To investigate the effects of primary PDAC on the activation of HSCs, sequencing data from human-derived (GSE71729) and mouse-derived (GSE148139, GSE160541) pancreatic cancer released mediators were obtained from the public database Gene Expression Omnibus (GEO) datasets (Fig. 1A). In GSE71729, 145 primary and 61 metastatic PDAC patients tumors were selected in our study. In GSE148139, an examination of mice primary HSCs RNA sequencing comparing the PDAC-exosomes treatment group with the control group was selected. And in GSE160541, bulk RNA sequencing was performed in 3 isolates of quiescent mouse HSCs; 3 isolates of HSC-derived CAF from Panc02; and 3 isolates of HSC-derived CAF from KPCY. Among these datasets, 25 genes were significantly upregulated (Fig. 1B; log[FoldChange] > 2, p < 0.05). Most of these genes were associated with hepatic fibrosis and cytokine secretion in HSCs (Fig. 1C). The highly expressed genes Col1a1, Acta2, and MMP3, which are related to HSCs activation, were identified from the GSE148139 and GSE71729 databases. Furthermore, gene-set enrichment analysis (GSEA) demonstrated that highly expressed genes in HSCs were predominantly enriched in protein digestion and absorption, collagen fibril organization, and Hippo signaling pathways induced by mediators secreted by PDAC cells. Additionally, the production of molecular mediators of the immune response was downregulated (Fig. 1D).

We then conducted single-cell sequencing analysis on PDAC liver metastases and normal liver tissue to investigate the presence of HSCs. We analyzed a total of 90,583 cells after filtering low-quality cells and visualized them using the Uniform Manifold Approximation and Projection (UMAP) plot. These cells were integrated well by removing the batch effect and clustered into 9 clusters (Fig. 1E). We defined the cell types of each cluster according to their signature genes and canonical cell-type markers curated from literature (Supplementary Fig. 1A). Our analysis revealed higher expression levels of IL1r1 in HSCs within the liver metastasis group compared to the normal group (Fig. 1F). In addition, we also detected the expression of IL-6 and YAP genes targeting a subpopulation of HSCs. These results fully corroborate the activation of HSCs in PDAC liver metastasis PMN (Fig. 1G). Enrichment analysis of single-cell sequencing data for PDAC liver metastases showed that HSCs-specific differential genes were enriched in collagen production, formation of an immunosuppressive environment, cytokine secretion, and activation of the Hippo signaling pathway (Fig. 1H). These results demonstrate that PDAC cells can promote hepatic fibrosis and activate HSCs.

Since exosomes play an important role in cell and organ communication [40], we hypothesize that PDAC-derived exosomes could activate HSCs during liver metastasis of pancreatic cancer. We next extracted exosomes of PDAC cell lines and subjected them to quality identification. Transmission electron microscopy (TEM) observation and exosome nanoparticle tracking analysis (NTA; Supplementary Fig. 1B, C) confirmed that the extracted exosomes had a diameter of approximately 80 nm and intact membranes. Exosomal signature proteins (CD9, CD81, CD63) were also detected to prove the successful extraction of the exosome (Supplementary Fig. 1D).

To investigate the specific effects of PDAC-derived exosomes on liver in vivo, PDAC-exo was injected into C57BL/6 mice via the tail vein. As shown in Fig. 2A, mice were treated with 40 μg of PDAC-exo every 3 days for 3 weeks, after which the degree of liver fibrosis was assessed, and the expression of α-SMA and Collagen I in the liver was detected. The results showed PDAC-exo-treated mice exhibited severe liver fibrosis (Fig. 2B, C), and increased expression of extracellular matrix protein (Fig. 2H).

