Matrix stiffness induces Ca²⁺-DCLK1-PIP5K1A mechanotransduction as a context-specific amplifier in pancreatic cancer progression and chemotherapy resistance

Abstract

Alterations in extracellular matrix (ECM) architecture and stiffness are hallmarks of aggressive pancreatic cancer progression. However, the mechanisms by which ECM biomechanical properties regulate malignant biological behavior remain unknown. Here, we reveal that calmodulin-dependent protein kinase doublecortin-like kinase 1 (DCLK1) integrates biomechanical signaling and promotes pancreatic cancer cell progression. DCLK1 expression and activation are selectively induced under conditions of high biomechanical stress mediated through the piezo-type mechanosensitive ion channel component 1 (PIEZO1)/calcium/hippocalcin-like protein 1 (HPCAL1) pathway. Consistently, in solid tumor experiments, DCLK1 overexpression under low stiffness conditions facilitates rapid tumor progression and chemoresistance, whereas using calcium inhibitors can partially reverse the adverse effects of DCLK1 overexpression. Conversely, under high stiffness conditions, DCLK1 knockdown inhibits tumor growth and enhances chemosensitivity but attenuates the sensitizing effect of combined calcium inhibitor treatment on chemotherapy efficacy. Mechanistically, DCLK1 interacts with phosphatidylinositol-4-phosphate 5-kinase type 1 alpha (PIP5K1A) by inhibiting its threonine phosphorylation, thereby facilitating PIP5K1A membrane localization. This activates the downstream phosphatidylinositol 3-kinase–protein kinase B (PI3K-AKT) signaling pathway, promoting cancer cell proliferation and chemoresistance. Collectively, these findings establish DCLK1 functions as a context-specific amplifier, exacerbating aggressive tumor progression and chemotherapy resistance in pancreatic cancer. Targeting the calcium/DCLK1 signaling axis may therefore enhance the efficacy of adjuvant therapies in pancreatic cancer.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40164-026-00756-6.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive malignancy with a dismal prognosis. It has a 5-year survival rate of 12% [1]. PDAC tumor microenvironment (TME) comprises an abundant extracellular matrix (ECM) and diverse stromal cells, including cancer-associated fibroblasts (CAFs), tumor endothelial cells, and various immune cells [2]. Those stromal cells actively sustain tumor progression, promote immune suppression, and mediate therapeutic resistance through multiple signal pathways across various cancer types [3–7]. ECM is crucial for maintaining cellular support during homeostasis and exerts a profound influence on cancer development and progression. Biochemical and physical ECM signals influence tumorigenesis, invasion, metastasis, and therapy resistance [8–10]. These ECM alterations can result in fibrosis, leading to stromal stiffness and exerting mechanical stresses on cells, known as stiffness or rigidity [11, 12]. ECM and ECM stiffening-induced mechanical stimuli can activate cell membrane receptors and mechanosensors, including integrin, piezo-type mechanosensitive ion channel component 1 (Piezo1), and transient receptor potential vanilloid 4 (TRPV4), consequently modulating the malignant phenotype of tumor and stromal cells [13]. Matrix stiffness is significantly correlated with poor prognosis in PDAC, where tissue stiffness increases slowly and gradually [14–16]. PDAC is characterized by substantial stromal components, rendering it one of the stiffest malignancies, with solid stresses exceeding 10 kPa [17].

In previous research, we established an adjustable extracellular matrix rigidity tumor model for in vivo and in vitro studies. We demonstrated that a biomechanical microenvironment characterized by high matrix stiffness substantially accelerates pancreatic cancer progression and activates cancer-related pathways, including PI3K-AKT and WNT. This setting is significantly associated with tumor immune suppression [16]. However, the precise influence of ECM stiffness and extracellular biomechanical cues in solid tumors remains unclear.

Doublecortin-like kinase 1 (DCLK1) possesses a serine/threonine kinase domain with Ca²⁺/calmodulin-dependent activity and two doublecortin motifs (DCX1 and DCX2) linked to microtubule function [18]. DCLK1 is closely associated with pancreatic cancer progression, chemotherapy resistance, and tumor immune suppression, as evidenced by our previous research [19–21]. Herein, we discover that DCLK1 can be selectively and significantly induced and activated under biomechanical stress, serving as a crucial context-specific amplifier in tumor progression and chemotherapy resistance mediated by biomechanical Ca²⁺ signals. We found that DCLK1 is selectively induced and activated under biomechanical stress, acting as a critical checkpoint in tumor progression and chemotherapy resistance via Ca²⁺ signaling. Our findings indicate that combining chemotherapy with interventions targeting Ca²⁺-DCLK1 signaling could improve treatment efficacy for pancreatic cancer.

Method

Cell lines and culture conditions

Human PDAC cell lines PANC1 (Kras G12D mutation) and BXPC3 (wild-type Kras) were obtained from the China Center for Type Culture Collection. The murine PDAC cell line, LSL-Kras(+/G12D); LSL-Trp53(+/R172H); Pdx1-Cre (KPC), characterized by Kras G12D and Trp53 R172H mutations, was graciously supplied by Fudan University Shanghai Cancer Center. The genetic mutations of the KPC cell line were confirmed using next-generation sequencing performed by Sangon Biotech (Shanghai, China) upon receipt. All cell lines were cultured in Dulbecco’s modified Eagle medium (BasalMedia, Shanghai, China) supplemented with 10% fetal bovine serum (Biological Industries, Israel) and 1% penicillin/streptomycin (BasalMedia, Shanghai, China). Cultures were maintained at 37 °C in a humidified atmosphere containing 5% CO₂. Mycoplasma contamination was ruled out utilizing the MycoBlue Mycoplasma Detector (Vazyme, Nanjing, China). The technique for cell transfection and lentivirus infection was consistent with those of previous studies [22–24]. All plasmids utilized in the study were synthesized by Generalbiol (Anhui, China), and the lentiviruses were procured from Genechem (Shanghai, China).

Preparation of gelatin methacrylate (GelMA) hydrogels

Hydrogels were synthesized utilizing GelMA with a 60% degree of substitution (DS) (GM60; EFL-GM-60, Yongqinquan Intelligent Equipment Co., Ltd., Suzhou, China) by dissolving it in phosphate-buffered saline (PBS) at concentrations of 5% and 20% (w/v) and incubating in a 55 °C water bath until completely dissolved. Similarly, 90% Degree of Substitution Gelatin Methacryloyl (GM90; EFL-GM-90, Yongqinquan Intelligent Equipment Co., Ltd., Suzhou, China) was dissolved at 10% (w/v) in PBS under the identical conditions. The GelMA solutions were mixed with the photoinitiator lithium phenyl-2, 4, 6-trimethyl-benzoyl phosphinate to attain a final concentration of 0.025% w/v. The mixture was heated to 37 °C in a water bath, sterilized using a 0.22-µm filter, aliquoted, and stored at − 20 °C.

In vitro 3D cell cultures

Consistent with previously published studies, 3D cell cultures employing visible-light (405 nm) crosslinked hydrogels were performed using 5% GM60 (1 kPa), 10% GM90 (10 kPa), and 20% GM60 (20 kPa) hydrogels, maintained at 37 °C in a light-protected metal bath. Two in vitro culture protocols were employed:

Protocol A: Cells were thoroughly mixed with GelMA solution and dispensed onto a 405 nm light curing unit (EFL-LS-1601-405, Yongqinquan Intelligent Equipment Co., Ltd., Suzhou, China), sealed with parafilm to prevent adhesion. Cell-climbing slices were placed on the cell suspension drops. GelMA-cell suspensions were crosslinked into hydrogels within 30 s using the light-curing unit. The slices were meticulously extracted and transferred to Petri dishes, followed by the addition of growth medium.

