Isolation of pancreatic cell populations with stemness features

To identify cell subpopulations within the mouse pancreas with stem-like potential, various stem cell-related cell surface markers (ie, EpCAM, Sca-1, cMet, CD34, CD49f and CD133) were tested,14–17 and only the combination of EpCAM and Sca-1, after stroma depletion (CD45^- and CD31^- cells), allowed for the separation of four distinct subpopulations (figure 1A, online supplemental figure S1A), which were subjected to spheroid and organoid formation assays. EpCAM⁺Sca-1⁺ cells displayed the highest sphere/organoid formation capacity (figure 1B,C), suggesting stem cell potential. We also detected by immunofluorescence EpCAM⁺Sca-1⁺ cells in normal pancreata sections (figure 1D) and also found that these cells expressed genes related to tissue stemness (online supplemental figure S1B). To test multilineage potential, we analysed lineage markers in spheroids and found amylase (acinar lineage), insulin (islet lineage), Hes1 and cytokeratin 19 (CK19) (both ductal lineage) single-positive cells, indicating that EpCAM⁺Sca-1⁺ cells can give rise to more differentiated cells (figure 1E). We also found that most EpCAM⁺Sca-1⁺ cells were also CD133⁺ (online supplemental figure S1C), and in functional assays, EpCAM⁺Sca-1⁺CD133⁺ cells (ie, triple positive) had the highest spheroid and organoid forming capacity (figure 1F,G). Remarkably, spheroids could be maintained across several generations, indicating extended self-renewal capacity (figure 1H).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Isolation of pancreatic populations with stemness features. (A) Top panel: representative flow cytometry plot showing the gating strategy for isolation of pancreatic cell populations after lineage depletion from an 8-week-old C57Bl/6J mouse. Cells were stained with anti-CD45 and anti-CD31 (lineage cocktail). Bottom panel: lineage depleted cells were stained with anti-EpCAM and anti-Sca-1 resulting into four populations. (B–C) Quantification of organoid forming capacity in Matrigel (B) or spheroid forming efficiency (C) for the different pancreatic cell subsets identified in (A). Shown are mean organoid numbers/10 000 cells±STDEV or mean spheroid numbers/10 000 cells±STDEV (n=4, p values determined by one-way analysis of variance (ANOVA), with Dunnett’s test). (D) Representative confocal images of normal murine pancreatic tissue showing rare cells positive for EpCAM (red), Sca-1 (green) and DAPI (nuclear marker, blue). Arrows indicate EpCAM⁺Sca-1⁺ populations. Scale=50 µm. (E) Representative confocal images of EpCAM⁺Sca-1⁺ cell-derived spheroids after 10 days of culture. Spheroids were stained with antibodies against pancreatic lineage markers amylase, Hes1, insulin and cytokeratin 19 (CK19) (all in green) and DAPI (nuclear marker, blue). (F–G) Mean spheroid numbers/10 000 cells±STDEV (F) or mean organoids/10 000 cells±STDEV (G) and representative bright field images (bottom) of spheroids or organoids (in Matrigel) generated from sorted EpCAM⁺Sca-1⁺CD133⁺ and EpCAM⁺Sca-1⁺CD133^– cells after 10 days of culture (n=4, p values as determined by unpaired t-test). (H) Mean spheroid numbers/10 000 cells±STDEV of EpCAM⁺Sca-1⁺CD133^- and EpCAM⁺Sca-1⁺CD133⁺ cell-derived spheroids across serial passages (top) (n=4, p values determined by unpaired t-test). Representative bright field images (bottom) show EpCAM⁺Sca-1⁺CD133⁺ cell-derived spheroids at each passage. DAPI, 4',6-diamidino-2-phenylindole; EpCAM, epithelial adhesion cell adhesion molecule; Sca-1, stem cell antigen 1; STDEV, standard deviation.

EpCAM⁺Sca-1⁺ cells expand during carcinogenesis and possess CSC features

In the KPC mouse model of PDAC,18–20 all epithelial cell descendants constitutively express active Kras and p53 mutant genes. However, the limited number of malignant lesions suggests that not all cells are equally tumourigenic. We studied if this process would affect the frequency of the above-identified populations. Notably at 8–9 weeks, when acinar-ductal metaplasia (ADM) can be detected, there was a significant increase in EpCAM⁺Sca-1⁺ cells in the pancreata of KPC mice compared with KP controls, while other populations remained mostly unchanged (figure 2A, online supplemental figure S1D, E). At later time points (16 weeks), when PDAC is detectable, expansion of EpCAM⁺Sca-1⁺ cells was more pronounced (figure 2B), and EpCAM⁺Sca-1⁺ cells could be partitioned using CD133 expression, with higher levels observed in KPC vs KP pancreata (figure 2C), translating into higher spheroid forming capacity (figure 2D), higher tumour propagating capacity in an extreme limiting dilution assay (ELDA) (figure 2E) and sustained long-term (ie, serial transplantation) tumorigenicity (figure 2F). Importantly, other cell populations did not give rise to tumours (online supplemental figure S1F), supporting that only triple-positive cells have tumour initiating capacity. Of note, this population could also be detected in ADM and in PDAC lesions from KPC pancreata (figure 2G). Finally, using a lineage tracing single-cell RNA sequencing (scRNA-seq) dataset based on a Ptf1a-CreER; LSL-KRas-G12D; LSL-tdTomato PDAC mouse model,21 we found that epithelial cells expressing Epcam, Ly6a/Sca-1 and Prom1/CD133 expand from early metaplastic states to full tumour cells (figure 2H,I), supporting the existence of these cells early on and throughout tumour development. Thus, triple-positive cells possess all the essential characteristics to be considered bona fide CSCs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