The activation of HSCs is an important cause of liver fibrosis. To investigate whether HSCs are activated by uptake exosomes, we isolated primary hepatic stellate cells (p-HSCs) from surgical specimens of patients’ livers in Tongji Hospital. PKH26-labeled exosomes were added to a culture system containing p-HSCs. After 12 hours, red fluorescence was observed in p-HSCs, indicating the uptake of exosomes by HSCs (Fig. 2D). The collagen gel contraction assay and transwell migration assay were performed. As expected, AsPC1- and BxPC3-derived exosomes had the greatest impact on the contractility and migration ability of p-HSCs (Fig. 2E, F, Supplementary Fig. 1E, F). Under fluorescence microscopy, PDAC-exo-treated p-HSCs showed the increasing fluorescence intensity of α-SMA, along with significant increases in cell volume and synapses (Fig. 2G). Increasing the level of PDAC-exo in p-HSCs and LX2 cells resulted in a significant upregulation of extracellular matrix proteins (fibronectin and collagen I) and HSC activation proteins (α-SMA and FAP) (Fig. 2I, Supplementary Fig. 1G). To further investigate the role of exosomes, PDAC cells were treated with the exosome-release inhibitor GW4869, and the supernatant was added to p-HSCs and LX2 respectively. Inhibition of HSC activation proteins and extracellular matrix protein expression was observed in both p-HSCs and LX2 cells (Fig. 2J, Supplementary Fig. 1H). Therefore, these findings confirm that PDAC-exo promotes hepatic stellate cell activation and induces liver fibrosis.

So far we have shown that PDAC-derived exosomes are the potential carrier and initiator of liver fibrosis. To further explore PDAC-exo contain tRFs in their cargo that may play a role in HSC activation, we collected peripheral blood supernatants from 5 PDAC patients with early metastasis (C1-5) and 5 healthy volunteers (N1-5) in Tongji Hospital. After extracting exosomes, we performed small RNA sequencing to identify differentially expressed tRFs. From the initial analysis of 431 tRFs, we selected the top 10 genes with the most significant differences for heat map presentation (Fig. 3A). Meanwhile, we examined the expression of tRFs in a variety of cancer tissues in the online dataset tdRFun. Remarkably, tRF-GluCTC-0005 was found to be significantly overexpressed in PDAC tissues (Suppl. Fig. 2A).

tRF-GluCTC-0005 (hereafter referred to as tRF-GluCTC) is a cleavage fragment of tRNA-GluCTC, 24 bases in length, formed by Dicer enzyme shearing from the 5’ end of tRNA, and contains the D-loop of the dihydrouridine region (Fig. 3B). We isolated exosome from the peripheral blood samples of 73 PDAC patients (cohort I, n = 73) collected at Tongji Hospital. We combined tRF-GluCTC with the traditional diagnostic marker CA19-9 to evaluate its diagnostic efficiency. The expression of tRF-GluCTC in peripheral blood serum exosomes showed higher sensitivity and specificity compared to CA19-9 (Supplementary Fig. 2B).

To validate the specificity of tRF-GluCTC as a potential biomarker for liver metastasis, we compared its expression in PDAC patients with or without liver metastases. Using RT-qPCR on peripheral blood exosomes, we detected the expression of 4 candidate tRFs in 5 PDAC patients with liver metastases and 5 PDAC patients without liver metastases. Notably, tRF-GluCTC displayed significant expression differences, indicating its potential association with liver metastasis (Fig. 3C). We then examined the expression of tRF-GluCTC and the control tRF in the serum and tissue of PDAC patients in cohort I and cohort II. The results demonstrated significantly higher expression of tRF-GluCTC in serum exosomes of patients with liver metastases compared to those without liver metastases in cohort I (OR = 4.125, p = 0.0292, Suppl. Fig. 2C). Similarly, patients with high tissue expression of tRF-GluCTC had a greater likelihood of liver metastasis in cohort II (OR = 2.982, p = 0.0334, Supplementary Fig. 2D).

To assess the prognostic value of tRF-GluCTC, we performed a Kaplan-Meier survival analysis. Patients with high tRF-GluCTC levels exhibited significantly shorter overall survival (OS) and disease-free survival (DFS) compared with low tRF-GluCTC levels in both cohort I (n = 73) and cohort II (n = 151, Fig. 3D). Furthermore, multivariate analyses using Cox proportional hazards regression demonstrated that tRF-GluCTC expression served as an independent prognostic factor for overall survival in PDAC patients (p = 0.026, Fig. 3E & Table 1). We then compared tRF-GluCTC levels in tissues according to different tumor stages in cohort I. The results revealed that tRF-GluCTC levels were significantly higher in advanced-stage tumors (stages IIB/III/IV) than in early-stage tumors (stages I/IIA; Fig. 3F).

