Gasdermin E mediates resistance of pancreatic adenocarcinoma to enzymatic digestion through a YBX1–mucin pathway

Pancreatic ductal adenocarcinoma (PDAC) originates from normal pancreatic ducts where digestive juice is regularly produced. It remains unclear how PDAC can escape autodigestion by digestive enzymes. Here we show that human PDAC tumour cells use gasdermin E (GSDME), a pore-forming protein, to mediate digestive resistance. GSDME facilitates the tumour cells to express mucin 1 and mucin 13, which form a barrier to prevent chymotrypsin-mediated destruction. Inoculation of GSDME−/− PDAC cells results in subcutaneous but not orthotopic tumour formation in mice. Inhibition or knockout of mucin 1 or mucin 13 abrogates orthotopic PDAC growth in NOD-SCID mice. Mechanistically, GSDME interacts with and transports YBX1 into the nucleus where YBX1 directly promotes mucin expression. This GSDME–YBX1–mucin axis is also confirmed in patients with PDAC. These findings uncover a unique survival mechanism of PDAC cells in pancreatic microenvironments.

B oth tumorigenesis and pathogenesis of pancreatic ductal adenocarcinoma (PDAC) remain incompletely understood. The exocrine portion of the pancreas, the origin of PDAC, constitutes the majority (>95%) of the pancreatic mass, which includes acinar and duct cells and secretes digestive enzymes 1,2 . Physiologically, the secreted pancreatic juice flows through the pancreatic duct into the duodenum and aids digestion. However, this juice is potentially dangerous and is able to destroy neighbouring pancreatic cells under certain conditions such as acute pancreatitis 3,4 . Therefore, normal structures of the pancreatic duct are organized in a strict and orderly manner to avoid self-digestion. During malignant development, it is highly probable that disorderly tumour growth inevitably obstruct the normal ductal space, which leads to pancreatic juice leaking out and destroying nearby cells 5,6 . This raises a fundamental question of how PDAC tumour cells can evade pancreatic enzymatic destruction and survive.
Recent studies have highlighted a pivotal role of the gasdermin family members in mediating inflammatory cell death through their pore-forming activity [7][8][9][10] . Gasdermin interdomain cleavage allows the amino-terminal domain to bind membrane phospholipids and oligomerize into a pore on the plasma membrane, which leads to rapid cellular swelling, large bubbles emerging from the plasma membrane and subsequent cell lysis. This gasdermin-mediated programmed necrosis is called pyroptosis. Among the members, gasdermin E (GSDME) is unique because its active form requires cleavage by caspase-3, an enzyme involved in tumour cell apoptosis 11,12 . GSDME is silenced in various tumour cell types due to high methylation of the GSDME promoter region [13][14][15] . However, GSDME is also present in some tumours and can induce tumour cell death 11 and enhance antitumour immunity 15 . But it is hard to explain why tumour cells would express GSDME to kill themselves. It may be that GSDME plays a tumour-promoting role under certain conditions. In the present study, we show that PDAC tumour cells express GSDME at high levels to mediate resistance to pancreatic enzymatic digestion through a GSDME-YBX1-mucin pathway, thereby playing a tumour-promoting role beyond the known pore-forming function. These findings provide a deeper understanding of the pathogenesis of PDAC with chronic inflammation 16 , which considers the potential enzymatic digestion of paracancerous parenchymal cells.

