Endosomal Trafficking Bypassed by the RAB5B-CD109 Interplay Promotes Axonogenesis in KRAS-Mutant Pancreatic Cancer.

abstract cc by

AbstractPerineural invasion (PNI) represents a unique biological feature associated with poor prognosis in pancreatic ductal adenocarcinoma (PDAC), especially in the presence of KRAS mutations. Extracellular vesicle (EV)‐packaged circular RNAs (circRNAs) function as essential mediators of tumor microenvironment communication, triggering PDAC cell invasion and distant metastasis. However, the regulatory mechanisms of EV‐packaged circRNAs in the PNI of KRAS‐mutant PDAC have not yet been elucidated. Herein, a KRASG12D mutation‐responsive EV‐packaged circRNA, circPNIT, which positively correlated with PNI in PDAC patients is identified. Functionally, KRASG12D PDAC‐derived EV‐packaged circPNIT promoted axonogenesis and PNI both in vitro and in vivo. Mechanistically, the circPNIT‐mediated Rab5B‐CD109 interplay bypassed traditional endosomal trafficking to anchor Rab5B to the lipid rafts of multivesicular bodies and packaged circPNIT into CD109+ EVs. Subsequently, CD109+ EVs delivered circPNIT to neurons by binding to TRPV1 and facilitating DSCAML1 transcription‐induced axonogenesis, which in turn enhanced the PNI by activating the GFRα1/RET pathway. Importantly, circPNIT‐loaded CD109+ EVs are established to dramatically promote PNI in a KRASG12D/+ Trp53R172H/+ Pdx‐1‐Cre mouse model. Collectively, the findings highlight the mechanism underlying how EV‐packaged circRNAs mediate the PNI of KRAS‐mutant PDAC cells through the Rab5B endosomal bypass, identifying circPNIT as an effective target for the treatment of neuro‐metastatic PDAC.

introduction cc by

Pancreatic ductal adenocarcinoma (PDAC) is a digestive malignancy with a high rate of KRAS oncogenic mutation,[ 1 ] resulting in a 5‐year survival rate of less than 12%.[ 2 ] Additionally, 90% of PDAC patients with KRAS mutations exhibit extensive perineural invasion (PNI) at an early stage, resulting in a nearly 23‐month reduction in their overall survival (OS).[ 3 ] Thus, PDAC patients with KRAS mutations and PNI exhibit a highly aggressive phenotype. However, most of these patients often face increased risk owing to the lack of effective treatment options available in clinical decision‐making. Therefore, elucidating the relationship between KRAS mutations and PNI in PDAC and exploring the underlying regulatory mechanisms are of theoretical significance for improving the prognosis of PDAC. PDAC with oncogenic KRAS mutations alters the tumor microenvironment to promote tumor metastasis by releasing extracellular vesicles (EVs),[ 4 ] a diverse group of bilayer membrane structures. Inhibiting the formation, release, or uptake of EVs is anticipated to be a pivotal strategy for blocking the progression of KRAS‐mutant PDAC.[ 5 ] Among these, EVs harboring oncogenic KRASG12D ‐specific short interfering RNAs (siRNAs) demonstrated sustained tumor‐suppressive effects in an orthotopic xenograft model of PDAC, addressing the challenge of targeting KRAS mutations.[ 5a ] Thus, EVs may be crucial regulatory mediators of metastasis, including PNI, in KRAS‐mutant PDAC. Consequently, investigating the regulatory mechanism of EVs is expected to lead to the identification of effective targets for the treatment of KRAS mutation‐induced PNI in PDAC patients. Tumors deliver bioactive substances to specific target cells via EVs, thereby establishing a microenvironment conducive to the growth of disseminated cells.[ 6 ] Therefore, blocking the secretion of tumor‐derived EVs is essential for inhibiting PDAC metastasis. The formation and secretion of EVs are complex biological processes that rely on endosomal maturation regulated by Rab GTPases and membrane fusion of multivesicular bodies (MVBs), which is mediated by the endosomal sorting complex required for transport (ESCRT).[ 7 ] However, the mechanism by which KRAS‐mutant PDAC drives EV secretion has not yet been elucidated. Therefore, exploring EV secretion in KRAS‐mutant PDAC holds significant translational value, as it may reveal strategies to block PNI in PDAC patients. In this study, we elucidated for the first time that KRASG12D ‐related PDAC cells exhibit severe PNI via the secretion of tumor‐derived CD109+ EVs, which deliver circPNIT to facilitate tumor‐associated axonogenesis and PNI. Mechanistically, circPNIT acts as a molecular scaffold, linking Rab5B to its novel effector, CD109, enabling specific anchoring on lipid rafts of the MVBs membrane. This interaction triggers lipid raft‐mediated EV secretion, revealing a bypass mechanism of endosomal trafficking mediated by the Rab5B‐CD109 interplay. Notably, CD109‐engineered EVs loaded with circPNIT effectively promoted neural invasion in a KRASG12D/+ Trp53R172H/+ Pdx‐1‐Cre (KPC) mice model. Our study provides an applicable strategy for treating KRASG12D ‐mutated PDAC using PNI.