EpCAM⁺Sca-1⁺CD133⁺ cells possess CSC properties and expand during tumour progression. (A) Left panel: representative flow cytometry plot showing Lin^- pancreatic cells stained with EpCAM and Sca-1 from 8-week-old control LSL-Kras^G12D/+; LSL-Trp53^R172H/+ mice (KP) versus LSL-Kras^G12D/+; LSL-Trp53^R172H/+; Pdx-1-Cre mice (KPC). Right panel: histogram plot showing the mean±SEM percentage of EpCAM⁺Sca-1⁺ cells in control KP and KPC animals (n=6, p values determined by unpaired t-test). (B) Left panel: representative flow cytometry plot displaying EpCAM and Sca-1 expression in Lin^- cells from a KPC mouse tumour at 16 weeks and corresponding pancreata from a control KP mouse. Right panel: histogram plot showing the mean±SEM percentage of EpCAM⁺Sca-1⁺ cells in KP and KPC groups (n=6, p values determined by unpaired t-test). (C) Representative flow cytometry plot showing the expression of CD133 within the EpCAM⁺Sca-1⁺ gate in a KP and KPC mouse at 16 weeks. (D) Mean±SEM of the number of spheroids/10 000 cells generated by the indicated populations after 10 days in sphere culture conditions. Cells were sorted from the pancreata of 16-week-old KPC mice (n=4, p values determined by one-way analysis of variance (ANOVA), with Dunnett’s test). (E) Panel detailing the tumorigenic potential (number of tumours formed/number of injections) of the indicated number of EpCAM⁺Sca-1⁺ cells injected in the flanks of athymic nude mice. Cells were sorted from KPC tumours and segregated by CD133 expression. Predicted frequency (freq.) of CSCs as a function of the evaluated dilutions are shown (p values determined by χ² analysis obtained using ELDA software). (F) Panel detailing secondary engraftment potential of EpCAM⁺Sca-1⁺CD133⁺ and EpCAM⁺Sca-1⁺CD133^- tumour cells isolated from parental tumours generated in (E). Predicted CSC frequencies (freq.) as a function of the evaluated dilutions are shown (no. of injections >5, p values determined by χ² analysis obtained by ELDA software). (G) Representative confocal images showing triple staining with antibodies against EpCAM (red), Sca-1 (green), CD133 (grey) and DAPI (nuclear marker, blue). Top panel: pancreata from a KPC mouse at 8 weeks where acinar-ductal metaplasia (ADM) can be observed. Bottom panel: tumour (PDAC) from a 16-week-old KPC mouse. Arrowheads indicate the triple-positive population. Scale=80 µm. (H) Left panel: UMAP representing the clusters from the lineage tracing single-cell RNA sequencing (scRNA-seq) dataset from Schlesinger et al.21 Middle panel: the metaplastic cell cluster is amplified with Epcam⁺Ly6a⁺Prom1⁺ cells highlighted in blue, and the percentage of triple-positive cells is indicated. Right panel: expression of the markers in the metaplastic population. (I) Left panels: UMAP corresponding to the different states of early tumorigenesis included in this dataset. Middle panels: amplification of the different states in early tumour evolution. Triple-positive cells are highlighted in blue. Right panels: violin plots showing the levels of expression of the indicated genes in the populations displayed. (J) Histogram representing the percentage of early metaplastic, late metaplastic and tumour cells expressing the three markers from the total in each population. CSC, cancer stem cell; ELDA, extreme limiting dilution assay; EpCAM, epithelial adhesion cell adhesion molecule; Ly6a, lymphocyte antigen 6A; PDAC, pancreatic ductal adenocarcinoma; Prom1, prominin-1; Sca-1, stem cell antigen 1; STDEV, standard deviation; UMAP, Uniform Manifold Approximation and Projection.

Triple-positive CSCs are enriched in stemness and immune evasion signatures

Transcriptomic analysis of triple-positive CSCs versus all other tumour cells was performed, and gene set enrichment analysis (GSEA) revealed a significant enrichment in stem cell signatures (figure 3A) and other pathways associated with inflammation and leucocyte migration (online supplemental figure S2A). Importantly, we observed that triple-positive CSCs were also enriched in immune modulation/evasion, innate immune response and tumour invasion signatures (figure 3B). While already described PDAC progression and/or stemness-related genes (ie, Mmp7, Cxcl5, Msln, Dckl1, Tspan8 and Lgr5) were among the most significantly upregulated in CSCs,22–26 we also identified, for the first time, Pglyrp1 as significantly upregulated (figure 3C). In mammals, PGLYRP1 is mainly produced by neutrophils in antibacterial granules, also regulating innate immunity,27 28 or by epithelial cells to modulate the microbiome.29–31 Recently, PGLYRP1 was associated with immune evasive mechanisms in T cells from certain tumours (ie, melanoma), positioning it as a potential target for cancer treatment.32 However, to date, PGLYRP1 has not been studied in PDAC or in PDAC CSCs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Transcriptomic analysis of triple-positive CSCs and validation of PGLYRP1 as a CSC marker. (A) Gene sets enriched in the transcriptional profile of the EpCAM⁺Sca-1⁺CD133⁺ population (CSCs) versus the rest of populations (non-CSC), showing an enrichment in stem cell signatures. Shown are the -log(FDR q-val) values for each pathway using the indicated published gene signatures, nominal p<0.05, FDR <15% (n=3 biological replicates). (B) GSEA plots showing enrichment of indicated signatures in the EpCAM⁺Sca-1⁺CD133⁺ population (CSC) versus all other tumour cells (non-CSC). (C) Volcano plot showing the significantly enriched genes (in red) in EpCAM⁺Sca-1⁺CD133⁺ CSCs versus all other tumour cells (n=3 biological replicates). (D) Quantification of PGLYRP1⁺ cells in the indicated populations in KPC PDAC tumours detected by flow cytometry, represented as percentage±SEM (n=7, p values determined by one-way ANOVA with Dunnett’s test). (E) Top panel: representative western blot of PGLYRP1 levels from protein extracts derived from adherent (ADH) and spheroid (SPH) cultures. GAPDH was used as an internal loading control. Bottom panel: mean fold change of PGLYRP1/GAPDH densitometric ratios±STDEV, with ADH set as 1.0 (n=3, p values determined by unpaired t-test). (F) Left panel: representative flow cytometry plots showing the percentage of PGLYRP1-FITC expressing cells in both cell lines cultured as ADH and SPH. Right panel: quantification represented as the mean fold change±SEM, with ADH set as 1.0 (n=6, p values determined by unpaired t-test). (G) Left panel: representative flow cytometry plots showing the expansion of the EpCAM⁺Sca-1⁺CD133⁺ CSC and triple-positive PGLYRP1⁺ compartments in KPC cell lines cultured as ADH and SPH. Right panels: frequency of triple-positive and PGLYRP1⁺ triple-positive cells in the indicated cell cultures. Data are represented as the mean fold change in the percentage of the indicated populations±SEM, with ADH set as 1.0 (n=6, p values determined by unpaired t-test). (H) Representative confocal images showing CSC marker expressing cells. EpCAM or Sca-1 (red), CD133 (pink), PGLYRP1 (green) and DAPI (nuclear marker, blue). Arrowheads indicate the CSC population. ANOVA, analysis of variance; CSCs, cancer stem cells; EpCAM, epithelial adhesion cell adhesion molecule; FDR, false discovery rate; FITC, fluorescein isothiocyanate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GSEA, gene set enrichment analysis; PDAC, pancreatic ductal adenocarcinoma; PGLYRP1, peptidoglycan recognition protein 1; Sca-1, stem cell antigen 1; STDEV, standard deviation.