Results
GSDME is required for orthotopic PDAC growth. GSDME was expressed at high levels in human PDAC tumour cell lines (Fig. 1a). In line with this, the GSDME promoter region was highly hydroxymethylated (Fig. 1b), and ten-eleven-translocation methylcytosine dioxygenase 2 (TET2) was strongly upregulated in PDAC cell lines and primarily found in the nucleus (Extended Data Fig. 1a,b). Knocking down TET2 downregulated GSDME expression in PDAC cells (Extended Data Fig. 1c,d), which suggests that PDAC cells use an altered epigenetic programme to express GSDME at high levels. Such a pattern of GSDME expression prompted us to explore the role of GSDME in PDAC cells by using the CRISPR-Cas9 technique to knockout GSDME. We subcutaneously or orthotopically injected BxPC-3 cells transfected with single guide RNAs (sgRNAs) targeting GSDME (GSDME-SGs) into NOD-SCID mice. GSDME deficiency did not affect subcutaneous tumour growth (Fig. 1c,d and Extended Data Fig. 1e), whereas orthotopic tumour growth was markedly inhibited in the pancreas (Fig. 1e). This phenomenon was not due to local immune surveillance because immunodeficient Gasdermin E mediates resistance of pancreatic adenocarcinoma to enzymatic digestion through a YBX1-mucin pathway mice were used. We speculated that the pancreas probably produces certain factors that are toxic to these GSDME knockout cells, thus disrupting their growth. Indeed, re-expressing GSDME conferred the ability of BxPC-3 cells transfected with GSDME-SGs to grow a tumour in the pancreas ( Fig. 1f and Extended Data Fig.  1f). Similar results were obtained in PANC-1 and AsPC-1 cells that had GSDME knocked out. These cells rapidly formed a subcutaneous tumour but exhibited weak tumorigenicity in the pancreas, but GSDME re-expression rescued orthotopic tumour growth (Extended Data Fig. 1g-j). Given the role of GSDME in tumour immunomodulation 15 , we used immunocompetent mouse models to verify this by constructing GSDME knockout mouse Pan02 cells and human AsPC-1 cells and orthotopically inoculating them into wild-type (WT) C57BL/6J and humanized mice, respectively. GSDME deficiency in these models similarly suppressed tumour growth (Extended Data Fig. 1k-n). Together, these results suggest that GSDME is required for the orthotopic growth of pancreatic tumours. GSDME mediates resistance of PDAC cells to digestive enzymes. Next, we explored whether GSDME mediates PDAC resistance to pancreatic enzymatic digestion, thus explaining the tumour suppression induced by GSDME deficiency. To test this, we used pancreatic lysates to treat PDAC cells. GSDME deficiency markedly decreased the viability of PDAC cells ( Fig. 2a and Extended Data Fig. 2a,b). Insulin-like growth factor 2, insulin and glucagon are not involved in GSDME-SG-mediated tumour cell death in vitro (Extended Data Fig. 2c,d); therefore we speculated that exocrine enzymes are involved in GSDME-deficient tumour cell death. Pancreatic exocrine enzymes are composed of trypsin, chymotrypsin, amylase and lipase. The addition of an amylase or a lipase inhibitor did not affect pancreatic-lysate-mediated tumour cell death (Extended Data Fig. 2e,f); however, cell death was blocked by either a chymotrypsin or a trypsin inhibitor (Fig. 2b,c). This result suggests that trypsin and/or chymotrypsin in the pancreatic digestive juice is involved in GSDME −/− PDAC cell death. To further confirm this result, we used trypsin and chymotrypsin to treat cells transfected with GSDME-SGs. Trypsin or chymotrypsin alone did not cause cell death in cells transfected with GSDME-SGs, but the combination of these two enzymes did (Fig. 2d,e). As chymotrypsin is produced as an inactive form in the pancreas and activated by trypsin, we treated cells with both trypsin and chymotrypsin and then added a trypsin inhibitor or a chymotrypsin inhibitor to the medium. Only the chymotrypsin inhibitor blocked cell death (Fig. 2f,g), which suggests that chymotrypsin exerts a direct cytotoxic effect on GSDME-SGs pancreatic tumour cells. Although previous reports have indicated that GSDME mediates pyroptosis, in this study, we found that pancreatic lysate did not induce pyroptosis of cells (Extended Data Fig. 2g). In addition, unlike malignant pancreatic cells, normal human