kras cc by

To investigate the correlation between KRAS mutations and the PNI in PDAC, we evaluated KRAS subtypes in 530 PDAC patients from multiple centers using Sanger sequencing (Figure S1A, Supporting Information, KRASWT : n = 71; KRASG12D : n = 266; KRASG12V : n = 144; KRASG12C : n = 49). The PNI score based on multiplex immunohistochemistry (mIHC) assays showed that the severity of PNI was significantly higher in KRASG12D PDAC patients than in those with other KRAS subtypes (KRASWT , KRASG12V , and KRASG12C ) (Figures 1A and S1B, Supporting Information). Because tumor‐associated axonogenesis, which manifests as increased nerve density,[ 8 ] represents the initial step that drives neuroplasticity and PNI, we conducted further analyses to determine whether KRAS is involved in axonogenesis in PDAC. Increased nerve density was observed in the KRASG12D PDAC cohort compared to that in the other KRAS‐mutant subtypes (Figure 1B), indicating that the KRASG12D mutation is associated with axonogenesis during PNI. We have previously revealed the specific abundant production of EVs in response to the KRASG12D mutation, which is involved in the early metastasis of PDAC.[ 4b ] Therefore, we hypothesized that KRASG12D ‐related PDAC cells trigger axonogenesis and PNI via EV transmission. To test this hypothesis, we isolated and characterized EVs from the culture media of PDAC cells with different KRAS subtypes (KRASWT : BxPC‐3; KRASG12D : PANC‐1, AsPC‐1; KRASG12C : MiaPaCa‐2; KRASG12V : Capan‐2). The isolated particle vesicles exhibited a typical double conical disc shape and ranged in size from 30 to 150 nm (Figure S2A–E, Supporting Information). Subsequently, to evaluate the effect of PDAC‐derived EVs on PNI, we found that compared with other KRAS subtypes, KRASG12D PDAC cells exhibited enhanced neurotropism rather than invasiveness (Figures 1C–E and S2F–I, Supporting Information). Subsequently, we induced PANC‐1 through EV and EV‐free supernatant from KRAS‐mutant PDAC cells, which showed that PANC‐1 induced by EV‐free supernatant from KRAS‐mutant PDAC cells did not alter PNI function (Figure S2J,K, Supporting Information). Conversely, PANC‐1 cells induced by EVs from KRASG12D PDAC exhibited enhanced neurotropism compared with EVs from other KRAS mutant PDACs (Figure 1F). Moreover, to confirm that EVs are indispensable for KRASG12D PDAC‐induced PNI, we inhibited EV secretion from KRASG12D cells by knocking out Rab27a. The results showed that the inhibition of EV secretion significantly decreased the neurotropism of PANC‐1 (Figure S2L–O, Supporting Information), indicating that EVs derived from KRASG12D PDAC cells facilitate neurotropism in these cells. To further explore the function of KRAS mutations in axonogenesis and PNI in PDAC in vivo, we established an orthotopic xenograft model via intrapancreatic injection of equivalent PDAC cells with different KRAS subtypes (Figure 1G; KRASWT : BxPC‐3; KRASG12D : PANC‐1; KRASG12C : MiaPaCa‐2; KRASG12V : Capan‐2). The PNI score indicated that KRASG12D ‐related PDAC was associated with increased nerve density and severity of PNI compared to other KRAS subtypes, whereas inhibition of EV secretion in Rab27a‐knock out (KO) mice suppressed the severity of neural infiltration in PANC‐1‐induced tumor tissues (Figures 1H–I and S2P, Supporting Information). Taken together, these results indicated that KRASG12D PDAC‐induced EV significantly promoted axonogenesis and PNI.

identification of circpnit in kras cc by

The aberrant synthesis of circular RNAs (circRNAs) has been proposed to be involved in KRAS mutation‐induced metastasis in PDAC.[ 9 ] Thus, to identify the crucial circRNAs involved in KRASG12D PDAC‐derived EV‐mediated PNI, we analyzed our previous next‐generation sequencing data (GSE234760) for three pairs of KRASG12D PDAC and normal adjacent tissues (NATs) (Figure 2A). A total of 50 circRNAs that were upregulated by more than 10‐fold were subjected to a larger cohort of 530‐case PDAC patients, and 11 circRNAs were upregulated in KRASG12D PDAC compared to those in NATs or PDAC with other KRAS subtypes (Figures 2B,C and S3A–L, Supporting Information). Further evaluation of the correlation between these 11 circRNAs and PNI grade in KRASG12D PDAC tissues revealed five circRNAs that were overexpressed in KRASG12D PDAC tissues with high‐degree PNI (high PNI) compared to those with low‐degree PNI (low PNI) (Figures 2D and S3M,N, Supporting Information). Subsequently, we analyzed the expression of these circRNAs in the serum EVs of 266‐case KRASG12D PDAC patients and paired healthy volunteers, confirming that hsa_circ_0007006 (a term for perineural invasion‐associated transcript, circPNIT) was significantly upregulated in EVs obtained from KRASG12D PDAC patients (Figures S3O–Q and S4A, Supporting Information). Therefore, circPNIT was selected for further analysis. Subsequently, we explored the molecular characteristics of circPNITs. CircPNIT is derived from the back‐splicing of exons 10 to 13 of the DYM gene, which has a length of 514 nucleotides (nt), as confirmed via Sanger sequencing (Figure S4B, Supporting Information). In addition, circPNIT was amplified using only spliced cDNA instead of the gDNA template (Figure S4C, Supporting Information). Furthermore, we amplified circPNIT from the reverse‐transcribed sequence using random primers rather than oligo‐dT primers, confirming that circPNIT has a circular structure without a poly A tail (Figure S4D, Supporting Information). Moreover, compared to DYM mRNA, circPNIT exhibited stronger resistance to RNase R, and the actinomycin D assay confirmed that circPNIT has a longer half‐life than DYM mRNA (Figure S4E–G, Supporting Information). These results demonstrated that circPNIT possesses a stable circular structure. Moreover, analysis of the pathological characteristics of KRASG12D PDAC patients revealed a positive correlation between circPNIT expression and PNI severity (Table S1, Supporting Information). circPNIT expression was significantly upregulated in KRASG12D PDAC tissues with high PNI and slightly increased in those with low PNI compared to NATs (Figure 2E). Furthermore, correlation analysis revealed that high circPNIT expression in KRASG12D PDAC tissues was significantly positively correlated with nerve density (Figure S4H, Supporting Information). Additionally, Kaplan–Meier analysis of KRASG12D PDAC patients revealed that high circPNIT expression was associated with shorter OS and disease‐free survival (DFS) (Figure S4I,J, Supporting Information). These results indicated that circPNIT was overexpressed in KRASG12D PDAC‐derived EVs and positively associated with PNI in KRASG12D PDAC patients.