To further confirm the expression of tRF-GluCTC in pancreatic cancer tissues, we obtained surgical specimens from PDAC patients. Immunofluorescence staining revealed increased fibrotic features and higher fluorescence intensity of alpha-smooth muscle actin (α-SMA) and collagen I in pancreatic cancer tissues compared to paracancerous tissues. Additionally, fluorescence in situ hybridization (FISH) staining demonstrated significantly higher fluorescence intensity of tRF-GluCTC in pancreatic cancer tissues compared to paracancerous tissues (Fig. 3G).

To further investigate the changes in the expression of tRF-GluCTC during different stages of PDAC development, including acinar-to-ductal metaplasia (ADM) and mouse pancreatic intraepithelial neoplasia (mPanIN), we used the Kras-mediated mouse model of spontaneous pancreatic cancer in LSL-Kras^G12D/+; Pdx1-Cre (KC) mice, as well as LSLKras^G12D/+; Trp53^fl/+; Pdx1-Cre (KPC) mice (Supplementary Fig. 2E). First, we employed positron emission tomography/computed tomography (PET/CT) imaging with ⁶⁸Ga-DOTA-FAPI on KC, KPC, and wild type (WT) mice. The ⁶⁸Ga-DOTA-FAPI Micro-PET data revealed that the KPC mice developed pancreatic cancer with significant liver metastases, whereas KC mice exhibited a slower progression of pancreatic cancer (The red arrows represent pancreatic cancer foci. Fig. 3H). Histological examination through hematoxylin and eosin (H&E) staining demonstrated that PDAC cells in KPC mice had infiltrated the basal lamina and formed diffuse tumors in the pancreas. Conversely, KC mice showed no penetration of the basal lamina, forming ADM or mPanIN (Fig. 3I).

Furthermore, the expression of α-SMA was significantly increased in liver issues of KPC mice compared with KC mice and WT mice. The FISH analysis revealed that the fluorescence intensity of tRF-GluCTC was highest in KPC mice compared to KC and WT mice (Fig. 3I). Additionally, we extracted peripheral blood serum exosomes from KPC, KC, and WT mice, and the RT-qPCR results showed that tRF-GluCTC expression was highest in the peripheral blood exosomes of KPC mice (Fig. 3J). Similarly, the expression of tRF-GluCTC was highest in the liver tissue of KPC mice (Fig. 3K). Taken together, these findings demonstrated that the expression of tRF-GluCTC in peripheral blood exosomes and liver tissues increased significantly with the progression of PDAC.

To investigate the specific molecular mechanisms of tRF-GluCTC in PDAC liver metastasis, firstly, we transfected mimics and inhibitors of tRF-GluCTC into AsPC1 cells and assessed the invasion and migration abilities of AsPC1 cells (Supplementary Fig. 3A). Next, we extracted exosomes from the supernatants of AsPC1 cells transfected with mimic and inhibitor. We observed that only the expression of tRF-GluCTC was altered after exosome treatment in p-HSCs, while the expression of other tRFs remained unaffected (Supplementary Fig. 3B).

To better understand the role of PDAC-exo-induced liver PMNs and metastasis, we established 2 mouse models in vivo (Fig. 4A). The first animal model was aimed to establish the PMNs of PDAC. We injected 40ug of Ctrl-exo, tRF-mimic-exo, or tRF-inhibitor-exo through the tail vein of mice every three days for a total of three weeks, a process referred to as “education”. We obtained mouse livers and detected the expression of HSCs-activated proteins and extracellular matrix proteins in mouse livers using immunohistochemical staining. We observed a significant increase in liver fibrosis in tRF-mimic-exo stimulated mice, whereas liver fibrosis in tRF-inhibitor exosome-stimulated mice was not significant (Supplementary Fig. 4A). Western blot analysis of mouse liver proteins confirmed that the expression of α-SMA and FAP was higher in the tRF-mimic group, and similarly the expression of extracellular matrix proteins was also increased (Supplementary Fig. 4B).