pancreatic cells, which express GSDME at low levels (Extended Data . e,f, BxPC-3 cells transfected with SGCTR, GSDME-SGs or GSDME-SG/Flag-GSDME (2.5 × 10 5 cells) were orthotopically injected into the pancreas of mice. Tumours were photographed (left) and weighed (right) (n = 6). Normal pancreas served as the control. Scale bars, 1 cm. In a and b, n = 3 biological independent experiments. P values were determined by one-way ANOVA Bonferroni's test (b-f). The data represent the mean ± s.d. Fig. 2h), were not resistant to trypsin or chymotrypsin (Extended Data Fig. 2i,j). Next, we used Prss1 −/− C57BL/6J mice (in which trypsinogen is knocked out) to further validate these in vitro results in vivo. Trypsin deficiency led to GSDME -/-Pan02 cells to grow in the pancreas, similar to the control cells (Extended Data Fig. 2k). Together, these results suggest that pancreatic tumour cells mobilize GSDME to resist exocrine digestive enzymes. GSDME induces mucins for resistance to digestive enzymes. Next, we investigated the molecular mechanism by which GSDME regulates the resistance to digestive enzymes. The mucosal epithelium can protect itself by expressing a layer of mucous molecules to retain particles and exogenous enzymes, which prompted us to hypothesize that GSDME promotes resistance to digestive enzymes by upregulating mucin expression. In line with this hypothesis, measurement of the membrane-associated mucins MUC1, MUC3, MUC4, MUC12, MUC13, MUC15 and MUC16 (ref. 17 ) showed that MUC1 and MUC13 are expressed at high levels in pancreatic tumour cells (Extended Data Fig. 3a). Moreover, knocking out GSDME in AsPC-1 or PANC-1 cells resulted in the downregulation of MUC1 and MUC13 ( Fig. 3a and Extended Data Fig. 3b,c), and GSDME overexpression upregulated MUC1 and MUC13 (Extended Data Fig. 3d). Knockout or inhibition of mucin decreased the resistance of PDAC cells to digestive enzymes ( Fig. 3b and Extended Data Fig. 3e-g). In addition, although GSDME −/− tumour cells lost the resistance to digestive enzymes, re-expression of exogenous MUC1 or MUC13 largely restored their resistance ( Fig. 3c and Extended Data Fig. 3h), and the co-expression of MUC1 and MUC13 achieved the greatest resistance (Extended Data Fig. 3i). Meanwhile, trypsin and chymotrypsin treatment upregulated the expression of GSDME, MUC1 and MUC13 ( Fig. 3d and Extended Data Fig. 3j). In contrast, normal human pancreatic cells expressed markedly low levels of MUC1 and MUC13 ( Fig. 3e) concomitant with the loss of resistance to digestive enzymes (Extended Data Fig. 2j,k). The carbohydrate chains on mucins are commonly initiated with N-acetylgalactosamine (GalNAc) by linking to a mucin serine or threonine hydroxyl group. This O-linked glycosylation of MUC1 and MUC3 was observed in PDAC cells (Extended Data Fig. 3k), which is consistent with previous reports 18, 19 . Moreover, trypsin and chymotrypsin treatment promoted such Try/Chy Fig. 2 | GSDME helps tumour cells to resist trypsin and chymotrypsin digestion. a, AsPC-1 cells transfected with SGCTR or GSDME-SGs were treated with lysate (20 μl ml -1 ) isolated from mouse pancreas for 72 h. Viable cells were determined by trypan blue (TB) staining. b,c, AsPC-1 cells transfected with SGCTR or GSDME-SGs were treated with or without lysate (20 μl ml -1 ) in the presence of PI-1840 (100 μM) or a trypsin inhibitor (TI; 50 μg ml -1 ) for 72 h. Cell viability was measured by TB staining (b) or an ATP cell viability assay (c). d,e, AsPC-1 cells transfected with SGCTR or GSDME-SGs were treated with PBS, trypsin (50 U ml -1 ), chymotrypsin (40 U ml -1 ), trypsin and chymotrypsin (Try/Chy) or lysate for 72 h. Cell viability was determined by TB staining (d) or an ATP cell viability assay (e). f,g, The same as b-e, except that some cells were pretreated with Try/Chy for 10 min and then treated with IP-1840 or a trypsin inhibitor for 72 h. For a-g, n = 3 biological independent experiments. P values were determined by one-way ANOVA Bonferroni's test (a-g). NS, not significant. The data represent the mean ± s.d. Fig. 3k), whereas blocking the glycosylation of mucins by the addition of benzyl-GalNAc, an inhibitor of N-acetylgalactosaminyltransferase, disrupted the resistance to trypsin and chymotrypsin (Fig. 3f). In