Pglyrp1 expression was first validated in primary KPC tumours (figure 3D, online supplemental figure S2B), confirming its expression predominantly in the triple-positive population. Next, we cultured primary KPC tumour-derived cell lines in adherence or as spheroids, the latter being a common methodology to enrich for CSCs,33 and performed transcriptomic analyses. Again, we observed the previously mentioned signatures and genes (online supplemental figure S2C, D), as well as increased total, secreted and surface-linked PGLYRP1 expression in spheroids from two KPC-derived (ID11 and ID95) cell lines (figure 3E,F, online supplemental figure S2E). Moreover, we found Pglyrp1 expression in metaplastic cells from the previously mentioned scRNA-seq dataset (online supplemental figure S2F). We also observed enrichment in both triple-positive and PGLYRP1⁺ triple-positive subpopulations in spheroids (figure 3G). In agreement with these data, PGLYRP1 colocalised with EpCAM⁺Sca-1⁺ and EpCAM⁺CD133⁺ cells in KPC tumours (figure 3H).

To evaluate the expression of PGLYRP1 during tumour evolution, we combined immunofluorescence analyses with interrogation of single-cell RNA-seq data. In histological sections of the healthy pancreas, we could not detect PGLYRP1 expression in epithelial cells (figure 4A), although in pancreatitis, some acini presented PGLYRP1 (figure 4B). In KPC transformed pancreas; however, PGLYRP1 was detectable as early as 10 weeks and throughout tumour evolution (figure 4C). PGLYRP1 was also expressed in metastatic lung lesions, colocalising with cytokeratins and EpCAM (figure 4D). While PGLYRP1 overexpression is associated with PDAC tumour cells, inflammation can slightly increase its expression in acini (figure 4E). In transcriptomic data, Pglyrp1 expression increased approximately four times from early metaplastic to late metaplastic cells (figure 4F). Interestingly, nearly half of the triple-positive CSCs expressed Pglyrp1 in early metaplastic states, increasing to almost 100% in later stages (figure 4G). Taken together, these results suggest that the expression of PGLYRP1 by CSCs and other tumour cells could be relevant for tumour initiation and progression.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

PGLYRP1 expression during tumour evolution. Representative H&E-stained images and confocal microscopy images of PanCK (pink), PGLYRP1 (green), EpCAM (red) and DAPI (nuclear marker, blue) in (A) a mouse healthy pancreata, (B) pancreatic tissue following cerulein-induced pancreatitis, (C) in the pancreas of a 10-week, 24-week and 35-week-old KPC mice and (D) in a lung metastasis (PGLYRP1 colocalises with EpCAM⁺PanCK⁺ lesions; PGLYRP1 staining can also be found in surrounding immune cells). For all images, brightness and contrast were adjusted with ImageJ. Scale bar=60 µm (H&E images) and scale bar=100 µm (fluorescence images). (E) Quantification of PGLYRP1 expression, comparing the mean grey value of the PGLYRP1 staining signal, shown as the mean fold change±STDEV, with healthy pancreas set as 1.0 (n=3, p values determined by one-way ANOVA with post hoc Tukey test). (F) Left panels: UMAPs showing Pglyrp1 expression in the Schlesinger et al dataset21 in early metaplastic (EM), late metaplastic (LM) and tumour cell (T) states. Right panel: histogram representing the percentage of Pglyrp1 expressing cells for each state. (G) Left panels: UMAPs showing Pglyrp1 expression in the triple-positive population in the Schlesinger et al dataset21 in early metaplastic (EM), late metaplastic (LM) and tumour cell (T) states. Right panel: histogram representing the percentage of Pglyrp1 expressing cells in triple-positive cell population for each state. ANOVA, analysis of variance; CSCs, cancer stem cells; DAPI, 4',6-diamidino-2-phenylindole; EpCAM, epithelial adhesion cell adhesion molecule; PanCK, pan-cytokeratin; PGLYRP1, peptidoglycan recognition protein 1; STDEV, standard deviation.

