Bag6 deficiency in tumor cells accelerated tumor growth and induced changes in the TME in PDAC mouse models

Addressing a putative role of the chaperone BAG6 in pancreatic cancer (PC) we measured the BAG6 protein level in plasma samples of PC patients (quantified using Olink Explore 3072 analysis). This analysis revealed that PC patients with low BAG6 expression levels in the serum have significantly shorter overall survival compared to patients with high expression (Fig. S1A). Of note, BAG6 expression declined in human and mouse PDAC cell lines under hypoxic conditions, which are indicative of PDAC [24] (Fig. S1B, C). The potential role of BAG6 in the progression and pathology of PDAC was investigated in the following experiments.

Pan02 Bag6 knock-out (Bag6 KO) and Pan02 wild-type tumor cells (Bag6 WT) were transplanted into immune-competent mice (Fig. 1A). The s.c. tumor growth in the absence of Bag6 was significantly increased according to the volume (Fig. 1B, C) and weight (Fig. 1D). Representative images of s.c. tumors are shown in (Fig. 1E). A similar tumor promotion was observed in the orthotopic model (Fig. 1F) and representative images of the tumors depict the differences in tumor volume (Fig. 1G, H). The Bag6 knockout did not change the growth kinetics in vitro (Fig. S3D).

Fig. 1

Loss of Bag6 accelerated pancreatic tumor growth and altered the TME in mouse models. A Reporter mice were s.c or orth. transplanted with WT or KO Pan02 cells. B Tumor growth of s.c. tumors. C Tumor volume at day 21. D Tumor weight (mean ± SEM, n = 6). E Representative images of s.c. tumors in each group. F Tumor volume in orth. model (mean ± SEM, n = 8–9). G, H Representative images of orth. tumor growth measured via ultrasound. I Representative images of immune marker expression in tumor tissue. J Immune cells counted per square millimeter of tumor area (mean ± SEM, n = 7–10). K Relative gene expression of immune markers normalized to Rpl32 in tumor tissue from WT and KO tumors (mean ± SEM, n = 8–14). L Scatter plot between tumor volume in the KO group and cell types as indicated (n = 9), nonparametric Spearman correlation test. Statistical significance: (B) two-way ANOVA followed by Bonferroni corrections for multiple comparisons test; ((C), (D), (F), (I), (J)) unpaired Mann–Whitney U test; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s. (not significant)

Immunohistochemistry staining revealed a significant reduction in the infiltration of Cd4+ T cells, Cd8+ T cells, and Cd56+ NK cells in the tumor tissue of the KO group, whereas αSma-expressing fibroblasts were more abundant (Fig. 1I, J). The gene expression of Cd4 and Cd8 was also reduced in the tumor tissue, and the expression of αSma was increased in KO tumors (Fig. 1K). However, no significant changes were noticed in the expression of the NK cell marker Ncr1/Nkp46 (Fig. 1K).

Furthermore, Spearman correlation analysis validated a negative correlation between the tumor volume and the number of infiltrated Cd4+ and Cd8+ T cells in KO tumor tissue, whereas the number of infiltrated αSma+ cells increased with tumor volume. NK cell infiltration did not correlate with the tumor volume (Fig. 1L). This phenotype was already observed at day 9 (Supplementary Fig. S1D, E), before the onset of significantly accelerated tumor growth suggesting that the altered TME is rather a prerequisite than a result of advanced tumor growth.

These results indicate that Bag6 conferred tumor-suppressing activity in the PDAC models. The absence of Bag6 allowed the establishment of a tumor-promoting TME characterized by reduced infiltration of T cells and enhanced abundance of fibroblasts, known to restrict or support PDAC progression [25], respectively.

