CD41+/CD62P+ platelets are an independent prognostic factor for PDAC
First, a meta-analysis of 12 published studies was performed to clarify the function of blood platelets in PDAC (Supplementary Table 1). The data showed that a high platelet count is associated with high mortality in PDAC patients (HR = 1.31) (Figure 1A). Next, blood samples from 274 stage I/II patients, 382 stage III/IV patients and 3105 healthy people were collected in our cohort. The three groups were matched in terms of age and sex (Supplementary Table 2). Blood platelets were collected before surgery and anticancer therapy. In contrast to the meta-analysis, we found that the blood platelet count was higher in healthy people (n=3105) than in PDAC patients (n=656) (P < 0.0001) (Figure 1B), while there was no significant difference between stage I/II (n=274) and stage III/IV (n=382) PDAC patients (Figure 1B, C), suggesting that total platelet count is not a valuable predictive marker in PDAC malignancy. Our previous study also showed that total platelet count could not predict PDAC malignancy.21 The subgroup of CD41+/CD62P+ platelets, representing a group of activated platelets, was reported to promote metastasis in lung cancer.11 We therefore explored whether the CD41+/CD62P+ platelets is correlated to PDAC malignancy. The flow cytometry analysis of platelets showed that the percentage of CD41+/CD62P+ platelets was higher in stage III/IV PDAC than in stage I/II PDAC (P < 0.001) (Figure 1D). Additionally, the percentage of CD41+/CD62P+ platelets in stage I/II PDAC patients was higher than that in healthy donors (P < 0.0001) (Figure 1D). The stage I/II group of PDAC patients undergoing radical pancreatectomy was divided into two groups according to the median percentage of CD41+/CD62P+ platelets. Elevated CD41+/CD62P+ platelets were significantly associated with several clinical tumor features, such as microvascular invasion (P = 0.002), 8th edition AJCC stage (P < 0.001), and high CA19-9 classification (P = 0.027) (Supplementary Table 3). We also found that elevated level of CD41+/CD62P+ platelets was significantly associated with several clinical tumor features in III/IV stage patients, such as 8th edition AJCC stage (P < 0.001), and metastasis (P < 0.001) (Supplementary Table 4). Kaplan-Meier (K-M) analysis showed that patients with higher CD41+/CD62P+ platelets had a shorter overall survival time (OS) and relapse-free survival time (RFS) than patients with lower CD41+/CD62P+ platelets after surgery (Figure 1E, F). However, higher level of CD41+/CD62P+ platelets was not associated with a shorter OS for those III or IV stage patients (Supplementary Figure 1). Receiver operating characteristic (ROC) curve analysis showed that the area under the ROC curve (AUC) for CD41+/CD62P+ platelets associated with 1-year OS was 0.776 and 1-year RFS was 0.733 in stage I/II patients after surgery (Figure 1G, H). The multivariate analysis indicated that CD41+/CD62P+ platelets represented an independent prognostic risk factor associated with OS (hazard ratio [HR]: 2.180, P < 0.001) and RFS (HR: 3.361, P < 0.001) after surgery (Supplementary Table 5).
PX1 is involved in the activation of CD41+/CD62P+ platelets
Next, we explored critical genes involved in the activation of CD41+/CD62P+ platelets. Platelets isolated from 6 healthy people were treated by thrombin to obtain activated platelets. Then RNA-seq was performed to find differentially expressed genes. We noticed that PX1 transcription level was significantly upregulated in CD41+/CD62P+ platelets (activated platelets) compared to CD41+/CD62P- platelets (non-activated platelets) (Figure 2A). Further validation using qRT-PCR verified that PX1 (not PX2 or PX3) was upregulated in CD41+/CD62P+ platelets than in CD41+/CD62P- platelets (Supplementary Figure 2A). Consistently, western blotting also proved that PX1 expression level was higher in CD41+/CD62P+ platelets than in CD41+/CD62P- platelets (Supplementary Figure 2B). Interestingly, the PX1 expression level in platelets was also higher in stage III/IV patients than in stage I/II patients and healthy people (Figure 2B). We next focused on the role of PX1 in platelet activation. The platelets isolated from PDAC patients were collected for further study. The extended pseudopodia shape of platelets usually represents activated platelets while the round shape represents non-activated platelets.22 Immunofluorescence (IF) and scanning electron microscopy (SEM) indicated that PX1+ platelets were more easily activated than PX1- platelets (Figure 2C).
