ULK1 knockout suppresses pancreatic cancer progression by inhibiting autophagy and enhancing antitumor immunity

Autophagy plays a dual role in cancer, acting as a tumor suppressor and promoter depending on tumor stage and context. While Atg5 and Atg7 are well established core autophagy genes, the role of Unc-51-like kinase 1 (ULK1)—a key autophagy initiator—remains poorly understood in pancreatic ductal adenocarcinoma (PDAC). Here we investigated the role of ULK1 using tissue-specific deletion in genetically engineered mouse models. Although ULK1 messenger RNA levels remained unchanged between normal and tumor cells in The Cancer Genome Atlas dataset, multiplex immunohistochemistry revealed elevated ULK1 activity, marked by pATG14, in high-grade human PDAC tissues. Genetic deletion of Ulk1 impaired autophagy and reduced cell proliferation, colony formation and invasiveness of pancreatic cancer cells. In vivo, both syngeneic orthotopic and KPC (LSL-KrasG12D/+; LSL-Trp53R172H/+; Pdx1-Cre) mouse models with tissue-specific Ulk1 deletion exhibited significant delayed tumor progression, reduced tumor burden and extended survival. Importantly, Ulk1 deficiency remodeled the tumor immune microenvironment by reducing tumor-promoting polymorphonuclear myeloid-derived suppressor cells and neutrophils while substantially enhancing recruitment of cytotoxic CD8+ T cells and major histocompatibility complex II+ antigen-presenting cells. Chemokine and cytokine profiling revealed that downregulation of Cxcl2, Ccl2 and G-CSF might lead to polymorphonuclear myeloid-derived suppressor cell and neutrophil recruitment and survival, with concurrent upregulation of GM-CSF for dendritic cell infiltration, thereby inducing antitumor immunity. These findings provide insights into the role of ULK1 in PDAC progression through tumor-intrinsic metabolic support by autophagy activation and immune modulation by tumor-derived cytokines. Targeting ULK1 may represent a promising therapeutic strategy by inhibiting autophagy and enhancing antitumor immune responses in pancreatic cancer.

Autophagy is a conserved intracellular degradation pathway that sustains cellular homeostasis by recycling macromolecules and clearing damaged organelles. Autophagy plays a crucial role in balancing normal cell functions and multiple pathophysiological conditions, including cancer. While it preserves genome stability in early cancer, it later supports tumor survival by maintaining organelles, enabling recycling and fulfilling metabolic demands1,2. The tumor-promoting role of autophagy has been validated in genetically engineered mouse (GEM) models. Oncogenic mutations in HRas or KRas increase basal autophagy, which is necessary for tumor cell survival during tumorigenesis3. Deletion of Atg5 or Atg7 in KRasG12D-driven spontaneous lung cancer models reduced tumor size compared with that in wild type (WT) mice4,5. Melanomas with BRAF mutations6 and pancreatic cancers harboring 95% KRas mutations also exhibit high dependency on autophagy genes such as Atg5 and Atg77,8. In addition, depleting Atg proteins such as FIP200, Atg16l or Atg4 suppresses tumor growth in various cancers9–11. Accordingly, autophagy inhibition has been considered a promising therapeutic strategy7,8. Unc-51-like kinase 1 (ULK1), a serine–threonine kinase, has gained attention as a druggable target, which plays a crucial role in the initiation step of autophagy12. ULK1 forms a complex with proteins such as ATG13, ATG101 and FIP200/RB1CC1, facilitated by the phosphorylation of ATG1313–15, and subsequently phosphorylates the components of the class III phosphoinositide 3)3 Kinase (PIK3C3)/Vps34 complex, including ATG14 and Beclin1 to drive autophagosome formation16–18. ULK1 activity is tightly controlled by upstream nutrient-signaling molecules such as mTORC1 and AMPK, which phosphorylate ULK1 at distinct residues to either inhibit or enhance its function19. The ULK family includes ULK1, ULK2, ULK3, ULK4 and STK368, with ULK1 and ULK2 showing the highest similarity20. However, ULK1 is the dominant initiator of autophagy, as seen in Ulk1−/− Ulk2−/− double-knockout (KO) mice, which exhibit severe neonatal mortality and autophagy defects similar to other core Atg gene deletions21–24. Although the cellular functions of ULK1 have been studied in several diseases, direct evidence of the role of ULK1 on the tumor progression of pancreatic ductal adenocarcinoma (PDAC) remains largely unexplored. Given the critical dependency of KRas-driven tumors including PDAC on autophagy4,7,8,25, pharmacological or genetic inhibition of KRas-downstream RAF, MEK and ERK pathways has been shown to enhance autophagy, which supports the rationale for combining mitogen-activated protein (MAP) kinase inhibitors with a derivative of chloroquine (CQ), a lysosomotropic agent26–28. However, the therapeutic potential of CQ derivatives is limited by the requirement for high inhibitory doses and poor selectivity, highlighting the need for more potent and selective autophagy modulators. Targeting ULK1, an autophagy-initiating kinase, has emerged as an alternative strategy for autophagy inhibition. A series of ULK1 inhibitors have demonstrated antitumor effects in vitro and in xenograft models29,30, and selective ULK1 inhibitors have shown synergistic tumor regression when combined with KRas or downstream effector inhibitors31. However, despite the increasing interest in ULK1-targeted therapies, the exact role of ULK1 in cancer progression remains poorly understood. Here, we investigate how ULK1 function influences PDAC progression by using syngeneic orthotopic and spontaneous cancer GEM models. Our findings reveal that tumor-intrinsic ULK1 deletion suppresses PDAC progression by impairing autophagy-mediated tumor adaptation and by reprogramming the tumor microenvironment (TME) through altered infiltration of distinct immune cell subtypes, thereby providing tumor-suppressive immune states. These results provide insights into ULK1 as a promising therapeutic target in pancreatic cancer.

MIA PaCa-2 and HEK293T cells were kindly provided by Drs. Yongdoo Choi and Jong Heon Kim (National Cancer Center Korea (NCC Korea)), which were originally purchased from the American Type Culture Collection (ATCC). KPC cell lines were a generous gift from Jong Heon Kim (NCC Korea) and were originally purchased from Ximbio; they were derived from PDAC tumors arising in KPC mice on a C57BL/6 background. All cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS; HyClone), 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco) and were maintained at 37 °C in a humidified incubator with 5% CO2. For amino acid starvation media (-A.A), Earle’s Balanced Salt Solution (EBSS, HyClone) or Hank’s balanced saline solution (HBSS) was supplemented with 10% dialyzed FBS, glucose, vitamins, HEPES and minerals at the same concentrations as in DMEM.

CRISPR–Cas9-mediated KO of Ulk1 was performed with lentiCRISPR v2 vector (a gift from Feng Zhang; Addgene plasmid no. 52961; RRID no. Addgene_52961)32. Single guide (sg)RNA oligo targeting mouse Ulk1 CRISPR–Cas9 guide RNA (U0448BI200-1) or non-target sequence (sgControl) were synthesized (Gene Script). The guide RNA sequence was selected using the CRISPICK online tool (Broad institute, https://portals.broadinstitute.org/gppx/crispick/public) and sgRNA sequences are presented in Supplementary Table 2. Viral transduction processes were followed by standard protocols detailed in the Supplementary Materials and Methods.