Protocol B: Cells were mixed with GelMA solution and dispensed directly into Petri dishes. GelMA-cell suspensions were crosslinked into hydrogels in 30 s using the light-curing unit, and growth medium was subsequently added.

Western blot analysis based on 3D cell cultures

Protein extraction in a 3D cell culture system was referred to as tissue protein extraction. Cells were harvested and reseeded in 3D culture in a six-well plate following the protocol as outlined previously. Following five days of cell culture, the media was removed and washed with PBS (2 × 2 mL). A cell spatula was utilized to gently lift the hydrogel and place it in a 1.5 mL microcentrifuge tube.

The samples were subsequently lysed in radioimmunoprecipitation assay lysis buffer and fully ground by hand using an electric grinding rod. After sonication and centrifugation at 12,000 g for 15 min at 4 °C, the supernatants were collected. The remainder of the procedure was performed as previously described [19–21].

Cell transfection

Cell transfection involved seeding cells in culture plates and maintaining them until they attained 70%–80% confluence. Transient transfection was conducted using Lipofectamine 2000 reagent (Invitrogen) following the manufacturer’s instructions. Briefly, plasmid DNA and Lipofectamine 2000 were diluted separately in Opti-MEM reduced-serum medium, mixed, and incubated for 20 min at room temperature to generate transfection complexes. The mixture was introduced dropwise to the cells. The culture medium was replaced with fresh complete medium after 4–6 h, and cells were harvested for subsequent analysis 24–48 h post-transfection.

Lentiviral transfection and generation of stable cell lines

Lentiviral constructs were acquired from GeneChem (Shanghai, China). Short hairpin RNAs (shRNAs) targeting human hippocalcin-like protein 1 (HPCAL1), PIEZO1, and DCLK1, and mouse PIEZO1, were utilized for gene silencing. Single guide RNAs (sgRNAs) targeting human PIP5K1A and mouse DCLK1 were utilized for CRISPR/Cas9-mediated knockout. Scrambled shRNAs or non-targeting sgRNAs served as negative controls. Cells were infected at the optimal multiplicity of infection in the presence of 5 µg/mL polybrene (GeneChem). Stable cell lines were established using selection with puromycin (2–5 µg/mL) or blasticidin S (5–10 µg/mL), depending on the vector resistance marker, for approximately 1 week.

Endoscopic ultrasound (EUS) elastography

Tumor stiffness was evaluated using EUS, a technique utilized during routine EUS examinations to visualize tissue elasticity and determine the elasticity modulus. The stiffness was quantified utilizing the strain ratio (SR), calculated by examining the color distribution in elastography images to compare the strain rates between the tumor and the surrounding normal pancreatic tissue. To ensure quantitative precision and reproducibility, the region of interest (ROI) was delineated to encompass the entire tumor lesion for the target measurement, whereas an equivalent area of adjacent normal pancreatic tissue was designated as the reference. The SR was calculated automatically by the system. All elastography examinations were conducted and evaluated jointly by at least two senior endosonographers experienced in endoscopic procedures to reduce inter-observer variability. This retrospective study of 63 patients with pancreatic cancer was conducted at Wuhan Union Hospital. The study protocol was reviewed and approved by the Ethics Committee of Wuhan Union Hospital [2021]IEC(0075 − 01). The requirement for informed consent was waived by the Ethics Committee due to the retrospective nature of the study and the use of de-identified data from a standardized clinical database.

Co-immunoprecipitation (Co-IP)

To investigate protein-protein interactions, we utilized a Co-IP protocol employing FLAG/HA affinity beads (Magneti-G Anti-DYKDDDDK Beads and Magneti-G Anti-HA Beads, ACE Biotechnology), following a standardized procedure optimized for high specificity and yield. Cells were cultured under ideal conditions and harvested at approximately 80% confluence. Post-harvest, cells were lysed with a lysis buffer (Affinibody LifeScience) to facilitate solubilization of cellular proteins and maintain protein interactions. The lysates were cleared by centrifugation at 12,000 g for 15 min at 4 °C to remove cellular debris. Bead Washing: Magnetic beads were thoroughly mixed, and 20 µL was transferred into a centrifuge tube. Wash buffer (500 µL) was added, and the tube was inverted 5–10 times to mix. The tube was then placed on a magnetic stand until the beads were fully captured, after which the supernatant was discarded. An additional 500 µL of wash buffer was added, the tube was inverted to mix 5–10 times, placed back on the magnetic stand, and the supernatant was discarded once the beads were captured. The washing steps were repeated thrice. Sample incubation: We added 500–1000 µL of cell lysate to the washed beads and incubated them for 60 min at room temperature or 4 °C, preferably on a rotational mixer to ensure thorough contact. After incubation, the tube was placed on the magnetic stand, and the supernatant was transferred to a new tube for further analysis. Wash buffer (500 µL) was subsequently added, the tube was gently inverted 5–10 times, placed on the magnetic stand, and the supernatant was discarded once the beads were captured. The washing step was repeated five times. Elution Buffer Elution: For every 20 µL of original bead volume, 40 µL of elution buffer was added. The mixture was incubated for 5–10 min at room temperature or 4 °C with gentle mixing 3–5 times. The tube was then placed on a magnetic stand, and the supernatant was collected into a new tube. The eluate was neutralized with a neutralization buffer and subsequently used for Western blot analysis.

In vivo experiments

All animal experiments were conducted in accordance with the ethical guidelines approved by the Institutional Animal Care and Use Committee at Tongji Medical College, Huazhong University of Science and Technology (Approval No. [2021] IACUC NUMBER: 2617). Tumor volume was calculated using the formula: volume = 0.5 × length × width².

Orthotopic tumor transplantation models in PDAC

Male C57BL/6J mice (8 weeks old) were obtained from Beijing Weitong Lihua Experimental Animal Technology Co., Ltd. The mice were housed under specific pathogen-free conditions at Tongji Medical College’s Experimental Animal Center. In each group (n = 8), mice were anesthetized with an intraperitoneal injection of 1.25% tribromoethanol (20 µL/g). The surgical area on the left flank was shaved and sterilized with 75% alcohol.

Hydrogels with varying rigidity (5% GM60 at 1 kPa and 20% GM60 at 20 kPa) were maintained at 37 °C, shielded from light. KPC cells (5 × 10⁵ cells/20 µL) were thoroughly mixed with GelMA solution and orthotopically injected into the pancreas through a left flank incision. The GelMA-KPC cell suspension was crosslinked in situ using a 405 nm light-curing unit for 30 s. The needle was withdrawn post-crosslinking, and the abdominal wall was sutured in layers with a 7–0 silk thread.

Animal elastography

Elastography was performed using the SuperSonic Aixplorer ultrasound system with an SL15-4 linear transducer. After pancreatic tumors were identified by ultrasound, a study box was positioned, and an ultrasound wave was applied at varying depths to compress the tissue. Three elastographic images in the plane of maximal diameter were captured for each lesion using the shear wave elastography (SWE) mode. SWE values were obtained by placing fixed ROIs over the entire lesion and adjacent tissue. All SWE values were recorded, and the mean maximal SWE values were utilized for evaluation.

