EVs from pancreatic cancer cells and tumour-derived PSCs promote increased migration of hSCs

To test the hypothesis that EVs secreted by pancreatic cancer cells can trigger the migration of hSCs, EVs were isolated from BxPC3, MiaPaCa2, Panc1, and normal human pancreatic duct epithelial cells (HPDE) by differential ultracentrifugation. Characterisation of pancreatic cancer cells was performed by western blot (Supplementary Fig. 2A, B). These EVs were first characterised by TEM, NTA and western blot. TEM images reveal the vesicles isolated from BxPC3 were spherical and membrane-encapsulated with an artificial cup-shaped morphology, which is one of the typical features of EVs (Fig. 1a). Secondly, NTA shows that Panc1 cells secreted a significantly higher number of particles (14.62 × 109 ± 3.47 particle/mL) compared to HPDE cells (3.10 × 109 ± 1.03 particle/mL), given the same cell seeding density. The mean diameter distribution of particles from 4 cell lines was found to be around 130 nm (Fig. 1b), supporting the size definition of small EVs ranging from 30 nm to 150 nm. On top of that, EVs were subjected to protein composition characterisation based on Minimal information for studies of extracellular vesicles 2018 (MISEV2018) guidelines [18]. For transmembrane proteins, Integrin β1, CD9, CD81, CD63 and EpCAM were detected exclusively in the 100, 000 x g pellet (P100) fractions from HPDE, BxPC3, MiaPaCa2 and Panc1 cells, but not in the supernatant (S100) (Fig. 1c, Supplementary Fig. 3). Similar observation can be noticed for the category of cytosolic proteins related to MVEs, such as Alix and TSG101, as well as the category of secreted proteins, fibronectin. It is noteworthy that the RASG12D protein was clearly evident in the P100 fraction from the Panc1 cell line, which is known to carry KRAS point mutation (G > A) [19]. This indicates the fact that mutated proteins from cancer cell lines can be packaged into EVs. In addition, the absence of calreticulin, a marker of endoplasmic reticulum (ER), and GADPH, which has been proven to be present in the non-vesicular components, in the P100 fractions suggest the purity of the isolated EVs. Overall, the extensive characterisation results demonstrate that the isolated EVs correspond to the current notion of the EV identity.

Fig. 1: EVs from human pancreatic cancer and tPSCs promote increased migration of hSCs.

a Representative negative-stain TEM images of BxPC3 pellet (P100) fractions. Scale bar indicates 200 nm. b NTA analysis of P100 fractions from HPDE, BxPC3, MiaPaCa2 and Panc1. Data represent the mean ± S.E.M. of six biological replicates. c Western blot analysis of EV and non-EV markers of proteins from whole cell lysates (Cells), EV-depleted supernatant (S100) and EV pellets from pancreatic cell lines (P100). Colour bars denote the EV marker categories from MISEV2018 guidelines: Category 1: Transmembrane proteins; Category 2: Cytosolic proteins; Category 4: Non-EV proteins; Category 5: Secreted proteins, e.g., cytokines, growth factors and ECM proteins. d Live-tracking analysis of PKH67-labelled Panc1-derived EV (green fluorescent spots) uptake by hSCs. Representative image of green fluorescent spots accumulated in the cells (red border) at time point 3. Line graphs illustrated the mean total intensity of Alexa Fluor 488 over 16 h. Measurements and analysis were performed by the Operetta CLSTM high-content analysis system. e Transwell migration assay of hSCs exposed to EVs or control for 24 h. Untreated hSCs seeded onto the inserts placed in the wells containing EVs or control. Mean fluorescence unit was measured with ImageJ software. The data are mean ± S.E.M. from three independent experiments. f, g 3D migration assay of hSCs mixed with EVs from pancreatic cancer cell lines (f), BJ, HPaStc, nPSCs or tPSCs (g) embedded in a single Matrigel drop and monitored for 4 days. Images (pseudo-colour) were taken by digital phase contrast of the Operetta system. Quantification was performed by measuring the area covered by cells (µm2). The data are mean ± S.E.M. from three independent experiments. Statistical difference for all data was analysed by two-tailed unpaired Student’s t-test. *p < 0.05, **p < 0.01. HPDE Human pancreatic duct epithelial cells, BJ Fibroblasts derived from normal foreskin, HPaStc human pancreatic stellate cells, nPSC Pancreatic stellate cells derived from normal adjacent tissues, tPSC Pancreatic stellate cells derived from tumour tissue.