Hoechst 33342 (H3570) was purchased from Thermo Fisher Scientific. Rapamycin (R0395) was purchased from Sigma-Aldrich.

Cell proliferation was monitored using the image-based cell proliferation analyzer IncuCyte (Essen Instruments). Cells were seeded in complete DMEM media and imaged throughout the indicated time period. An IncuCyte automated cell proliferation detector was used to measure cell confluence over time. Cell viability was determined by Annexin V and PI staining following standard protocols at the indicated time periods (556547; BD Biosciences), and analyzed using the FACS Verse analyzer (BD Biosciences). Viability was assessed on the basis of double-negative populations, and dead cells were defined by Annexin V+ and/or PI+ staining.

KPC cells (5 × 102 cells per well) were seeded in 12-well plates in complete media (10% FBS in DMEM). After 7 days, cells were fixed for 10 min in 10% formalin (HT501128-4L; Sigma-Aldrich) and stained with 0.25% crystal violet (C6158; Sigma-Aldrich). For three-dimensional (3D) spheroids culture, KPC cells (1 × 103 cells per well) were seeded in 96-well round-bottom Ultra Low Attachment plates (7007; Corning) mixed with 1% Matrigel (354234; Corning) and complete media. Spheroid growth was monitored for 7 days and imaged using HCS system Operetta CLS (PerkinElmer). Quantification was performed by using Harmony software (PerkinElmer).

Invasion assay was performed to ascertain cell invasion using a 0.8-μm Transwell apparatus (Corning) with coated Matrigel (354234; Corning) on a 24-well plate. KPC cells (1 × 105) suspended in 100 μl serum-free medium were seeded into the upper chamber and complete DMEM was added to the lower chamber. After incubation at 37 °C for 24 h, invaded cells were stained using Diff-Quik reagents (Sysmex Corporation) and counted under an inverted microscope at a 4× and 20× magnifications.

KrasG12D/+ (B6.129S4-Krastm4Tyj, strain no. 01XJ6) was received from the National Cancer Institute (NCI) mouse repository. Trp53R172H/+ (129S-Trp53tm2Tyj/J, strain no. 008652) and Pdx1-Cre (B6.FVB-Tg (Pdx1-Cre)6Tuv/J, strain no. 014647) mice were purchased from the Jackson Laboratory. These mice were backcrossed more than six times with C57BL/6 and were subsequently used for generating KPC (KrasG12D/+; Trp53R172H/+; Pdx1-Cre) mice. The KPC model of PDAC was first described in 2005 and incorporates, through Cre-lox technology, the conditional activation of mutant endogenous alleles of the Kras and Trp53 gene33. Ulk1fl/fl (B6.129-Ulk1tm1Thsn/J, stock no. 017976) mice were purchased from The Jackson Laboratory, which were generated and donated by Craig B. Thompson (Memorial Sloan Kettering Cancer Center)24. The Ulk1fl/fl and KPC mice were crossed, and after several rounds of breeding, KPC;Ulk1fl/fl mice were successfully obtained. All animal procedures were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee (IACUC) of the National Cancer Center (NCC). The NCC is an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility that abides by the guidelines of the Institute of Laboratory Animal Resources Guide and Usage Committee. The methods applied in this study were performed in accordance with the approved guidelines.

Here, 6-week-old female C57BL/6 mice (Orient Bio) were orthotopically injected with 2 × 105 KPC sgSC or KPC sgUlk1 cells suspended in 1:1 Matrigel (50 µl) into the pancreatic parenchyma. After injection, the peritoneum and skin were closed with a 6-0 suture (639 G; Ethicon). A total of 3 weeks after the allograft, the pancreas was isolated from the mice and used for the indicated experiments.

Pancreatic tissues from indicated orthotopically transplanted mice or KPC;Ulk1+/fl and KPC;Ulk1fl/fl GEM mice were isolated from 12–17-week-old mice, respectively. The tissues were fixed in 10% neutral buffered formalin (BN019, Biosolution) and paraffin embedded. Sections (4 μm thick) of mouse pancreas were deparaffinized, rehydrated and incubated in boiling CC1(pH 9) or CC2(pH 6) for antigen retrieval by Benchmark TX (950-123 or 124; Ventana Medical Systems) or Discovery XT (Ventana Medical Systems). Immunohistochemical (IHC) staining was performed with the indicated primary antibodies (Supplementary Table 1) and then detected with a DAB detection kit (Ventana Medical Systems) according to the manufacturer’s instructions, followed by counterstaining with hematoxylin (Ventana Medical Systems). The stained images were acquired using the Vectra Polaris (PerkinElmer/Akoya Biosciences) and quantified using inForm Tissue Analysis Software (Akoya Biosciences). We calculated the H-score as follows: H-Score = (1 × % of 1+ staining spot) + (2 × % of 2+ staining spot) + (3 × % of 3+ staining spot). Statistical analyses were performed using Graphpad Prism 8.0.1.

Tissue microarray (TMA) sections from patients with PDAC (PA241e, TissueArray.Com) and the formalin-fixed paraffin-embedded (FFPE) pancreatic tissue from KPC Ulk1+/fl and KPC Ulk1fl/fl mice were subjected to multiplex immunohistochemistry (multi-IHC) with indicated antibodies (Supplementary Table 1) and automated staining Leica Bond RX (Leica Biosystems) using an Opal 7 kit (nos. NEL871001KT and ARR1001KT; Akoya Biosciences). Tyramide signal amplification fluorophores were used to combine with each antibody separately according to manufacturer’s instructions. Imaging and quantification were performed by Vectra Polaris Automated Quantitative Pathology Imaging System (PerkinElmer) and inForm image analysis software (Akoya Biosciences). Data were compiled and analyzed after merging and consolidating each case by R studio (version 2021.09.2.0).

For tumor immunophenotyping, tumor tissues were dissociated in a fully automated way by using the gentleMACS Octo Dissociator with Heaters (130-096-427; Miltenyi Biotec) with collagenase P (C7657, Merck) and the Tumor Dissociation Kit (130-096-730; Miltenyi Biotec), which are optimized for epitope preservation. Digested tissues were filtered through a 70-μm nylon cell strainer (352350; BD Falcon). Cells were blocked with antimouse CD16/CD32 antibody (Mouse BD Fc Block) (BD553142, clone 2.4G2; BD Biosciences) and stained with the indicated antibodies for 20 min at 4 °C. Intracellular staining was performed using a Fixation/Permeabilization Kit (BD554714; BD Bioscience). Samples were analyzed on a LSRFortessa Cell Analyzer (BD Biosciences).

Cytokine profiles were assessed using the Proteome Profiler Mouse Cytokine Array Kit (ARY006; R&D systems) following the manufacturer’s instruction. Detailedly, mouse pancreatic cancer tissues were homogenized by Tissue Lyser II (QIAGEN) in PBS containing protease inhibitor cocktail (Roche) and then lysed with 1% Triton X-100. Cellular debris were removed by centrifugation at 10,000g for 5 min at 4 °C, and protein was quantified using BCA Protein Assay (23227; Thermo Fisher Scientific). Conditioned media (CM) were collected from cells cultured under identical conditions for 24 h. Media were then normalized to the total cellular protein content, as determined from corresponding cell lysates. Membranes incubated with CM were developed and visualized using the FUSION SOLO S imaging system (Vilber) with exposure times ranging from 2 to 6 min.

