KPC cells were originally established from the Tuveson LSL-Kras G12D/+; LSL-Trp53R172H/+; Pdx-1-Cre model (27) and were grown in DMEM + 10% FBS (Gibco) in a humidified incubator at 37°C. ATGListatin (20 µM), drug vehicle control DMSO, docosahexanoic acid (DHA) (60.7 µM) (all from Sigma-Aldrich; Merck KGaA) were used as indicated for various experiments. Western blots were repeated twice, biochemical assays were performed in four independent biological replicates. All treatments were performed at 37°C.

HILPDA KO cell lines were generated using a double nickase strategy. Two gRNAs targeting Hilpda: A (5′-TCTAACAAAGATGGAAAGCA-3′) and B (5′-GGAGTCTCTGGGAGGCTTAC-3′) were individually cloned in pX462-Cas9n backbone [pSpCas9n(BB)-2A-Puro V2.0, Addgene: 62987] using BbsI restriction site. Constructs targeting Hilpda were sequence-verified and used to create Hilpda KO cell lines. Cells were transfected with 2 µg DNA using Lipofectamine 2000 (Thermo Fisher Scientific). Single clones were selected by antibiotic resistance for 3 days and further expansion for 2 weeks, screened by westerns blotting and 7 successful KO clones were combined to generate the KO pool. The pIRES-neo-Hilpda-myc-flag expression vector has been described previously (26). Cells were transfected with pIRES-neo-Hilpda-myc-flag or empty vector pIRES-neo (Origene) and underwent G418 (Sigma-Aldrich; Merck KGaA) selection at 2 mg/ml for 2 weeks before being screened for transgene expression.

KPC cells were lysed in RIPA buffer (150 mM NaCl, 1% NP 40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0) supplemented with 100X Halt protease inhibitor cocktail (Thermo Scientific Fisher), 100X phosphatase inhibitor cocktail (Cell Signaling Technology) and 1 mM PMSF (Thermo Fisher Scientific). Lysates were cleared by centrifugation for 5 min at 12,000 × g and at 4°C. Protein concentrations were measured with a bicinchoninic acid (BCA) protein kit (Thermo Scientific Fisher). Then, 20–30 µg of whole protein lysates were resolved in an 11% acrylamide gel and transferred onto PVDF membrane. For immunodetection, primary antibodies used were: Custom-made rabbit anti-Hilpda 1:50 (21), rabbit a-Perilipin2 (1:1,000; Origene, cat. no. TA321279), mouse a-myc (1:1,000; Cell Signaling Technology, cat. no. 2276S), and rabbit a-ATGL (1:1,000; Cell Signaling Technology, cat. no. 2138S), and mouse a-tubulin (1:1,000; Invitrogen; Thermo Fisher Scientific, cat. no. PIMA516308). Primary antibodies were detected using Licor goat anti-mouse (1:2,000, Licor, cat. no. 926-68070) or goat anti-rabbit (1:2,000; Licor, cat. no. 926-32211) secondary antibodies, and visualized using a Licor Odyssey CLx near infrared imager.

Cells were grown on glass coverslips, treated as required and fixed with 4% paraformaldehyde. Lipid droplets were stained with 0.1 µg/ml Nile Red (Santa Cruz Biotechnology) for 20 min at room temperature. Nuclei were counterstained with 10 µg/ml Hoechst-33342, and samples were mounted with Slowfade Diamond antifade mounting media (Life Technologies, cat. no. S36968). The slides were imaged on a Zeiss Axioskope widefield microscope at the OSUCCC microscopy and imaging facility. Lipid droplet images were visualized with ImageJ.

Cells were grown on 10 cm culture dishes in regular DMEM media, and fatty acid was loaded (using DHA), or starved of FBS for 24 h. Where indicated, ATGListatin was added at the beginning of treatment. Triglycerides were measured using the colorimentric Triglyceride Quantification Assay Kit (Abcam; cat. no. ab65336) as per the manufacturer's recommendation (sensitivity >2 µM).

All animal experiments were approved by the Ohio State University's Institutional Animal Care and Use Committee. Five hundred thousand cells in PBS were injected subcutaneously on the back of non-anesthetized 7- to 9-week-old female nu/nu mice (18–21 g) obtained from the OSUCCC Target Validation Shared Resource (n=11/group). Groups of 3–4 animals were housed in autoclaved cages, were fed ad libitum, and maintained on a 12-h light/dark cycle. Room temperature was maintained at 22°C and humidity at 30%. Cages were randomly assigned to experimental groups. Tumor growth was measured using calipers. Tumor volume was calculated using the formula: S × S × W × 0.52. Animals were euthanized by CO₂ asphyxiation followed by cervical dislocation according to the approved protocol. Maximal tumor dimensions at the time of sacrifice were 12 and 8 mm in the WT and KO groups, respectively.