PGLYRP1 is not relevant for stemness but essential for tumour formation

To determine if PGLYRP1 is involved in stemness, PGLYRP1 KO cells were generated in KPC ID11 and ID95 cells (figure 5A,B). To maintain the heterogeneity of the primary culture and the CSC compartment, CRISPR KO validation was performed on the cell pool instead of from a single-cell-derived culture. A consistent reduction in PGLYRP1 expression was confirmed. We next performed an ELDA of wild-type (WT) and KO cells in C57Bl/6J immunocompetent mice. In addition, PGLYRP1 overexpressing (OE) cells were generated (online supplemental figure S3A) and included. While no relevant differences in tumour numbers or weights between WT and OE cells were observed (figure 5C), OE cell-derived tumours grew significantly faster (figure 5D, online supplemental figure S3B, C), and many presented a less differentiated appearance in histological analyses (online supplemental figure S3D). Strikingly, PGLYRP1 KO significantly impaired tumour take (figure 5E). For ID11, no tumours formed when 10³ KO cells were injected, and for ID95, KO cells were unable to form tumours at all dilutions (figure 5D,E, online supplemental figure S3B–D). To discard differences in PGLYRP1 KO cell proliferation, we quantified cell numbers in vitro over 3 days and observed no differences compared with WT cells (figure 5F). Importantly, using ID95 cells expressing Cas9 and a control Scramble guide, we could discard that the effects observed for PGLYRP1 KO cells in vivo were due to a CRISPR/Cas9 immunogenic effect as no differences between WT and Scramble cells were observed (online supplemental figure S3E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Tumorigenic and metastatic potential of PGLYRP1 knockout and overexpression models. (A) Left panel: representative western blot of intracellular PGLYRP1 levels in protein lysates from wild-type (WT) and CRISPR knockout (KO) KPC ID11 and ID95 cell lines. GAPDH was used as an internal loading control. Right panel: mean fold change of the PGLYRP1/GAPDH densitometric ratios±STDEV, with each WT set as 1.0 (n=3, p values determined by unpaired t-test). (B) Left panel: representative flow cytometry plots of PGLYRP1 expression in both WT and KO cell lines. Right panel: mean fold change in PGLYRP1 expressing cells±SEM, with WT set as 1.0 (n=5, p values determined by unpaired t-test). (C) Left panel: images of tumours obtained from a subcutaneous extreme limiting dilution assay (ELDA). Shown are tumours extracted from immunocompetent C57Bl/6J mice 3 weeks postinjection of 100 000 WT, PGLYRP1 overexpressing (OE) or KO cells derived from the ID11 cell line. Right panel: mean tumour weights (mg)±SEM (no. of injections=8, p values determined by one-way ANOVA with post hoc Tukey test). (D) Growth curves indicating the mean tumour volume (mm³)±SEM over 20 days following injection of 100 000, 10 000 or 1000 ID11 wild-type (WT), PGLYRP1 OE or KO cells. The slopes for each group were compared using the ‘comparing slopes tool’ (GraphPad v8), and the p value presented was calculated by comparing all slopes. P values to compare between groups were calculated by two-way ANOVA. (E) Panels detailing the tumorigenic potential of the indicated numbers of injected ID11 and ID95 WT, PGLYRP1 OE or KO cells in immunocompetent C57Bl/6J mice. Each column shows the number of tumours formed/number of injections. (F) Proliferation (no. cells) of KPC ID11 and ID95 WT and PGLYRP1 KO at 24, 48 and 72 hours (h) after seeding, represented as the mean fold change±STDEV with WT 24 hours set as 1.0 (n=3, per condition and time, p values determined by unpaired t-test). (G) Images of tumours at the time of sacrifice from orthotopic injection of 10⁴ ID11 or ID95 WT or OE cells in immunocompetent C57Bl/6J mice. KO cells did not succeed in forming tumours. (H) Mean pancreata weight (mg)±SEM (n=6, p values determined by one-way ANOVA with post hoc Tukey test). (I) Representative H&E-stained sections of orthotopic tumours and pancreata from (G). Scale=60 µm. (J) Mean fold change in the percentage of mCherry⁺ tumour cells in digested livers±STDEV, with WT set as 1.0, as detected by flow cytometry (n=4 for ID11 and n=6 for ID95, p values determined by unpaired t-test). (K) Left panel: images of the lungs at the time of sacrifice from intravenous injection of 10⁶ KPC WT, PGLYRP1 OE or KO cells in immunocompetent C57Bl/6J mice. Right panel: Mean lung weight (mg)±SEM (n=5, p values as determined by one-way ANOVA with post hoc Tukey test). (L) Representative H&E-stained sections of lung metastases from (K). Left images: zoom 10× (scale=1000 µm or 200 µm). Right images: zoom 40× (scale=80 µm). ANOVA, analysis of variance; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PGLYRP1, peptidoglycan recognition protein 1; STDEV, standard deviation.

To investigate the potential influence of the immune system and further study the possible role of PGLYRP1 in PDAC stemness, we performed an additional tumour formation assay in immunodeficient NOD.CB17-Prkdc^scid/scid/Rj (NOD.SCID) mice injecting 10³ cells, the limiting dilution obtained above. In this setting, KO cells generated the same number of tumours compared with WT and OE cells (online supplemental figure S3F). Although ID11 KO cells grew significantly slower than WT cells, tumours did not show significant diferences in final average weight and volume, which was replicated with the ID95-derived cell lines, suggesting that PGLYRP1 does not impair tumour initiation/stemness and that adaptive immune cells and fully functional innate immune cells were likely responsible for impeding KO cell growth in the immunocompetent setting (online supplemental figure S3G). Moreover, these data discarded a possible role for PGLYRP1 in stemness, although they suggest that PGLYRP1 is an important putative immune evasion protein used by CSCs. Following this idea, we investigated if PGLYRP1 expression favoured primary tumour formation and metastatic colonisation after orthotopic implantations in C57Bl/6J mice pancreata. PGLYRP1 overexpression increased primary tumour size (figure 5G,H) and induced a more dedifferentiated phenotype as determined by histological analysis (figure 5I). Moreover, disseminated cells from OE tumours colonised the liver more efficiently (figure 5J). Again, KO cells did not form tumours or metastases (figure 5G–J). To bypass tumour formation as a requisite for distant organ colonisation, 10⁶ WT, OE and KO cells were intravenously injected in C57Bl/6J mice. PGLYRP1 WT and OE cells succeeded in forming metastases in the lungs, while KO cells did not (figure 5K,L). Altogether, these data support the hypothesis that PGLYRP1 is necessary for initiation and metastasis in immunocompetent mice.

PGLYRP1 confers immune protection and alters the immune tumour microenvironment

In line with an immune evasive role, we found that PGLYRP1⁺ cells also coexpressed higher levels of PD-L1 and CD86 compared with PGLYRP1^- cells (figure 6A,B), both of them immune checkpoint ligands that engage with PD-1 and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) on T cells, respectively, and inhibit T cell activation and antitumour responses,34 indicating an overlap between PGLYRP1 and established immune checkpoint ligands. To directly test the immunoevasive capacity, WT, OE and KO cells were cocultured with MΦs and activated T cells to evaluate phagocytosis and cytotoxicity, respectively. We found that KO cells were phagocytosed more efficiently and were more vulnerable to cytotoxic killing than WT cells, while OE cells were more resistant to immune cell attack (figure 6C–E). Supporting these results, anti-PGLYRP1 antibodies increased KPC WT susceptibility to MΦ-mediated phagocytosis (figure 6D). Thus, PGLYRP1 expression seems to have an active role in immune evasion in vitro and may condition the immune landscape in vivo. To test the latter, we performed immune phenotyping in the orthotopic tumours established above and observed a significant decline in immune cell infiltration, particularly in the myeloid cell, monocyte and neutrophil subsets in OE versus WT tumours (figure 6F). However, there were no significant changes in MΦs, CD206⁺ MΦs, general T cells, and CD4⁺ and CD8⁺ T-cell subsets in OE tumours (figure 6F), indicating that PGLYRP1 does not likely influence the infiltration or retention of these immune subsets. In PGLYRP1 KO cell-injected pancreata (in which tumours did not form), we could only evaluate the immune cells that remained after tumour cell clearance, which we compared against the pancreata of control mice. All the immune populations analysed, except for CD206⁺ TAMs, augmented in comparison with healthy controls, indicating the presence of a robust immune infiltration/response against injected KO cells (figure 6G). Thus, PGLYRP1 expression seems to modulate the immune response during tumour formation by altering immune cell infiltration and conferring resistance to activated T cells and MΦs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