Pancreas, spleen and blood from the abdominal aorta were collected and dissociated as indicated above. Polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC)/neutrophils (CD45+CD11b+Ly6G+), CD8+ T cells (CD45+CD3+CD8+) and dendritic cells (DCs) (CD45+CD11c+) were sorted using Melody (BD Biosciences) and then cultured in the presence of 50% CM from KPC sgSC or KPC_sgUlk1 cells on a 384- or 96-well black plate (6057308 and 6055302; PerkinElmer). After incubation for the indicated time, live cell numbers were assessed using calcein-AM (C1430; Invitrogen) staining. Imaging and quantification were performed using the Operetta CLS system and Harmony software (PerkinElmer).

The Cancer Genome Atlas (TCGA)-pancreatic adenocarcinoma (PAAD) patients were stratified with the autophagy score using the surv_cutpoint function from the survminer package (version 0.4.9). Kaplan–Meier overall survival plots and the P value between patient groups were calculated using the log-rank test implemented in the survival package (version 3.5-7). Cox hazard ratios and confidence intervals were calculated using the coxph function of the survival package. The autophagy score was calculated using a modified version of ssGSEA2.0, which is an adaptation of the GSEA algorithm for single-cell analysis34,35. The list of genes used for the autophagy score calculation is provided in the Supplementary Information. Default parameters for ssGSEA were used to calculate scores for TCGA-PAAD patients and for cells in single-cell data from patients with pancreatic cancer (CRA001160)35,36. To stratify patients into high and low autophagy groups, we used the surv_cutpoint function from the survminer package in R. This method identifies the optimal threshold that best separates patient groups on the basis of survival outcome. The resulting cutoff value was 6.975026, with patients scoring greater than or equal to this value classified as ‘high autophagy’.

We initially downloaded 183 PAAD samples directly from the TCGA_PAAD cohort, as described in the Methods. Of these, 178 patients had valid overall survival information and 162 patients had both survival information and complete somatic mutation single-nucleotide variant (SNV) profiles. For the survival analyses, we included only those patients with sufficient clinical metadata. Specifically, the 162 patients with both survival data and complete mutation profiles were used in cases where mutation-based subgrouping or stratification was considered.

Immunoblot quantifications were performed using ImageJ software version 1.50i (NIH). Statistical significance was calculated using Student’s t-test in GraphPad Prism 8. P < 0.05 was considered statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001). Data are expressed as standard error of the mean (s.e.m.), which are from at least three independent experiments.

To assess the clinical relevance of autophagy proteins in pancreatic cancer, we analyzed the expression of 25 core autophagy proteins in PAAD from TCGA. These autophagy gene sets were significantly upregulated in tumor tissues compared with adjacent normal tissues (P < 0.05; Supplementary Fig. 1a–c), and high expression levels of autophagy gene sets correlated with poor 5-year overall survival (Supplementary Fig. 1d), supporting the tumor-promoting role of autophagy in PDAC. Interestingly, the messenger RNA level of ULK1 was not significantly different between tumor and normal tissues (Supplementary Fig. 1e). However, multi-IHC revealed elevated ULK1 protein levels and activity, marked by phosphorylated ATG14 (pATG14), particularly in high-grade (grade 3) pancreatic tumors compared with normal tissues (Fig. 1a–d and Supplementary Fig. 2a,b).Fig. 1ULK1 expression and activities are upregulated in pancreatic cancers.a,b Representative images (a) and quantitative data (b) of ULK1 (yellow) and pATG14 (red) in human PAAD TMA using multi-IHC. Tissues from six patients were analyzed, each including matched adjacent normal and tumor tissue (four cores per patient). Images from representative grade 3 PAAD cases are shown. C7 and C8 cores indicate adjacent normal pancreas tissue (control (CTL)), while C5 and C6 represent adenocarcinoma (Adeno) tumor regions. Scale bars are indicated in the figures. Error bars indicate the mean ± s.e.m. from two independent cores (ULK1 P = 0.2295; pATG14 P = 0.0050). c,d Representative images (c) and quantification (d) of pATG14 by grade using IHC staining in adenocarcinoma cores from human PAAD TMA, stratified by tumor grade (grade 1 versus 3 P = 0.0016). e Immunoblot analysis of ULK1, pATG14, ATG101 and β-Actin (loading control) in human pancreatic cancer cell lines. f Representative images of hematoxylin and eosin (H&E) staining and IHC staining for CK19, Ulk1 and pAtg14 in pancreas tissue from K, KC and KPC mice. Scale bars are indicated in the figures. All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01). Similarly, increases in ULK1 and pATG14 levels were observed in human pancreatic cancer cell lines compared with normal human pancreatic epithelial and fibroblast (IMR90) cells (Fig. 1e). Moreover, elevated expression of ULK1 and pATG14 was detected in transformed pancreatic intraepithelial neoplasia in Kras+/LSL-G12D; Pdx1-Cre (KC) mice and in advanced PDAC lesions from Kras+/LSL-G12D; Trp53R172H/+; Pdx1-Cre (KPC) mice (Fig. 1f). These results suggest that ULK1 activity rather than its mRNA levels is markedly elevated in malignant pancreatic cancers, supporting its potential role in tumor progression.