Accelerated tumor growth and remodeling of the TME were mediated by tumor cell-released EVs

Both, Pan02 WT and Bag6 KO cells released EVs, in which the number of EVs from KO cells was slightly higher (Fig. 2A). The inhibition of the in vivo synthesis and release of EVs using GW4869, an EV inhibitor previously described [26] resulted in the reduction of tumor growth and weight (Fig. 2B, C) of Bag6 KO but not WT tumors. This demonstrated the crucial role of EVs in the Bag6 phenotype and tumor progression. The treatment of cells with GW4869 had no toxic effect on tumor cells (Fig. S3D).

Fig. 2

Monitoring of EV uptake in vivo via Cre-LoxP and single-cell sequencing. A Quantification of particles isolated from the supernatant (SN) of WT/KO Pan02 cells (mean ± SEM, n = 10). B, C s.c. tumor growth and weight of Bag6 WT/KO groups upon treatment with GW4869 (n = 3–6). D Confocal images of tumor tissues. GFP+ cells correspond to cre recombination events and cre-negative Bag6 KO tumors were used as negative control. E UMAP depiction of cell characterization based on cell markers. F UMAP projections of cre+ KO (red, left panel) and cre+ WT tumors (blue, left panel). Recombination events in WT (middle) and KO (right) are highlighted in green. Statistical significance: (A, C) unpaired Mann-Whitney U test; (B) or two-way ANOVA followed by Bonferroni corrections for multiple comparisons test; *P < .05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s. (not significant); UMAP Uniform Manifold Approximation and Projection

EVs from KO and WT cells did not show any differences in size distribution, percentage of tetraspanin positive events (Cd9, Cd63, Cd81), or EV marker expression (Tsg101, Alix, Flotillin-1, Hsp70) (Fig. S2A–D). According to the MISEV guidelines [27], transmission electron microscopy of total EV preparations was performed to validate the expected EV morphology and purity (Fig. S2E).

A more comprehensive proteome characterization using mass spectrometry revealed however distinct protein loading (Fig. S3A, B) with 69 over- and 277 under-represented proteins in KO EVs. WASH complex subunit 4 (Washc4) one of the top hits detected in KO EVs is involved in endosomal sorting and actin cytoskeleton remodeling [28] and other factors contributing to EVs biogenesis, cargo sorting, as well as cancer-related proteins were differently expressed (see Fig. S3A, B).

To investigate the in vivo EV transfer, we employed a Cre-LoxP reporter system [29]. We first validated Bag6 expression and cre-mRNA expression in tumor cells by Western blotting and PCR, respectively (Fig. S3C, E). The uptake of the cre recombinase mRNA delivered via vesicles from transplanted tumor cells is expected to be translated to induce the expression of the GFP reporter gene in the recipient non-malignant host cells (Fig. S4A). To validate this approach, Bag6 KO and WT cells, both releasing EVs containing cre-mRNA, and Bag6 KO cells without cre were transplanted orthotopically into mice. Immunofluorescence staining of GFP unraveled recombination events in the tumor tissue of animals transplanted with tumor cells releasing cre-positive EVs, while absent in mice transplanted with cre-negative tumor cells (Fig. 2D). These data suggest that the method is eligible to identify EV-recipient cells within the tumor tissue.

Next, Bag6 KO- and WT-derived orthotopic tumors (n = 2) were resected, digested, and single-cell suspensions subjected to targeted single-cell sequencing (Fig. S4B). Cell clusters were annotated according to their surface markers expression, including B cells, Cd4+ T cells, Cd8+ T cells, fibroblasts, plasmacytoid dendritic cells (pDCs), macrophages, NK cells, and MCs (Fig. 2E). These cells were further divided into cells derived from KO or WT tumors (Fig. 2F, left panel) and GFP-positive/negative cells (Fig. 2F, middle and right). Overall recombination and vesicle uptake were observed in macrophages, neutrophils, and MCs. Strikingly, the MC cluster in which recombination was observed in almost 100% of the cells was exclusively detectable in KO but absent in WT tumor tissues. This finding suggests that MCs may contribute to the aggressive phenotype of KO tumors. Of note, the phenotype of fibroblasts, which did not show much recombination was also different between KO and WT, portending that fibroblasts are indirectly affected, potentially via MCs. In line, the analysis of the sequencing data using the CellChat Cell-Cell Communication Atlas Explorer [30], unraveled that MCs communicate predominantly with fibroblasts, tumor cells, Cd4+ T cells, and macrophages (Fig. S4C).