line with this result, mRNA expression of sialyltransferases (ST6GALNAC4, ST3GAL1, ST3GAL2 and ST3GAL5) was upregulated by the addition of trypsin and chymotrypsin (Extended Data Fig. 3l). To validate these results in vivo, MUC1-SG, MUC13-SG or MUC1/MUC13-SG tumour cells were inoculated into the pancreas of mice. Knockout of either MUC1 or MUC13 retarded tumour growth, but the knockout of both achieved the greatest tumour regression (Fig. 3g). Similarly, the use of a mucin inhibitor inhibited tumour growth and prolonged the survival of mice (Extended Data Fig. 3m,n). In addition, overexpression of MUC1 or MUC13 rescued GSDME-deficiency-retarded tumour growth (Extended Data Fig. 3o). Together, these results suggest that GSDME induces the expression of MUC1 and MUC13, which in turn promotes the resistance of PDAC cells to digestive enzymes. GSDME interacts with YBX1 to express mucin. Next, we explored the molecular pathway that GSDME uses to regulate MUC1 and MUC13. GSDME is typically found in the cytoplasm where it is cleaved by caspase-3 to exert a pore-forming function. Notably, GSDME was localized in the nucleus in pancreatic tumour cells, which was enhanced by the presence of trypsin and chymotrypsin (Extended Data Fig. 4a). To verify whether nuclear GSDME is involved in the resistance of pancreatic tumour cells to digestive enzymes, cells transfected with GSDME-SGs were forced to express either nuclear localization sequence (NLS)-GSDME or nuclear export sequence (NES)-GSDME. NLS-GSDME restored the resistance of cells trasnfected with GSDME-SGs to trypsin and chymotrypsin, whereas NES-GSDME had no such effect ( Fig. 4a and Extended Data Fig. 4b). Consistently, only NLS-GSDME induced the upregulation of MUC1 and MUC13 (Extended Data Fig. 4c). Such mucin upregulation could not be attributable to the binding of GSDME to MUC1 or MUC13 mRNA (Extended Data Fig. 4d) or to GSDME-mediated mucin stabilization (Extended Data Fig. 4e). On the basis of these results, we speculated that unlike the pore-forming effect in the cytosol, GSDME in the nucleus acts as a transcriptional regulator. Following the pull-down of a Flag-GSDME fusion protein, the immunoprecipitants were analysed by mass spectrometry. Although more than 1,000 GSDME-binding proteins were detected in the PDAC tumour cells, among the transcriptional regulatory proteins, transcription factor Y-box-binding protein 1 (YBX1) drew our attention (Extended Data Fig. 4f and Supplementary Table 1). As the most prominent member of the YBX family, YBX1 has been associated with multiple cancer-related processes 20 . Co-immunoprecipitation of the nuclear fraction showed that GSDME bound to YBX1 (Fig. 4b), and ultra-high super-resolution microscopy showed that GSDME colocalized with YBX1 in the nucleus of AsPC-1 cells in which endogenous GSDME had been deleted (Fig. 4c). GSDMB, which has similar molecular weight and isoelectric point to GSDME 21,22 , did not colocalize with YBX1 in the nucleus (Extended Data Fig. 4g), which suggests that binding of GSDME to YBX1 is specific. A bio-layer interferometry assay further confirmed a direct interaction between GSDME and YBX1 (Fig. 4d). The aspartate residue 270 (D270) of human GSDME has been identified as the cleavage site for caspase-3 (ref. 11 ). Mutating this aspartate to alanine (D270A) resulted in the disruption of the binding of GSDME to YBX1 (Extended Data  Fig. 4h); however, GSDME N-terminal fragments (N320, N394 and
N419) did not lose the binding ability, as evidenced by proximal ligation assay results (Extended Data Fig. 4h). In addition, GSDMB could not bind YBX1 (Extended Data Fig. 4h). Using AsPC-1 cells in which GSDME was knocked out, we further demonstrated that only WT GSDME (not GSDME-D270A or WT GSDMB) rescued the GSDME-deficiency-retarded tumour growth in the pancreas (Fig. 4e). In line with this result, the in vitro assay showed that neither GSDME-D270A nor GSDMB was able to rescue the resistance of GSDME knockout AsPC-1 cells to enzymatic digestion (Fig. 4f). Meanwhile, YBX1 is expressed in PDAC cells (Extended Data Fig. 4i), and YBX1 knockout resulted in the loss of the resistance to trypsin and chymotrypsin, which could be rescued by the re-expression of YBX1 (Fig. 4g and Extended Data Fig. 4j,k).