To investigate the function of PX1 in platelets in vivo, PX1 knockout (PX1-/-) mice were established as previously described.23 PX1 expression in platelets was confirmed by western blotting as shown in Supplementary Figure 3A. The blood concentration of platelets between WT (wild-type) and PX1-/- mice was not significantly different, while the bleeding time was significantly extended in PX1-/- mice (Supplementary Figure 3B, C). In addition, the platelet aggregates from the PX1-/- mice were markedly weaker than those of the WT mice (Supplementary Figure 3D-F). These results indicated that PX1 knockout did not affect platelet production but blocked the function of platelets. WT and PX1-/- platelets isolated from WT mice and PX1-/- mice were treated with thrombin for 15min or co-culture with PDAC cells for 24 hours (Panc02), we noticed that WT platelets were more easily activated (higher CD41+/CD62P+ counts) than PX1-/- platelets (Figure 2D). These data suggested that PX1 plays an important role in the activation of CD41+/CD62P+ platelets, and the interaction between tumor cells and platelets may play a role in this phenomenon. However, whether PDAC cells can directly recruit platelets is largely unknown, although our previous study showed that platelet infiltration into tumors was an independent prognostic factor for patients with PDAC.21 Here, time-lapse live cell imaging revealed that platelets tended to cluster around pancreatic cancer cells (Figure 2E), while in control platelets (without tumor cells), they were evenly distributed (Supplementary Figure 3G). These observations indicated that the communication between platelets and tumor cells may play a role in platelet activation.
PX1 in platelets promoted the migration and invasion of PDAC cells via the EMT pathway
We co-cultured platelets with PDAC cells to determine the effect of PX1 expression in platelets on PDAC migration and invasion. Consistent with previous studies,24 platelets promoted tumor invasion and metastasis in vitro compared to that without co-culture of platelets (Supplementary Figure 4A-D). Compared to the co-cultured WT platelets, the migration and invasion of Panc02 cells were significantly inhibited with the co-culture of PX1-/- platelets (Figure 3A, B). For further validation, we isolated platelets from PDAC patients and used the PX1-specific inhibitor 10PX1 to block the PX1 on the platelets before co-culture with PDAC cells. 10PX1-treated platelets were washed and centrifuged to remove the remaining 10PX1 to exclude the direct effect of 10PX1 on PDAC cells (Supplementary Figure 4E, F). We found that platelets with 10PX1 treatment significantly inhibited the migration and invasion of MIA PaCa-2 cells (Figure 3A, B), suggesting that PX1 expression in platelets determines the pro-tumor role of platelets in PDAC tumor.
In addition, co-culture of PX1-/- platelets decreased the PDAC cell-cell adhesion and spindle-like appearance compared to those observed in WT platelets (Supplementary Figure 4G). We then investigated whether PX1 expression could upregulate the EMT pathway in PDAC tumor cells. qRT-PCR showed that WT platelets significantly increased the mRNA levels of snail, vimentin, and N-cadherin, the hallmarks of mesenchymal cells, and reduced the mRNA level of E-cadherin, the hallmark of epithelial cells in PDAC cells, compared to their levels in PX1-/- platelets (Figure 3C). Similar results were observed when human platelets were treated by 10PX1 (Figure 3C). Murine WT platelets and human platelets also increased the protein expression of EMT markers in Panc02 and MIA PaCa-2 cells compared to the loss function of PX1 in platelets (Figure 3D, E).
PX1 in platelets promotes PDAC invasion and metastasis in vivo
In agreement with previous studies,24 we found that platelets promoted tumor invasion and metastasis in vivo (Supplementary Figure 5). To confirm the effect of PX1 in platelets on PDAC invasion and metastasis, we established an adoptive platelet transfusion mouse model (Supplementary Figure 6A), in which the endogenous platelets were neutralized and depleted by anti-CD41 antibodies before administration of PX1-/- platelets or WT platelets. We found that the number of platelets in mouse was effectively depleted by anti-CD41 injection (Supplementary Figure 6B), and the transfused platelets accounted for 50% of the total platelet counts after exogenous platelet transfusion (Supplementary Figure 6B, C). These data suggested that we successfully established an adoptive platelet transfusion mouse model. In the subcutaneous tumorigenesis model, Panc02 cells co-cultured with PX1-/- platelets and WT platelets were implanted subcutaneously into the flank of adoptive platelet transfusion nude mice. The results showed that the tumors co-cultured with PX1-/- platelets grew more slowly than those co-cultured with WT platelets (Figure 4A). In WT platelet-educated subcutaneous tumors, more positive vimentin, snail, and N-cadherin staining and less E-cadherin were observed compared to the results in PX1-/- platelet-educated tumors (Supplementary Figure 6D). Furthermore, another human