Multiplex immunofluorescence

Paraffin-embedded tissue sections were deparaffinized and rehydrated using xylene and ethanol washes. Antigen retrieval was performed by heating in a citrate buffer (pH 6.0) for 20 min. Non-specific binding was blocked with 5% bovine serum albumin. Sections were incubated with the first primary antibody overnight at 4 °C, followed by washing with PBS and incubation with a horseradish peroxidase-conjugated secondary antibody for 1 h at room temperature. The TSA kit was utilized to amplify the signal with fluorophore-conjugated tyramide. Bound antibodies were stripped using a low pH buffer while simultaneously retaining the fluorescence signal. This process was repeated sequentially for additional targets using different fluorophores. Nuclei were counterstained with DAPI, and the sections were mounted with antifade mounting medium. Images were acquired using a fluorescence microscope. Colocalization and distribution were analyzed using appropriate software. Immunohistochemistry and immunofluorescence staining of tumors were performed as previously described. The average integrated optical density and fluorescence intensities were quantified for five randomly selected fields per group using FIJI ImageJ software (https://imagej.net/software/fiji/). This section of the experiment was conducted with support from PINUOFEI BIOLOGICAL.

Von Kossa staining

We performed Von Kossa staining to visualize calcium deposition. Paraffin-embedded tissue sections were deparaffinized, rehydrated, and washed with distilled water. Slides were incubated in a 5% silver nitrate (AgNO₃) solution under ultraviolet light for 30 min. Unreacted silver was removed by treatment with 5% sodium thiosulfate for 2 min. Sections were counterstained with hematoxylin for 2–3 min to visualize nuclei, dehydrated, and mounted. Calcium deposits appeared as black precipitates, whereas nuclei stained blue. This experiment was conducted with technical support from PINUOFEI BIOLOGICAL.

Masson’s trichrome staining

We performed Masson’s trichrome staining to evaluate collagen deposition and fibrosis. Paraffin-embedded tissue sections were deparaffinized, rehydrated, and washed. Nuclei were stained with Weigert’s iron hematoxylin for 5–10 min. Sections were subsequently stained with Biebrich scarlet-acid fuchsin solution to visualize cytoplasm and muscle fibers. After differentiation with phosphomolybdic-phosphotungstic acid, the sections were stained with Aniline Blue solution for 5 min to label collagen fibers. Furthermore, slides were treated with 1% acetic acid, dehydrated, and mounted. Collagen fibers appeared blue, nuclei were black/dark blue, and cytoplasm appeared red. This experiment was conducted with technical support from PINUOFEI BIOLOGICAL.

Bulk RNA-seq assay

Bulk RNA sequencing was performed by Wuhan Seqhealth Co., Ltd. Total RNA was extracted from murine tumor tissues using Trizol and sequenced on an Illumina platform. Post-sequencing quality checks were conducted, followed by a standard analysis pipeline to extract biological signals. Differential expression analysis and functional enrichment were performed utilizing the RNAseqStat2 package (Zeng J, Xia Y, 2022). Gene set variation analysis (GSVA) was performed with the GSVA package (version 1.42.0) utilizing Hallmark gene sets (mh.all.v2022.1.Mm.symbols.gmt).

Single-cell RNA-seq assay

Single-cell RNA sequencing datasets were obtained from GSE186960. Downstream analysis and visualization of single-cell data were conducted using the SCP package (Zhang H 2023. SCP: Single Cell Pipeline. R package version 0.5.6). Quality control was conducted to filter out low-quality cells and potential doublets utilizing metrics including mitochondrial gene percentage and unique gene counts. Normalization and scaling were performed to ensure comparability across cells. Subsequent analyses included dimensionality reduction utilizing PCA, followed by clustering with the Louvain algorithm. Visualization was achieved using UMAP plots to illustrate cellular diversity and identify distinct cell types. Differentially expressed genes were visualized and clustered using the ClusterGvis package (Jun Zhang 2022. ClusterGVis: One-step to Cluster and Visualize Gene Expression Matrix. https://github.com/junjunlab/ClusterGVis). Differential expression analysis was performed to identify marker genes for each cluster, providing insights into the functional characteristics and biological significance of the identified cell populations.

Statistical analysis

Data are expressed as mean ± standard deviation. GraphPad Prism software (version 9.0) was utilized for statistical analysis, employing Student’s t-test and one-way analysis of variance. A P < 0.05 was considered statistically significant.

Results

Mechanical stress is associated with an unfavorable prognosis for tumor progression

EUS elastography is a technique for visualizing tissue elasticity modulus during a conventional EUS examination, utilized for assessing tissue elasticity and stiffness (Fig. 1A). By collecting and analyzing EUS-E data from 63 patients with pancreatic cancer, we observed that high strain ratio (SR) levels were associated with primary tumor progression (T), lymph node metastasis (N), and distant metastasis (M) (Fig. 1B-C).

Fig. 1.

DCLK1 specifically associates with High Matrix Stiffness Mechanical Microenvironment-induced Rapid Tumor Progression. A. Schematic of endoscopic ultrasound-elastography for pancreatic cancer (created by BioRender). B. Representative images of endoscopic ultrasound-elastography for pancreatic cancer. C. Violin plot of strain ratio (SR) among patients with various TNM stages (AJCC 8th edition). D. Schematic of in vitro matrix stiffness adjustable 3D culture system for pancreatic cancer (created by Figdraw). E. Immunoblotting analysis of DCLK1 expression in PDAC cells cultured in different matrix stiffness 3D culture systems (1 kPa–20 kPa); F. Schematic of adjustable extracellular matrix rigidity orthotopic transplantation murine models for pancreatic cancer (created by Figdraw). G and H. Representative images and quantification of mIHC analysis of DCLK1. (DCLK1 in green, DAPI in blue) in adjustable extracellular matrix rigidity KPC tumors (1 kPa vs. 20 kPa); I and J. Representative images of ultrasound elastography and quantification of matrix stiffness analysis in tranditional KPC orthotopic transplantation tumors models (nc vs. oeDCLK1); K and L. Representative images of tumor and quantification of tumor volume analysis in tranditional KPC orthotopic transplantation tumors models (NC vs. oeDCLK1)

DCLK1 specifically associates with high matrix stiffness and mechanical microenvironment-induced rapid tumor progression

DCLK1 encodes a protein belonging to the protein kinase superfamily and the doublecortin family. The protein encoded by this gene contains two N-terminal doublecortin domains that bind to and modulate microtubule polymerization, potentially impacted by mechanical stress [25]. To determine whether mechanical force signaling (MFS) modulates DCLK1 expression, we established an in vitro experimental model using GelMA gels with variable stiffness (1, 10, and 20 kPa) [16]. Western blot analysis demonstrated that increased MFS can upregulate DCLK1 expression (the in vitro culture model, Fig. 1D-E). Additionally, we established an in vivo animal model using GelMA gels with varying stiffness (1 and 20 kPa) [16]. The murine PDAC KPC cells were derived from KPC mice. Immunofluorescence staining revealed that DCLK1 expression significantly increases with a higher matrix stiffness mechanical microenvironment in vivo (the in vivo animal model, Fig. 1F–H). Furthermore, the enhancing effects of DCLK1 on tumor progression were validated utilizing orthotopic KPC mouse models based on Matrigel. Ultrasound elastography results indicate that DCLK1 overexpression marginally increased tumor stromal stiffness, though not significantly (Fig. 1I-J). Consistent with previous studies, DCLK1 overexpression enhances pancreatic tumor progression significantly in mouse models (Fig. 1K-L). We assessed whether DCLK1 expression was associated with specific KRAS mutation types (G12D, G12C, or G12V) in a panel of PDAC cell lines. No evident correlation was observed between DCLK1 protein levels and the KRAS genotype (Fig. S1).

