote LN metastasis of BCa in vivo.
Neutrophil contributes to lymphangiogenesis in BCa
Since neutrophils recruited to sites of inflammation coordinate lymphangiogenesis 25, we hypothesized that infiltrated TANs might contribute to lymphangiogenesis in tumor. To validate this hypothesis, we performed IHC analysis of LYVE1-a lymphatic endothelial cell-specific marker. Neutrophil depletion obviously decreased lymphangiogenesis in footpad tumor (Fig. 2A and Supplementary Fig. 2D). Moreover, a positive correlation was observed between the infiltration neutrophils to the aggregation of lymphatic endothelial cells around them (Fig. 2C). In addition, IF staining of MPO and LYVE1 revealed that MPO+ neutrophils and microlymphatic vessel density (MLD) were all stronger in LN-metastatic BCa than in LN-negative BCa (Fig. 2B). Furthermore, we observed significant positive correlation between neutrophil infiltration and MLD in human BCa tissues (Fig. 2C). To further investigate whether TANs directly regulate lymphangiogenesis, we isolated neutrophils from peripheral blood of patients with BCa (Supplementary Fig. 2E) and proceeded tube formation assay of HLECs in vitro. Interestingly, neutrophil culture medium (CM) slightly increased tube formation of HLECs, while T24 or UM-UC-3 cells-stimulated-neutrophil CM significantly enhanced this effect (Fig. 2D). Taken together, those results demonstrate that infiltrated neutrophils promote lymphangiogenesis and LN metastasis of BCa.
TANs-derived VEGFA and MMP9 are essential for lymphangiogenesis
To understand the mechanism how infiltrated TANs facilitate lymphangiogenesis, we examined VEGFs expression in isolated neutrophils. Unexpectedly, quantitative real-time PCR (qRT-PCR) releveled that detectable VEGFA, while little VEGFC and VEGFD were measured in freshly isolated neutrophils (Fig. 2E). In addition, expression of VEGFA and MMP9 in neutrophils was significantly increased, but not VEGFC or VEGFD after stimulating by CM of T24 or UM-UC-3 cells (Fig. 2F). Enzyme-linked immunosorbent assay (ELISA) analysis revealed production of VEGFA and MMP9 in untreated neutrophils were much lower than in BCa cells (Fig. 2G). However, production of VEGFA and MMP9 in neutrophils were significantly increased to 2-3 folds of BCa cells after stimulating by CM of T24 or UM-UC-3 cells (Fig. 2G). Furthermore, IF staining of MPO, MMP9 and VEGFA revealed that main of the MMP9 and almost half of the VEGFA were produced by TANs (Fig. 2H-L). Previous reports identified that neutrophils contributed to inflammatory lymphangiogenesis by increasing VEGFA bioavailability and bioactivity via the secretion of MMP9 25. Consistently, blocking VEGFA by its neutralizing antibody or inhibiting MMP9 activity by its inhibitor effectively suppressed UM-UC-3 cells-stimulated-neutrophil-mediated tube formation of HLEC (Fig. 2M), suggesting neutrophil-derived VEGFA and MMP9 are essential for lymphangiogenesis.
CXCL1/8-CXCR2 axis is crucial for neutrophil-mediated lymphangiogenesis
It is reported that CXCL1/8-CXCR2 and ERK/JNK pathway are relevant with neutrophil-derived VEGFA 26, 27, while CXCL1 and CXCL8 are abundant in BCa 28-31, we hypothesized that BCa cells-derived CXCL1 or CXCL8 might cooperates with CXCR2 of neutrophil to activate ERK/JNK pathway, leading to increase the production of VEGFA and MMP9. Expression of VEGFA or MMP9 and production of VEGFA or MMP9 were measured after challenged with CM of UM-UC-3 cells, with or without inhibitors of ERK and JNK. ERK or JNK inhibition significantly counteracted the increase in expression of VEGF-A and MMP9 induced by CM of UM-UC-3 cells, with comparable efficacy (Fig. 3A, B). In addition, phosphorylation of ERK and JNK were induced by CM of T24 or UM-UC-3 cells (Fig. 3C). Moreover, CXCR2 inhibitor (SB225002) markedly impaired T24 or UM-UC-3 cells-induced expression of VEGFA or MMP9 and production of VEGFA or MMP9 (Fig. 3D, E). Consistently, both CXCL1 and CXCL8 significantly induced expression of VEGFA and MMP9, while inhibition of CXCR2 blocked this effect (Fig. 3F, G). Consistent with the results from gene induction analysis, phosphorylation of ERK and JNK induced by BCa cells CM or CXCL1/8 were decreased by inhibitor of CXCR2 (Fig. 3H, I). Moreover, inhibition of CXCR2 also impaired BCa cells or CXCL1/8-stimulated-neutrophils induced tube formation of HLECs (Fig. 3J, K). These data collectively indicate that CXCL1/8-CXCR2 axis is crucial for neutrophil-mediated lymphangiogenesis by promoting VEGFA and MMP9 expression.