Deletion of Bag6 induced MC activation and infiltration in the TME

The KO-specific MC cluster was characterized by high expression of MC markers including c-Kit (Cd117), Cpa3, Fcera1, and Lat2 (Fig. 3A). Additionally, these MCs showed increased Gzmb and cytokine expression (Cxcl7, Il13, Il4, Lif, and Il6) (Fig. 3B, Fig. S5A), indicative of activated MCs.

Fig. 3

KO tumors were infiltrated by activated MCs. A, B Violin plots depicting MC markers and cytokines produced by activated MCs in KO tumor tissue. C Representative Cd117+ cells immunohistochemistry staining in tumor tissue of WT/KO groups. D Absolute Cd117+ cell counts per square millimeter (mean ± SEM, n = 6). E Spearman correlation analysis between tumor size and Cd117+ cell expression in the KO tumor tissue (mean ± SEM, n = 6). F Relative gene expression of MC markers (Cd117 and Cpa3) in WT/KO tumor tissue normalized to Rpl32 (mean ± SEM, n = 8–11). G Spearman correlation analysis between MC signature gene expression (CD117, FCERA1, and CPA3) and BAG6 in PDAC tumors (GEPIA2 analysis 28). Statistical significance: (D, F) unpaired Mann–Whitney U test; *P < 0.05; **P < 0.01

The immunohistochemistry of tumor tissue revealed an increase in infiltrated MCs (Cd117+ cells) (Fig. 3C, D) which correlated with the volume of the KO tumors (Fig. 3E). In line, RT-qPCR revealed the upregulation of MC-specific genes Cd117 and Cpa3 in the tumor tissue (Fig. 3F). Of note, the human MC gene signature (CD117, FCERA1, CPA3) and BAG6 expression correlate inversely in human PDAC tissue (Fig. 3G), in accordance with a putative role for BAG6 in regulating MC infiltration in humans.

To directly test the contribution of MCs to KO tumor progression we applied imatinib, a tyrosine kinase inhibitor that targets Cd117 and depletes MCs [31]. We observed a significant reduction in KO tumor growth and weight after imatinib treatment (Fig. 4A–C) demonstrating the critical role of MCs in tumor progression. The tumor growth and weight of WT tumors remained unaffected after imatinib treatment (Fig. 4A, B) and imatinib had no or minimal direct toxic or inhibitory effects on tumor cells in vitro (Fig. S5B).

Fig. 4

MC depletion reduced tumor growth. A Tumor growth curves of Bag6 WT/KO tumors (s.c.) treated twice weekly with imatinib or DMSO control, presented as volume. B Tumor weight (mean ± SEM, n = 3–6 mice). C Representative images of resected tumors from imatinib- and DMSO-treated animals. D Representative MC and BAG6 immunohistochemistry staining of Tissue Microarrays (TMAs). E Quantification of CD117+ cells in BAG6 high and low samples (mean ± SEM, n = 5). Statistical significance: (A) two-way ANOVA followed by Bonferroni corrections for multiple comparisons test; (B, E) unpaired Mann–Whitney U test; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s. (not significant)

To address MC infiltration in human PDAC, we used immunohistochemistry to stain both MCs (CD117) and BAG6 in TMAs (n = 69). We assessed MC infiltration and then selected 5 patients with the highest and 5 patients with the lowest MC infiltration prior to classification into high or low BAG6 expression. Representative images revealed that high MC infiltration corresponded to low BAG6 protein intensity, and vice versa (Fig. 4D, E; Fig. S5C) suggesting that patients with high MC infiltration and low BAG6 expression might benefit from imatinib therapy.