kras cc by

After demonstrating the enrichment of circPNIT in EVs and its clinical correlation with axonogenesis and PNI in KRASG12D ‐related PDAC, we further investigated the functional role of EV‐packaged circPNIT in KRASG12D ‐related PDAC‐induced axonogenesis and PNI. First, we found that applying siRNAs targeting circPNIT or circPNIT overexpression plasmids to KRASG12D PDAC cells efficiently reduced or increased circPNIT expression in EVs, respectively (Figure S5A,B, Supporting Information). To verify the tumor‐nerve interaction, we constructed a Matrigel/dorsal root ganglia (DRG) model. The results showed that the migration and extension of neural axons were inhibited after culturing with EVs from circPNIT‐knockdown KRASG12D PDAC cells, and these effects were accompanied by decreased neurotropism in PDAC cells (Figures 2F and S5C, Supporting Information). Conversely, incubation with EVs from circPNIT‐overexpressing KRASG12D PDAC cells enhanced neural axon migration and extension in combination with increased neurotropism in PDAC cells compared to incubation with control EVs (Figures 2G and S5D, Supporting Information), indicating a positive relationship between axonal changes and neural invasion conferred by KRASG12D PDAC‐derived EV‐packaged circPNIT. To determine the effects of EV‐packaged circPNIT on neural axons, we cultured DRG cells with KRASG12D PDAC‐derived EVs and analyzed neuronal axonogenesis. The nerve density of DRG cells decreased after treatment with EVs from circPNIT‐knockdown KRASG12D PDAC cells, whereas culturing with EVs from circPNIT‐overexpressing KRASG12D PDAC cells increased the nerve density of DRG cells, indicating the positive effect of KRASG12D PDAC‐derived EV‐packaged circPNIT on axonogenesis (Figure 2H–K). Further validation was performed using DRG matrix assays to explore whether axonogenesis mediated by EV‐packaged circPNIT was involved in the PNI of KRASG12D ‐related PDAC (Figure 2L). The results showed that the invasion of PDAC cells into the DRG was inhibited after treatment with EVs of circPNIT‐knockdown KRASG12D PDAC cells compared to treatment with control EVs, whereas treatment with EVs from circPNIT‐overexpressing KRASG12D PDAC cells significantly enhanced the neurotropism of KRASG12D PDAC cells (Figures 2M,N and S5E,F, Supporting Information). Taken together, these results indicated that EV‐packaged circPNIT triggered PNI in KRASG12D PDAC cells by promoting axonogenesis in vitro.

ev‐packaged circpnit promotes neural metastasis of kras cc by

To explore whether EV‐packaged circPNIT is involved in the PNI of KRASG12D PDAC cells in vivo, we established an orthotopic xenograft model and treated mice with control or circPNIT‐overexpressing EVs (Figures 3A and S5G,H, Supporting Information; PANC‐1‐EVVector or PANC‐1‐EVcircPNIT). PNI score analysis showed that the severity of neural invasion within the primary tumor was significantly higher in the PANC‐1‐EVcircPNIT group than in the PANC‐1‐EVVector group (Figure 3B–D). Additionally, the density of nerves increased dramatically in PANC‐1‐EVcircPNIT‐induced primary tumors (Figure 3E,F), suggesting that EV‐packaged circPNIT is involved in axonogenesis and PNI in KRASG12D ‐related PDAC cells. To further explore the effect of EV‐packaged circPNIT‐induced neurons on the neural metastasis of KRASG12D PDAC cells, we constructed an in vivo model of neural infiltration to simulate the process of PDAC cell invasion along the nerves, which facilitated quantification of the distance PDAC cells invaded into the nerves (Figure 3G). An equal amount of PANC‐1 was injected into the sciatic nerves of the nude mice. After 1 week, the mice were randomly divided into three groups (six mice per group), and EV was injected into the sciatic nerve every 3 days (PBS, PANC‐1‐EVVector, or PANC‐1‐EVcircPNIT). The nerve function score confirmed that nude mice in the PANC‐1‐EVcircPNIT group showed significant dysfunction in the affected hind limb compared to the PBS and PANC‐1‐EVVector groups (Figure 3H). Moreover, magnetic resonance imaging (MRI) and HE analyses confirmed that PANC‐1‐EVcircPNIT enhanced the capacity of KRASG12D PDAC to metastasize from the injection site along the sciatic nerve compared to the PBS and PANC‐1‐EVVector groups (Figure 3I–K). Furthermore, the volume of the invaded sciatic nerve was significantly higher in the PANC‐1‐EVcircPNIT group than that in the PBS and PANC‐1‐EVVector groups (Figure 3L,M). Taken together, these results suggested that EV‐packaged circPNIT facilitates axonogenesis and PNI in KRASG12D ‐mutant PDAC in vivo.

circpnit binds to rab5b and cd109 to form a ternary complex cc by

To explore the mechanism underlying the package of circPNIT into KRASG12D PDAC cell‐derived EVs to induce PNI, we first analyzed the subcellular location of circPNIT in KRASG12D PDAC cells by performing fluorescence in situ hybridization assays, which showed that circPNIT was mainly localized in the cytoplasm of KRASG12D PDAC cells (Figure S6A,B, Supporting Information). Cytoplasmic circRNAs predominantly serve as miRNA sponges to suppress the AGO2‐dependent miRNA‐mediated degradation of downstream mRNAs and exhibit biological functions.[ 10 ] Nevertheless, we observed that altering the expression of circPNIT in AGO2‐silenced KRASG12D PDAC cells affected the expression of circPNIT in the corresponding EVs, which was consistent with the results in KRASG12D PDAC cells without the AGO2‐KO (Figure S6C, Supporting Information). Moreover, circPNIT overexpression in EVs derived from AGO2‐silenced KRASG12D PDAC cells facilitated axonogenesis (Figure S6D,E, Supporting Information), suggesting that circPNIT acts as an miRNA sponge to regulate its loading into KRASG12D PDAC‐derived EVs and subsequently induces axonogenesis. As the interaction of circRNAs with proteins is another important regulatory mechanism, we further investigated the binding partners of circPNIT in KRASG12D PDAC. RNA pull‐down assays with biotinylated circPNIT revealed two prominent bands at 130–180 kDa and 20–35 kDa (Figure 4A), which were confirmed to be CD109 and Rab5B, respectively, through mass spectrometry and western blot analysis (Figures 4B,C and S6F, Supporting Information). Fluorescence staining revealed the colocalization of circPNIT, CD109, and Rab5B in the cytoplasm of KRASG12D PDAC cells (Figure 4E). Consistently, RNA immunoprecipitation (RIP) assays showed that circPNIT was enriched with CD109 and Rab5B (Figures 4D and S6G, Supporting Information), confirming that circPNIT interacts with CD109 and Rab5B, respectively. To further identify the binding sites of CD109 and Rab5B on circPNIT, sequential deletion assays were performed with truncated circPNIT sequences, which revealed that both CD109 and Rab5B were enriched in the 401–514 nt region of circPNIT (Figure 4F). Furthermore, sequence analysis using CatRapid predicted the potential binding site of CD109 in the 451–502 nt region of circPNIT and the potential Rab5B binding site in the 426–477 nt region of circPNIT (Figure 4G). Mutation of these two sites in circPNIT significantly reduced circPNIT enrichment by CD109 and Rab5B (Figures 4H,I and S6H, Supporting Information), indicating that these specific sequences are essential for circPNIT interaction with CD109 and Rab5B. Considering the overlap in the binding sites of CD109 and Rab5B on circPNIT, we explored whether circPNIT was involved in the interaction between CD109 and Rab5B (Figure 4J). Co‐immunoprecipitation (co‐IP) assays showed that circPNIT knockdown significantly inhibited the interaction between CD109 and Rab5B, whereas circPNIT overexpression promoted CD109 binding to Rab5B in KRASG12D PDAC cells (Figures 4K,L and S6I,J and S6L–N, Supporting Information). Moreover, fluorescence staining revealed the colocalization of CD109 and Rab5B in the cytoplasm of KRASG12D PDAC cells, which was abolished by circPNIT knockdown (Figure 4M). Importantly, mutation of the 426–502 nt region in circPNIT impaired the interaction between CD109 and Rab5B (Figures 4N and S6K, Supporting Information), indicating that binding with circPNIT is crucial for the CD109‐Rab5B interaction. Taken together, these results indicated that circPNIT formed a stable ternary complex by linking Rab5B and CD109 in KRASG12D PDAC cells.