PDAC cell line, MIA PaCa-2, was co-cultured with human platelets with/without 10PX1 blockade and then implanted subcutaneously into the flanks of adoptive platelet transfusion nude mice. Tumors co-cultured with 10PX1-treated platelets grew more slowly than those without 10PX1-treated platelets (Supplementary Figure 6E). Similar result was observed when PDAC patient-derived platelets were co-cultured with MIA PaCa-2 cells compared to the platelets isolated from healthy donors (Supplementary Figure 6F). In the mouse lung metastasis model, Panc02 cells co-cultured with PX1-/- platelets and WT platelets were injected into adoptive platelet transfusion nude mice via the tail vein. We found that more metastatic lung nodules were formed in the WT platelet-injected mice than in the PX1-/- platelet-injected mice (Figure 4C). To visualize the localization of platelets in lung, WT platelets and PX1-/- platelets were fluorescently labeled by PKH26 and injected into mice via the tail vein. We found that more WT platelets were present in lungs (Figure 4D), while PX1-/- platelets mainly remained in the blood (Supplementary 6G). Interestingly, more Panc02 cells were observed to be present in the lung when mixed with WT platelets compared to that with PX1-/- platelets (Figure 4D). These data suggested that PX1 expression in platelets promotes PDAC tumor metastasis. To further confirm the effect of PX1 in platelets on PDAC metastasis, the adoptive platelet transfusion WT mice was used to study. Compared to the transfusion of WT platelets, transfusion of PX1-/- platelets resulted in less platelet activation and less lung metastasis in the WT mice after Panc02 cell injection with the tail vein (Figure 4E, F). Orthotopic mouse model of pancreatic cancer (Panc02 tumor cells) were established in adoptive platelet transfusion WT mice. After 14 days, all mice were assessed by bioluminescence imaging (BLI). The results showed that the WT mice with the transfusion of PX1-/- platelets had lower average BLI tumor intensities than were found in mice transfused with WT platelets (Figure 4G). In addition, the WT mice transfused with PX1-/- platelets had longer survival times (Figure 4H). Furthermore, PX1-/- platelets alleviated the deterioration of the whole body of mice and bloody ascites caused by tumor invasion in adoptive platelet transfusion WT mice. (Supplementary Figure 7A-C). These data indicated that PX1 expression in platelets promoted the aggressiveness of PDAC tumors.
PX1 induced the synthesis and secretion of IL-1β in platelets that promoted PDAC cell invasion and metastasis
Cytokines and chemokines are cell-cell communication factors and essential coordinators of tumor metastasis. We performed a high-throughput multiplex cytokine screening of blood liquid from an established orthotopic PDAC model in WT and PX1-/- mice. The data showed that the protein level of IL-1β was significantly higher in the blood of WT mice compared to that of PX1-/- mice after establishing the PDAC orthotopic model (Figure 5A). Other chemokines, such as MCP-3 (CCL7), MIP-1β (CCL4), MIP-2 (CXCL2) and RANTES (CCL5), were not significantly different between WT and PX1-/- mice after the tumor orthotopic establishment (Supplementary Figure 8A). Several other cytokines, such as IL-10, IL-18, M-CSF, and TGF-β, also did not show any differences (data not shown). In the co-culture of PDAC cells in vitro, the expression of IL-1β was significantly elevated in supernatants of cells when WT platelets rather than PX1-/- platelets were co-cultured with Panc02 or treated by thrombin (Figure 5B), suggesting that PDAC cells stimulates the WT platelet-derived IL-1β secretion. We also observed that the serum IL-1β levels were higher in stage III/IV patients (n=20) than in stage I/II patients (n=20) and healthy people (n=20) (Figure 5C). Furthermore, immunofluorescence and flow cytometry results showed that most of the CD41+CD62P+ platelets isolated from PDAC patients were IL-1β positive, while CD41+CD62P- platelets were IL-1β negative (Figure 5D, E). Next, the wound healing assay and transwell assay were performed to test the role of IL-1β secreted by platelets in PDAC metastasis. The data revealed that anti-IL-1β antibodies (IL-1βAb) significantly impeded the migration ability and EMT pathway of PDAC cells co-cultured with WT platelets (Figure 5F-I), while recombinant IL-1β (rmIL-1β) significantly promoted the migration ability and EMT pathway in PDAC cells co-cultured with PX1-/- platelets (Figure 5F-I). For further control, we used IL-1βAb and rmIL-1β to treat PDAC cells without the co-culture of platelets. The data showed that rmIL-1β promoted PDAC cell invasion and migration in vitro (Supplementary Figure 8B, C). In vivo, IL-1βAb administration in WT mice suppressed lung metastasis compared to that in WT mice injected with PBS (Figure 5J). In contrast, rmIL-1β increased lung metastasis in PX1-/- mice compared to that in PX1-/- mice injected with PBS (Figure 5J). These results demonstrated that PX1 induced the synthesis and secretion of IL-1β in platelets, which promoted PDAC cell invasion and metastasis.