Bag6-deficient tumor cells release Il33-presenting EVs which induced MC activation

Single-cell sequencing of KO tumors revealed a notable and MC-specific expression of Il1rl1 (ST2/Il33 receptor) (Fig. 5A), which was confirmed by higher overall expression of Il1rl1 in KO tumor tissues (Fig. 5B). Thus, we tested whether MC activation was driven via Il33/Il1rl1 interaction. First, EVs were collected from mouse Pan02 and human PANC-1 cells with either WT or KO genetic background to perform an IL33/Il33-specific ELISA. The amount of EV-associated IL33/Il33 protein was elevated in KO-EVs (Fig. 5C). IL33 was not increased in the soluble fraction of the EVs purification from Pan02 WT or KO cells (WT-sol, KO-sol), whereas a moderate increase was observed in the corresponding sample of PANC-1 cells. Thus, the contribution of soluble IL33 cannot be excluded. Next, to explore the binding interaction between Il33 and Il1rl1, MCs were incubated with EVs or crude supernatant (SN) from WT or KO Pan02 cells, and Il1rl1 receptor expression on the MCs was assessed using flow cytometry (Fig. 5D). A significant reduction in Il1rl1 level upon incubation with KO (EVs and SN) was observed, indicating blocking of the receptor. The Il1rl1 signal was not changed when incubating MCs with WT samples (EVs and SN). The quenching was specific for I1rl1 since the detection of Cd117 remained unchanged upon pre-incubation with EVs.

Fig. 5

EV-associated IL33 induced MC activation. A Violin plots and UMAP depiction of Il1rl1 (Il33 receptor) expression in KO tumors. Expression of Il1rl1 was low on Treg cells and similar between KO and WT (Fig. S5D). Surface expression of Il1rl1 measured using flow cytometry was detectable on MC/9 cells but absent on WT and KO Pan02 cells (Fig. S5E). B RT-qPCR of Il1rl1 gene expression in tumor tissue normalized to Rpl32 (mean ± SEM, n = 7–9). C ELISA to detect mouse and human IL33 in EVs and EV purified soluble fractions (-sol) from WT and KO PDAC cells (mean ± SEM, n = 3). D Murine MCs pre-treated with EVs or crude supernatant (SN) from KO and WT Pan02 cells were analyzed for Ilrl1(IL33 receptor) expression. Cd117 expression was used as a control (mean ± SEM, n = 4–5). E Cytokine gene expression analysis of mouse and human MCs stimulated with KO- or WT-EVs isolated from Pan02 and PANC-1 cells, respectively. Data were normalized to Rpl32 (mean ± SEM, n = 4–6, 2 independent experiments). F Microbead assay to measure Il33 expression on WT-/KO-EVs from Pan02 cells (mean ± SEM, n = 6). Statistical significance: (A–F) unpaired Mann–Whitney U test t-test; *P < 0.05; **P < 0.01

The stimulation of mouse or human MCs with KO-EVs, but not with WT-EVs induced the expression of cytokines including Il6, Lif, and Tnfα (Fig. 5E), cytokines that are inducible via the Il33 signaling cascade [32]. Of note, inhibition of this signaling pathway using anti-IL33 blocking antibodies diminished the expression of these cytokines (Fig. 5E). In line, CRISPR/Cas9-mediated knock-out of Il33 in Bag6 KO cells, or MC pre-treatment with a blocking anti-Il1rl1/ST2 receptor, abrogated the EV-mediated MC activation (Fig. S6A, B).

To validate vesicle-associated Il33 expression directly, Il33 was detected using flow cytometry on beads-coupled EVs. The expression level of Il33 was significantly higher on EVs from KO cells (Fig. 5F), which is in line with the ELISA data (Fig. 5C). These data demonstrate that IL33/Il1rl1 engagement drives MC activation.