The second animal model focused on liver metastasis of pancreatic cancer. Following three weeks of PDAC-exosome “education” in the mice, 2*10⁶ Panc02 cells were injected into the portal vein. The IVIS Lumina imaging system was used to monitor the liver metastasis weekly. In the fourth week, we found the fluorescence intensity in mouse livers was higher in the tRF-mimic-exo education group, compared with the Ctrl-Exo education group and tRF-inhibitor-exo education group (Fig. 4B). We quantified the proportion of liver metastases that occurred in the different groups, and the results demonstrated that the tRF-GluCTC mimic group had a higher propensity for liver metastasis compared to the control group (Fig. 4C). Subsequently, mice were sacrificed, and their livers were harvested for IHC staining. The degree of fibrosis and rates of metastases were significantly higher in mice with the tRF-mimic-exo education group compared to those in the Ctrl-exo education group (Fig. 4D). These results suggested that exosomal tRF-GluCTC promotes liver metastasis induced by PDAC derived exosomes.

The results of the collagen gel contraction assay and transwell migration assay showed that increased expression of tRF-GluCTC via exosome in p-HSCs and LX2 significantly enhanced cell contractility and migration ability (Fig. 4E, F). Under fluorescence microscopy, we found that the addition of PDAC-exo transfected with tRF-mimic resulted in increased cell volume and synapse formation in p-HSCs. Moreover, the fluorescence intensity of α-SMA and collagen I was also elevated (Fig. 4G). Furthermore, in both p-HSCs and LX2 cells, tRF-mimic upregulated both α-SMA and FAP, as well as the extracellular matrix protein. Conversely, the expression of these proteins was also downregulated in the tRF-inhibitor group (Fig. 4H).

In order to investigate the mechanism by which exosomal tRF-GluCTC promotes liver metastasis in PDAC, we utilized bioinformatics analysis through the online tool mirmap to identify the binding targets of tRF-GluCTC. We found that the 3’ untranslated region (UTR) of WDR1 mRNA contains a sequence complementary to tRF-GluCTC, spanning eight bases (Fig. 5A).

To confirm the direct interaction between tRF-GluCTC and WDR1 mRNA, we applied an RNA pulldown assay with biotin-labeled WDR1 mRNA probe (Supplementary Fig. 5A). Interestingly, the results of qRT-PCR indicated a higher enrichment of tRF-GluCTC in the WDR1 mRNA probe group compared to the control probe group (Fig. 5B). The dual-luciferase reporter assay was also applied in 293 T cells. The full-length WDR1 mRNA-WT and a mutant version without tRF-GluCTC binding sites were subcloned into GP-miRGLO plasmids. The results indicated the luciferase activity of the WDR1 mRNA-WT group significantly increased in the tRF-GluCTC mimic group compared with the NC group. There was minimal difference between tRF-GluCTC mimics and the NC group in the luciferase activity of the WDR1 mRNA-Mut group (Fig. 5C). These results suggest a direct interaction between tRF-GluCTC and WDR1 mRNA in HSCs.

To further identify the potential pathways involved in HSC activation during WDR1-mediated facilitation of PDAC liver metastasis, we analyzed proteins associated with WDR1. The STRING analysis showed that the proteins associated with WDR1 are cytoskeleton-associated proteins (Fig. 5D). GSEA analysis was also performed in the PADC liver metastasis cohort from the Cancer Genome Atlas (TCGA) based on the transcriptional expression of WDR1. The results indicated the upregulation of ECM and collagen formation in the group with high WDR1 expression, which is known to be associated with PDAC liver metastasis (Fig. 5E).