DCLK1 promotes pancreatic cancer progression by activating the PI3K-AKT pathway through interaction with PIP5K1A

To investigate the downstream regulator of MFS that facilitates DCLK1-mediated tumor progression, we performed RNA-seq analysis of PDAC cells cultured under varying matrix stiffness conditions in 3D culture (20 kPa versus 1 kPa) or with DCLK1 overexpression (oeDCLK1 versus NC). KEGG functional enrichment analysis revealed that differentially expressed genes were significantly enriched in the PI3K/AKT signaling pathway (Fig. 2A-B). Western blot analysis corroborated the result of RNA-sequencing; DCLK1 overexpression activates the PI3K/AKT signaling pathway (phosphorylation of PI3K and AKT) (Fig. 2C).

Fig. 2.

DCLK1 activated the PI3K-AKT pathway through interaction with PIP5K1A. A. Bar plot depicting upregulated pathways in pancreatic cancer under high matrix stiffness 3D culture conditions (20 kPa vs. 1 kPa, via KEGG enrichment analysis); B. Bar plot depicting upregulated pathways in pancreatic cancer with DCLK1 over expression (oeDCLK1 vs. NC, via KEGG enrichment analysis); C. Immunoblotting analysis of DCLK1, phosphorylated (p) and total PI3K and AKT transfected with a control lentivirus (NC) or DCLK1-overexpression lentivirus (oe DCLK1). D. Venn diagram comparing the mass spectrometric analysis of the protein candidates pulled down by Flag-DCLK1 and Flag alone. E. Co-immunoprecipitation (Co-IP) experiments were performed with anti-Flag antibody on BXPC-3 cell lysates expressing Flag-DCLK1 and Flag alone. Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-DCLK1 (top) and anti-PIP5K1A (bottom) antibodies shows co-immunoprecipitation of DCLK1 with PIP5K1A. F. Immunoblotting analysis of DCLK1, PIP5K1A, phosphorylated (p) and total PI3K and AKT expression in PDAC cell lines transfected with control lentivirus (NC), DCLK1-knockdown lentivirus (kd DCLK1), and DCLK1-overexpression lentivirus (oe DCLK1); Immunoblotting analysis of DCLK1, PIP5K1A, phosphorylated (p) and total PI3K and AKT expression in PDAC cell lines cultured in different matrix stiffness 3D culture systems (1 kPa–20 kPa). G. Immunoblotting analysis of DCLK1, PIP5K1A, phosphorylated (p) and total PI3K and AKT expression in PDAC cell lines which were infected with control lentivirus (NC), DCLK1-knockdown lentivirus (kd DCLK1), and DCLK1-overexpression lentivirus (oe DCLK1) together with control lentivirus (NC) or PIP5K1A-knockout lentivirus (ko PIP5K1A)

To examine the role of DCLK1 in the activation of the MFS-mediated PI3K/AKT signaling pathway, we conducted co-immunoprecipitation followed by mass spectrometry to identify proteins associated with DCLK1. The observation indicates a potential interaction between DCLK1 and PIP5K1A (Fig. 2D), which was subsequently confirmed by Co-IP assays (Fig. 2E). PIP5K1A synthesizes PtdIns-4,5-P2 (PIP2), which was subsequently converted by PI3K into PtdIns-3,4,5-P3 (PIP3). PIP3 is crucial for the activation of AKT, a key player in cellular growth and survival pathways [26]. Subsequent immunoblotting analysis verified that DCLK1 overexpression elevated the phosphorylation levels of PI3K and AKT, whereas DCLK1 knockdown exhibited the opposite effects. As expected, MFS (1–20 kPa) enhanced DCLK1 expression and PI3K/AKT signaling pathway activation. Notably, neither DCLK1 expression nor MFS affected the PIP5K1A expression levels (Fig. 2F).

To investigate if PIP5K1A is involved in DCLK1-mediated PI3K/AKT signaling pathway activation, we developed PIP5K1A knockout (Ko PIP5K1A) PDAC cell lines. Post-knockout, DCLK1 overexpression or knockdown did not significantly alter PI3K and AKT phosphorylation levels, indicating that PIP5K1A is essential for DCLK1-mediated activation of the PI3K/AKT signaling pathway (Fig. 2G).

The serine/proline-rich linker domain of DCLK1 binds to the phosphatidylinositol phosphate kinase (PIPK) domain of PIP5K1A, promoting the dephosphorylation and membrane localization of PIP5K1A

We conducted truncation mapping to delineate essential interaction domains of DCLK1 and PIP5K1A. Six Flag-tagged truncated DCLK1 constructs were developed: Doublecortin domain 1 (Δ1-160 aa), doublecortin domain 2 (Δ160–280 aa), serine/proline-rich domain (Δ280–390 aa), non-kinase domain (Δ1-378 aa), serine/threonine protein kinase domain (Δ379–696 aa), and autoinhibit domain (Δ600–740 aa). Furthermore, two His-tagged truncated PIP5K1A constructs were generated: The PIPK domain (Δ1-500 aa) and the non-kinase domain (Δ400–562 aa) (Fig. 3A). Co-IP revealed that the serine/proline-rich domain of DCLK1 (Δ280–378 aa) was essential for its interaction with full-length PIP5K1A, excluding the involvement of the doublecortin or serine/threonine protein kinase domains (Fig. 3B–D). The PIPK domain of PIP5K1A (Δ1-400 aa) is essential for interaction with full-length DCLK1. We utilized AlphaFold2 to predict the DCLK1-PIP5K1A complex model and identified potential interacting residues between DCLK1 (Δ280–378 aa) and PIP5K1A (Δ1-400 aa) utilizing LigPlus software (Fig. S2A).

Fig. 3.

The serine/proline-rich linker domain of DCLK1 binds to the PIPK domain of PIP5K1A, promoting the dephosphorylation and membrane localization of PIP5K1A. A. Schematic representation of the domains of DCLK1 and PIP5K1A and the structure of DCLK1 and PIP5K1A truncations used in this work. B. Co-immunoprecipitation (Co-IP) experiments were performed with anti-Flag antibody on BXPC-3 cell lysates expressing Flag-DCLK1 truncations). Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-Flag (bottom) and anti-PIP5K1A (top) antibodies shows co-immunoprecipitation between DCLK1 truncations and PIP5K1A. C. Co-immunoprecipitation (Co-IP) experiments were performed with anti-his antibody on BXPC-3 cell lysates co-expressing his-PIP5K1A and Flag-DCLK1 truncations (DCLK1 1-160, DCLK1 160–280, DCLK1 280–390, DCLK1 1-378, DCLK1 379–696 and DCLK1 600–740) of input and immunoprecipitated (IP) samples with anti-Flag (bottom) and anti-PIP5K1A (top) antibodies shows co-immunoprecipitation between PIP5K1A and DCLK1 truncations. D. Co-immunoprecipitation (Co-IP) experiments were performed with anti-his antibody on BXPC-3 cell lysates expressing HA-PIP5K1A truncations (PIP5K1A 1-500, and PIP5K1A 400–560). Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-His (bottom) and anti-Flag (top) antibodies shows co-immunoprecipitation between PIP5K1A truncations and DCLK1. E and F. Confocal fluorescent imagings of PIP5K1A with control (NC), or DCLK1-overexpression (oe DCLK1). The line charts display the fluorescence intensity values of DCLK1(green) and PIP5K1A (red), as determined along the path of the white line depicted in the corresponding inset panel. Pearson’s colocalization coefficient (PCC) was used to quantify the degree of colocalization. G. Co-immunoprecipitation (Co-IP) experiments of the PIP5K1A phosphorylation were performed with anti-His antibody on BXPC-3 were stably infected with His-PIP5K1A-lentivirus with control lentivirus (NC), and DCLK1-overexpression lentivirus (oe DCLK1). Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-DCLK1 antibody, anti-PIP5K1A antibody, Anti-Phospho-Serine/Threonine/Tyrosine antibody, Anti-Phospho-Serine antibody, Anti-Phospho-Threonine antibodies, Anti-Phospho-Tyrosine antibodies. H. Co-immunoprecipitation (Co-IP) experiments were performed using DCLK1 truncation constructs to identify the crucial functional domain of DCLK1 involved in the interaction with and phosphorylation modulation of PIP5K1A. Cells were co-expressed with wild-type PIP5K1A and various DCLK1 truncates. Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-DCLK1 antibody, anti-PIP5K1A antibody, Anti-Phospho-Serine/Threonine/Tyrosine antibody, Anti-Phospho-Serine antibody, Anti-Phospho-Threonine antibodies, Anti-Phospho-Tyrosine antibodies