After EV characterisation, the next step was to examine if hSCs could uptake EVs from pancreatic cancer cells. Hence, live-tracking analysis of Panc1-derived EVs labelled with PKH67 green fluorescent cell linker added onto hSCs was first conducted. It was observed that the green fluorescence signal increased throughout the treatment time of 15 h (Fig. 1d, Movie S1) and eventually accumulated at the perinuclear region of the cells (indicated as a red border) (Fig. 1d, left). Together with the phase contrast images and immunofluorescence (IF) staining of hSCs after the exposure to PKH67-labelled Panc1-EVs (Supplementary Fig. 4), these results confirm the time-dependent uptake of cancer cell-derived EVs by hSCs. Next, to investigate the functional effects of cancer cell-derived EVs on hSCs, a transwell migration assay was employed when untreated hSCs were placed in the insert of the transwell whereas pancreatic cell-derived EVs were added to the bottom of the wells. It was observed that there was a significant increase in the number of hSCs (illustrated by the increase in mean fluorescence unit) detected at the bottom of the inserts when Panc1- and BxPC3-derived EVs were present compared to control and HPDE-derived EVs, implying the chemoattraction displayed by the EVs (Fig. 1e). However, none of the human pancreatic cancer cell-derived EVs had an impact on the cell viability of hSCs across 3 different time points, i.e., 24 h, 48 h and 72 h (Supplementary Fig. 5). To allow for investigation in a more physiologically relevant condition, a novel 3D migration assay incorporated with EVs was first established in this study, in which pancreatic cell-derived EVs were mixed with hSCs in a single Matrigel drop. Consistent with the results from the 2D transwell migration assay, hSCs treated with Panc1- and BxPC3-derived EVs demonstrated the strongest migratory ability and were significantly different from the HPDE-derived EV-treated hSCs, as evidenced by larger area covered by cells (µm2) trained by the Harmony software associated with the Operetta CLSTM high-content analysis system (Fig. 1f). Furthermore, only the hSCs exposed to tPSC-derived EVs increased the migratory ability significantly, but not the EVs derived from BJ, primary human pancreatic stellate cells (HPaStc) and PSCs derived from adjacent tissues (nPSCs) (Fig. 1g). The increased migration of hSCs exerted by tPSC-derived EVs was similar to that of the Panc1-derived EVs (Fig. 1g). These results show a specific cellular regulation in the migration of hSCs exerted by EVs from pancreatic cancer cell- and tumour-derived PSCs. Altogether, these data provide an indication that the migration of hSCs towards pancreatic cancer cells may be triggered by cancer cell- and PSC-derived EVs from pancreatic cancer and its tumour microenvironment.