In vivo, we employed an adeno-associated virus (AAV) to target interference with WDR1, and we observed a significant reduction in WDR1 expression in the liver of mice (Supplementary Fig. 4C, D). Subsequently, we injected 5*10⁶ Panc02 cells into the portal venous and observed the liver metastasis of PDAC after 6 weeks. The proportion of PDAC liver metastasis lesions was significantly smaller in the sh-WDR1 group compared to the NC group, and the size of the liver metastasis was also reduced (Fig. 5F, G). By immunohistochemical staining, we found that the fibrosis of the liver was obviously ameliorated in the sh-WDR1 group, accompanied by decreased expression of extracellular matrix proteins and HSC activation proteins (Fig. 5H).

To further confirm our finding, we transfected tRF-mimic and tRF-inhibitor in p-HSCs and LX2 cells, and found that tRF-GluCTC enhances the expression of WDR1, whereas the inhibitor of tRF-GluCTC represses WDR1 expression (Fig. 5I). Moreover, we added tRF-mimic-exo and tRF-scramble-exo in p-HSCs and LX2 cells. We observed an increase in WDR1 expression upon the addition of tRF-GluCTC (Fig. 5J). Subsequently, we added AsPC1-exo after regulating the expression of the WDR1 gene in LX2 cells. The results demonstrated that WDR1 could upregulate the expression of fibronectin and α-SMA in response to exosome stimulation, while the inhibition of WDR1 decreased the expression of these proteins (Fig. 5K & Supplementary Fig. 4E, F). Moreover, the ability of tRF-GluCTC to upregulate fibronectin and α-SMA was decreased after interfering with WDR1 expression. In contrast, overexpression of WDR1 rescued the inhibitory effect of tRF-inhibitor (Fig. 5L).

It was reported that WDR1 encodes a protein containing 9 WD repeats, which are approximately 30- to 40-amino acid domains containing several conserved residues, mostly including a trp-asp at the C-terminal end. These WD domains are known to be involved in protein-protein interactions [41]. In the current study, aiming to clarify the potential regulatory mechanism of WDR promoting PDAC liver metastasis, the CoIP/MS technology was adopted to identify WDR1-interacting proteins in p-HSCs cells, and a total of 205 proteins were identified (Top 60 genes were shown in Supplementary Table S3). The overall profiles of the biological functions were analyzed for the candidate proteins via GO and pathway enrichment. The analysis revealed that WDR1 is involved in several cancer-associated pathways, including the AMPK signaling pathway, TGF-β signaling pathway, and Hippo signaling pathway (Fig. 6A). Meanwhile, the GEO RNA-seq database (GSE148139) showed increased expression of α-SMA and multiple ECM-related proteins after PDAC-exo treatment of HSCs (Fig. 6B).

Based on these findings, we next sought to validate the interaction of WDR1 and critical signaling pathway proteins using western blot assay repetitively. The western blot experiment showed that YAP was detected in the WDR1 pulled-down samples but not in the control group (Fig. 6C). This was further confirmed by interfering or overexpressing WDR1 in p-HSCs and LX2. The results revealed an increase in YAP expression and a decrease in YAP phosphorylation with WDR1 overexpression (Fig. 6D, E).

Next, we aimed to elucidate the crucial protein domains that are responsible for the interaction between WDR1 and YAP. Molecular docking was employed to investigate interactions between vital active compounds and major targets. The binding affinity lower than –5.0 kcal/ mol indicates that the conformations have good interactions. In this study, molecular docking results showed that the conformations of WDR1 and YAP protein targets contain good binding interactions, and the interactions were also reliable (Fig. 6F). Specifically, the residues SER135 and SER138 in the WDR1 protein were found to interact with the YAP protein. Based on their protein structures, truncated mutation plasmids of WDR1 [wild type (WT), WD-Domain1, WD-Domain2, WD-Domain3, and WD-Domain4] were constructed as indicated (Fig. 6G upper). CoIP experiments were conducted after transfecting these plasmids into 293 T cells. The result revealed that the deletion of the WD-Domain1 in WDR1 significantly disrupted the interaction (Fig. 6G lower). Additionally, the influence of the N-terminals of WDR1 in promoting the expression of a-SMA in p-HSCs cells was also investigated. To this end, exogenous N-terminals, M-terminals and C-tail of WDR1 were transfected into p-HSCs cells. The western blot assay showed that the expression of α-SMA was not increased in cells treated only with WDR1 M-terminals and C-tail, but it was enhanced significantly within the existing N-terminals and full-length WDR1 (Supplementary Fig. 4G).