In addition, although YBX1 knockout induced PDAC cell death, it did not induce GSDME cleavage following pancreatic lysate treatment (Extended Data Fig. 4l). Notably, YBX1 −/− PDAC cells barely grew a visible tumour in the pancreas of mice (Extended Data Fig. 4m). Furthermore, exogenous expression of MUC1 and MUC13 rescued the ability of YBX1 −/− PDAC cells to grow a tumour in the pancreas (Fig. 4h). Together, these results suggest that GSDME regulates MUC1 and MUC13 expression by interacting with the transcription factor YBX1. GSDME transports YBX1 into the nucleus for mucin expression. Next, we asked whether YBX1 in the nucleus transcriptionally regulates the expression of MUC1 and MUC13. Overexpression of were treated with Try/Chy for 72 h. Viable cells were measured by TB staining. c, AsPC-1 cells treated with Try/Chy for 24 h were collected for ChIP-qPCR assay with anti-YBX1 and specific primers for MUC1 (left) or MUC13 (right). d, HEK-293T cells were co-transfected with MUC1 (left) or MUC13 (right) promoter luciferase reporter PGL3 and YBX1 plasmid for 24 h. Cells were then treated with Try/Chy for another 24 h, followed by analysis of luciferase activity. e, Immunostaining images (left) and quantification (right) of YBX1 from AsPC-1 cells transfected with SGCTR or GSDME-SGs and treated with Try/Chy for 36 h. Scale bar, 5 μm. f, The cell viability of AsPC-1 cells transfected with SGCTR, GSDME-SG or GSDME-SG/NLS-YBX1 was determined by TB staining. g, AsPC-1 cells transfected with SGCTR, YBX1-SG, YBX1-SG/NLS-YBX1, YBX1-SG/NES-YBX1 or YBX1-SG/NLS-GSDME (2.5 × 10 5 cells) were injected into the pancreas of mice. Tumours were photographed (left) and weighed (right) (n = 6 per group). Scale bar, 1 cm. For a-f, n = 3 biological independent experiments. P values were determined one-way ANOVA Bonferroni's test (b-g). The data represent the mean ± s.d.