To further confirm the role of EVs in SC-increased migration, exogenous addition of EV inhibitors, heparin and EIPA (5-[N-ethyl-N-isopropyl] amiloride), was employed to evaluate if uptake of EVs by hSCs was mediated through surface binding to receptor or/and macropinocytosis, a major route of EV uptake [20]. It has been reported that the heparan sulfate proteoglycans (HPSGs)-dependent uptake route is highly relevant for the internalisation and functional activity of EVs, as HPSGs function as receptors for EVs [21,22,23]. As for EIPA, an inhibitor for macropinocytosis, it inhibits the uptake of EVs through the blocking of membrane ruffling by decreasing cytosolic pH and inactivation of Rac1 and Cdc42 GTPases [24]. Expectedly, following exposure to Panc1-derived EVs pre-treated with different doses of heparin, the EV uptake by hSCs decreased significantly in a dose-dependent manner. Increasing doses of heparin from 20 μg/mL to 100 μg/mL resulted in a decreased proportion of PKH67-stained EVs by 20.19% ± 4.95 to 40.74% ± 4.49, respectively, as compared to Panc1-derived EV uptake without heparin (Fig. 2a). As for EIPA pre-treatment on the hSCs, although the EV uptake decreased significantly in the presence of 50 μM and 100 μM by 34.26% ± 2.98 and 45.68% ± 4.42, respectively (Fig. 2a), the cell viability of hSCs reduced significantly after treated with EIPA for 24 h by 38.9% ± 4.9 and 64.5% ± 5.2, respectively (Fig. 2b, right), indicating that the EIPA treatment was too toxic for the hSCs. The cell viability of hSCs remained intact in the presence of heparin, regardless of the concentration (Fig. 2b, left). It was also observed that combined heparin (20 μg/mL) and EIPA (20 μM) together did not significantly reduce the uptake of Panc1-EVs (reduction of 27.47%) compared to the single treatment (reduction of 20.19% and 18.05% by heparin 20 μg/mL and EIPA 20 μM, respectively), suggesting that targeting at two different uptake routes of Panc1-EVs in the case of hSCs does not lead to a synergistic effect. Using the concentration of 20 μg/mL of heparin to perform a 3D migration assay, the results show that heparin significantly reduced the area covered by hSCs triggered by Panc1-derived EVs, indicating that the migratory ability of hSCs was inhibited by heparin. These results suggest that Panc1-derived EV uptake by hSCs was partially regulated by cell surface HPSGs and these EVs were, in part, responsible for the migratory effect on hSCs.

Fig. 2: Impact of EV inhibitors on EV-mediated increased migration of hSCs.

a hSCs were treated with PKH67-labelled Panc1-derived EVs pre-incubated with different concentrations of heparin or EIPA for 1 h or throughout the experiments (24 h). Images were taken by the Operetta CLSTM high-content analysis system. Quantification was performed by measuring the intensity of Alexa Fluor 488. The error bars depict mean ± S.E.M. from three independent experiments. b Cell viability of hSCs was measured by PrestoBlue reagent after exposure to Panc1-EVs, heparin or EIPA for 24 h. Data were obtained from three independent experiments and shown as mean ± S.E.M. c 3D migration assay of hSCs mixed with Panc1-EVs pre-incubated with heparin for 1 h before embedding in a Matrigel drop and monitored for 4 days. Images (pseudo-colour) were taken by the digital phase contrast function of the Operetta system. Quantification was performed by measuring the area covered by cells (µm2). Statistical difference of all data was analysed by two-tailed unpaired Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

Murine neuroinvasive cancer cell-derived EVs influence SC phenotype by promoting increased migration and are enriched with p75NTR