ETV4 regulates CXCL1/8 transcription in BCa cells
Since BCa cell-derived CXCL1 and CXCL8 are essential for neutrophil-modulated lymphangiogenesis, we hypothesized that their regulator gene might also play an important role. We analyzed promoter and enhancer of CXCL1 and CXCL8 using GeneHancer, a regulatory element database within the GeneCards® Suite (https://www.genecards.org/). We found 14 and 33 transcription factors might bind to promoter or enhancer of CXCL1 and CXCL8, respectively. By analyzing RNA-sequencing date from TCGA database, we identified that only ETV4, a potential CXCL8 regulator, was highly expressed in tumor tissues compared to normal adjacent tissues (NATs) (Supplementary Fig. 3A, B). Similarly, the analysis of Lee Bladder Cohort from the Oncomine database also indicated that ETV4 was upregulated in tumor tissues compared to normal adjacent tissues (Supplementary Fig. 3C). RT-qPCR and western blot analysis confirmed ETV4 overexpression in BCa tissues from patients compared with the corresponding NATs (Supplementary Fig. 3D, E). Analysis of the paired primary tumors and metastatic LNs from the same patients further confirmed obvious elevation of ETV4 in metastatic LNs (Fig. 4A). Next, we investigated whether ETV4 was a CXCL8 regulator. We designed two siRNAs of ETV4, both of which potently downregulated the mRNA and protein levels of ETV4 in UM-UC-3 or T24 cells (Fig. 4B, C). Knockdown of ETV4 by these two siRNAs significantly inhibited expression of CXCL8 in UM-UC-3 or T24 cells (Fig. 4B). Unexpectedly, knockdown of ETV4 also impaired expression of CXCL1, rather than other neutrophil recruitment-associated chemokines or cytokines, including CXCL2, CXCL5, CXCL6, IL1B, IL17A and TNFA(Supplementary Fig. 3F). In addition, knockdown of ETV4 decreased CXCL1 and CXCL8 secretion, as determined by ELISA (Fig. 4D). To further confirm that ETV4 drives transcription of CXCL1 and CXCL8, we cloned 200 bp around ETV4 binding sites of CXCL1 and CXCL8 into the pGL3-Basic luciferase vector, and made constructs with mutations in the ETV4 binding sites, and performed luciferase reporter assays. Interestingly, ETV4 potently activated the luciferase activity of CXCL1 and CXCL8 promoter, which was substantially impaired by mutation of ETV4 binding site (Fig. 4E). Results from chromatin immunoprecipitation (ChIP) assay demonstrated that ETV4 directly bound to promoter of CXCL1 and CXCL8, while knockdown of ETV4 decreased the enrichment (Fig. 4F). These results suggest that ETV4 directly binds to promoter of CXCL1 and CXCL8, and drives transcriptional activation of CXCL1 and CXCL8 gene.
Knockdown of ETV4 inhibits TANs recruitment and TANs-mediated lymphangiogenesis in vitro
Next, we explored the role of ETV4 in BCa cells on TANs. Knockdown of ETV4 decreased UM-UC-3 or T24 cells-induced expression of VEGFA and MMP9 in TANs (Fig. 4G, H and Supplementary Fig. 4A, B). Consistently, tube formation assay of HLECs revealed that knockdown of ETV4 impaired UM-UC-3 cells-stimulated-TANs induced lymphangiogenesis in vitro (Fig. 4I). Since the secretion of CXCL1 and CXCL8 by cancer cells promoted TANs recruitment to the tumor sites, we hypothesized that ETV4 might promote TANs accumulation. TANs were added to the upper chamber, while the supernatant of UM-UC-3 cells was added to the lower chamber. As expected, the migratory response of TANs to the supernatant of ETV4-knockdown cells was suppressed (Fig. 4J). Collectively, these data reveal that knockdown of ETV4 in BCa cells decrease TANs recruitment and TANs-mediated lymphangiogenesis in vitro.