circpnit/cd109/rab5b mediates the packaging of circpnit into evs cc by

Given that Rab proteins regulate vesicle targeting, docking, and fusion by binding to specific effectors,[ 11 ] we explored whether CD109 functions as an effector of Rab5B via circPNIT (Figure 4O). Immunofluorescence assays showed that blocking CD109 decreased the localization of Rab5B in CD63‐labeled MVBs (Figure 4P) and promoted the localization of Rab5B to early endosomes (EEA1‐positive), late endosomes (Rab7‐positive), and lysosomes (LAMP2‐positive), whereas circPNIT overexpression contributed to Rab5B localization in MVBs (Figure S7A,B, Supporting Information). Moreover, knockdown of CD109 impaired the localization of Rab5B and CD63 induced by circPNIT overexpression (Figure S7C, Supporting Information), indicating that circPNIT‐linked CD109 promoted the localization of Rab5B to MVBs. To clarify the binding of Rab5B to CD109, we constructed dominant active (Rab5BQ79L) and negative (Rab5BD136H) variants of Rab5B and verified the combination of Rab5B and CD109 in the active and negative states.[ 12 ] The results showed a higher enrichment of CD109 in Rab5BQ79L than in Rab5BD136H (Figure S6O, Supporting Information). Conversely, si‐circPNIT effectively reduced CD109 enrichment induced by Rab5BQ79L (Figure S6P, Supporting Information), indicating that Rab5B binding to CD109 occurs in a RabGTP‐activated state and is regulated by circPNIT. Considering that CD109 is a glycosylphosphatidylinositol‐linked glycoprotein that is mainly enriched in lipid rafts,[ 13 ] we investigated whether lipid rafts are required for CD109‐mediated Rab5B localization in MVBs. Disrupting lipid rafts structure using methyl‐β‐cyclodextrin[ 14 ] reversed the colocalization of Rab5B and CD63 induced by CD109 overexpression (Figure S7D, Supporting Information). Flotillin 1 (FLOT1), the core constitutive protein in lipid rafts,[ 15 ] was further examined to determine CD109‐mediated Rab5B localization in MVB lipid rafts. Co‐IP assays showed that knockdown of CD109 significantly inhibited the ability of Rab5B to recognize FLOT1 (Figures 4Q and S7E, Supporting Information), whereas overexpression of CD109 promoted the interaction between Rab5B and FLOT1, which was reversed by circPNIT knockdown (Figures 4R and S7F, Supporting Information). Furthermore, disruption of the colocalization between Rab5B and FLOT1‐tagged lipid rafts was observed in KRASG12D PDAC cells treated with CD109 or circPNIT knockdown (Figure 4S,T). These findings demonstrated that circPNIT‐linked CD109 mediated the translocation of Rab5B to MVBs through interactions with lipid rafts. Next, we investigated whether the localization of Rab5B to MVBs contributes to the packaging of circPNIT into EVs. The results showed that Rab5B knockdown inhibited circPNIT expression in EVs, whereas Rab5B overexpression promoted circPNIT enrichment in EVs (Figures 4U and S7G, Supporting Information). Inhibiting lipid rafts by knocking down FLOT1 or MβCD significantly decreased the expression of circPNIT in EVs (Figure 4V), confirming that Rab5B mediated the loading of circPNIT into EVs in a lipid raft‐dependent manner. Considering that formation of the circPNIT/CD109/Rab5B complex triggers trafficking to MVBs and subsequent EV production, we further evaluated whether CD109 is also located in EVs. The inhibition of Rab5B significantly reduced the expression of CD109 in EVs (Figure 4W), whereas Rab5B overexpression promoted CD109 in EVs (Figure S7H, Supporting Information). In addition, blockade of lipid rafts by FLOT1 knockdown and MβCD reduced CD109 secretion in EVs (Figure 4X). Collectively, these results indicated that the circPNIT/CD109/Rab5B complex mediated the packaging of circPNIT and CD109 into EVs.