PGLYRP1 promotes immune evasion and alters immune infiltration. (A) Representative flow cytometry plots of PD-L1 and CD86 expression in PGLYRP1^- or PGLYRP1⁺ populations. (B) Violin plots representing the quantification of the mean percentage of PD-L1⁺ and CD86⁺ cells in the PGLYRP1^- or PGLYRP1⁺ populations for the ID11 cell line (left) and the ID95 cell line (right) determined by flow cytometry (n=9, p values determined by unpaired t-test). (C) Quantification of MΦ-phagocytosed cells (ID11, left; ID95, right), determined by flow cytometry as double-positive (mCherry/BFP⁺ and CD11b⁺) live cells, represented as mean fold change±SEM, with WT set as 1.0 (n=14 for ID11 and n=6 for ID95, p values determined by one-way ANOVA with post hoc Tukey test). Bone marrow-derived primary MΦs were obtained from three different donor mice. (D) Quantification of MΦ-phagocytosed PGLYRP1 WT cells (Ctrl) without or with anti-PGLYRP1 (αPGLYRP1) treatment (0.5 µg/mL; 24 hours), determined by flow cytometry as double-positive (EpCAM⁺ and CD11b⁺) live cells, represented as mean fold change±SEM, with Ctrl set as 1.0 (n=19, p values determined by unpaired t-test). (E) Quantification of dead cells (TO-PRO-3⁺) in the tumour population (mCherry⁺/BFP⁺) for ID11 (left) and ID95 (right) cells, determined by flow cytometry and represented as mean fold change±SEM, with WT set as 1.0 (n=12 for ID11 and n=6 for ID95, p values determined by one-way ANOVA with post hoc Tukey test). T cells were obtained from lymph nodes and spleen of three different donor mice. (F) Percentage of the different indicated immune cell populations in the TME from WT or OE tumours, determined by flow cytometry and represented as box plots (n=3 mice per condition and cell line, p values determined by unpaired t-test). (G) Percentage of the different indicated immune cell populations in the pancreata microenvironment from control or KO-cell-injected pancreata, determined by flow cytometry and represented as box plots (n=3 mice per condition, p values as determined by unpaired t-test). (H) Left panel: representative images of PGLYRP1^low and PGLYRP1^high tumour areas from spontaneous KPC tumours obtained by 3D quantitative confocal microscopy. Top images show a conventional IF staining, while bottom images show a postprocessed image where neutrophils (light blue) and macrophages (burgundy) are annotated through Imaris spots function for quantification. Right panels: number of indicated immune cells±SEM (neutrophils or macrophages) per µm³ in both tumour areas based on PGLYRP1 expression (low or high) (n=9 tumours for neutrophils and n=3 tumours for macrophages, p values determined by unpaired t-test). ANOVA, analysis of variance; EpCAM, epithelial adhesion cell adhesion molecule; IF, immunofluorescence; KO, knockout; MΦ, macrophage; OE, overexpressing; PD-L1, programmed cell death ligand 1; PGLYRP1, peptidoglycan recognition protein 1; TME, tumour microenvironment; WT, wild type.

To assess the expression and impact of PGLYRP1 on the PDAC immune compartment, the stroma cell populations expressing PGLYRP1 were analysed using the previously mentioned scRNA-seq dataset.21 Neutrophils, monocytes, CD4⁺ and CD8⁺ T cells were found to express Pglyrp1, although in a lower proportion compared with tumour cells (online supplemental figure S4A, B). Histological sections of KPC tumours confirmed neutrophils as the primary immune cell expressing PGLYRP1 in PDAC (online supplemental figure S4C). Although there is a reduction in neutrophil numbers in OE tumours, a transcriptional analysis of these tumour-associated neutrophils (TANs) revealed that OE-infiltrated TANs express lower levels of migration-related genes (eg, Cxcr2 and Cxcr4) and, interestingly, higher levels of immunosuppressive molecules, including Pglyrp1 (online supplemental figure S5A), compared with WT tumours. Reduced neutrophil motility in a transwell migration assay (online supplemental figure S5B) and decreased surface expression of CXCR2 and CD95 by flow cytometry in recombinant (r)PGLYRP1-stimulated neutrophils supported these findings (online supplemental figure S5C). Additionally, PGLYRP1 alone enhanced neutrophil survival similar to lipopolysaccharide (LPS), and conditioned medium (CM) from KPC OE cells increased neutrophil survival compared with KPC WT CM (online supplemental figure S5D). M0 MΦs exposed to rPGLYRP1 did not polarise to an M1 or M2 phenotype (online supplemental figure S5E), but when M1-polarised or M2-polarised MΦs were treated with rPGLYRP1, Cd86 expression was reduced in all MΦ subtypes and Arg1 expression increased in M2 MΦs (online supplemental figure S5F). In MΦs, CD86 serves as a costimulatory molecule that activates T cells via the interaction with CD28. Consequently, rPGLYRP1 may diminish the immunogenicity of these cells by downregulating CD86 expression. However, the presence of CTLA-4 on T cells alters the function of CD86 towards inducing T cell anergy. Despite its reduced expression, CD86 remains active in MΦs. Moreover, PGLYRP1 exposure also induced the expression of Camp, which has been shown to activate CSCs,35 reduced the expression of Tnf (online supplemental figure S5G) and increased MΦ migration (online supplemental figure S5H), despite no significant effects on MΦ infiltration in PGLYRP1 OE tumours in vivo. In light of the aforementioned results, we used 3D quantitative confocal microscopy to further analyse the spatial distribution of neutrophils and MΦs in KPC spontaneous tumours. Areas with high PGLYRP1 expression exhibited reduced neutrophil but higher MΦ infiltration (figure 6H). These findings support the in vitro migration results and open up the question regarding the role of MΦs in PGLYRP1 expression, given their increased presence in PGLYRP1^high areas.

Regarding T cells, exposure of activated and non-activated T cells to rPGLYRP1 showed no variation in exhaustion markers (online supplemental figure S5I). However, activated T cells exhibited reduced viability compared with non-activated T cells (online supplemental figure S5J), highlighting a possible mechanism by which PGLYRP1 protects cells from cytotoxic T-cell mediated death. Overall, these data support the hypothesis of an immunomodulating role for PGLYRP1 promoting an immunosuppressive environment through reduced migration of neutrophils, increased presence of protumoral macrophages and possibly direct induction of activated T cell death.