To investigate further the underlying mechanisms of increased SC migration, non-neuroinvasive murine cancer cells, KPC, and neuroinvasive murine cancer cells, TPAC that were established previously [14, 25] were employed in this study to perform horizontal 3D coculture assay in the presence or absence of heparin (Supplementary Fig. 6A). As described in the current literature, p75NTR/TrkA/NGF signalling pathway plays an important role in NI [7, 15, 16]. Hence, the lysates from mSCs were probed for p75NTR, TrkA and NGF antibodies. p75NTR expression levels were strongly increased in the mSCs facing TPAC cells compared to those confronted with KPC cells (Supplementary Fig. 6B), which is a profile of dedifferentiated SCs during nerve injury or repair. On top of that, the increased p75NTR levels in the mSCs facing TPAC cells decreased significantly in the presence of heparin (Supplementary Fig. 6B), implying that the increased p75NTR levels may be EV-mediated. Apart from that, EVs collected from the coculture assay show that p75NTR levels in the EVs increased significantly when mSCs confronted with TPAC (TPAC-mSC) compared to TPAC alone and mSCs confronted with KPC cells (KPC-mSC) using Alix as an EV marker for normalisation (Supplementary Fig. 6C), suggesting that the interaction between mSCs and neuroinvasive TPAC cells promoted the enrichment of p75NTR expression in the EVs. To study the effect of EVs from murine cancer cell lines on mSCs, EVs were first isolated by differential ultracentrifugation, characterised by NTA and western blot. As illustrated by the size distribution graph, KPC cells (1.10 × 1011 ± 1.35 × 1010 particle/mL) secreted significantly higher particle concentration compared to TPAC cells (6.78 × 1010 ± 7.23 × 109 particle/mL) while there was no significant difference in terms of the particle size (KPC: 146.2 ± 1.46 nm; TPAC: 145.5 ± 0.98 nm) (Fig. 3a). Similar to the characterisation of EVs from human pancreatic cancer cell lines, Alix, TSG101 and CD9 were detected whereas calreticulin and GAPDH were undetected in the P100 fractions from KPC and TPAC cells after ultracentrifugation (Fig. 3b, Supplementary Fig. 7). Next, live-tracking analysis showed that mSCs took up significantly higher amount of TPAC-EVs than KPC-EVs within the same time frame (endpoint mean total intensity = KPC-EVs: 1.09 × 107 ± 8.76 × 105; TPAC-EVs: 1.73 × 107 ± 9.20 × 105; p-value = 0.0005) (Fig. 3c), suggesting the differential EV uptake efficiency in the mSCs. For the functional effect, TPAC-EVs increased the migration of mSCs significantly compared to KPC-EVs, as demonstrated by the 3D migration assay (Fig. 3d). On top of that, the increased migration of mSCs after exposure to TPAC-EVs was reduced significantly in the presence of heparin (20 μg/mL), suggesting the possibility of EV-mediated migration of the mSCs (Fig. 3d).

Fig. 3: EVs from neuroinvasive cancer cells promote increased migration of mSCs and enriched with p75NTR.

a NTA analysis of P100 fractions from KPC and TPAC cells. Data represent the mean ± S.E.M. of ten replicates. b Western blot analysis of EV and non-EV markers of proteins from whole cell lysates (Cells), EV-depleted supernatant (S100) and EV pellets from pancreatic cell lines (P100). c Live-tracking analysis of PKH67-labelled KPC- and TPAC-EV (green fluorescent spots) uptake by hSCs. Representative image of green fluorescent spots accumulated in the cells (red border) at time points 0 and 19. Line graphs illustrated the mean total intensity of Alexa Fluor 488 over 19 h. Measurements and analysis were performed by the Operetta CLSTM high-content analysis system. d 3D migration assay of mSCs mixed with EVs from KPC or TPAC cells embedded in a single Matrigel drop and monitored for 4 days. Images (pseudo-colour) were taken by digital phase contrast of the Operetta system. Quantification was performed by measuring the area covered by cells (µm2). The data are mean ± S.E.M. from three independent experiments. e qRT-PCR analysis of the targets in KPC and TPAC cells as well as their respective EVs. The heatmap (left) depicting neurotrophin-related targets in KPC and TPAC cells was generated based on 2-ΔΔCT relative to KPC cells. The heatmap (right) showing the raw CT values of the targets in KPC- and TPAC-derived EVs. Log of the raw values was used for the generation of a heatmap using R software version 4.1.2. Data were obtained based on four independent replicates. f Western blot analysis of p75NTR and NGF expression levels in whole cell lysate (Cells), EV-depleted supernatant (Sup) and EVs (EV) from KPC and TPAC cells. Data were obtained based on four independent replicates and shown as mean ± S.E.M. Statistical difference of all data here was analysed by a two-tailed unpaired Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