To investigate the functional role of ULK1 in pancreatic cancer progression, we generated Ulk1 KO KPC cells using CRISPR–Cas9, which was confirmed by genomic sequencing and quantitative PCR with reverse transcription (Supplementary Fig. 3a,b). Ulk1 KO cells exhibited significantly impaired cell growth compared with control cells, as measured by IncuCyte live-cell imaging (Fig. 2a) and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Fig. 2b). ULK1 loss also reduced colony formation (Fig. 2c) and markedly impaired 3D spheroid growth (Fig. 2d,e). In addition, Transwell invasion assay showed a substantial decrease in invasion of Ulk1 KO cells compared with sgScrambled (sgSC) control (Fig. 2f,g), indicating a significant reduction in both proliferative and invasive potential.Fig. 2Ulk1 regulates tumorigenesis in an orthotopic pancreatic cancer model.a Relative cell proliferation between sgSC, sgUlk1 KPC clone 1 (c1) and clone 2 (c2) cells analyzed by IncuCyte over 48 h (P < 0.0001 for all time points after 2 h). b Relative cell viability between sgSC, sgUlk1 c1 and c2 KPC cells analyzed by MTT assay over 72 h (sgSC versus sgUlk1 c1 P = 0.0015, sgSC versus sgUlk1 c2 P = 0.0022). c The colony formation assays using sgSC, sgUlk1 c1 and c2 KPC cells. d,e Bright field images (d) and quantification (e) showing 3D spheroid formation using sgSC, sgUlk1 c1 and c2 KPC cells. Image analysis and quantification were performed using the ImageJ program (sgSC versus sgUlk1 c1 P = 0.1049, sgSC versus sgUlk1 c2 P = 0.1705). f,g Representative images captured at 4× and 20× magnification (f) and quantification (g) of Transwell migration and Matrigel invasion assays analyzed using sgSC, sgUlk1 c1 and c2 KPC cells (sgSC versus sgUlk1 c1 P = 0.0001, sgSC versus sgUlk1 c2 P = 0.0001). (h) Immunoblot analysis of Ulk1, pAtg14, Atg14, LC3 and β-Actin (loading control) in sgSC or sgUlk1 KPC cells cultured in complete medium (COM) or replaced by amino acid-deprived media for 2 h with or without 100 nM bafilomycin A1. i,j, Representative images of GFP-LC3 puncta (i) and quantification (j) in MIA PaCa-2 cell line harboring shControl (shCTL) or shULK1 after treatment with either 4 µM of rapamycin (Rapa) or vehicle (Veh) for 4 h (area, P = 0.0002; number, P < 0.0001). k,l Representative images (k) and quantitative data (l) of H&E staining and Ki67 and Ulk1 by IHC in pancreatic tissues obtained from orthotopic models of PDAC injected by KPC sgSC or KPC sgUlk1 (Ki67 P = 0.0008/H-score P = 0.5206, Ulk1 P < 0.0001/H-score P < 0.0001). (m) Survival graph illustrating orthotopic mouse models of PDAC, with each group consisting of n = 5 mice (P = 0.0214). The models were generated using KPC cells carrying either sgSC or sgUlk1, and the observation period extended up to 119 days. Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for three independent images. All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). To determine whether these phenotypes were linked to autophagy, we assessed autophagic flux by analyzing LC3 processing and localization. Compared with Ulk1 WT cells, Ulk1-deficient cells showed LC3-I accumulation and reduced conversion to LC3-II, particularly under amino acid-deprived conditions. This was further supported by a significantly lower LC3-II:I ratio in bafilomycin A1-treated cells, indicating defective autophagic flux upon Ulk1 deletion (Fig. 2h and Supplementary Fig. 3c). In addition to the reduced LC3-II conversion rate, the absolute LC3-II levels, normalized by loading control, were also significantly lower in Ulk1-deficient cells even with the treatment of bafilomycin A1 (Fig. 2h and Supplementary Fig. 3c). Rapamycin-induced GFP-LC3 puncta formation was also significantly impaired in ULK1 KO human pancreatic cancer cells (Fig. 2i,j), further confirming autophagy suppression and defect in viability (Supplementary Fig. 3d). These findings indicate that ULK1 is required for nutrient-responsive autophagy and that its depletion impairs both the growth and invasion of pancreatic cancer cells.

To extend our in vitro findings, we investigated the in vivo role of ULK1 in tumor progression using a syngeneic mouse model. An orthotopic PDAC mouse model was established by injecting Ulk1 KO (sgUlk1) or control (sgSC) KPC cells into the pancreas of C57BL/6J mice. Histological analysis 3–5 weeks after injection revealed reduced Ki67 staining in Ulk1-deficient tumors compared with controls, consistent with decreased Ulk1 expression (Fig. 2k,l). Moreover, mice bearing Ulk1 KO tumors showed significantly extended survival compared with control KPC tumor-injecting mice (Fig. 2m). These findings support a critical role of Ulk1 in promoting cell proliferation and tumor progression in anorthotopically transplanted in vivo model.

To further investigate the role of ULK1 in pancreatic cancer progression, we used the LSL-KrasG12D/+; LSL-Trp53 R172H/+; Pdx1-Cre (KPC) mouse model33, which spontaneously develops pancreatic tumors that mimic the pathology of human PDAC. These mice were crossbred with previously established Ulk1fl/fl mice24 to generate KPC;Ulk1fl/fl mice, enabling pancreas-specific Ulk1 deletion via Cre recombinase (Fig. 3a). In this GEM model, the pancreas-specific expression of Cre recombinase, driven by Pdx1-Cre, induces a constitutively active Kras mutation (G12D) and a dominant negative Trp53 mutation (R172H) (Supplementary Fig. 4a). Immunoblotting confirmed efficient Cre-mediated Ulk1 deletion in pancreatic tissues (Fig. 3b). Histological analysis showed only crucial abnormalities in pancreas rather than other tissues, although occasional lung pathology was observed in positive control (KPC;Ulk1fl/+) mice, suggesting that Ulk1 deletion specifically impacts pancreatic tumor development (Fig. 3c and Supplementary Fig. 4b). However, pancreas weights were significantly reduced in KPC;Ulk1fl/fl mice compared with that of KPC control mice (KPC;Ulk1fl/+) (Fig. 3d). Moreover, 18F-fluorodeoxyglucose positron emission tomography–computed tomography (18F-FDG-PET-CT) imaging revealed a reduced tumor burden and lower maximum standardized uptake values (SUVmax) in KPC;Ulk1fl/fl mice, compared with KPC;Ulk1fl/+ control mice (Fig. 3e and Supplementary Fig. 4c). Notably, increased fluorodeoxyglucose (FDG) uptake in the lungs, indicative of distant metastasis, was observed in KPC;Ulk1fl/+ control mice but diminished in KPC;Ulk1fl/fl mice. These findings suggest that Ulk1 deletion primarily impacts pancreatic tumor development in the KPC model, with occasional lung metastasis in the late stages of cancer.Fig. 3Pancreas-specific Ulk1 deletion suppresses pancreatic cancer development in KPC GEM model.a Breeding strategy for generation of KPC;Ulk1fl/fl mice. b Immunoblot analysis of ULK1, pATG14 and β-Actin (loading control) in the indicated organs of KPC;Ulk1fl/fl mice compared with controls (C;Ulk1fl/+ or KPC;Ulk1+/+). c, Macroscopic pancreas morphology (top) and H&E staining (bottom) of various organs from normal CTL, KPC;Ulk1fl/+ and KPC;Ulk1fl/fl mice. d Quantification of pancreas weight (n = 6; P = 0.0207). Error bars indicate the mean ± s.e.m. e Positron emission tomography–computed tomography imaging showing tumors indicated by arrows from CTL, KPC;Ulk1fl/+ and KPC;Ulk1fl/fl mice. Next, further histopathological results from the pancreas showed that all KPC control mice (KPC;Ulk1fl/+ or KPC;Ulk1+/+) developed moderate-to-malignant PDAC within 18 weeks, whereas only 12% (1/8) of KPC;Ulk1fl/fl mice (pancreas-specific Ulk1 KO KPC) developed low-grade pancreatic intraepithelial neoplasia lesions, with the majority displaying normal pancreatic architecture with healthy acini cells (Fig. 4a,b). IHC staining further revealed significantly lower levels for cytokeratin 19 (CK19) and Ki67 in KPC;Ulk1fl/fl pancreatic tissues compared with those from KPC control (KPC;Ulk1fl/+) mice, consistent with suppressed epithelial transformation and proliferative capacity (Fig. 4c,d). These findings aligned with reduced expression of Ulk1 and pAtg14 in the pancreatic ductal regions of KPC;Ulk1fl/fl mice, which was more pronounced in the transformed regions of control KPC mice (Fig. 4c,d). In agreement with impaired autophagic flux, LC3B, an autophagy marker, was significantly decreased in the pancreas of KPC;Ulk1fl/fl mice (Supplementary Fig. 4d,e).Fig. 4Ulk1 activity is associated with pancreatic tumor transformation in KPC mice.a,b Representative images of H&E staining (a) and quantification of tumor area (b) in pancreatic tissues from KPC;Ulk1fl/+ and KPC;Ulk1fl/fl GEM mice (n = 6; P = 0.0087). Error bars indicate the mean ± s.e.m. for six independent experiments (P = 0.0087). c,d Representative images (c) and quantitative data (d) of H&E staining and CK19, Ki67, Ulk1 and pAtg14 by IHC in pancreas tissues from KPC;Ulk1fl/+ and KPC;Ulk1fl/fl mice (CK19 H-score P = 0.017, Ki67 H-score P = 0.0035, Ulk1 H-score P = 0.0386 and pAtg14 H-score P = 0.0187). e,f Multi-IHC images (e) and quantification (f) for H&E, CK19 (white), Ulk1 (yellow) and pAtg14 (red) in pancreas tissues from KPC;Ulk1fl/+ and KPC;Ulk1fl/fl mice (CK19 P = 0.0898, Ulk1 P = 0.5825, pATG14 P = 0.0255). Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for over four independent experiments. g,h Spatial analysis of CK19+ and pAtg14+ cells with quantification of their proximity in tumor tissues. Heat map showing median distance (g, left) and cell count (g, right) and quantitative data (h) representing pAtg14+ cells near or far from CK19+ malignant cell phenotypes across the mouse cohort. Error bars indicate the mean ± s.e.m. for three independent experiments. (at 10 µm, P = 0.1313). i, Kaplan–Meier survival analysis of KPC;Ulk1fl/+ (n = 10) and KPC;Ulk1fl/fl (n = 9) mice (P = 0.0011). All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). Multi-IHC further confirmed that Ulk1 and pAtg14 levels were significantly enriched in CK19+ epithelial cells within tumors from KPC;Ulk1fl/+ mice, while these signals were nearly absent in the tissue from KPC;Ulk1fl/fl mice (Fig. 4e,f). Spatial analysis supported a strong association between CK19+ adenocarcinoma cells in and pAtg14+ autophagy-active cells in KPC;Ulk1fl/+ tumors, which was largely diminished in Ulk1-deficient mice (Fig. 4g,h). Importantly, KPC;Ulk1fl/fl mice exhibited a significantly prolonged median overall survival—approximately 10 weeks longer than that of KPC;Ulk1fl/+ controls (Fig. 4i). Collectively, these data demonstrate that Ulk1 is critical for PDAC progression in a physiologically relevant spontaneous PDAC model and support its function as a key effector of both primary tumor growth and metastatic potential.