ev‐packaged circpnit specifically targets neurons via cd109 cc by

To evaluate the recipient cells of EV‐packaged circPNIT derived from KRASG12D PDAC tumors in the tumor microenvironment (TME), EVs were labeled with PKH67 and co‐cultured with different stromal cells. Flow cytometry confirmed that compared to other stromal cells, PKH67‐labeled EVs were mainly taken up by the DRG in the TME (Figures 5A and S7I, Supporting Information). However, circPNIT‐KO in KRASG12D PDAC cells abolished the internalization of the corresponding EVs in the DRG (Figures 5B and S8A–C, Supporting Information). Accordingly, the fluorescence intensity of DRG cells incubated with PKH67‐labeled EVs was significantly higher than that of other stromal cells, and the internalization of EVs in DRG cells was significantly decreased in KRASG12D PDAC with circPNIT‐KO (Figures 5B and S8D, Supporting Information). EVs’ surface proteins serve as recognition signals, enabling their targeted transmission to specific cells.[ 5 , 16 ] Herein, we found that CD109 was located on the surface of KRASG12D PDAC‐derived EVs, whereas Rab5B overexpression promoted the secretion of CD109+ EVs (Figure 5C). Moreover, using immunoelectron microscopy, we observed that circPNIT downregulation suppressed the enrichment of CD109 on the membrane of EVs derived from KRASG12D PDAC cells (Figure 5D). This impaired the ability of these EVs to induce circPNIT expression in DRG cells. In contrast, PANC‐1‐EVcircPNIT and EVs derived from circPNIT‐overexpressing AsPC‐1 cells (AsPC‐1‐EVcircPNIT) markedly upregulated circPNIT expression in DRG cells (Figure 5E,F). To confirm whether CD109 is involved in the targeting of KRASG12D PDAC‐derived EVs to DRG cells, we blocked CD109 with a neutralizing antibody and demonstrated that the overexpression of circPNIT promoted the internalization of KRASG12D PDAC‐derived EVs and their packaged circPNIT by DRG cells, whereas the blockade of CD109 significantly inhibited the delivery of EV‐packaged circPNIT to DRG cells (Figures 5G–K and S8E, Supporting Information). Next, we verified neural targeting of CD109 in an orthotopic xenograft model in vivo. Similarly, compared to the PANC‐1‐EVvector, PANC‐1‐EVcircPNIT was significantly enriched in the nerves of primary tumor tissue, and the CD109 neutralizing antibody effectively inhibited the accumulation of EV in nerves (Figure 5L,M), indicating that CD109 mediated the targeted delivery of KRASG12D PDAC‐derived EVs to neurons. As the targeting specificity of EVs to cells is achieved by the specific binding of EV surface proteins to cell membrane receptors,[ 17 ] we further explored the receptors recognized by CD109 on the surface of DRG cells. Confocal microscopy confirmed that the GST‐labeled recombinant CD109 protein was localized on the surface of the DRG cell membrane (Figure 5N). GST pull‐down assays confirmed an evident band in the 71–117 kDa range, which was identified as TRPV1 (Figure 5O,P). Co‐IP assays confirmed the interaction between CD109 and TRPV1 (Figure 5Q). Moreover, CD109 and TRPV1 co‐localized on the membrane of DRG cells, while TRPV1 knockdown in DRG cells inhibited the internalization of KRASG12D PDAC‐derived EVs (Figure 5R,S and S9A, Supporting Information). Taken together, these results demonstrated that KRASG12D PDAC‐derived CD109+ EVs achieved targeted delivery of circPNIT to neurons by recognizing TRPV1 on the membrane surface.

ev‐packaged circpnit upregulates dscaml1 expression in neurons to promote axonog cc by

We further explored the mechanism by which KRASG12D PDAC EV‐packaged circPNIT regulates axonogenesis after internalization into DRG cells. NGS analysis was performed on three pairs of independent DRG cells induced by PANC‐1‐EVcircPNIT or PANC‐1‐EVVector. Gene Ontology analysis revealed enrichment of the axonogenesis pathway in DRG cells after PANC‐1‐EVcircPNIT incubation, and 32 genes that were differentially expressed in these pathways were enriched (Figure 6A,B and S8F,G, Supporting Information). Subsequently, using qRT‐PCR, we found that DSCAML1 was the most upregulated gene in EV‐packaged circPNITs (Figure 6C,D and S8H,I, Supporting Information). DSCAML1 is a crucial regulator of neural development that participates in the growth and branching of neuronal axons.[ 18 ] To further explore the mechanisms of EV‐packaged circPNIT‐induced DSCAML1 overexpression in DRG cells, a series of truncated sequences of DSCAML1 promoters spanning from ‐2000 to +200 bp relative to the transcriptional start site were cloned and inserted into luciferase reporter genes. Dual‐luciferase assays revealed that EV‐packaged circPNIT promoted transcriptional activity when DRG cells were transfected with the ‐800 to ‐1200 bp region of the DSCAML1 promoter (Figures 6E and S8J, Supporting Information). Moreover, Chromatin Isolation by RNA Purification assays confirmed that EV‐packaged circPNIT directly interacted with ‐1063 to ‐1073 bp of the DSCAML1 promoter (referred to as DSCAML1‐P3) in DRG cells (Figures 6F,G and S8K, Supporting Information). Mutation of the DSCAML1‐P3 region reduced the luciferase activity induced by EV‐packaged circPNIT (Figures 6H and S8L, Supporting Information), indicating that DSCAML1‐P3 was crucial for EV‐packaged circPNIT‐induced DSCAML1 upregulation in DRG cells. Specific patterns of histone modification in promoter regions are hallmarks of transcriptional activation.[ 19 ] We found that the enrichment of H3K9Ac at the DSCAML1 promoter was strongly associated with EV‐packaged circPNIT expression (Figures 6I,J and S8M,N, Supporting Information), indicating that EV‐packaged circPNIT increased H3K9Ac levels at the DSCAML1 promoter to drive its transcriptional activation. Next, we assessed whether DSCAML1 was indispensable for EV‐packaged circPNIT‐induced axonogenesis. The Matrigel/DRG model revealed that EV‐packaged circPNIT overexpression enhanced the migration and extension of neural axons toward PDAC cells, which was impaired by the downregulation of DSCAML1 (Figure 6K,L). Moreover, incubation with EV‐packaged circPNIT alone promoted axonal outgrowth of neurons, whereas blocking DSCAML1 reversed EV‐packaged circPNIT‐induced axonogenesis (Figure 6M,N). Collectively, these results suggested that EV‐packaged circPNIT promoted the transcriptional activation of DSCAML1 by recruiting H3K9Ac to the ‐1063 to ‐1073 bp region of DSCAML1, thereby promoting axonogenesis in KRASG12D PDAC cells.

ev‐packaged circpnit‐induced axonogenesis facilitates pni by activating gfrα1/re cc by

Next, we explored how EV‐packaged circPNIT‐induced axonogenesis promotes PNI in KRASG12D PDAC tumors. Extension of axons is reportedly accompanied by the expression of receptors that promote cancer‐neuron crosstalk to achieve local invasion and metastasis.[ 3 , 20 ] Thus, the expression levels of core molecules involved in PNI‐related pathways in neurons were evaluated, and the results showed that the expression of only GFRα1 and RET was positively correlated with the expression of EV‐packaged circPNIT in DRG cells (Figure S8O, Supporting Information). Given that the GFRα1/RET complex is considered the major pathway leading to neural tracking and adhesion via downstream MAPK/ERK signaling during PNI,[ 20 , 21 ] we further assessed the activation of the GFRα1/RET pathway. Western blot assays showed that knockdown of EV‐packaged circPNIT significantly decreased GFRα1 and RET expression and reduced the phosphorylation of c‐RAF, MEK1/2, and ERK1/2, while overexpressing EV‐packaged circPNIT significantly promoted GFRα1 and RET expression and c‐RAF, MEK1/2, and ERK1/2 phosphorylation (Figures 6O,P and S8P,Q, Supporting Information). Moreover, downregulation of DSCAML1 reversed the increase in GFRα1 and RET expression and MAPK/ERK pathway activation induced by EV‐packaged circPNIT (Figures 6P and S8Q, Supporting Information), suggesting that EV‐packaged circPNIT activated the GFRα1/RET pathway in DRG cells by triggering DSCAML1‐induced axonogenesis. To further elucidate whether GFRα1/RET pathway activation is essential for EV‐packaged circPNIT‐induced PNI in KRASG12D PDAC, transwell assays were performed, which revealed that DRG cells precultured with EV‐packaged circPNIT significantly promoted the invasion of KRASG12D PDAC cells into DRG cells, while treatment with the RET inhibitor AST‐487 reversed this effect (Figure 6Q). Moreover, a DRG‐matrix assay further confirmed that overexpression of EV‐packaged circPNIT promoted the neurotrophic ability of KRASG12D PDAC cells, whereas AST‐487 inhibited EV‐packaged circPNIT‐induced neurotropism in KRASG12D PDAC cells (Figure 6R). In conclusion, EV‐packaged circPNIT‐induced axonogenesis promotes the PNI of KRASG12D PDAC by activating the GFRα1/RET pathway.