To explore the EV cargoes from KPC and TPAC cells, several neurotrophic factors including p75NTR and other potential candidates related to NI were examined by qRT-PCR and western blot. As illustrated by the heatmap, the expression levels of Npy and Ngfr in the TPAC cells were significantly increased compared to the KPC cells whereas Ngf is significantly expressed in the KPC cells compared to TPAC cells (Fig. 3e, left). As for the EV levels, since there is currently no universal loading control used for the normalisation of the CT values, the CT values were used for the heatmap illustration. The mean CT values of all the genes in KPC-EVs were above 30 (Npy = 33.35 ± 0.76; Nft5 = 35.26 ± 0.39; Ngfr = 35.90 ± 1.84; Ngf = 31.09 ± 0.23; Bdnf = undetermined, 40 was used for the heatmap generation; Gfra1 = 33.87 ± 0.18), suggesting that they were lowly expressed in the KPC-EVs. On the other hand, the mean CT values of these genes, especially Npy, Ngfr and Ngf were significantly lower compared to KPC-EVs (Npy = 28.15 ± 0.31; Nft5 = 31.15 ± 0.26; Ngfr = 27.32 ± 0.18; Ngf = 29.65 ± 0.18; Bdnf = 31.31 ± 0.64; Gfra1 = 31.65 ± 0.09), suggesting the enrichment of these genes in the TPAC-derived EVs (Fig. 3e, right). Interestingly, the protein levels of p75NTR were 8-fold enriched in the TPAC-derived EVs compared to the KPC-derived EVs, but not its ligand, NGF (Fig. 3f). This is in agreement with another study showing the absence of NGF in the EVs [13], implying the selective packaging of EVs by the cells.

Tumour tissue- and plasma-derived EVs from patients with Pn1 promote increased SC migration

To validate the findings obtained from the murine cancer cell model, human pancreatic tissues were obtained to isolate the EVs from the freshly obtained tissues (direct), adjacent tissues (explant model) and tumour tissues (explant model). As shown by the HE staining, the tissues remained intact after 24 h of incubation (Supplementary Fig. 8A). Although the particle concentration from the adjacent tissues (4.58 × 1011 ± 1.05 × 1011 particle/mL) and tumour tissues (5.74 × 1011 ± 1.01 × 1011 particle/mL) did not differ significantly, the mean diameters distribution of the particles from the tumour tissues (157.3 ± 2.58 nm) were significantly larger than those from the adjacent tissues (144.3 ± 2.30 nm) (Supplementary Fig. 8B). TEM images and western blot analysis of EV markers confirmed the identity of EVs, in addition to showing the different subtypes of EVs from different tissues (Supplementary Fig. 8C, D). For the functional effect of these patient tissue-derived EVs, hSCs were treated with the tissue-derived EVs in the 3D migration assay. Expectedly, hSCs increased migration significantly after exposure to the tumour tissue- and tumour tissue direct-EVs compared to untreated and that of the adjacent tissue-EVs (Fig. 4a). In particular, when the patients were categorised based on Pn status in the pathology reports, it could be observed that hSCs increased migration significantly after exposure to the tumour tissue-EVs from the PDAC patients diagnosed with Pn1 compared to those with Pn0 (Fig. 4b). To increase the depth of NI analysis, further specification of NI, i.e tumour focal or circular appear in nerve sheet and tumour intraneural, was analysed with HE and IHC staining of PDAC tissue sections (Supplementary Fig. 9). The tissues sections displaying tumour focal NI secreted EVs that increased the migration of hSCs significantly compared to tumour tissue-derived EVs from patients tissues with Pn0 (Fig. 4c). To study the effect of tumour tissue-EVs on SC phenotype, qRT-PCR was performed with the RNAs from the hSCs treated with tumour tissue-EVs in the 3D migration assay. The results show that the increased hSC migration after exposure to the tumour tissue-EVs from those patients diagnosed with Pn1 also corresponded to a higher expression level of NGFR in the hSCs (Fig. 4d), suggesting the potential effect of neuroinvasive EVs on the dedifferentiation of hSCs, thereby increasing their migration ability. Despite a limited number of patients, it was shown that tumour tissue EVs from patients with Pn1 seem to contain higher protein expression levels of p75NTR than those EVs from Pn0 patients (Supplementary Fig. 10), suggesting that p75NTR may be enriched in the neuroinvasive tissue-derived EVs that could influence the SC behaviour. This observation warrants further validation with more patient samples. To verify the possible involvement of p75NTR in the increased migration of hSCs after exposure to tissue-derived EVs, a p75NTR antagonist, THX-B was used in the 3D migration assay. The results show that the increased migration of hSCs in the presence of tumour tissue-EVs (p-value = 0.003) reduced significantly with the addition of 15 μM THX-B (p-value = 0.023, Fig. 4e), a dose that was not toxic to the hSCs (Supplementary Fig. 11). This result support the hypothesis that tumour tissue-EVs may contain p75NTR that mediates the increased migration of hSCs. To understand the effect of EVs from tumour, peritumoral microenvironment and systemic environment on SC behaviour, hSCs were treated with EVs from PDAC patients with matched adjacent, tumour tissues and plasma in the 3D migration assay (Fig. 4f). Although tumour tissue-EVs increased the migration of hSCs significantly more than plasma-EVs, it should be noted that plasma-EVs from patients with Pn1 did promote stronger migration of hSCs relative to those from Pn0 patients (Fig. 4f). This suggests that PDAC with NI may alter the systemic environment such as distant organs and immune cells to produce EVs that have an impact on SC migration. The characteristics of PDAC patients contributing tissues- and plasma-derived EVs for EV characterisation and 3D migration assay were summarised in Supplementary Table 1.