The colocalization analysis using confocal imaging confirmed the interaction between DCLK1 and PIP5K1A. The fluorescence signals of DCLK1 and PIP5K1A revealed colocalization, demonstrating their close proximity within the cells. Furthermore, DCLK1 overexpression markedly diminished the colocalization of DCLK1 and PIP5K1A (Fig. 3E). High-magnification confocal imaging reveals that the membrane localization of PIP5K1A was significantly enhanced through DCLK1 overexpression (Figs. 3F and S2B). As a kinase protein, we hypothesized that DCLK1 may play a crucial role by phosphorylating PIP5K1A. Consequently, we conducted additional tests to determine whether DCLK1 overexpression could influence the phosphorylation level of PIP5K1A. Unexpectedly, the overexpression of DCLK1 resulted in a decrease in PIP5K1A phosphorylation levels, with the most significant reduction observed at the threonine residues within PIP5K1A (Fig. 3G). PIP5Ks and PIP4Ks possess conserved structural features, comprising a lipid kinase domain and a dimerization domain. Previous studies have confirmed that dephosphorylation and membrane localization of PIP4Ks serve as markers of their activation [27–29], hence supporting our findings. Co-IP assays employing these truncated constructs confirmed that the serine/proline-rich domain of DCLK1 represents the minimal region necessary for PIP5K1A binding. Notably, further investigation revealed that this specific DCLK1 fragment (Δ280–378 aa), which harbors the serine/proline-rich domain, markedly mediated PIP5K1A dephosphorylation, specifically at its threonine residues (Fig. 3H). Our results indicated that the serine/proline-rich linker domain of DCLK1 binds to the PIPK domain of PIP5K1A, enhancing the dephosphorylation and membrane localization of PIP5K1A, thereby activating PIP5K1A.

High matrix stiffness mechanical microenvironment promotes calcium influx by activating PIEZO1

We subsequently determined how MFS regulates DCLK1. Gene ontology (GO) enrichment analysis indicates that MFS is associated with cation channel activation (Fig. 4A). DCLK1 belongs to the protein kinase superfamily, specifically the CAMK Ser/Thr protein kinase family and CaMK subfamily. We hypothesized that MFS may modulate DCLK1 expression through calcium ions. We assessed MFS-mediated Ca²⁺ signals in PDAC cells cultured on a 3D culture system with variable stiffness and observed that the Ca²⁺ influx was enhanced in PDAC cells cultured on the highest stiffness matrix (Fig. 4B). Von Kossa staining results for animal and human pancreatic tumor tissues reveal a significant increase in intracellular calcium ions in pancreatic cancer cells within a high-stiffness microenvironment (Figs. 4C-D). Mechanical stimuli are transduced to cells through several mechanosensitive ion channels, including the PIEZO family. Single-cell RNA sequencing analysis revealed that PIEZO1 expression is prominent among several cell subpopulations inside the pancreatic cancer microenvironment (Fig. 4E-F). Immunohistochemical staining assays revealed that PIEZO1 expression was higher in PDAC tissues than in paracarcinoma tissues and was primarily localized in ductal-like cancer tissues (Fig. 4G). We subsequently aimed to determine whether MFS facilitated calcium signaling through PIEZO1 in PDAC. Yoda1 has been recognized as a specific agonist for PIEZO1 but not PIEZO2 [30]. The results revealed that Yoda1 can enhance Ca²⁺ influx under a low stiffness matrix, whereas PIEZO1 knockout impedes MSF-mediated calcium signaling (Fig. 4H). Ajay Rana’s study developed PANC1-GR, a gemcitabine-resistant derivative of the PANC1 cell line, and compared it with PANC1 cells treated with the calmodulin inhibitor W-7, the calcium chelator BAPTA-AM, or the calcium channel blocker amlodipine, followed by single-cell RNA sequencing (GSE186960) [31]. Cells were clustered according to transcriptional similarity. When visualized, Panc1-GR cells treated with the calmodulin inhibitor W-7, calcium chelator BAPTA-AM, or the calcium channel blocker amlodipine each formed five nonoverlapping clusters (Fig. 4I). Notably, antagonizing calcium signaling significantly reduces the proportion of Cluster 5 (Fig. 4J). Compared to other clusters, KEGG enrichment analysis revealed a significant upregulation of the PI3K-AKT pathway in Cluster 5. Treatments with calcium signaling inhibitors (calmodulin inhibitor W-7, calcium chelator BAPTA-AM, and calcium channel blocker amlodipine) suppressed PI3K-AKT pathway activity in pancreatic cancer cells (Fig. 4K).

Fig. 4.

High matrix stiffness mechanical microenvironment promotes calcium influx by activating PIEZO1. A. Stick plot depicting upregulated pathways in pancreatic cancer under high matrix rigidity 3D culture conditions (20 kPa vs. 1 kPa, via GO enrichment analysis); B. Fluorescence microscope images of intracellular calcium fluorescence staining (Fluo-4 stained) in pancreatic cancer cells or cell spheres on day 3 under various matrix rigidity 3D culture conditions (1 kPa vs. 20 kPa, ). The quantification of fluorescence intensities were presented in heatmaps. C. The representative images of calcium staining of in adjustable extracellular matrix rigidity KPC tumors (1 kPa vs. 20 kPa). Calcified deposits were brown to black in the von Kossa stain (von Kossa, ×20 and ×40). D. The representative images of Massion staining and calcium staining of PDAC tissue chip (low matrix rigidity vs. high matrix rigidity). Calcified deposits were brown to black in the von Kossa stain (von Kossa, ×5 and ×40). E and F. Analysis of single-cell RNA sequencing in pancreatic cancer tissues. UMAP of subpopulation after cell annotation (E). UMAP plot of PIEZO1 and PIEZO2 gene expression. G. Expression of PIEZO1 in carcinoma tissue and para-carcinoma tissue (IHC, ×5 and ×40). H. Fluorescence microscope images of intracellular calcium fluorescence staining (Fluo-4 stained) in pancreatic cancer cell spheres on day 3 under various matrix rigidity 3D culture conditions (1 kPa vs. 20 kPa) together with Yoda1 treatment or PIEZO1 sgRNA transfected. The quantification of fluorescence intensities were presented in heatmaps. I and J. Analysis of single-cell RNA sequencing in PANC1 cell treated with amlodipine, BAPTA-AM and W-7. UMAP of subpopulation after cell clustering (all groups) combined and UMAP plot of Single-cell trajectory results (I). UMAP of subpopulation after cell clustering (each group) and the bar plots show the percentage of each subpopulation (J). K. Differential gene heat map and enrichment analysis results of sequencing differential gene KEGG in each cluster or each group

Calcium ions facilitate the interaction between HPCAL1 and DCLK1, activating the DCLK1-PIP5K1A axis