Given the role of autophagy in modulating tumor immunity37,38, we next examined whether Ulk1 deletion alters immune cell composition within the TME. IHC analysis revealed that cancer-associated fibroblast (CAF) markers (αSMA, PDPN and FAP) and immunosuppressive M2 macrophage marker CD204 were more abundant in tumors from KPC;Ulk1fl/+ control mice, particularly in CK19+ adenocarcinoma lesions, whereas these markers were notably reduced in the pancreas of KPC;Ulk1fl/fl mice (Fig. 5a,b). By contrast, cytotoxic T cell (CD8α) and natural killer (NK) cell (NCR1) staining tended to be substantially higher in Ulk1 KO tumors compared with controls, suggesting a possible shift toward a more antitumor immune profile. M1 macrophage markers (CD86 and iNOS) and CD4+ T cell levels did not differ significantly between groups (Fig. 5a,b).Fig. 5Pancreas-specific Ulk1 depletion alters tumor immune microenvironment in KPC mice.a,b Representative images (a) and quantitative data (b) of H&E staining and αSMA, PDPN, FAP, CD86, iNOS, CD204, CD4, CD8α and NCR1 in pancreas tissue from KPC;Ulk1fl/+ and KPC;Ulk1fl/fl GEM mice. Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for over four independent experiments (αSMA P = 0.0486, PDPN P = 0.0296, FAP P = 0.0421, CD86 P = 0.2302, iNOS P = 0.4515, CD204 P = 0.0350, CD4 P = 0.0791, CD8α P = 0.2718 and NCR1 P = 0.1920). All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). Multi-IHC further confirmed that, CD204+ M2 macrophages were more abundant in control tumor (KPC;Ulk1fl/+) mice, while CD8+ T cells were substantially more enriched in the pancreas of KPC;Ulk1fl/fl mice than in control KPC mice, although the results were not statistically significant (Supplementary Fig. 5a–c). Collectively, these findings indicate that tumor Ulk1 deletion remodels the pancreatic TME primarily by reducing immunosuppressive immune cell populations, along with a modest increase in cytotoxic lymphocyte infiltration.