To assess the role of YAP stability through WDR1 regulation, we manipulated the expression of WDR1 in p-HSCs cells and treated the cells with cycloheximide (CHX), a protein synthesis inhibitor. The half-life analyses revealed that YAP was more stable in WDR1-adequate cells (Fig. 6H, I). Since YAP was previously reported to be subjected to proteasomal degradation activated by upstream kinase, we sought to further confirm that WDR1 regulates the turnover of YAP. Upon treating WDR1-deficient cells with proteasome inhibitor MG132, we restored the protein levels of YAP which were previously decreased by WDR1 depletion. Also, the degradation process of YAP protein was significantly inhibited by adding MG132 in WDR1-adequate cells (Fig. 6J, K).

The formation of a PMN is considered a crucial step in the establishment of a metastatic lesion. Previous studies have reported that exosomes derived from PDAC can promote the infiltration of MDSCs in the liver, leading to the formation of PMNs [42]. To explore the relationship between the tRF and the cytokines secreted by HSCs, we transfected WDR1 siRNA and tRF inhibitor into p-HSCs cells respectively, and analyzed the expression of various cytokines in the supernatants using protein microarrays. As compared to cells transfected with scramble RNA, p-HSCs which down-regulated WDR1 or the tRF-inhibitor decreased the release of IL-1a, and IL-10 (Fig. 7A). We also quantified the concentration of IL-6, IL-10, and IL-1a in the supernatant of p-HSCs after ASPC-1 derived exosome treatment using ELISA. The ASPC-1 exosome in p-HSCs significantly could increase the secretion of IL-6, IL-10, and IL-1a (Fig. 7B). To further investigate the stimulation of tRF-GluCTC in PMN immunosuppressive environment via cytokines, we quantified the concentration of IL-6, IL-10, and IL-1a in the supernatant of p-HSCs after regulation the expression of tRF-GluCTC. As expected, tRF-GluCTC can promote IL-6 and IL-1a release in p-HSCs (Fig. 7C). Furthermore, the expression of IL-6, IL-10, and IL-1a in the supernatant of p-HSCs is also up-regulated by the overexpression of WDR1 (Fig. 7D).

To explore the relationship between the transcriptional co-activator YAP and the cytokines secreted by HSCs, we analyzed the correlations among YAP and cytokines in TCGA data using the GEPIA website The results showed that YAP had positive correlations with IL-1a, IL-6, and IL-11 (Fig. 7E). We also quantified the concentration of IL-6 in the supernatant of p-HSCs after manipulating the expression of YAP using ELISA. Overexpression of YAP in p-HSCs significantly increased the secretion of IL-6 (Fig. 7F).

To investigate the role of WDR1 in promoting infiltration of MDSC within the liver, we harvested liver tissues from AAV-shWDR1 mice. Flow cytometry results revealed that interference with WDR1 expression significantly reduced MDSC infiltration in liver tissues (Fig. 7G), as well as decreased infiltration of macrophages (Fig. 7H). To determine whether CD11b⁺Gr1⁺ MDSCs infiltrate the liver in response to tRF stimulation in PDAC-exo, exosomes containing scramble RNA, tRF-mimic and tRF-inhibitor were injected into the peritoneal cavity of mice (Supplementary Fig. 5B). Flow cytometry analysis revealed that the tRF-GluCTC contained in PDAC-exo significantly increase the MDSC population in mouse liver tissue (Fig. 7I). Moreover, flow cytometry analysis revealed that tRF-GluCTC promoted the infiltration of macrophages in liver tissue (Supplementary Fig. 5C).

Overall, our results highlight the potential involvement of HSCs, specifically the upregulation of WDR1 and the secretion of cytokines like IL1, in the process of liver metastasis in pancreatic cancer. These findings contribute to a better understanding of the complex cellular interactions within the liver microenvironment during metastatic progression.