The molecular mechanisms underlying DCLK1 activation encompass three potential ways to alleviate its autoinhibition: (A) Activation of DCLK1 by a physiological activator protein such as HPCAL1; (B) mutational dysregulation that interferes with the C-terminal autoinhibitory domain (AID) binding, resulting in constitutive DCLK1 activation; (C) upregulation of DCLK1 expression levels, which amplifies overall kinase activity [32]. We subsequently investigated Ca²⁺ regulation of DCLK1 by generating BXPC-3 and PANC1 cell lines stably expressing Flag-DCLK1 and His-PIP5K1A using lentiviruses. Western blot and Co-IP revealed that BAPTA-AM treatment suppresses DCLK1 expression and inhibits its serine phosphorylation, thereby inhibiting its function at expression and phosphorylation levels. Furthermore, BAPTA-AM significantly inhibits PI3K-AKT pathway activation. Consistent with these findings, BAPTA-AM treatment enhances inhibitory threonine phosphorylation of PIP5K1A, resulting in its suppressed activity (Fig. 5A-B). Furthermore, to examine the specific regulatory mechanism of Ca²⁺ signaling on PIP5K1A phosphorylation, we performed mutational analysis on potential threonine residues. Given the observed complex PIP5K1A regulation by threonine phosphorylation, we generated a series of PIP5K1A point mutants at conserved threonine sites and assessed the effect of the Ca²⁺ inhibitor (BAPTA-AM) on their phosphorylation levels. Notably, when the threonine 482 site (T482) was mutated to alanine PIP5K1A^T482A, the promoting effect of the Ca²⁺ inhibitor on PIP5K1A phosphorylation was entirely abolished (Fig. 5C). This key finding demonstrates that the Ca²⁺ signaling pathway specifically inhibits the PIP5K1A phosphorylation at the threonine 482 site (T482), and that the threonine 482 site (T482) phosphorylation is essential for PIP5K1 activation. Therefore, the Ca²⁺-DCLK1 axis regulates PIP5K1A function through a complex mechanism involving site-specific threonine phosphorylation.

Fig. 5.

Calcium ions activating and stabilize DCLK1. A and B. Co-immunoprecipitation (Co-IP) experiments were conducted using anti-Flag and anti-His antibodies to examine the phosphorylation levels of DCLK1 and PIP5K1A in BXPC-3 and PANC1 cells stably co-infected with His-PIP5K1A-lentivirus and Flag-DCLK1-lentivirus, either in the presence or absence of 10 µM calcium chelator BAPTA-AM for 24 h. Immunoblotting (IB) analyses included detection of DCLK1, PIP5K1A, phosphorylated and total PI3K and AKT, as well as phospho-Serine/Threonine/Tyrosine, phospho-Serine, phospho-Threonine, and phospho-Tyrosine levels in input samples. Immunoprecipitated (IP) samples were probed with anti-DCLK1, anti-PIP5K1A, anti-phospho-Serine/Threonine/Tyrosine, anti-phospho-Serine, anti-phospho-Threonine, and anti-phospho-Tyrosine antibodies. C. Co-immunoprecipitation (Co-IP) experiments were performed using anti-His antibody on PDAC cells stably infected with His-PIP5K1A-lentivirus (Wild-Type) or His-PIP5K1A threonine point mutants. Immunoprecipitated (IP) samples were probed with anti-DCLK1, anti-PIP5K1A, anti-phospho-Serine/Threonine/Tyrosine, anti-phospho-Serine, anti-phospho-Threonine, and anti-phospho-Tyrosine antibodies. D. Co-immunoprecipitation (Co-IP) experiments were performed with anti-Flag antibody on BXPC-3 cell lysates expressing Flag-DCLK1. Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-DCLK1 (top) and anti-HPCAL1 (bottom) antibodies shows co-immunoprecipitation of DCLK1 with HPCAL1. E. The schematic diagram of the molecular mechanism of PI3K-AKT pathway activation through the Ca²⁺-HPCAL1-DCLK1-PIP5K1A axis. F, G, and H. BXPC-3 cells stably expressing Flag-DCLK1 were transfected with plasmids encoding HA-Ub or its mutants for 24 h and then incubated in the presence or absence of 10 µM calcium chelator BAPTA-AM for another 24 h. Co-immunoprecipitation (Co-IP) experiments were performed with anti-Flag antibody and immunoblotting analysis with anti-HA antibody (F). Heatmap representation of quantitative analysis of DCLK1 ubiquitination levels (G). Immunoblotting (IB) analyses included detection of Flag and HA (H). I. Venn diagram comparing the mass spectrometric analysis of the Ubiquitination-related proteins candidates pulled down by Flag-DCLK1 and Flag alone. J. Table lists 11 proteins potentially interacting with DCLK1. K. Co-immunoprecipitation (Co-IP) experiments were performed with anti-Flag antibody on BXPC-3 cell lysates expressing Flag-DCLK1. Immunoblotting (IB) of input and immunoprecipitated (IP) samples with anti-DCLK1, anti-Flag, anti-ANAPC5 and PSMA7 antibodies shows co-immunoprecipitation of DCLK1 with ANAPC5 and PSMA7. L. The schematic diagram of the molecular mechanism underlying the ubiquitin-mediated degradation of DCLK1 through the ANAPC5/PSMA7 axis

Previous studies have identified the neuronal calcium sensor HPCAL1 as an activator of DCLK1. The AID of DCLK1 competitively inhibits the ATP-binding site with ATP. HPCAL1 directly binds to the AID in a Ca²⁺-dependent manner, hence releasing the autoinhibition [32]. Co-IP assays demonstrated that HPCAL1 interacts with DCLK1 in pancreatic cancer cells (Fig. 5D). To further substantiate the PIEZO1/Ca²⁺/HPCAL1 axis’s role in modulating DCLK1 and PIP5K1A membrane localization, we conducted immunofluorescence analysis following PIEZO1 and HPCAL1 knockdown. PIEZO1 or HPCAL1 knockdown significantly reduced DCLK1 expression and inhibited the membrane localization of PIP5K1A. However, DCLK1 overexpression was able to partially rescue the impaired membrane localization of PIP5K1A induced by HPCAL1 knockdown (Fig. S3). To conclusively demonstrate that DCLK1 induces PI3K activation, we directly measured the levels of the lipid product PI(3,4,5)P3 (PIP3) using an enzyme-linked immunosorbent assay. Consistent with the phosphorylation data, DCLK1 or PIP5K1A overexpression significantly increased cellular PIP3 levels. Notably, PIP3 elevation induced by DCLK1 overexpression was effectively abrogated by the knockdown of upstream regulators PIEZO1 or HPCAL1. Moreover, the PIEZO1 agonist Yoda1 stimulated PIP3 production; however, this effect was inhibited in PIP5K1A-knockout cells (Fig. S4). These findings offer direct evidence that the mechanotransduction axis facilitates the conversion of lipids to produce PIP3, subsequently activating the PI3K signaling cascade. This is the first study to discover that calcium signaling inhibition can suppress DCLK1 activation by inhibiting DCLK1 expression and serine phosphorylation. This enhances threonine phosphorylation of downstream PIP5K1A, inhibits PIP5K1A activity, and consequently suppresses the PI3K-AKT pathway activity (Fig. 5E).