Bag6 regulates Il33 protein level via degradation and inhibition of release

Given the role of Bag6 in protein biogenesis and endoplasmic reticulum-associated degradation (ERAD) [33] we tested whether kifunensine (KIF), an inhibitor of BAG6-mediated ERAD [34], affected IL33 protein level. Treatment of cells with KIF resulted in an accumulation of Il33 in the lysates of Pan02 WT cells, an effect that was not seen in Pan02 KO cells, indicating that Il33 is degraded via the proteasomal pathway in a Bag6-dependent manner (i). Notably, increased cellular Il33 protein levels in WT cells did not result in Il33 release, which was only detectable in KO cells suggesting that Bag6 counteracts Il33 secretion (Fig. 6A). Secretion of Il33 in association with exosomes was recently observed in the context of chronic airway disease and shown to be mediated via the neutral sphingomyelinase 2 pathway (nSmase) [35], an enzyme also involved in the Bag6-mediated loading and release of EVs via the endosomal sorting complexes required for transport (ESCRT) pathway [10]. We therefore investigated whether Il33 release is mediated via the nSmase pathway from Pan02 cells. Bag6 WT/KO cells were treated with the nSmase inhibitor GW4869. Remarkably, this led to Il33 accumulation in KO cells, while WT cells remained unaffected (Fig. 6B), indicating that EV-associated Il33 release in KO cells is mediated through the nSmase pathway, which is inactive in WT cells. Finally, we rescued the expression of Bag6 in KO cells, which reproduced the WT Il33 phenotype (Fig. 6C). In summary, these data show that Bag6 promoted Il33 degradation (i) and inhibited the nSmase-dependent release of Il33 (ii).

Fig. 6

Bag6-mediated regulation of Il33 protein level and release. A, B ELISA to detect IL33 in cell lysate or supernatant (SN) of WT/KO Pan02 cells pretreated with or without Kifunensine (KIF) (150 µM) in (A) or with GW4869 (20 µM) in (B). Treatment was performed for 24 h followed by replacement of medium and 24 h incubation (n = 3 ± SEM). C ELISA to detect IL33 in cell lysate and SN after rescue of the Bag6 expression in Bag6 KO Pan02 cells (n = 3 ± SEM). Statistical significance: (A–C) unpaired t-test; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

The secretome of EV-activated MCs promoted tumor growth

Next, the secretome of human MCs exposed to KO-EVs derived from PANC-1 cells was collected and released proteins were quantified using antibody-based proximity extension assay (Fig. S6C). Of note, the 15 top-upregulated proteins (Fig. 7A) reflect a spectrum of regulatory mediators, such as CD73, IL13, and TGFβ, which influence immune cell and fibroblast functions (for global distribution see Fig. S6D, full list Supplementary Table S1). Furthermore, we observed high levels of factors associated with tumor cell proliferation (CDCP1, EpCAM, EGFR, and PDGFB). This MC secretome signature of the up-regulated factors is associated with shorter survival in PDAC patients (TCGA expression database, Fig. 7B) further indicating the critical role of the MC secretome in promoting tumor progression in PDAC patients.

Fig. 7

The secretome of MCs activated with KO-EVs promoted tumor growth. A The top 15 upregulated proteins in the secretome of human MCs pre-treated with PANC-1 KO-EVs (n = 2) are depicted. Normalized protein expression values (NPX) were averaged and the effect size calculated as compared to a PBS control is indicated (full protein list in Supplementary Table S1 and Fig. S6D for the global distribution of the effect size for the 15 upregulated proteins). B Kaplan Meier analysis of the top 15 upregulated proteins correlated with survival (TCGA data via GEPIA2). C Representative images of KPC mouse and PDAC human organoids after treatment with MC secretome from mouse/human MCs that were pre-treated with WT- or KO-EVs from Pan02 or PANC1, respectively as well αIL33 and αPDGF as indicated. D Quantification of organoid sizes determined via cell titer (mean ± SEM, n = 4–6). E The organoids were stained with anti-Ki67 antibodies (red), with Hoechst 33342 (nuclei, blue) and Alexa 546 phalloidin (actin, green) to visualize cellular structures. Scale bars: 10 µm. Statistical significance: (D) unpaired Mann–Whitney U test; *P < 0.05; **P < 0.01; ***P < 0.001