PX1 enhanced platelet-derived IL-1β secretion through phosphorylating p38 MAPK
The p38 MAPK pathway has been reported to regulate NLRP3 and IL-1β expression in macrophages.25 Furthermore, our RNA-Seq data showed that PX1 was associated with the NLRP3/IL-1β and p38 MAPK pathway (Supplementary Figure 9A). Other pathways, such as the WNT/β-catenin, PI3K/AKT, and ERK/MAPK pathways, did not show significant differences between the WT platelets and PX1-/- platelets (Supplementary Figure 9B). We next tried to determine the relationship between PX1 and the p38 MAPK pathway. After 15 min of thrombin treatment, we found that the phosphorylated p38 MAPK (p-p38 MAPK) level was upregulated only in WT platelets but not in PX1-/- platelets (Figure 6A). We found that the p-p38 MAPK level in human platelets was inhibited by the 10PX1 peptide (Figure 6A). Furthermore, the expression of p-p38 MAPK in patients’ platelets was significantly higher than that in platelets isolated from healthy people after thrombin treatment (Supplementary Figure 9C). Immunofluorescence also showed a higher p-p38 MAPK expression in WT platelets compared to that in PX1-/- platelets (Figure 6B). After thrombin stimulation for 24 h, P38 MAPK was not upregulated in either WT or PX1-/- platelets. By contrast, only in WT platelets were NLRP3 and IL-1β expression as well as their mRNA levels increased (Figure 6C, D). Consistent with this observation, the expression of NLRP3 and IL-1β was significantly decreased only in human platelets with 10PX1 treatment (Figure 6C, D). After blocking the p38 MAPK pathway by SB203580, the expression of NLRP3 and IL-1β was significantly decreased in WT-derived platelets and human platelets (Figure 6E). These data suggested that PX1 regulated the secretion of IL-1β via activating the p38 MAPK pathway. Consistently, in vivo, the serum IL-1β level in WT mice was significantly reduced by using SB203580 (Figure 6F). Furthermore, SB203580 inhibited the activation of WT platelet and suppressed PDAC cell invasion and metastasis (Figure 6G, H). Of note, SB203580-treated platelets were washed and centrifuged to remove the remaining SB203580 for excluding the direct effect of SB203580 on PDAC cells (Supplementary Figure 9D, E). We also found that PX1 deficiency inhibited ATP release from platelets after the thrombin treatment (Supplementary Figure 9F), which is similar to the findings of a previous study that PX1 is associated with ATP release in macrophages.23 The structure analysis showed that ATP could bind to the ATP binding site on p38 MAPK and thereby promoted the phosphorylation of p38 MAPK (Figure 6I).
PX1 is a potential therapeutic target for PDAC
Our above results indicated that the PX1-p38 MAPK-IL-1β axis was activated in CD41+/CD62P+ platelets and promoted PDAC cell invasion and metastasis. Thus, we investigated whether PX1 on CD41+/CD62P+ platelets could serve as a novel therapeutic target for PDAC tumors. P-selectin glycoprotein ligand-1 (PSGL-1) was reported to specifically bind to CD62P, which is a marker of activated platelets and is expressed on their surface.26 Carbenoxolone is a PX1 inhibitor.19 These two components were ligated by chemical synthesis to generate the new PX1 inhibitor (Figure 7A). 1H-nuclear magnetic resonance (NMR) and mass spectrometry indicated that PSGL-1 and carbenoxolone formed a perfect combination (Figure 7B, C), and the molecular ion of the newly generated inhibitor was observed at 63435 (Figure 7C). We therefore designated this novel drug as PC63435. Compared to the WT mice injected with PBS, the WT mice injected with PC63435 had a prolonged bleeding time (Supplementary Figure 10A) and impaired platelet aggregation (Supplementary Figure 10B). Flow cytometry showed that the number of CD41+/CD62P+ platelets in the WT mice injected with PC63435 was lower than that in the PBS-injected mice (Figure 7D, E). Additionally, PC63435 reduced serum IL-1β levels in WT mice (Figure 7F). Then, the anti-tumor effect of PC63435 was tested in the orthotopic PDAC mouse model and lung metastasis model in WT mice with Panc02 injection. The data showed that PC63435 suppressed PDAC invasion and metastasis in WT mice compared to those in the control group (Figure 7G-I). In addition, PC63435 significantly prolonged the survival of WT mice by approximately 2-fold compared to that in PBS-treated mice in the orthotopic PDAC mouse model (Figure 7J).
In conclusion, we demonstrated that PX1 expression in platelets activates the p38 MAPK pathway, enhances platelet-derived IL-1β secretion, and promotes PDAC invasion and metastasis. Specifically, we developed a novel drug, PC63435, that is able to block PX1 on CD41+/CD62P+ platelets and therefore suppress the invasion and metastasis potential of PDAC (Figure 8).