ETV4 enhances LN metastasis of BCa in vivo via increasing TANs recruitment
To further investigate the role of ETV4 in regulating LN metastasis in vivo, the UM-UC-3/luc cells with the indicated lentiviral transfection were inoculated into the footpads of nude mice. Knockdown of ETV4 resulted in significant reduction in popliteal LN metastasis and shortened animal survival (Fig. 5A and Supplementary Fig. 5A). Consistently, decreased volume of popliteal LN and reduced LN metastasis rate were observed in ETV4-knockdown group (Fig. 5B and Supplementary Fig. 5B). These findings demonstrated knockdown of ETV4 alleviated the LN metastasis of BCa cells. Moreover, flow cytometry analyses showed that ETV4 knockdown decreased the percentages of CD11b+Ly6G+ TANs, but not CD11b+F4/80+ TAMs in the footpad tumor (Fig. 5C and Supplementary Fig. 5C). The changes of neutrophils in footpad tumor were not caused by altered biogenesis, since their abundance in the spleen of the mice was negligibly changed (Supplementary Fig. 5D). Noticeably, knockdown of ETV4 significantly decreased lymphangiogenesis in footpad tumor (Fig. 6D). To further clarify these, we constructed the stably ETV4-overexpressing UM-UC-3 cell line by lentiviral transfection. We found that overexpression of ETV4 wild type (WT) enhanced LN metastasis of UM-UC-3 cells, TANs recruitment and lymphangiogenesis in vivo. However, when TANs were specifically depleted in the mice by the anti-Ly6G antibody before BCa cells injection, the pro-LN metastasis effect of ETV4 was significantly attenuated, including the LN metastasis rate and the volume of popliteal LNs (Fig. 5E, F and Supplementary Fig. 5E). In addition, overexpression of ETV4 enhanced lymphangiogenesis in footpad of mice, which was mostly inhibited by depleting TANs (Fig. 5G). It suggests that the promoting function of ETV4 in lymphangiogenesis is also mediated by increased TANs recruitment.
ETV4 promotes migration and invasion of BCa cells in vitro
Interestingly, the volume of popliteal LN (mean volume 12.58 mm3) and LN metastasis rate (66.67 %) in (ETV4 OE + aLy6G) group were still larger and higher than the volume of popliteal LN (mean volume 7.76 mm3) and LN metastasis rate (50 %) in (CON. + IgG) group (Fig. 5F and Supplementary Fig. 5E), suggesting the pro-LN-metastatic effect of ETV4 was not only depending on TANs. To further investigate the pro-LN-metastatic effect of ETV4, we proceeded cell metastasis assay in vitro. We observed that ETV4 knockdown inhibited, while ETV4 overexpression enhanced migration and invasion abilities of BCa cells (Supplementary Fig. 6A, B and Supplementary Fig. 7E, F). Meanwhile, annexin V/PI apoptotic assay revealed that ETV4 knockdown did not affect the apoptosis of BCa cells (Supplementary Fig. 6C, D). Western blotting, and immunofluorescence analyses revealed that E-cadherin was upregulated, whereas the expression levels of N-cadherin and SNAIL were downregulated in ETV4-knockdown cells (Supplementary Fig. 6E, F). Similarly, immunoblotting analyses revealed that ETV4 overexpression upregulated the expression levels of N-cadherin and SNAIL, whereas downregulated the expression of E-cadherin (Supplementary Fig. 6G). Collectively, those results suggest that ETV4 promotes migration and invasion of BCa cells in vitro.
Y392 of ETV4 is required for the pro-LN-metastatic effect
Next, we interrogated the mechanisms underlying ETV4-regualted CXCL1/8 transcription. Since post-translational modifications are critical to the function of proteins, we analyzed the types of modifications that might occur in ETV4 through PhosphoSitePlus (https://www.phosphosite.org/). ETV4 might be phosphorylated at S140, S149 and Y392, and ubiquitinated at K96, K226, K260, K322 and K441 (Fig. 6A). To explore the potential effect of those sites on ETV4, we made constructs with mutations at those sites, respectively. RT-qPCR analyses showed that overexpression of ETV4 WT, K96R, K226R, K260R, K322R, K441R, S140A and S149A all improved expression of CXCL1 and CXCL8, but not Y392F (Fig. 6B), suggesting Y392 was critical for ETV4. Analysis of primary sequences of Homo. sapiens ETV4 and its homologues in 11 species suggested that Y392 of ETV4 was evolutionarily conserved (Supplementary Fig. 7A). Interestingly, overexpression of ETV4 WT but not Y392F in BCa cells increased CXCL1 and CXCL8 secretion of BCa cells, expression of VEGFA and MMP9 in neutrophils, neutrophil-induced tube formation of HLECs, and chemotaxis of neutrophils in vitro (Fig. 6C-F). Furthermore, in vivo assays shown that overexpression of Y392F mutated ETV4 could not promoted LN metastasis of UM-UC-3 cells, TANs recruitment and lymphangiogenesis compared with ETV4 WT, leading to the prolonged animal survival (Fig. 6G-J and Supplementary Fig. 7B-D). Besides, overexpression of ETV4 WT but not Y392F enhanced migration and invasion abilities of BCa cells (Supplementary Fig. 7E, F). Consistently, E-cadherin was downregulated, whereas the expression levels of N-cadherin and SNAIL were upregulated in ETV4-overexpressed cells, but this effect was disappeared in Y392F mutated ETV4 cells (Supplementary Fig. 7G). Taken together, these data demonstrate that Y392 is critical for ETV4 to promote LN metastasis of BCa cells.