Fig. 4: Tumour tissue- and plasma-derived EVs from patients with Pn1 enhance the migratory capacity of hSCs.

a 3D migration assay of hSCs mixed with EVs from adjacent tissues, tumour tissues and tumour tissue direct embedded in a single Matrigel drop and monitored for 4 days. Images (pseudo-colour) were taken by digital phase contrast of the Operetta CLSTM high-content analysis system. Quantification was performed by measuring the area covered by cells (µm2). The error bars depict mean ± S.E.M. Patient number for adjacent normal tissue-EVs = 10; Tumour tissue-EVs = 20; Tumour tissue direct-EVs = 13. b Analysis of the results from 3D migration assay in (a) hSCs treated with tumour tissue-EVs or tumour tissue direct-EVs from PDAC patients without NI (Pn0) and with NI (Pn1). The error bars depict mean ± S.E.M. n for Pn0 = 5; Pn1 explant = 16; Pn1 direct = 11. c Analysis of the results from 3D migration assay in (a) hSCs treated with tumour tissue-EVs from PDAC patients without NI (Pn0) and with NI (Pn1) categorised by focal NI and intraneural NI. The error bars depict mean ± S.E.M. n for Pn0 = 5; Pn1 explant = 16; Pn1 direct = 11. d qRT-PCR analysis of NGFR expression levels in the hSCs treated with tumour tissue-derived EVs harvested from 3D migration assay shown in (a). n for Pn0 = 4; Pn1 = 14. e 3D migration assay of hSCs mixed with tumour tissues-EVs embedded in a single Matrigel drop cultured in medium containing no THX-B, or 5 μM, 10 μM, 15 μM THX-B for 4 days. Images (pseudo-colour) were taken by digital phase contrast of the Operetta CLSTM high-content analysis system. Quantification was performed by measuring the area covered by cells (µm2). The error bars depict mean ± S.E.M. Patient number for tumour tissue-EVs = 6. f Representative images of 3D migration assay of hSCs treated with plasma-, adjacent tissue-, and tumour tissue-EVs from matched patients. Statistical difference between hSCs treated with plasma-EVs and tumour tissues-EVs was analysed by Wilcoxon matched-pairs signed rank test. For hSCs treated with plasma-derived EVs from patients with Pn0 (n = 8) and Pn1 (n = 15), statistical differences were analysed by two-tailed unpaired Student’s t-test after passing the normality test by the Shapiro–Wilk method. Statistical differences for (a) and (c) were analysed by the Mann–Whitney U-test. Statistical differences for (b), (d) and (e) were analysed by two-tailed unpaired Student’s t-test after passing the normality test by the Shapiro–Wilk method. *p < 0.05, **p < 0.01, ***p < 0.001.