cd109 cc by

Given that KRASG12D PDAC‐derived CD109+EVs deliver circPNIT to neurons and trigger axonogenesis‐induced PNI, we sought to determine the targeted engagement of circPNIT in the primary tumor tissue generated by KRASG12D PDAC. To achieve the targeted engagement of circPNIT in neurons, we constructed engineered EVs to simulate neuron‐targeted CD109+EVs derived from KRASG12D PDAC cells, which were confirmed to be loaded with CD109 on the surface through nanoflow cytometry and western blot analyses (Figure 7A,B). Moreover, fluorescence staining confirmed that engineered EVcircPNIT enhanced spot fluorescence in nerves in an orthotopic xenograft model compared to that in control EV and engineered EVEmpty (Figure 7C,D), suggesting that engineered EVcircPNIT has significant neuronal targeting and delivers circPNIT to nerves in the primary tumor tissue. The PNI score showed that engineered EV containing circPNIT, compared with control EV and engineered EVEmpty, promoted the severity of PNI and nerve density in the primary tumor tissue constructed by PANC‐1 (Figure 7E). Meanwhile, the survival time of engineered EVcircPNIT‐treated nude mice was significantly shorter than those of the control and engineered EVEmpty (Figure 7F). Further validation was performed using the KPC model to explore the PNI effect of engineered EV‐packaged circPNIT in KRASG12D mutation‐induced tumor tissues. Once the primary tumor was detected, PANC‐1‐EVVector, PANC‐1‐EVcircPNIT, engineered EVEmpty, and engineered EVcircPNIT were injected into the pancreas every 3 days for 6 weeks (Figure 7G). The PNI score showed that PANC‐1‐EV‐loaded circPNIT or engineered EV‐packaged circPNIT exacerbated the severity of PNI in mouse tumor tissue compared to that in the control group, indicating that circPNIT triggers KRASG12D PDAC‐derived EV to induce PNI (Figure 7H–K). Conversely, engineered EV‐loaded si‐circPNIT (loading efficiency: 3.06% ± 0.48%) significantly reduced PNI severity and nerve density in comparison to those of control EV and engineered EVsi‐NC (Figure S9B–E, Supporting Information). Moreover, survival time after EV treatment was longer in the engineered EVsi‐circPNIT group than in the control group (Figure S9F, Supporting Information). Taken together, these results indicate that KRASG12D PDAC‐derived CD109+EV promote PNI in KRASG12D mutant tumor tissues by transferring circPNIT.

clinical relevance of ev‐packaged circpnit in kras cc by

We further evaluated the clinical relevance of EV‐packaged circPNIT in KRASG12D PDAC. The expression of serum EV circPNIT in KRASG12D PDAC patients with high PNI was significantly higher than that in patients with low PNI, indicating a positive correlation between serum EV circPNIT and the PNI of KRASG12D PDAC (Figure 7L). Moreover, Kaplan–Meier analysis revealed that KRASG12D PDAC patients with higher serum EV circPNIT expression had shorter OS and DFS rates (Figure 7M,N). Univariate and multivariate analyses revealed that higher circPNIT expression was an independent predictor of poor survival of KRASG12D PDAC patients (Tables S2 and S3, Supporting Information). Furthermore, IHC analysis revealed that patients with upregulated serum EV circPNIT expression had elevated DSCAML1 levels, which were associated with increased PNI severity (Figure 7O–Q). In conclusion, our results indicated that serum EV circPNIT has potential diagnostic and predictive value for neuroinvasive KRASG12D PDAC (Figure 7R).