TNFα induces PGLYRP1 expression, protecting CSCs from immune clearance

Our RNA microarray data highlighted an enrichment in the TNFα-TNFR1 signalling pathway in triple-positive CSCs (figure 7A and online supplemental figure S6A), including Tnf and Tnfrsf1a upregulation, which we validated at the mRNA, protein and/or cell surface levels using CSC-enriched spheres or PGLYRP1⁺ triple-positive cells (figure 7B–E, online supplemental figure S6B). Interestingly, a functional connection between TNFα and PGLYRP1 has been described recently.36 Bushal et al have shown that in microglia, TNFα induces the expression of PGLYRP1. In agreement, another study using oesophageal cancer cells showed that PGLYRP1 modulates TNFα-tumour necrosis factor receptor 1 (TNFR1) signalling by binding TNFR1.37 Thus, we hypothesised that TNFα might regulate PGLYRP1 in PDAC CSCs. Indeed, The Cancer Genome Atlas (TCGA) data38 showed a positive correlation between TNF and PGLYRP1 expression, supporting a putative functional connection (figure 7F). To experimentally dissect this plausible link, we exposed KPC cells to recombinant (r)TNFα, resulting in increased Pglypr1 mRNA expression (figure 7G) and frequency of PGLYRP1⁺ cells (figure 7H) along with an expansion in PGLYRP1⁺ CSCs, which was counteracted by the addition of rPGLYRP1 (figure 7I), likely via its binding to TNFR1. Importantly, the non-CSC population did not increase PGLYRP1 expression on rTNFα stimulation (figure 7J). The observed competition with TNFα was confirmed by assessing the expression of TNFα-induced genes in KPC cells treated with rPGLYRP1 (online supplemental figure S6C) and was further validated by using infliximab (TNFα inhibitor) to reduce basal PGLYRP1 expression in spheroid conditions (online supplemental figure S6D). Thus, the sum of these data links PGLYRP1 induction in PDAC CSCs with increased TNFα-TNFR1 signalling.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

TNFα induces PGLYRP1 expression, protecting CSCs from immune clearance. (A) GSEA plot representing enrichment of the TNF-TNFR signalling pathway from transcriptomic data generated from tumour-isolated triple-positive CSCs versus non-CSCs. (B) RT-qPCR analysis of Tnf mRNA expression in KPC cells in adherent monolayer (ADH) or spheroid (SPH) culture conditions. Shown is the mean fold change±SEM, with ADH set as 1.0 (n=4 for ADH and n=5 for SPH, p values determined by unpaired t-test). (C) RT-qPCR analysis of Tnfrsf1a mRNA expression in KPC ADH or SPH cultures. Shown is the mean fold change±SEM, with ADH set as 1.0 (n=3, p values determined by unpaired t-test). (D) Box plots representing the mean levels of soluble TNFα (pg/ml) as determined by ELISA comparing KPC ADH versus SPH cell conditioned medium (n=6, p values determined by unpaired t-test). (E) Quantification of flow cytometric analyses of TNFR1⁺ KPC cells grown in ADH or SPH conditions. Shown is the mean fold change±SEM, with ADH set as 1.0 (n=7, p values determined by unpaired t-test). (F) Correlation dot plot of TNF and PGLYRP1 expression in the TCGA database. P value and r were calculated employing Pearson’s correlation. (G) RT-qPCR analysis of Pglyrp1 mRNA expression in KPC cells after recombinant (r)TNFα stimulation (20 ng/mL) for 6 hours. Shown is the mean fold change±SEM, with Ctrl set as 1.0 (n=6 for controls and n=12 for TNFα treated, p values determined by unpaired t-test). (H) Quantification by flow cytometry of PGLYRP1⁺ cells after treatment with rTNFα (20 ng/mL, 6 hours). Shown is the mean fold change±SEM, with Ctrl set as 1.0 (n=6 for control and n=12 for rTNFα treated, p values determined by unpaired t-test). (I) Quantification by flow cytometry of PGLYRP1⁺ triple-positive CSCs after treatment with rTNFα (20 ng/mL; 6 hours) and rTNFα + rPGLYRP1 (20 ng/mL+1 µg/mL; 6 hours). Shown is the mean fold change±STDEV, with Ctrl set as 1.0 (n=3 for controls and n=4 for treated group, p values determined by one-way ANOVA with post hoc Tukey test). (J) Quantification by flow cytometry of PGLYRP1⁺ non-CSCs after treatment with rTNFα (20 ng/mL; 6 hours) and rTNFα + rPGLYRP1 (20 ng/mL+1 µg/mL; 6 hours). Shown is the mean fold change±STDEV, with Ctrl set as 1.0 (n=3 for controls and n=4 for treated group, p values determined by one-way ANOVA with Dunnett’s test). (K) Quantification of dead cells (TO-PRO-3⁺) after 6 hours of TNFα (20 ng/mL) treatment of KPC WT, PGLYRP1 OE or KO cultures, determined by flow cytometry. Data are represented as the mean fold change±SEM, with WT set as 1.0 (n=5, p values determined by one-way ANOVA). (L) Quantification of MΦ-phagocytosed PGLYRP1 KO cells without or with infliximab (IFX) treatment (10 µg/mL; 24 hours), determined by flow cytometry as double-positive (BFP⁺ and CD11b⁺) live cells, represented as mean fold change±SEM, with KO set as 1.0 (n=12 for KO and n=14 for KO+IFX, p values determined by unpaired t-test). (M) Quantification of MΦ-phagocytosed PGLYRP1 KO cells without or with rPGLYRP1 treatment (1 µg/mL; 24 hours), determined by flow cytometry as double-positive (BFP⁺ and CD11b⁺) live cells, represented as mean fold change±SEM, with KO set as 1.0 (n=19, p values determined by unpaired t-test). (N) Quantification of MΦ-induced cell death in WT and PGLYRP1 KO cells, determined by flow cytometry as EpCAM⁺ or BFP⁺ dead cells, in basal conditions or KO cells treated with rPGLYRP1, infliximab (IFX) or WT cells treated with anti-PGLYRP1 antibody (αPGLYRP1) represented as percentage±SEM (n=20 for WT and KO, 9 for KO+rPGLYRP1 and IFX, and 7 for WT+ αPGLYRP1, p values determined by unpaired t-test). ANOVA, analysis of variance; CSCs, cancer stem cells; GSEA, gene set enrichment analysis; KO, knockout; MΦ, macrophage; rPGLYRP1, recombinant peptidoglycan recognition protein 1; RT-qPCR, reverse-transcription quantitative PCR; STDEV, standard deviation; TCGA, the cancer genome atlas; TNFα, tumour necrosis factor alpha; TNFR, tumour necrosis factor receptor; WT, wild type.