Clinical relevance of p75NTR for PDAC patients with NI

While it is accepted that NI is associated with a reduced OS and recurrence-free survival (RFS) for pancreatic cancer patients, there are some publications with opposing results [26, 27]. To confirm the impact of NI on PDAC patient OS and RFS, Kaplan–Meier curve analyses were performed with PDAC patients from UHD based on their Pn status. The patients with Pn1 were significantly correlated with reduced OS and RFS compared to patients with Pn0 (log-rank p-value for OS = 0.0038; RFS = 0.014) (Fig. 5a, b). The clinical data of PDAC patients contributing to the OS and RFS analyses were presented in Supplementary Table 2. Since EVs hold great potential as liquid biopsy, it was of high interest to evaluate the clinical relevance of EV p75NTR as a biomarker for PDAC patients with NI. As plasma remains the most accessible source for liquid biopsies and is depleted of platelet-derived EVs, western blot analysis of p75NTR expression in the plasma-derived EVs from a total of 165 PDAC patients (Supplementary Table 3) was performed to evaluate the potential of plasma-derived EV p75NTR for diagnosis and prognosis of NI in patients with PDAC. Since syntenin was demonstrated to be present in plasma-derived EVs ubiquitously [28], the expression levels of p75NTR were normalised to syntenin as an EV marker internal control. p75NTR was significantly higher in the plasma-derived EVs from patients with Pn1 (n = 140; mean fold change = 3.186 ± 0.218) compared to Pn0 (n = 25; mean fold change = 2.136 ± 0.291) (p-value = 0.004) (Fig. 5d and Supplementary Fig. 12). Furthermore, Kaplan–Meier analysis shows that patients with high expression of plasma-derived EV p75NTR displayed a reduced OS (log-rank p-value = 0.030) compared to those with low p75NTR expression (Fig. 5e). While the commonly used diagnostic and prognostic biomarker for PDAC, carbohydrate antigen 19-9 (CA19-9) had no significant prognostic ability in the current cohort (Supplementary Fig. 13), the combination of EV p75NTR and CA19-9 demonstrated a superior prognostic performance (log-rank p-value = 0.011) than the p75NTR or CA19-9 alone (Fig. 5f), suggesting that the EV p75NTR could complement to the current prognosis of PDAC patients. However, both EV p75NTR and CA19-9 did not show significant prognostic value for RFS in the current study (Supplementary Figs. 13, 14). Based on the univariate and multivariate Cox regression analyses, the EV p75NTR was demonstrated to be an independent prognostic factor (log-rank p-value = 0.004) after considering age, gender, tumour stage, grading, resection margin, neoadjuvant therapy, lymphatic invasion, venous invasion and Pn status (Table 1), suggesting that plasma-derived EV p75NTR may be used for risk stratification to predict the prognosis of PDAC patients.

Fig. 5: Clinical relevance of p75NTR for PDAC patients with NI.

a, b Kaplan–Meier analyses of OS (a) and RFS (b) of PDAC patients from University Hospital Dresden based on NI status (Pn0: no NI; Pn1: with NI). c Representative samples of western blot analysis of p75NTR and EV markers in plasma-derived EVs from PDAC patients (n for Pn0 = 25; Pn1 = 140). Quantification of p75NTR was performed based on the normalisation of syntenin. d The prognostic correlation of plasma EV p75NTR from PDAC patients was assessed by Kaplan–Meier analyses for OS. Median cut-off was used to divide the patients into p75NTR high expression and low expression. e The prognostic performance of the combination of EV p75NTR and CA19-9 for PDAC patients. Statistical differences for (a), (b), (d) and (e) were analysed by Log-rank test with p < 0.05 considered as statistically significant. Statistical differences for (c) were analysed by the Mann–Whitney U-test. *p < 0.05, ***p < 0.001.

Table 1 Univariate and multivariate Cox regression analysis of the EV p75NTR.