Finally, the impact of the MC secretome on tumor cell proliferation was assessed (Fig. 7C, D). Initially, human and mouse MCs were stimulated with either WT- or KO-EVs, in the presence or absence of anti-IL33 antibodies (Fig. S6E). Subsequently, the obtained secretomes were co-cultured with human or mouse PDAC organoids. An increase in size, number, and cell titer of organoids when cultured with the secretome from MCs stimulated with KO-EVs was observed in both models (Fig. 7C, D). This effect was absent when human/mouse MCs were pretreated with anti-IL33 antibodies. Notably, we observed no changes in the organoids when cultured with secretomes obtained from MCs stimulated with WT-EVs or PBS, regardless of the presence or absence of anti-IL33.

Due to the important role of PDGF in promoting tumor growth [36] and its high abundance in the MC secretome (Fig. 7A), the secretome of human/mouse MCs treated with KO-EVs was incubated with anti-PDGF antibody during organoid stimulation. Anti-PDGF antibodies effectively blocked the growth of organoids and significantly reduced the cell titer (Fig. 7C, D). Whole-mount staining of organoids allowed the visualization of single cells and confirmed the increased proliferation after KO-secretome pretreatment (Fig. 7E).

Further analysis revealed an increase in gene expression of cell cycle-associated factors in the KO-EV group and the corresponding pathways, which is in line with the induction of organoid proliferation (Fig. S6F, G). These findings suggest the crucial role of EV-associated IL33 as an upstream stimulator for MCs in the absence of BAG6, and PDGF as a downstream stimulator promoting tumor growth.

The activated MC secretome-induced iCAF polarization

Given the differences in the phenotype and infiltration of fibroblasts in the KO/WT tumor tissues (Figs. 1I–K and 2F (left UMAP), we investigated the impact of the MC secretome on mouse pancreatic stellate cells (PSCs) in vitro. PSCs were seeded in Matrigel drops and maintained in starvation medium to induce quiescence (q)PSCs. Culturing qPSC with the secretome from MCs stimulated with KO-EVs resulted in an iCAF phenotype, characterized by the expression of Saa1/2, Cxcl1, Il6, Lif, and αSma genes (Fig. 8A, upper panel). In contrast, qPSC cultured with the secretome from MCs stimulated with WT-EVs exhibited a myofibroblast-like cancer-associated fibroblast (myCAF) phenotype (upregulation of Col1a and Col4a) (Fig. 8A, lower panel). In line with these findings, single-cell sequencing data and gene expression analysis of tumor tissues from WT and KO groups displayed an upregulation of Col1a in the WT and αSma in the KO group, respectively (Fig. 8B, C) corresponding to the fibroblasts-1 and -2 clusters (Fig. 2E, F). The genes are differentially regulated in vivo and in vitro in fibroblasts upon MC secretome treatment and are summarized in (Fig. 8D).

Fig. 8

The secretome from activated MCs induced an iCAF phenotype. A mPSCs were treated with MC secretome of MCs pre-treated with WT- or KO-EVs. Gene expression of iCAF markers and mCAF markers was determined by RT-qPCrR. (mean ± SEM, n = 6). B UMAP projection of fibroblast markers Col1a and aSMA in WT/KO tumors. C Col1a expression in KO and WT tumors determined by RT-qPCR. Data were normalized to Rpl32 (mean ± SEM, n = 9–14). D Graphical summary of mPSC/fibroblast phenotypes in vivo or after MC secretome treatment. Statistical significance: unpaired Mann–Whitney U test; *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001

In summary, this study unravels that the loss of BAG6 provoked the release of EV-associated IL33 which activated MCs. The secretome of these activated cells (i) directly stimulated tumor cell proliferation and (ii) exerted an impact on fibroblasts supporting a tumor-promoting environment. Depletion of MCs using imatinib restricted tumor development in a Bag6-deficient mouse model, indicating that patients with low BAG6 expression and high MC infiltration may benefit from imatinib therapy (see graphical abstract for summary).