We next explored TNFα as a potential cytotoxic mediator shaping tumour evolution and the role of PGLYRP1 in regulating/counteracting this effect. Exposure of WT, PGLYRP1 OE and KO cells to rTNFα decreased KO cell viability compared with WT cells, but PGLYRP1 OE cells were significantly protected (figure 7K). While WT cells were also susceptible to the cytotoxic effect of rTNFα, the addition of rPGLYRP1 counteracted this phenotype (online supplemental figure S6E). These results suggested that PGLYRP1 is not only induced by TNFα, but it also serves as a protective mechanism against TNFα-derived cytotoxic effects, promoting tumour cell survival. If this hypothesis is true, the blockade of TNFR1 signalling with the TNFα inhibitor infliximab or the presence of rPGLYRP1 should have an impact on MΦ-mediated phagocytosis of PGLYRP1 KO cells. Indeed, both approaches reduced phagocytosis (figure 7L,M). These results were further supported by the reduced viability of PGLYRP1 KO KPC cells when treated with MΦ CM (online supplemental figure S6F), which contains MΦ-secreted TNFα.39 Indeed, when KPC PGLYRP1 KO cells were cocultured with MΦs in the presence of rPGLYRP1 or infliximab, reduced cell death was observed, while when KPC WT cells were cocultured with MΦs in the presence of anti-PGLYRP1 antibodies, increased cell death was observed (figure 7N). In summary, PDAC cells, particularly CSCs, upregulate PGLYRP1 in response to inflammatory signals, notably TNFα from both tumour and tumour microenvironment (TME) cells (eg, MΦ), as a strategy to mitigate its cytotoxic effects by blocking TNFα/TNFR1 signalling.

PGLYRP1 is present in human PDAC cells, enhances immune evasion and is a potential PDAC biomarker

To validate the results obtained in PDAC murine models in human samples, we evaluated the expression of PGLYRP1 in patients. First, we assessed, by RNA-seq,40 the expression of PGLYRP1 in EPCAM⁺ cells from the Espinet et al dataset.40 Overall, there was an increase in PGLYRP1 expression in tumours compared with healthy adjacent tissue; however, many tumours expressed low levels of PGLYRP1 (figure 8A), as would be expected of a marker (ie, CSC marker) whose expression is restricted to a subpopulation of cells. Thus, we analysed a scRNA-seq dataset obtained from primary tumours of patients with PDAC41 and found that PGLYRP1 was expressed in few cells, mainly in the tumour compartment (figure 8B,C). PGLYRP1⁺ cells were enriched in genes related with stemness, aggressiveness and chemoresistance (figure 8D), in agreement with our results in KPC mice. We next validated the results obtained employing patient-derived xenograft (PDX)-derived primary PDAC cells (ie, PANC185 and PANC354) in vitro. As expected, human PDAC spheroid cultures expressed higher PGLYRP1 levels (figure 8E,F). Human PDAC cells are EPCAM⁺ but do not express Sca-1, as it is a murine protein with no human homologue. Thus, CD133 and CXCR4, which are markers linked with a metastatic CSC subpopulation,4 were used to verify the enrichment of PGLYRP1 expression in human CSCs ex vivo and in vitro (figure 8G,H). We also confirmed PGLYRP1 expression in epithelial tumour cells across a series of tissue samples including PDXs and freshly resected tumours and its absence in healthy tumour-adjacent pancreas (online supplemental figure S7A), with results similar to those obtained in our mouse models. To further validate these results, we analysed PGLYRP1 expression in a tissue microarray (TMA) containing cores from 113 patients. Although most of the patient samples expressed PGLYRP1 in the TME, a smaller fraction expressed PGLYRP1 in tumour cells (online supplemental figure S7B). Interestingly, PGLYRP1 expression was detected in 26.19% of samples with poor or undifferentiated histological tumour differentiation (grade 3–4) and in 11.94% with well to moderate differentiation (grade 1–2). Consequently, the grade 3–4 samples, associated with aggressive tumour behaviour, presented higher levels of PGLYRP1 than grade 1–2 samples, although it was not significant probably due to the limited number of samples (online supplemental figure S7C). Concerning immune evasion, we confirmed that PGLYRP1⁺ human cells also co-expressed PD-L1 and CD86 (figure 8I, online supplemental figure S7D). PGLYRP1 KO and OE cell lines were also generated (online supplemental figure S7E–G), and WT, OE and KO cells were cocultured with human MΦ and activated T cells to measure their capacity to avoid phagocytosis (figure 8J) and cytotoxic killing (figure 8K), confirming the immune evasive role of PGLYRP1 in the human setting. We also analysed immune infiltration ex vivo and in silico. In our TMA, patients with detectable levels of PGLYRP1 in the tumour compartment presented lower infiltration of immune cells (online supplemental figure S7H), although not significant, again probably due to the limited number of samples. Employing TCGA data, we found that PGLYRP1 correlated negatively with CD45 and CD11b genes (PTPRC and ITGAM, respectively) and positively with ARG1 and MPO (figure 8L).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