To elucidate the molecular mechanisms underlying tumor regression following Ulk1 depletion, we performed RNA-sequencing analysis of Ulk1 WT and Ulk1 KO KPC cells. Differential expression gene analysis identified 4,097 genes with altered expression, including 1,071 upregulated and 1,538 downregulated genes in Ulk1 KO cells, visualized by heat maps and volcano plots (Supplementary Fig. 6 and Supplementary Table 6). Hierarchical clustering of differentially expressed genes showed distinct profiles between the genotypes under both amino acid-complete (com) and -deprived conditions (-A.A). (Supplementary Fig. 6a). Volcano plots were also visualized using log2 fold changes (log2FC| > 2), and P values (P < 0.05) derived from comparisons between WT-com and KO-com (Supplementary Fig. 6b) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of upregulated gene sets in Ulk1 KO cells revealed significant enrichment in ‘MAP kinase signaling’, ‘PI3K–Akt signaling’ and ‘cytokine–cytokine receptor interaction’ pathways (Supplementary Table 3). Among these pathways, the gene set of cytokine–cytokine interaction pathway upregulated in Ulk1 KO cells was indicated in the volcano plot (Supplementary Fig. 6b). In addition, proteomic profiling of KRasG12D-expressing human pancreatic epithelial cells treated with CQ revealed a similar upregulation of immune-related pathways, with ‘cytokine–cytokine receptor interaction’, ‘JAK–STAT signaling’, ‘TGF-β signaling’ and ‘mitophagy’ among the most enriched pathways (Supplementary Table 4). Complementing these results, gene set enrichment analysis (GSEA) of TCGA pancreatic cancer datasets34 revealed that lower ULK1 expression correlated with elevated immune and mitochondrial gene signatures. Gene ontology analysis of gene clusters negatively co-expressed with ULK1 revealed strong enrichment for terms related to ‘antigen processing and presentation’, ‘immune system processes’ and ‘mitochondrial metabolism’, suggesting a link between ULK1 suppression and activation of immune and mitochondria metabolic pathways (Supplementary Table 5). To further assess whether tumor-intrinsic Ulk1 deletion affects immune cell composition in vivo, we used syngeneic orthotopic models by injecting KPC sgSC (Ulk1 WT) and KPC sgUlk1 (Ulk1 KO) cells into the pancreas of immunocompetent C57BL/6J mice. After 3 weeks, the mice were killed and subjected to immunophenotyping by fluorescence-activated cell sorting (FACS) analysis using the gating strategy shown in Supplementary Fig. 7. Although the overall proportions of tumor-infiltrating CD45+ immune cells and CD45+CD11b+ myeloid cells remained unchanged between the two groups, Ulk1 KO-derived tumors showed a marked reduction in the proportion of neutrophils (CD45+CD11b+Ly6G+) and PMN-MDSCs (CD11b+Ly6G+F4/80−) compared with tumors derived from Ulk1 WT KPC cells, alongside a significant increase in major histocompatibility complex (MHC)-II+ antigen-presenting cells (APCs) (CD45+CD11b+ MHC-II+) in tumors derived from Ulk1 KO KPC cells (Fig. 6a,b).Fig. 6Tumor Ulk1 deletion modulates myeloid cell recruitment in both orthotopic and KPC GEM models.a,b Representative flow cytometry plots (a) and quantification (b) of the expression by the myeloid immune cell types analyzed by FACS in pancreas tissue from the orthotopic mouse model (n > 3; CD45+ P = 0.376302, myeloid cell P = 0.742007, MHC-II+ P = 0.003583, PMN-MDSC P = 0.000378 and neutrophil P = 0.000079). The 7-week-old female C57BL/6 mice were injected with KPC sgSC or KPC sgUlk1 cells and then killed 3 weeks later. c,d Representative flow cytometry plots (c) and quantification (d) of the expression by the myeloid immune cell types analyzed by FACS in pancreas tissue from the ~12–14-week-old KPC;Ulk1fl/+ and KPC;Ulk1fl/fl GEM mice (n > 4; CD45+ P = 0.378793, myeloid cell P = 0.038968, MHC-II+ P = 0.391091, PMN-MDSC P = 0.007721 and neutrophil P = 0.005358). Error bars indicate the mean ± s.e.m. e,f Representative images (e) and quantitative data (f) of Ly6G, OPN/SPP1 and VEGFA in pancreas tissue from KPC;Ulk1fl/+ and KPC;Ulk1fl/fl GEM mice. Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for over three independent experiments (Ly6G P = 0.0121, OPN/SPP1 P = 0.1019 and VEGFA P = 0.0482). All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). In the lymphoid compartment, Ulk1 KO cell-derived tumors exhibited a substantial increase in the proportion of CD8+ T cells compared with Ulk1 WT tumors; however, this difference did not reach statistical significance (Supplementary Fig. 8a). CD4+ T cell proportions were moderately decreased in Ulk1 KO tumors, resulting in an overall increase in the CD8+:CD4+ T cell ratio relative to Ulk1 WT tumors. Notably, the effector memory subset among CD8+ T cells (CD44+, CD62L− in CD8+ cells) was significantly higher portion in Ulk1 KO tumors, despite the modest change in total CD8+ T cell levels. This enrichment of effector memory CD8+ T cells may be associated with the enhanced recruitment of MHC-II+ APCs observed in Ulk1 KO tumors. To validate these findings in a spontaneous PDAC model, we performed similar immunophenotyping in KPC control and KPC;Ulk1fl/fl mice (Fig. 6c,d). Similar to the orthotopic model, we observed increased proportions of MHC-II+ APCs and significantly reduced neutrophils and PMN-MDSCs in KPC;Ulk1fl/fl mice. IHC analysis further confirmed that immune-suppressive neutrophils and PMN-MDSC markers (Ly6G, OPN/SPP1 and VEGFA) were markedly reduced in the pancreas of KPC;Ulk1fl/fl mice (Fig. 6e,f). Notably, well-established protumorigenic neutrophil markers, OPN/SPP1 and VEGFA were significantly lower in KPC;Ulk1fl/fl tumors compared with KPC;Ulk1fl/+tumors, consistent with the reduced Ly6G+ cell infiltration observed in KPC;Ulk1fl/fl. These results suggest that protumorigenic neutrophil subsets are more abundant in KPC;Ulk1fl/+tumors than in KPC;Ulk1fl/fl tumors. Within lymphoid populations, KPC;Ulk1fl/fl mice showed a modest increase in CD8+ T cell proportions, whereas the CD4+ T cell proportion remained comparable to those in KPC;Ulk1fl/+ control mice, which was less pronounced than that observed in the syngeneic orthotopic model (Supplementary Fig. 8b). Collectively, these data from two independent in vivo models indicate that tissue-specific Ulk1 depletion enhances antitumor immune responses by markedly reducing immunosuppressive neutrophils and PMN-MDSCs, promoting APCs and substantially increasing CD8+ T cell infiltration, thereby contributing to tumor regression.