Phosphorylation of ETV4 at Y392 by PTK6 regulates the nuclear translocation of ETV4
Then, we investigated whether ETV4 was phosphorylated at Y392. Interestingly, ETV4 Y392F showed a markedly reduced level of tyrosine phosphorylation compared with ETV4 WT (Fig. 7A). Moreover, nuclear translocation of ETV4 Y392F was markedly decreased (Fig. 7B, C), which suggested that ETV4 phosphorylation at Y392 had a possible role in regulating the nuclear translocation of ETV4. We set out to identify the tyrosine kinase that is responsible for ETV4 phosphorylation. Cytoplasmic proteins were separated to perform co-Immunoprecipitation (Co-IP) assays and mass spectrometry (MS). Tyrosine kinase PTK6 was identified as a potential ETV4-interacting protein (Supplementary Fig. 8A). Co-IP assays suggested that ETV4 constitutively interacted with PTK6 (Fig. 7D). Moreover, PTK6 overexpression promoted, while PTK6 knockdown reduced the phosphorylation of ETV4 in UM-UC-3 cells (Fig. 7E, F). Furthermore, knockdown or inhibition of PTK6 did not significantly change nuclear translocation of ETV4 Y392F, but did suppress nuclear translocation of ETV4 WT (Fig. 7G, H and Supplementary Fig. 8A, B). Consistently, PTK6 knockdown or inhibition of PTK6 by Tilfrinib in UM-UC-3 cells did not reduce ETV4 Y392, but did reduce ETV4 WT-induced expression of CXCL1 or CXCL8 in UM-UC-3 cells and expression of VEGFA or MMP9 in TANs (Fig. 7I, K and Supplementary Fig. 8C, D). In addition, PTK6 knockdown in UM-UC-3 cells did not affect ETV4 Y392F, but did inhibit ETV4 WT-associated tube formation of HLECs and chemotaxis of TANs in vitro (Fig. 7J, L). Collectively, those results suggest that PTK6-mediated ETV4 phosphorylation at Y392 promotes nuclear translocation of ETV4, which leads to promoted LN metastasis-related function of ETV4.
ETV4 expression is associated with neutrophil infiltration and poor prognosis in BCa
We then assessed the clinical significance of ETV4 in LN metastasis of BCa. As shown by IHC, the protein expression levels of ETV4 were significantly upregulated in the LN-metastatic BCa tissues, slightly elevated in LN-negative BCa tissues, compared with normal adjacent tissues (Fig. 8A). Additionally, ETV4 was positively correlated with muscle-invasive and high grade BCa (Supplementary Table 7). Moreover, Kaplan-Meier analysis demonstrated that high ETV4 expression was significantly associated with decreased OS and DFS (OS: cohort 3, HR=2.931, 95% CI=1.817 to 4.728, p<0.001; DFS: cohort 3, HR=2.406, 95% CI=1.539 to 3.76, p<0.001.) (Fig. 8B, C). Importantly, univariate and multivariate Cox proportional hazards analyses showed that ETV4 expression was an independent prognostic factor for OS and DFS of BCa patients (Supplementary Table 8, 9). Similarly, the analysis of GSE31680 Cohort from GEO database indicated that higher levels of ETV4 expression was correlated with shortened OS (Fig. 8D). Furthermore, our pathology results showed that ETV4 expression levels positively correlated with infiltrated MPO+ or CD66b+ TANs levels in the tumor tissues (Fig. 8E, F). Consistently, we further observed ETV4 mRNA expression levels positively correlated with CXCL1 or CXCL8 mRNA expression levels in the tumor tissues (Fig. 8G, H). These results further verify that expression of ETV4 is positively correlated with CXCL1/8 expression and TANs infiltration, which predicts high rates of LN metastasis and poor prognosis outcome of BCa.