Calcium ions stabilize DCLK1, inhibiting its ubiquitin-mediated degradation via the ANAPC5/PSMA7 axle

Ubiquitination is a cellular process that modifies proteins for degradation through the proteasome complex. Treatment with the calcium chelator increased the ubiquitination of DCLK1 levels. Further experiments revealed that calcium chelator treatment catalyzes K11-linked and K27-linked polyubiquitination of DCLK1 (Figs. 5F-G). Hence, we performed Co-IP combined with mass spectrometry analysis to identify ubiquitin-related proteins that may bind to DCLK1 (Fig. 5I). We identified the top five proteins with the highest Mascot scores, all of which are integral to the ubiquitin-proteasome system, for subsequent validation (Fig. 5J). Among these candidates, Co-IP results indicate that the ANAPC5 (an E3 ubiquitin ligase) and PSMA7 (a 20 S proteasome subunit) demonstrated the most significant interaction with DCLK1 in reciprocal Co-IP assays, serving as critical nodes linking ubiquitination to proteasomal degradation. Consequently, we focused on these two proteins to elucidate the mechanism of DCLK1 degradation (Fig. 5K). This study demonstrated that calcium ions stabilize DCLK1, inhibiting its ubiquitin-mediated degradation through the ANAPC5/PSMA7 axis (Fig. 5L).

Ca²⁺-DCLK1 functions as a biomechanical axis to aggravate rapid tumor progression and chemotherapy resistance in pancreatic cancer

Furthermore, in vivo animal studies were conducted to clarify the function of the Ca²⁺-DCLK1 axis in biomechanically mediated pancreatic cancer progression and chemotherapy resistance. To ascertain if DCLK1 modulates rapid pancreatic cancer progression induced by a high-matrix-stiffness mechanical microenvironment, we established stable DCLK1 overexpression and knockdown KPC pancreatic cancer cell lines using lentiviral transfection. Further in vivo validation was performed using our developed adjustable extracellular matrix rigidity orthotopic tumor transplantation murine models of PDAC. Nab-paclitaxel combined with gemcitabine constitutes the first-line chemotherapy regimen for pancreatic cancer. We examined the tumor volume and tumor growth inhibition value between groups treated with this chemotherapy regimen and those treated with additional calcium ion inhibitors (Fig. 6A). In low stiffness conditions (NC, 1 kPa, +Blank), tumor progression was slow, and chemotherapy sensitivity was high (NC, 1 kPa, +CTH); the addition of Ca²⁺ inhibitors did not yield a significant enhancement of the effect (NC, 1 kPa, +CTH+Ca²⁺ inhibitors). DCLK1 overexpression under low stiffness (oeDCLK1, 1 kPa) significantly accelerated tumor progression and increased chemotherapy resistance (oeDCLK1, 1 kPa, and + CTH). However, the combination of Ca²⁺ ion inhibitors with DCLK1 overexpression somewhat reversed the associated chemotherapy resistance. compared to the low stiffness group (NC, 1 kPa), the high stiffness group (NC, 20 kPa, and +Blank) exhibited increased tumor volume and greater chemotherapy resistance (NC, 20 kPa, and + CTH). However, the addition of Ca²⁺ inhibitors (NC, 20 kPa, and + CTH+Ca²⁺ inhibitors) partially alleviated the chemotherapy resistance caused by elevated matrix stiffness. Moreover, under high stiffness conditions, DCLK1 knockdown (sgDCLK1, 20 kPa, +Blank) suppressed tumor growth and increased chemotherapy sensitivity (sgDCLK1, 20 kPa, and + CTH). The addition of Ca²⁺ inhibitors (sgDCLK1, 20 kPa, and + CTH+Ca²⁺ inhibitors) did not further enhance chemotherapy sensitivity following DCLK1 knockdown (Fig. 6B–F). Although CTH treatment diminished tumor volume in the high stiffness group, the overall tumor growth inhibition was markedly attenuated compared to lower stiffness conditions, indicating reduced chemotherapy sensitivity under biomechanical stress. The critical role of the Ca²⁺-DCLK1 axis in biomechanically mediated tumor progression was confirmed through targeted pharmacological and genetic interventions in vivo. Specifically, activating Ca²⁺ influx utilizing the PIEZO1 agonist, Yoda1, markedly enhanced tumor progression and volume in low stiffness conditions (1 kPa, +Yoda1), effectively simulating the adverse effects of a high stiffness microenvironment (Fig. 6G–I). However, under high stiffness conditions (20 kPa), the inhibition of DCLK1 using the selective inhibitor DCLK1-IN-1 or genetically reducing Ca²⁺ influx through PIEZO1 knockdown (shPIEZO1) led to a marked suppression of tumor growth. Collectively, these in vivo findings strongly confirmed the Ca²⁺-DCLK1 axis, modulated by PIEZO1, as a critical context-specific amplifier that accelerates rapid tumor progression and determines chemotherapy resistance in pancreatic cancer.

Fig. 6.

Combined calcium ion inhibitors can suppress DCLK1-mediated chemoresistance. A and B. Schematic of adjustable extracellular matrix rigidity orthotopic transplantation murine models for pancreatic cancer (1 kPa and 20 kPa, KPC cell lines) (created by Figdraw). KPC cell lines established from KPC spontaneous mouse model were injected subcutaneously in C57BL/6 mice, were stably infected with control lentivirus (NC), DCLK1-overexpression lentivirus (oe DCLK1) or DCLK1-knockdown lentivirus (sg DCLK1). Treated with chemotherapy alone (CTH, gemcitabine + nab-paclitaxel ) or chemotherapy + Ca²⁺ inhibitor (CTH + Ca²⁺ inhibitor, gemcitabine + nab-paclitaxel +BATPA-AM). C. Representative images of ultrasound elastography of animal models; D. Representative images of tumor in adjustable extracellular matrix rigidity KPC orthotopic transplantation tumors models. E and F. Comparison of final tumor volume and TGI values among groups with different matrix stiffness (1 kPa vs. 20 kPa), DCLK1 manipulation (DCLK1 overexpression (oeDCLK1), DCLK1 knockout (sgDCLK1)), and treatments (Blank, Chemotherapy (CTH), and CTH + Ca²⁺ Inhibitor). G. Treatment schedule for DCLK1-IN-1and Yoda1 intraperitoneal (IP) injection in orthotopic tumor models. H and I. Representative images of tumors and quantification of tumor volume analysis in orthotopic KPC tumor models (1 kPa vs. 20 kPa) treated with DCLK1-IN-1, Yoda1, or with PIEZO1 knockdown (shPIEZO1)

To further validate the correlation between the DCLK1-mediated mechanotransduction pathway and matrix stiffness, we employed multicolor immunofluorescence analysis to identify the expression of critical proteins. Masson staining demonstrated varied levels of fibrosis, which had a positive correlation with matrix stiffness. Consistent with in vitro cell experiments, staining of human pancreatic cancer tissue sections revealed that as matrix stiffness increased, intracellular calcium levels rose (von Kossa staining), DCLK1 expression increased, and the AKT pathway was activated (pAKT/AKT). Furthermore, there was evident colocalization of HPCAL1, DCLK1, and PIP5K1A, with high expression in duct-like cancerous lesions (Fig. 7). Quantitative analysis further confirmed these observations, indicating that increased matrix stiffness significantly correlated with higher expression of DCLK1 and HPCAL1, an elevated p-AKT/AKT ratio, and enhanced colocalization of DCLK1 with PIP5K1A and HPCAL1 (Fig. S5). These quantitative findings strongly support the central role of the matrix stiffness-DCLK1-PIP5K1A-AKT axis in the PDAC microenvironment.

Fig. 7.