To further dissect how ULK1 shapes the immune landscape in pancreatic tumors, we profiled cytokine and chemokine expression in pancreatic tissues from KPC control (KPC;Ulk1fl/+) and KPC Ulk1 KO (KPC;Ulk1fl/fl) mice. Several protumorigenic cytokines and chemokines including ICAM-1/CD54, TIMP1, interleukin (IL)-1RN (IL-1ra) and IL-1634,39–41 were upregulated in KPC control pancreatic tissues compared with those in KPC;Ulk1fl/fl mice (Supplementary Fig. 9). While Ccl2 was modestly downregulated in KPC;Ulk1fl/fl tissues, Cxcl12 levels remained unchanged. Cytokine and chemokine profiling of CM from Ulk1 WT and Ulk1 KO KPC cells revealed distinct secretory profiles, reinforcing the tumor-promoting immune landscapes (Fig. 7a,b). The most dominantly altered cytokines and chemokines between the two groups are shown in Fig. 7a,b. Icam-1/CD54 and Timp1 levels, which showed difference in pancreas tissues from GEM mice, were either undetectable or comparable in CM from KPC sgSC and KPC sgUlk1 cells, suggesting that these factors probably originate from stromal fibroblasts rather than from tumor epithelial cells. By contrast, protumorigenic chemokines Cxcl2 and Ccl2 —known to promote the recruitment of PMN-MDSCs and M2 macrophages42—were markedly reduced in Ulk1 KO cells, consistent with previous immunophenotyping data from pancreatic tissues of spontaneous cancer model (Fig. 7). Moreover, the secreted levels of G-CSF and Ccl2 were notably reduced, but GM-CSF levels were significantly elevated in in CM from Ulk1 KO cells, indicating that Ulk1 determines to regulate tumor-intrinsic G-CSF, Ccl2 and GM-CSF secretion oppositely (Fig. 7a,b). These findings were confirmed by quantitative PCR (qPCR), which reduced expression of Ccl2, IL-1rn, Cxcl2 and Csf3 (G-CSF) in Ulk1 KO cells but increased Csf2 (GM-CSF) expression (Fig. 7c). Enzyme-linked immunosorbent assay (ELISA) results also exhibited the concentration of Ccl2, G-CSF and GM-CSF secreted from both genotyped cells, which were consistent with their gene expression levels (Fig. 7d).Fig. 7Ulk1 regulates the expression and secretion of cytokines and chemokines critical to the TME.a,b Cytokine array images (a) and quantification (b) of CM from sgSC or sgUlk1 KPC cells. Key differences were observed in G-CSF (P = 0.0002), GM-CSF (P = 0.0040), IL-1ra (P = 0.0048), Ccl2 (P = 0.0022), Cxcl2 (P = 0.0104) and Timp1 (P = 0.0064). Red squares indicate cytokines with notable differences between two groups, and bar graphs show pixel densities analyzed by ImageJ. c qPCR analysis of cytokine and chemokine mRNA in sgSC or sgUlk1 KPC cells. (Ccl2 P < 0.0001, Il1ra P = 0.539, Cxcl1 P = 0.0142, Cxcl2 P = 0.0201, Cxcl12 P = 0.7400 and Csf3 P = 0.0258). d The secreted levels of cytokines from sgSC or sgUlk1 KPC cells using ELISA analysis (G-CSF P = 0.0017, GM-CSF P < 0.0001 and Ccl2 P = 0.0362). Each type of cell was seeded, incubated for 24 h and then harvested for either RNA extraction of cells or CM. Error bars indicate the mean ± s.e.m. for over three independent experiments. e Viability analysis of sorted immune cells cultured with CM from sgSC or sgUlk1 cells. Neutrophils including PMN-MDSCs (CD45+CD11b+Ly6G+ cells), cytotoxic T cells (CD45+CD8+ cells) and DCs (CD45+CD11c+ cells) from the pancreas in ~15–16-week-old KPC;Ulk1+/fl mice were sorted and then cultured with CM from KPC sgSC or KPC sgUlk1 (sgSC CM or sgUlk1 CM) for 24 h, which were measured calcein-stained live cells using Operetta CLS. Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for over two independent wells (PMN-MDSCs + neutrophils in pancreas count P = 0.0668, normalized P = 0.0124; cytotoxic T cells in pancreas count P = 0.6024, normalized P = 0.3923; DCs in pancreas count P = 0.0637, normalized P = 0.0298). f,g Representative images (f) and quantitative data (g) of CD163 (orange), CD8α (cyan) and MHC-II (magenta) by multi-IHC in grade 3 human PAAD TMA. C7 and C8 cores represent cancer-adjacent pancreas tissue or adjacent normal pancreas tissue (CTL), but C5 and C6 represent adenocarcinoma (Adeno). Scale bars are indicated in the figures, and error bars indicate the mean ± s.e.m. for two independent cores (CD163 P = 0.0453, CD8α P = 0.0004 and MHC-II P = 0.0115). All values were considered statistically significant by Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). Together, these data demonstrate that Ulk1 depletion suppresses the secretion of key cytokines and chemokines that support the recruitment of tumor-promoting immune cell subsets. However, Ulk1 loss oppositely increased a crucial cytokine such as GM-CSF that promotes the tumor-suppressive immune cell populations, through tumor-intrinsic signaling, ultimately leading to a less immune-suppressive TME and further reduced tumor progression.

To investigate whether tumor-derived factors directly influence tumor-infiltrated immune cell survival, we isolated freshly immune cells from either pancreas tumor tissues or the blood or spleen of tumor-bearing KPC mice at 15 weeks old and sorted neutrophils including PMN-MDSC (CD45+CD11b+Ly6G+), cytotoxic T cells (CD45+CD3+CD8+) and DCs (CD45+ CD11c+), respectively. These sorted immune cells were cultured in CM from either Ulk1 WT or KO KPC cells for 24 h, and we then assessed their viability using calcein-AM staining. Neutrophils including PMN-MDSCs from tumors showed significantly higher viability cultured in CM from KPC WT cells (Fig. 7e and Supplementary Fig. 10a). By contrast, the viability of CD8+ T cells and DCs from tumors was relatively higher in Ulk1 KO CM than in KPC WT. Interestingly, these viability differences were not observed in immune cells isolated from the blood or spleen of tumor-bearing KPC mice when cultured in the same CM, suggesting that Ulk1-dependent tumor-secreted factors specifically affect immune cells within the tumor context (Supplementary Fig. 10b,c). Collectively, these results reveal that ULK1 orchestrates a tumor-intrinsic cytokine and chemokine program that shapes an immunosuppressive microenvironment by supporting the survival of neutrophils and PMN-MDSCs and suppressing DC and cytotoxic T cell function—mechanisms that are disrupted upon ULK1 deletion. Finally, to evaluate the translational relevance of these findings, we analyzed TMAs from human PDAC samples. Grade 3 adenocarcinoma regions displayed higher CD163+ M2 macrophage staining, while the infiltration of CD8+ cytotoxic T cells and MHC-II+ APCs was lower in tumor lesions (Adeno) than in adjacent normal tissue (CTL) (Fig. 7f,g). This staining pattern was also observed in TMA of grade 1 and 2 (Supplementary Fig. 11a,b), consistent with high ULK1 activity observed in human pancreatic cancers (Fig. 1a and Supplementary Fig. 2a). Taken together, these findings indicate that Ulk1 deletion remodels the tumor immune microenvironment by reducing tumor-promoting myeloid cells such as M2 macrophages and neutrophils or PMN-MDSCs, and increasing cytotoxic T cells and APCs, leading to enhanced antitumor immunity and suppressed tumor growth.