YBX1 led to the upregulation of MUC1 and MUC13 in PDAC cells, whereas knockout of YBX1 downregulated MUC1 and MUC13, even when trypsin and chymotrypsin was used ( Fig. 5a and Extended Data Fig. 5a-c). In addition, in YBX1 knockout tumour cells, replenishment of NLS-YBX1 but not NES-YBX1 upregulated MUC1 and MUC13 (Extended Data Fig. 5d,e). Moreover, only NLS-YBX1 rescued the resistance of YBX1-deficient cells to trypsin and chymotrypsin (Fig. 5b), which suggests that YBX1 regulates MUC1 and MUC13 in the nucleus. Chromatin immunoprecipitation (ChIP) with quantitative PCR (ChIP-qPCR) confirmed that YBX1 indeed bound to the promoters of MUC1 and MUC13 (Fig. 5c). HEK-293T cells, which are commonly used for exogenous gene expression 23 , also express GSDME (Extended Data Fig. 5f,g). Luciferase assays using HEK-293T cells showed that YBX1 induced MUC1 and MUC13 expression (Fig. 5d). Notably, trypsin and chymotrypsin treatment promoted the entry of YBX1 into the nucleus, which could be abolished by GSDME knockout (Fig. 5e). In contrast, YBX1 knockout did not affect GSDME in the nucleus (Extended Data Fig. 5h), which, however, was blocked by the addition of the nuclear pore inhibitor wheat germ agglutinin (Extended Data Fig. 5i). Meanwhile, transfection of NLS-YBX1 into GSDME −/− AsPC-1 and PANC-1 cells rescued the resistance to trypsin and chymotrypsin ( Fig. 5f and Extended Data Fig. 5j); however, the transfection of NLS-GSDME into YBX1 −/− tumour cells did not have such an effect (Extended Data Fig. 5k). When YBX1 −/− AsPC-1 cells transfected with NLS-YBX1, NES-YBX1 or NLS-GSDME were inoculated into the pancreas of mice, the transfection of NES-YBX1 or NLS-GSDME barely rescued YBX1 −/− tumour cells to allow tumour growth. Conversely, the transfection of NLS-YBX1 favoured YBX1 −/− tumour cells and promoted tumour formation (Fig. 5g). In addition, GSDME N-terminal fragments (N320, N394 and N419) entered the nucleus, as shown by nuclear staining (Extended Data Fig. 5l). Nucleoporin 153 (NUP153) is a nuclear pore complex-associated basket protein involved in the nuclear import of proteins 24,25 . GSDME N-fragments colocalized with NUP153. However, GSDME-D270A was barely present in the nucleus, which indicates that the N-terminal part is required for the entry of GSDME into the nucleus (Extended Data Fig. 5l). Together, these results suggest that GSDME acts as a transporter to mediate the entry of YBX1 into the nucleus to promote mucin expression. . g, Schematic of the GSDME-YBX1-mucin pathway to regulate the resistance of PDAC to digestive enzymes. P values were determined by two-tailed Mann-Whitney test (a-c) or two-sided Pearson's correlation test (d-f). The data represent the mean ± s.d. GSDME is correlated with worse prognosis in patients with PDAC. Finally, we sought to validate the above-described results in patients with PDAC. Using the Gene Expression Profiling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku. cn/), we analysed PDAC tumour tissues (n = 179) and the adjacent tissues (n = 171), and found that GSDME, YBX1, MUC1 and MUC13 mRNAs are expressed at high levels in tumour tissues compared to the adjacent tissues ( Fig. 6a and Extended Data Fig. 6a). In line with these bioinformatics results, immunohistochemical staining of clinical samples from patients also revealed much higher levels of GSDME expression in PDAC tumour tissues but not in the adjacent paracancerous acinar or ductal cells, and that GSDME is present in the nucleus of the tumour cells (Fig. 6b). Consistently, YBX1, MUC1 and MUC13 were upregulated in tumour tissues of patients compared to the paracancerous tissues ( Fig. 6c and Extended Data Fig. 6b,c). Moreover, using data from The Cancer Genome Atlas (TCGA) database (https://tcga-data.nci.nih.gov/), Kaplan-Meier analysis of the survival of patients with PDAC showed that the level of GSDME methylation positively correlated with overall survival (P = 0.045) (Fig. 6d). In addition, the expression of YBX1 and MUC1 was inversely correlated with overall survival of patients (P = 0.0054 and P = 0.008, respectively) (Fig. 6e,f). Together, these results suggest that the GSDME-YBX1-mucin axis in human pancreatic cancer is crucial in protecting tumour cells from enzymatic destruction.