PGLYRP1 in human PDAC CSCs. (A) Mean fold change±SEM of PGLYRP1 expression in freshly sorted EPCAM⁺ cells from human tumours (tumour) or healthy adjacent pancreatic tissue (healthy tissue), determined by RNA-seq (Espinet et al dataset40) and with healthy tissue set as 1.0 (n=14 for healthy tissue and n=62 for tumours, p values determined by unpaired t-test). (B) Left panel: UMAP of the different cell clusters present in the Hwang et al scRNA-seq dataset of human PDAC.41 Right panel: UMAP of PGLYRP1 expressing cells in the different clusters. (C) Representation of PGLYRP1 expressing events in each cluster. (D) Left panel: volcano plot showing the significantly enriched genes (in red) in tumour PGLYRP1 expressing versus PGLYRP1 non-expressing cells. Genes related to stemness, tumour aggressiveness and chemoresistance are labelled and coloured in green (significant genes, FC >0.25 and p<0.05). Right panel: PGLYRP1 expression in tumour cells according to PROM1 expression. (E) Mean fold change±STDEV of PGLYRP1 mRNA expression in two PDX-derived primary human PDAC cell lines (PANC185 and PANC354) cultured as adherent monolayers (ADH) or spheroids (SPH), determined by RT-qPCR and with ADH set as 1.0 (n=3, p values determined by unpaired t-test). (F) Left panel: representative flow cytometry plots of the percentage of PGLYRP1⁺ cells in ADH or SPH cultures. Right panel: quantification of the mean fold change±SEM of PGLYRP1⁺ cells in ADH or SPH cultures, with ADH set as 1.0 (n=3, p values as determined by unpaired t-test). (G) Representative confocal images of human PDAC tumour stained with CD133 (red), PGLYRP1 (green) and DAPI (blue). Arrowheads indicate double-positive cells. Scale=100 µm. (H) Left panel: representative flow cytometry plots of the percent of CXCR4⁺CD133⁺ CSCs (top) or CXCR4⁺CD133⁺PGLYRP1⁺ (bottom) cells in human PDAC cells grown as ADH or SPH cultures. Right panels: quantification of the indicated populations in PANC185 and PANC354 cultures. Data shown as the mean fold change±SEM, with ADH set as 1.0 (n=3, p values determined by unpaired t-test). (I) Violin plots representing the mean percentage of PDL1⁺ or CD86⁺ cells in the PGLYRP1^- or PGLYRP1⁺ populations in PANC185 cells (n=6 for all groups, p values determined by unpaired t-test). (J) Mean fold change±SEM of MΦ-phagocytosed cells determined by flow cytometry as double-positive (mCherry⁺/BFP⁺ and CD45⁺) live events for both PANC185 and PANC354, with WT set as 1.0 (n=8, p values determined by one-way ANOVA). Primary MΦs were obtained from five different healthy donors. (K) Mean fold change±SEM of dead cells (TO-PRO-3⁺) in the tumour population (mCherry⁺/BFP⁺) for both PANC185 and PANC354, determined by flow cytometry and represented with WT set as 1.0 (n=9, p values determined by one-way ANOVA with post hoc Tukey test). T cells were obtained from five different healthy donors. (L) Correlation dot plot of PTPRC (CD45), ITGAM (CD11b), MPO, ARG1 and PGLYRP1 expression in the TCGA database. P value and R were calculated employing Pearson’s correlation. (M) Violin plots representing the mean levels of soluble PGLYRP1 (pg/ml) as determined by ELISA comparing non-PDAC controls (n=58) with patients having PDAC (n=87) (p values determined by unpaired t-test). (N) Overall survival (OS) probability curve of patients with stage 1 PDAC from seven publicly available datasets analysed via OSpaad online software. At stage 1, the upper tercile of patients according to PGLYRP1 expression has poorer OS than the lower tercile. P-values determined by log-rank test. ANOVA, analysis of variance; ARG1, arginase1; CSCs, cancer stem cells; EpCAM, epithelial adhesion cell adhesion molecule; ITGAM, integrin subunit alpha M; MΦ, macrophage; MPO, heme protein myeloperoxidase; PDAC, pancreatic ductal adenocarcinoma; PDX, patient-derived xenograft; PGLYRP1, peptidoglycan recognition protein 1; PTPRC, protein tyrosine phosphatase receptor type-C; RT-qPCR, reverse-transcription quantitative PCR; scRNA-seq, single-cell RNA sequence; STDEV, standard deviation; TCGA, the cancer genome atlas; UMAP, Uniform Manifold Approximation and Projection; WT, wild type.

To further study the interaction between human tumour cells and immune cells in vivo, we used a Danio rerio xenograft model, which has a functional innate immune system from birth.42 PDAC WT and KO cells were injected into 48-hour postfertilisation embryos, and tumour formation and metastatic capacity were measured up to 6 days postinjection (dpi). Although the growth rate of PANC185 cells was not suitable for quantification, PANC354 KO cells showed an impaired growth capacity compared with WT cells (online supplemental figure S7I), possibly due to immune clearance by MΦs. Concerning metastatic capacity, PGLYRP1 KO in both cell lines significantly reduced the number of tail micrometastases (online supplemental figure S7J,K), supporting the hypothesis that PGLYRP1 has a role in tumour formation and metastasis.

Next, we analysed the META,43 Moffitt,44 Janky45 and TCGA databases to additionally validate PGLYRP1 expression in patient samples. The Moffitt dataset (GSE71729) revealed that 45 genes, including PGLYRP1, were commonly upregulated across primary tumour, liver and lung metastasis compared with healthy tissue (online supplemental figure S8A) and that most of the signalling pathways associated with these genes were related with the immune system and antibacterial defence (online supplemental figure S8B). We also interrogated PGLYRP1^high patients from the Janky, TCGA and META datasets, revealing downregulation of gene signatures such as inflammatory response or TNFα-signalling via nuclear factor kappa B (online supplemental figure S8C), highlighted in red). These findings demonstrate that PGLYRP1 is upregulated in PDAC and that immunomodulation of the TME appears to be a consistent feature associated with PGLYRP1 expression. Additionally, we explored a more general link between PGLYRP1 expression and CSC features by assessing PGLYRP1 expression in different cancer cell lines and observed an enrichment in PGLYRP1 expression in spheroids in all cell lines tested, except for colorectal cancer cells, suggesting that PGLYRP1 could also be important in CSCs of other tumour types, but it is not a pan-CSC marker (online supplemental figure S8D).

Finally, considering PGLYRP1 secretion, we assessed the potential utility of PGLYRP1 as a PDAC serum biomarker. 145 samples were analysed (58 non-PDAC controls, including 18 healthy controls, 25 high-risk controls, 19 patients with non-tumoral pancreatic diseases and 87 patients with PDAC), and PGLYRP1 levels were found to be significantly higher in patients with PDAC (figure 8M, online supplemental figure S8E), suggesting that PGLYRP1 could be a useful biomarker in differential PDAC diagnosis. Nevertheless, no differences in PGLYRP1 serum levels were found among different tumour stages (online supplemental figure S8F).

We also evaluated the relationship between PGLYRP1 expression and overall survival (OS) with Ospaad, an online tool published by Zhang et al which combines data from seven different patient cohorts (eg, TCGA and ICGC).46 Interestingly, patients at stage 1 could be stratified according to PGLYRP1 expression with the upper tercile exhibiting worse OS compared with the lower tercile (figure 8N). These data indicate that PGLYRP1 could potentially serve as a valuable and specific biomarker for detecting PDAC in liquid biopsies of patients and for stratifying patients at early stages in clinical settings, although more cohort studies are needed.