Autophagy plays a context-dependent role in cancer, functioning as a tumor suppressor during early tumorigenesis but promoting tumor survival in established malignancies. In PDAC, elevated autophagy activity supports tumor progression by providing essential metabolic substrates for cancer cell survival. Although the importance of autophagy in PDAC has been demonstrated by the tumor-suppressive effects of Atg5 and Atg7 deletion7,8, the role of ULK1, a key initiator of autophagy, remains underexplored in in this context. Here, we suggest ULK1 as a critical regulator of pancreatic tumor progression through both cell-intrinsic and immune-modulatory mechanisms. Our findings support a role for ULK1 in generating a protumorigenic immune microenvironment and suggest that targeting ULK1 may offer a dual therapeutic benefit by impairing autophagy and promoting antitumor immunity. Although ULK1 mRNA expression is comparable between healthy patients and those with tumors from human PDAC datasets (TCGA) (Supplementary Fig. 1), our multi-IHC and immunoblot analysis revealed elevated ULK1 activity—as marked by phospho-Atg14 levels (pAtg14)—in high-grade human PDAC tissues and cells (Fig. 1), suggesting that ULK1 activity, rather than its expression, is important for tumor progression. The functional significance of ULK1 in tumor progression is further supported by our findings that ULK1 depletion in both mouse (KPC) and human (MIA PaCa-2) pancreatic cancer cells significantly impaired their proliferation and invasion, supporting its role in tumor aggressiveness (Fig. 2 and Supplementary Fig. 3). Consistent with its crucial role in autophagy activation, ULK1-depleted cells showed defective autophagosome formation, as indicated by reduced LC3-II/I ratios, LC3-II levels and GFP-LC3 puncta under stress conditions, confirming that ULK1 is critical for maintaining autophagy flux and metabolic adaptation for cancer growth. Importantly, our in vivo studies using both syngeneic orthotopic allografts and spontaneous KPC GEM models provide compelling evidence that Ulk1 is required for pancreatic tumor development in a physiological setting. Tumor-intrinsic deletion of Ulk1 significantly delayed tumor onset, reduced PDAC burden and extended survival (Figs. 2–4). This study provides direct evidence to demonstrate a tumor-promoting function of Ulk1 in a spontaneous PDAC model, suggesting its potential as a promising therapeutic target. Beyond its canonical role in autophagy-dependent tumor survival, we found an immune-modulatory function of ULK1 in the PDAC TME. Analysis of TCGA datasets revealed a negative correlation between ULK1 expression and MHC class II-mediated antigen presentation pathways (Supplementary Table 5), implicating that ULK1 may suppress antitumor immunity by inhibiting antigen processing in tumors and suppressing APCs34,37. While autophagy has been implicated in immune evasion, particularly through MHC-I degradation of tumor cells37, our data further revealed that Ulk1-deficient tumors markedly increased the infiltration of MHC-II+ APCs into tumors, potentially enhancing tumor antigen presentation and T cell priming (Fig. 5 and Supplementary Figs. 5, 7 and 8). Moreover, genetic deletion of Ulk1 in tumors significantly altered the composition of the TME across both orthotopic and spontaneous KPC models. We observed a marked reduction in immunosuppressive PMN-MDSCs and neutrophils. This decrease aligns with the IHC staining showing that intratumoral neutrophils enriched for SPP1 and VEGFA—markers of terminally differentiated, pro-angiogenic neutrophils—were abundant in KPC tumors but markedly reduced in KPC;Ulk1fl/fl tumors, consistent with their established role in promoting tumor vascularization and growth43,44. Such neutrophil populations are known to adapt to metabolic constraints in the TME, including hypoxia and glycolytic reprogramming, features we observed more prominently in KPC tumors relative to their Ulk1-deficient counterparts. Interestingly, although MHC-II+ APCs were significantly increased in Ulk1 depletion, CD8+ cytotoxic T cells showed only a substantial and not a statistically significant increase, with a notable expansion of the effector memory subset in Ulk1 KO orthotopic tumors (Figs. 5 and 6 and Supplementary Figs. 5, 7 and 8). These consistent observations across models strongly suggest that ULK1 contributes to immune evasion by orchestrating the recruitment and maintenance of suppressive myeloid populations while restricting the expansion of cytotoxic immune cells. Given that DCs were markedly increased in Ulk1 KO tumors, we assume that effective DC-mediated T cell priming may also require additional factors and distinct TME niches. Mechanistically, cytokine and chemokine profiling revealed that Ulk1-deficient tumors secreted lower levels of Ccl2, Cxcl2 and G-CSF, key factors that recruit immunosuppressive myeloid cells45–54. Interestingly, among the cytokines, GM-CSF was the most prominently elevated in Ulk1 KO tumors (Fig. 7). Although there are reports of dual roles of GM-CSF in the TME by promoting or inhibiting immune cell subtypes55–57, recently, GM-CSF shows antitumor immune responses by activating M1 macrophages and enhancing DC differentiation58–60. The upregulation of DC populations may explain the increased viability of APCs and may have further activated T cells observed in Ulk1-deficient tumors (Figs. 5 and 6 and Supplementary Figs. 5 and 6). These tumor-intrinsic cytokine changes were recapitulated in vitro using CM from Ulk1 KO KPC cells, confirming that Ulk1-mediated tumor signaling influences immune cell recruitment as determined by the types of cytokine and chemokine (Fig. 7a–d). Functional assays further demonstrated that CM from Ulk1 KO cells impaired the survival of tumor-infiltrating neutrophils including PMN-MDSCs, while enhancing the viability of DCs and CD8+ T cells (Fig. 7e–g and Supplementary Fig. 10). These results suggest that Ulk1-dependent cytokine secretion dictates immune cell fate within the tumor niche. Overall, these findings support the role of Ulk1 in maintaining an immunosuppressive TME, promoting tumor growth not only through autophagy-driven metabolic support but also by modulating the balance of immune cell compositions and dynamics, the latter of which requires further mechanistic investigation under physiological TME conditions. Previous studies have shown that autophagy promotes the secretion of protumorigenic factors—including IL-6, IL-8 and MMP2—in Ras-driven cancers and facilitates IL-1β secretion through unconventional secretory pathways61. In line with these studies, our data indicate that ULK1 plays an important role in regulating the secretion of cytokine and chemokines that reprogram the tumor immune landscape. While the precise molecular mechanisms by which ULK1 regulates cytokine expression and trafficking remain to be fully elucidated, our findings strongly implicate ULK1-mediated tumor signals as playing a crucial role in coordinating tumor–immune interactions through selective modulation of immune-regulatory chemokines and cytokines. A growing number of evidence suggests that ULK1 exerts autophagy-independent functions that also contribute to tumor progression. Beyond its canonical role in autophagy initiation, ULK1 regulates various cancer-related processes—including metabolic reprogramming, redox homeostasis, cell cycle progression and immune modulation—through direct phosphorylation of diverse downstream targets62,63. For instance, ULK1/2 phosphorylates key glycolytic enzymes such as hexokinase 1, phosphofructokinase 1 and enolase 1, thereby diverting glucose flux toward the pentose phosphate pathway to maintain redox balance under metabolic stress, independent of autophagy64. More recently, ULK1 was shown to directly phosphorylate lactate dehydrogenase A, enhancing lactate production and promoting Vps34 lactylation, which is associated with cell homeostasis and growth mediated by noncanonical autophagy induction65. These findings highlight that ULK1 acts as a key integrator of metabolic reprogramming and autophagic signaling. ULK1 also contributes to the cell cycle to facilitate cell proliferation. It phosphorylates the spindle assembly checkpoint protein Mad1, facilitating its recruitment to kinetochores during mitosis and ensuring chromosome segregation fidelity66. Furthermore, phosphorylation of ULK1 and its binding protein by cyclin-dependent kinase 1/cyclin B promotes mitotic entry and cell cycle progression, as indicated by impaired cancer growth upon genetic deletion of ULK1 or ATG1367. As evidence of cell death regulation, ULK1 directly phosphorylates RIPK1, thereby inhibiting the formation of the necrosome complex and protecting cells from tumor necrosis factor-induced necroptosis68. In terms of innate immune regulation, ULK1 shows context-dependent, autophagy-independent roles. ULK1 negatively regulates antiviral responses by phosphorylating STING and dampening type I interferon signaling69. Conversely, ULK1 enhances antiviral immunity by promoting type I interferon-stimulated gene expression via activation of p38 MAP kinase70. These apparently opposing roles imply the importance of ULK1 in balancing antiviral defense with protection against chronic inflammation. Collectively, these findings indicate the multifaceted, autophagy-independent roles of ULK1 in cancer progression. Taken together with our findings—that ULK1 promotes pancreatic cancer progression by supporting tumor metabolism and orchestrating an immunosuppressive TME through cytokine-mediated immune cell modulation—we suggest that ULK1 serves as a key regulator of both tumor-intrinsic and immune-regulatory pathways, even independent of its role in autophagy. Therefore, targeting ULK1 may offer a dual therapeutic strategy in PDAC by disrupting metabolic support and enhancing antitumor immunity.

Supplementary Information