Multiplex immunofluorescence staining validates the correlation between the DCLK1-mediated mechanotransduction pathway and matrix stiffness

Discussion

Malignant progression is increasingly recognized not only as a cell-autonomous process but also as a co-evolutionary dynamic between tumor cells and their surrounding microenvironment. The physical architecture of the extracellular matrix (ECM), marked by abnormal remodeling and biomechanical stiffness, is a crucial factor influencing tumor growth, mechanical resistance, and the evasion of antitumor immunity [33, 34]. This complex ecosystem simultaneously promotes phenotypic plasticity and the acquisition of stem-like traits, frequently facilitated by epithelial-mesenchymal transition (EMT), and maintains spatial heterogeneity that fuels therapeutic tolerance and disease recurrence [35, 36]. In addition to these physical barriers, the immunosuppressive characteristics of the tumor microenvironment (TME) present a significant challenge to clinical efficacy, yet recent findings indicate that specific stromal and immune patterns within this niche possess substantial predictive significance for patient outcomes [37, 38].

High matrix stiffness, caused by the dense desmoplastic stroma, is a hallmark feature of PDAC that facilitates aggressive phenotypes [39]. DCLK1 is well-characterized as a tuft cell marker and a promoter of tumorigenesis; nevertheless, its functional role in sensing and transducing mechanical signals remains unclear. This study demonstrates that DCLK1 serves as a critical mechano-transducer connecting extracellular physical signals to intracellular lipid kinase signaling [19, 20]. Based on our clinical observations correlating tumor stiffness with advanced staging, we mapped a novel axis where stiff substrates activate the mechanosensitive channel PIEZO1. This initiates a calcium-dependent cascade involving HPCAL1 that alleviates DCLK1 autoinhibition. Upon activation, DCLK1 recruits and modulates PIP5K1A, thereby enhancing PIP3 synthesis and sustaining AKT hyperactivation (Fig. 8). These findings establish a direct mechanistic connection between the physical microenvironment and the metabolic signaling networks driving PDAC progression [40].

Fig. 8.

Schematic illustration of the matrix stiffness-regulated Ca2+-DCLK1-PIP5K1A signaling cascade in pancreatic cancer progression

Mechanistically, our study clarifies the exact sequence of events connecting mechanical stress to lipid signaling. We identified PIEZO1 as the principal mechanosensor that perceives matrix stiffness, instigating the calcium influx necessary to initiate this cascade [41, 42]. Crucially, we reveal a novel, non-canonical role of DCLK1 downstream of this calcium signal. Our observation indicates that DCLK1 functions not only as a microtubule-associated kinase but also as a molecular scaffold that physically interacts with PIP5K1A. Notably, instead of phosphorylating PIP5K1A, DCLK1 facilitates the dephosphorylation at the Threonine 482 site. Although our findings implicate DCLK1 in facilitating this process through a scaffolding mechanism, the specific protein phosphatase responsible for executing PIP5K1A dephosphorylation remains undefined and lies beyond the scope of this study. The dephosphorylation process functions as a molecular switch, facilitating the translocation of PIP5K1A from the cytosol to the plasma membrane. Once localized to the membrane, PIP5K1A generates the PI(4,5)P2 substrate necessary for PI3K activation [43, 44]. This finding provides a unique insight into how DCLK1 integrates calcium signaling to spatially modulate lipid kinase activity, thereby activating the PI3K-AKT cell survival pathway in a stiffness-dependent manner.

Our study further reveals that the regulation of DCLK1 by the mechanical microenvironment functions through a dual mechanism encompassing both kinase activation and protein stability. Structurally, DCLK1 is regulated by a C-terminal AID that inhibits its intrinsic activity [32]. We illustrate that HPCAL1 functions as the essential calcium sensor that alleviates this autoinhibition. Upon matrix stiffness-induced calcium influx, HPCAL1 associates with the AID region, facilitating DCLK1 serine phosphorylation and its subsequent activation. However, beyond this classical activation, we discovered a novel layer of regulation governing DCLK1 protein stability. We found that limiting calcium influx suppresses DCLK1 activity and triggers its rapid degradation through the ubiquitin-proteasome system. Notably, low calcium levels specifically facilitate the interaction of DCLK1 with the E3 ubiquitin ligase ANAPC5 and the proteasome subunit PSMA7, resulting in the formation of K11- and K27-linked ubiquitin chains and subsequent proteasomal clearance. Thus, calcium signaling functions as a “two-pronged” regulator: it initiates signal transduction by activating DCLK1 through HPCAL1 and simultaneously maintains the signal’s duration by protecting DCLK1 from ANAPC5-mediated degradation.

Our study bridges these mechanistic insights with therapeutic implications for overcoming stiffness-induced chemoresistance. The dense, fibrotic stroma of PDAC has long been acknowledged as a physical impediment to drug delivery; nevertheless, our in vivo studies indicate that it also actively conversely signals that enhance survival [45]. We demonstrate that the Ca²⁺-DCLK1 axis is a central driver of this stiffness-mediated resistance. Notably, we observed that the ability of calcium inhibitors to sensitize tumors to chemotherapy is highly context-dependent: the synergistic effect was markedly diminished in tumors characterized by low matrix stiffness or DCLK1 depletion. Furthermore, our supplementary experiments using small-molecule inhibitors targeting this axis demonstrated significant antitumor efficacy, essentially disrupting the stiffness-induced protective mechanism. This finding is crucial, as it recognizes DCLK1 not merely as a downstream target but as the essential effector that transduces mechanical calcium signals into a survival advantage. Therefore, therapeutic strategies targeting the DCLK1 signaling axis may provide a specific approach to dismantle the stiffness-induced protection mechanism, potentially rendering aggressive, fibrotic tumors vulnerable to standard chemotherapeutic regimens.

Despite these findings, this study has limitations. First, although we validated our findings in patient tissues, the clinical cohort was retrospective. Large-scale prospective studies are required to confirm the correlation between the DCLK1-PIP5K1A axis and patient survival. Second, although we identified PSMA7 as a regulator of DCLK1 stability, the specific dynamics of this ubiquitination process in vivo necessitate further investigation. Subsequently, given the promising results with small-molecule inhibitors in our preclinical models, future efforts should focus on refining these compounds for clinical application and formulating clinical trials to evaluate their efficacy in patients with PDAC with high stromal stiffness. Notwithstanding these limitations, our findings collectively suggest that DCLK1 overexpression and activation are essential in linking matrix stiffness to aggressive tumor progression and chemotherapy resistance.

Despite these limitations, our findings collectively suggest that DCLK1 overexpression and activation are essential in connecting matrix stiffness to accelerated tumor progression and chemotherapy resistance.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Haoxiang Zhang, Jiaoshun Chen, and Chuangbin Zhao: Study design; methodology; collection, analysis, interpretation of data; visualization; writing-original draft. Xiaoqing Hu: Methodology; collection, analysis, interpretation of data; visualization; writing-original draft. Jianwei Bai: Collection, analysis, interpretation of data; resources; validation; formal analysis. Long He and Zanglong Deng: Resources; visualization. Tao Yin: Supervision; funding acquisition; conceptualization; methodology; validation; writing-review & editing; writing-original draft.

Funding

This work was supported by the National Natural Science Foundation of China (81772564, 82173196, 82472989 and 32401082), Key Research and Development Program of Hubei (2022BCA012), the Joint Funds for the Innovation of Science and Technology, Fujian Province (2023Y9311), Fujian Provincial Natural Science Foundation of China (2024J08244) and and Fujian Provincial Health Technology Project (2024GGA004).

Data availability

All data used to support the findings of this study are available from the corresponding author upon request.

Declarations

Ethics approval and consent to participate

All in vivo experiments involving animals in this research followed the ethical standards established by the Institutional Animal Care and Use Committee of Tongji Medical College, Huazhong University of Science and Technology.

Competing interests

The authors declare no competing interests.