Discussion
Following cleavage by caspase-3, GSDME can induce tumour cell pyroptosis and enhance the antitumour immunity of T cells 15 . Thus, GSDME is considered to have a tumour-suppressing function. Consistently, GSDME is silenced in most cancer cells 11 , which can be achieved through epigenetic suppression or loss-of-function mutations in GSDME 15 . However, our previous studies found that GSDME is expressed at high levels in B leukaemia cells and mediates the cytokine release syndrome induced by chimeric antigen receptor T-cell therapies 26 . In this study, we further demonstrated that PDAC tumour cells express GSDME at high levels. GSDME appears to function as a transporter to mediate the entry of the transcription factor YBX1 to the nucleus, where it promotes the expression of MUC1 and MUC13. These membrane-associated mucins protect PDAC cells from cytolysis caused by digestive enzymes secreted by acinar and duct cells. Thus, GSDME can also play a tumour-promoting role. In addition to preventing enzymatic digestion, GSDME plays an important part in mediating the pathogenesis of PDAC. A notorious pathological feature of PDAC is its strong desmoplastic stroma, which poses a formidable obstacle for treatments 27,28 . It is unclear how the desmoplastic stroma is triggered at the beginning of PDAC formation. The oncogenic KRAS mutation is thought to be a driving force for the PDAC desmoplasia [29][30][31] . Chronic inflammation is also thought to play a critical role in the stromal formation of PDAC 32 . However, what triggers and maintains pancreatic inflammation is unclear. Our present study may provide a clue, and we propose the following scenarios: (1) KRAS-transformed pancreatic epithelial cells rapidly proliferate and obstruct the normal ducts, which leads to pancreatic juice leaking to surrounding cells; (2) digestive enzymes in the pancreatic juice are activated and cytolyse normal cells; (3) transformed cells develop an increased GSDME phenotype, which facilitates their resistance to enzymatic digestion; and (4) normal cell lysis releases damage-associated molecular patterns, which trigger innate immune responses and inflammation. Future investigation is warranted to substantiate these ideas.
As the most ancient gasdermin, the conservation of GSDME in the lancelet and even earlier metazoa implies that it probably exerts a highly specific function for cell survival and propagation 11,33 . The original identification of GSDME as DFNA5 (deafness, autosomal dominant 5) was from a family with autosomal-dominant progressive hearing loss 34 . The connection of deafness to the known pore-forming function of GSDME is difficult to reconcile. However, our findings provide some answers. GSDME may have 'moonlighting' functions that are pyroptosis-independent and cleavage-independent but involved in transcription regulation. Our cell viability assays showed that knockout of GSDME abrogated the resistance of PDAC tumour cells to pancreatic digestive enzymes (Fig. 4f,g), which suggests that GSDME is required for cell survival rather than the known pore-formation role that causes cell death. Intriguingly, the D270 site, which is required for the cleavage of GSDME by activated caspase-3 to generate the N-terminal active form for membrane pore formation 11 , is also required for GSDME binding to YBX1 (Fig. 4e,f and Extended Data Fig. 4h). Based on previous studies of GSDME and our findings here, we propose that the previously known GSDME cleavage site (D270) may biologically act as a switch to guide GSDME towards either the pore-forming pathway or the YBX-1-binding pathway, which depends on the status of caspase-3 and YBX1 in the cells. In support of this moonlighting function of GSDME, it has been reported that another gasdermin (GSDMB) is localized in the nucleus of bronchial epithelial cells related to asthma-related gene expression, which suggests that gasdermins might have transcriptional roles 35 . Providing stronger data should be useful to support this notion. Together, a comprehensive understanding of the function of GSDME is worthy of further investigation.
In summary, the data in this study show that GSDME, by virtue of its upregulation in PDAC tumour cells, mediates the transcription factor YBX1 to enter the nucleus where it promotes mucin expression, which in turn enables tumour cells to escape pancreatic enzymatic digestion (Fig. 6g). These findings shed light on potential innovative strategies to target PDAC. One is to target the GSDME-YBX1-mucin pathway to deprive PDAC cells of their resistance to enzymatic digestion; another is to motivate the pore-forming activity of GSDME to trigger tumour cell pyroptosis and to activate antitumour immune responses.