discussion cc by

PDAC with KRAS mutations is accompanied by nerve infiltration in the early stages of tumor progression, which is a core driving force for postoperative recurrence and poor prognosis.[ 22 ] However, the mechanisms underlying the PNI in KRAS‐mutant PDAC remain unclear. Herein, we reported for the first time that the KRASG12D mutation was positively associated with a high PNI in PDAC patients via a multicenter analysis. We identified KRASG12D PDAC derived CD109+ EVs that transmit circPNIT to foster tumor‐associated axonogenesis and PNI, both in vitro and in vivo. In KRASG12D PDAC, circPNIT formed a ternary complex by binding to Rab5B and its effector—CD109. Subsequently, CD109 directed Rab5B anchoring to lipid rafts on MVBs, bypassing the traditional endosomal trafficking pathway and facilitating circPNIT secretion as a cargo of CD109+ EVs in a lipid raft‐dependent manner. Moreover, we showed that KRASG12D PDAC‐derived CD109+ EVs delivered circPNIT to neurons by recognizing TRPV1 on the neuronal membrane, thus triggering axonogenesis induced by DSCAML1 activation. Next, neurons accompanied by axonogenesis expressed high levels of GFRα1/RET, which in turn enhanced the neurotropism of PDAC. Our study highlighted the mechanism bridging KRAS mutation‐mediated EVs and PNI and identified CD109+ EV‐packaged circPNIT as an effective target for early intervention in PDAC patients with PNI. Although progress has been made in the diagnosis and treatment of PDAC, PNI remains the leading cause of disease recurrence and poor survival among these patients.[ 23 ] Surgical strategies based on nerve ablation have a profound impact on controlling tumor progression, including delaying tumor recurrence and metastasis.[ 24 ] Therefore, exploring the regulatory mechanism of PNI in PDAC has great clinical value for improving the prognosis of KRASG12D PDAC. In this study, we identified tumor‐associated axonogenesis as the key determinant of PNI in KRASG12D PDAC and demonstrated that axonogenesis is driven by the transcriptional activation of DSCAML1 in neurons, which is regulated by EV‐packaged circPNIT from KRASG12D . Moreover, DSCAML1‐induced axonogenesis of neurons was accompanied by elevated expression of GFRα1/RET, the critical receptor mediating PNI. Moreover, blockage of DSCAML1 effectively suppressed KRASG12D PDAC‐derived EV‐packaged circPNIT‐induced axonogenesis and PNI, implying the potential clinical utility of DSCAML1 as a therapeutic target for preventing neural invasion of PDAC. Traditionally, the production of EVs relies on multistep transport through the endosomal system via Rab GTPases and ESCRT complexes, which are highly energy‐intensive processes.[ 25 ] Dysfunction of the Rab GTPase‐mediated endosomal trafficking pathway leads to altered endocytosis and secretion of EVs.[ 26 ] However, the regulatory mechanisms driving EV secretion in KRASG12D PDAC have not yet been elucidated. Herein, we demonstrated an alternative mechanism that bypasses the traditional endosomal trafficking pathway involving Rab5B, thereby providing an efficient and energy‐saving pathway for EVs production in KRASG12D PDAC. By specifically recognizing effectors, Rab5B directly localized to lipid rafts on MVBs under the guidance of CD109, thereby bypassing the conventional trafficking pathways of early and late endosomes. Our findings reveal a novel mechanism of EV production involving Rab5B in KRASG12D PDAC that bypasses traditional endosomal transport, expanding our knowledge of the regulatory process by which PDAC cells adopt a more energy‐efficient mode of metastasis. EVs have shown extensive application prospects in tumor therapy.[ 27 ] Engineered EVs, which are essential carriers for tumor therapy, have shown satisfactory efficacy in treating a variety of tumors.[ 28 ] This study showed that CD109+ EVs exhibited strong neuro‐targeting potential and are expected to constitute a breakthrough in the development of targeted therapies for neuroinvasive PDAC. Notably, CD109+ EVs harboring si‐circPNIT effectively inhibited axonogenesis and PNI in mouse models of spontaneous tumorigenesis, thus providing a feasible strategy for PNI treatment in KRASG12D PDAC. Radical surgery in conjunction with chemotherapy is the prevailing therapeutic approach for PDAC. However, chemotherapy resistance is an unavoidable consequence of prolonged treatment in PDAC patients and is highly correlated with the PNI. This therapeutic strategy of targeting nerves is expected to improve chemoresistance and provide an adjuvant therapeutic option for comprehensive treatment of pancreatic cancer. Our study highlighted that CD109+EVs are essential for targeting the PNI of KRASG12D PDAC, providing a modifiable tool for treating PDAC nerve infiltration. Currently, validated assessment measures for PNI diagnosis in PDAC patients are lacking. Imaging modalities, such as CT and MR, cannot detect the occurrence of PNI. Therefore, the diagnosis of nerve metastases in PDAC patients remains challenging. In recent years, serum EVs have been widely used as an important measure in liquid biopsy for dynamic tumor detection and response assessment.[ 29 ] In this study, we found that circPNIT was upregulated in the serum EVs of PDAC patients with severe PNI and that the detection of serum EV‐circPNIT can effectively differentiate PDAC patients with PNI, thus filling the gap in the diagnosis of PDAC neuroinfiltration. In summary, this study revealed a bypass mechanism via the circPNIT/Rab5B/CD109 complex, distinct from conventional endosomal trafficking, highlighting the critical role of KRASG12D PDAC‐derived EV‐packaged circPNIT in PNI. Moreover, we propose that CD109+ EVs, which effectively target neurons and promote the occurrence of PNI in KRASG12D PDAC via circPNIT delivery, are attractive drug delivery carriers for the treatment of PNI.

patients and specimens cc by

Data were collected from 530 PDAC patients who underwent pancreatic surgery at Guangdong Provincial People's Hospital, Sun Yat‐sen Memorial Hospital, and Sixth Affiliated Hospital, Sun Yat‐sen University. All patients were fully informed about and consented to participate in the study. PNI was diagnosed by two independent pathologists in a blinded manner. Collected PDAC tissues and paired NATs were stored in −80 °C for RNA or protein extraction and in formalin for mIHC), immunohistochemistry IHC, and in situ hybridization (ISH). Corresponding peripheral serum samples from PDAC patients and matched healthy volunteers were collected for subsequent experiments. The PNI score was used to assess the severity, distribution, and frequency of PNI. Based on the location of PNI occurrence, it was categorized as 0‐no PNI 1‐PNI in the peritumoral region and 2‐PNI in the intratumoral region. Based on the frequency of PNI, it can be classified as 0‐no PNI, 1‐low, 2‐frequent, and 3‐excessive. The two scores were multiplied to calculate the severity of PNI in PDAC. The score < 4 was considered the “low PNI” group and the score ≥ 4 was considered the “high PNI” group.[ 8 , 30 ] Nerve density was quantified as described previously.[ 31 ]

cell lines and cell culture cc by

Human PDAC cell lines (KRASG12D : PANC‐1, AsPC‐1, KRASG12V : Capan‐2, KRASG12C : MiaPaCa‐2, KRASWT : BxPC‐3) and Human Umbilical Vein Endothelial Cells (HUVECs) were purchased from the American Type Culture Collection (ATCC, USA). Primary cancer‐associated fibroblasts (CAFs) and tumor‐associated macrophages (TAMs) were isolated from pancreatic cancer tissues using a Human Tumor Dissociation Kit (130‐095‐929, Miltenyi Biotec, Germany) following the manufacturer's instructions. T and B cells were isolated from the peripheral blood of healthy volunteers and cultured in RPMI 1640 medium (Invitrogen, USA). Human lymphatic endothelial cells (HLECs) were obtained from ScienCell Research Laboratories (Carlsbad, CA, USA, Cat#2500). PANC‐1, HUVECs, CAFs, and MiaPaCa‐2 cells were cultured in DMEM (Invitrogen, USA). Moreover, AsPC‐1, TAMs, and BxPC‐3 cells were cultured in RPMI 1640 (Invitrogen, USA). HLECs were cultured in an endothelial cell medium (ECM; ScienCell Research Laboratories, Cat#1001). Capan‐2 cells were cultured in McCoy's 5A medium (Invitrogen). DRG were cultured in DMEM supplemented with 10% FBS. All cells were cultured in a humidified incubator with 5% CO2 at 37 °C, and the authentication and mycoplasma testing were qualified through STR DNA profiling.

matrigel/drg model cc by

As previously reported, DRG explants and tumor co‐culture experiments were used to evaluate the paracrine effects of tumor cells on neurons.[ 32 ] Nude mice (female, 4‐weeks‐old, 18–20 g) were purchased from Guangdong Experimental Animal Center. DRG or 1 × 105 DRG cells were isolated from nude mice and seeded in six‐well plates containing 10 µL of Matrigel (BD Biosciences, NY, USA). Moreover, 1 × 105 PDAC cells were resuspended in 10 µL of Matrigel, seeded 2 mm from the edge of the DRG, and cultured in DMEM with 10% FBS at 37 °C in 5% CO2. Next, 20 µg of EVs was added to the culture medium. Cancer cell and DRG migration were evaluated 7 days later, as previously reported.[ 32a ] The distance between PDAC cells and the DRG (c), the migration of PDAC cells (α), and the migration of DRG cells (β) were defined. The invasion index (α/c) and outgrowth index (β/c) were used to quantify the neurotropism of PDAC cells and nerve outgrowth, respectively.

drg‐matrix assay cc by

Nude mice (female, 4‐weeks‐old, 18–20 g) were purchased from Guangdong Experimental Animal Center. The DRG matrix was established as previously described.[ 33 ] The DRG of the nude mice was isolated and washed with pre‐cooled PBS. The cells were fixed in six‐well plates containing 2 µL of growth factor‐reducing Matrigel (BD Biosciences, NY, USA). After 3 days of culture in 5% CO2, 1 × 105 GFP‐tagged PDAC cells were added, and tumor cell invasion into the DRG was observed 2–3 days later, as previously reported.[ 34 ] The fluorescent signals of the tumor cells invading the Matrigel were quantified using ImageJ (ImageJ, RRID:SCR_003070).

orthotopic xenograft model cc by

Nude mice (female, 4‐weeks‐old, 18–20 g) were purchased from Guangdong Experimental Animal Center. The mice were anesthetized via pentobarbital inhalation and placed in the right lateral decubitus position. The left ventrolateral side was incised to expose the pancreas. Subsequently, 1 × 106 PDAC cells were injected into the pancreas of each mouse. Six weeks later, the primary tumors were collected. HE and mIHC analyses were performed to assess nerve density and severity of PNI within the tumor tissues.

in vivo model of neural infiltration cc by

Nude mice (female, 4‐weeks‐old, 18–20 g) were purchased from Guangdong Experimental Animal Center. GFP‐labelled PDAC cells (1 × 105) were injected into the sciatic nerve. One week later, the mice were randomly divided into different groups (n = 6) and injected with PBS or 50 µg of EVs into the right sciatic nerve for every 3 days. Sciatic nerve function was monitored weekly after injection. Six weeks later, MRI was performed, and the sciatic nerves were collected for HE staining. Seven days after cell inoculation, sciatic nerve function was assessed using a neurological function score[ 35 ] and graded from 5 (normal) to 0 (complete paralysis) based on the hind limb response to active body extension. The grading system was as follows: 5: normal, symmetrical hind limbs, response to hand movement to pull the body; 4: transient abnormal response of the right hind limb to manual traction; 3: frequent abnormal response to hand movement and traction of the right hind limb; 2: persistent abnormal response of the right hind limb to hand movement and traction; 1: significant paralysis of the right hind limb; 0: inability to bear weight on the hind limbs or tumor diameter ≥ 20 mm.

genetically engineered models cc by

KRASG12D/+ Trp53R172H/+ Pdx‐1‐Cre (KPC) mice were purchased from Shanghai Model Biology. Pancreatic cancer formation in KPC mice was monitored weekly using MRI. Once tumors were detected, the mice were randomly divided into different groups (n = 12), followed by intratumoral injection with 50 µg EVs every 3 days. The primary tumor internal nerves were analyzed via HE and IHC staining at the end of the experiment.

construction of engineered ev cc by

The CD109 and LAMP2B CDSs were digested with BamHI and EcoRI and subsequently purified on a gel. The CDS of CD109‐LAMP2B was ligated into the eukaryotic expression vector pCDNA3.1(+)‐GNSTM‐HA using T4 DNA ligase (Vazyme, Nanjing, China). Next, 293T cells were transfected with engineered EVs. At 72 h after cell transfection, the cell supernatant was collected and EVs were collected through ultracentrifugation for subsequent experiments.

statistical analyses cc by

All the quantitative data were calculated as the means ± SD of at least three independent experiments. Kaplan–Meier analysis was performed to calculate the OS and DFS. Independent prognostic factors were determined using multivariate Cox proportional hazard models. Statistical analyses were performed based on the nature of the variables, assumptions of data distribution, and efficacy. Two‐tailed Student's t‐test or one‐way ANOVA was performed to evaluate parametric variables. Chi‐square test or Fisher's exact test was performed to evaluate nonparametric variables. The threshold for statistical significance was set at P < 0.05. All the statistical tests were performed with GraphPad Prism 9 (version 9.4.1; GraphPad Software, USA) and SPSS (version 26.0.0; IBM SPSS Statistics, USA).

study approval cc by

All 530 PDAC tissues and NAT samples were obtained from patients (311 males and 219 females) who underwent pancreatic surgery at Guangdong Provincial People's Hospital, Sun Yat‐sen Memorial Hospital, and The Sixth Affiliated Hospital of Sun Yat‐sen University. This study was approved by the Committees for Ethical Review of Research involving Guangdong Provincial People's Hospital, Sun Yat‐sen Memorial Hospital, and the Sixth Affiliated Hospital of Sun Yat‐sen University (approval no. 2016 (136), approval no. KY‐D‐2021‐127‐01, approval no. 2022ZSLYEC‐295). The animal study was approved by the Guangdong Provincial People's Hospital (approval no. KY‐Z‐2021‐211‐02).

conflict of interest cc by

The authors declare no conflict of interest.

author contributions cc by

D.Z., Y.Lu, Y.Li, and Z.F. contributed equally to this work. C.C., R.C., and Y.L. participated in the study design. D.Z., Y.L., Z.F., and M.Y. performed in vitro and in vivo experiments. Z.F. and Q.X. performed clinical data analysis. M.A., H.Z., and C.Y. performed immunofluorescence and mIHC experiments. Z.W. and J.Y. interpreted data and revised the manuscript. D.Z. and Y.L. wrote the manuscript. All the authors have read and approved the final version of the manuscript.