Using SPSS v25, data were screened for normality and homogeneity of variance using the Shapiro-Wilk and Levene tests, respectively. When normality and equal variance was met, a Student's t-test was used. When normality and equal variance was not met, a non-parametric Mann-Whitney U test was used. Data were considered to be statistically significant if P<0.05. Kaplan-Meier curves were compared by the Xena Browser using the log-rank test (www.xenabrowser.net).

We and others have shown that conditions that increase lipid flux can induce Hilpda protein (17,24,26). To determine whether similar mechanisms exist in murine pancreatic tumor cells, we exposed KPC cells to regular normoxic conditions in DMEM media, to oxygen deprivation (1% O₂) or to exogenous fatty acid (docosahexaenoic acid) for 24 h and examined Hilpda protein expression by western blot analysis (Fig. 1A). As has been reported in other tumor types, both hypoxia and fatty acid loading increased Hilpda levels in KPC cells. Expression of ATGL was detectable under all conditions but was not stress-responsive. To evaluate the possible impact of HILPDA expression in the clinical behavior of human pancreatic cancers we assessed the TCGA PDAC dataset using the Xena functional genomics explorer (xenabrowser.net). PDAC tumors with the highest quartile HILPDA expression had a significantly shorter overall survival than those with the lowest expression (Fig. 1B), suggesting that HILPDA may be associated with more aggressive cancers.

We genetically manipulated Hilpda in KPC cells and asked whether it is necessary or sufficient for LD growth in vitro under different growth conditions. The impact of Hilpda on the ability of cells to form lipid droplets appears to be cell-type specific. First, Hilpda KO cells were generated by CRISPR-Cas9 gene editing and single clones were screened for successful gene deletion (data not shown). A pool of 7 KO clones was established and loss of Hilpda protein expression was confirmed (Fig. 2A). In parallel, a KPC cell line stably overexpressing myc-Flag tagged Hilpda driven by the CMV promoter was generated (Fig. 2B). Next, we determined whether the engineered cells have perturbations in LD dynamics. Cells were incubated in different nutritional states that are known to increase lipid flux: Exogenous fatty acid supplementation and lipid deprivation through serum removal. After 24 h, cells were fixed with 4% paraformaldehyde and lipid droplets were visualized by fluorescence microscopy after staining with the neutral lipid dye Nile Red. Hilpda overexpression led to an increase in LD abundance compared to empty vector cells, under all conditions. Conversely, under all environmental conditions, the Hilpda KO cells had smaller and fewer lipid droplets, suggesting that Hilpda positively regulates lipid droplet abundance.

Hilpda promotes triglyceride storage in KPC cells, independently of ATGL inhibition. Qualitative and quantitative differences in the constitution of LD's neutral lipid core have been identified (28). To ascertain whether the differences in LD abundance caused by Hilpda loss in KPC results from deregulated triglyceride metabolism, we biochemically quantified triglyceride levels under basal- and fatty acid-loaded conditions (Fig. 3A). In accordance with the LD staining results, the Hilpda KO cells were significantly impaired in their maximum triglyceride storage capacity. In basal conditions there was a trend towards lower triglycerides in the KOs but did not reach statistical significance.

We and others have shown that in certain tissue contexts Hilpda promotes triglyceride storage by inhibiting the rate-limiting lipase ATGL/PNPLA2 (24–26). In order to establish if Hilpda functions as a molecular ATGL inhibitor in murine pancreatic tumors, we pharmacologically inhibited ATGL in Hilpda WT and KO cells with the small molecule inhibitor ATGListatin (ATGLi), and quantified triglycerides (Fig. 3B). Notably, the chemical inhibition of ATGL was not able to correct the defect in the KOs and to restore their triglyceride content to the level of the Hilpda WT cells. This finding suggests that, in KPC cells, decreased lipid storage following Hilpda ablation is not caused by elevated ATGL activity and enhanced lipolysis.

Hilpda promotes model tumor growth. Loss of Hilpda-dependent ATGL regulation has been shown to be growth inhibitory in model tumors (24,26). Owing to the in vitro findings of ATGL-independent Hilpda functions in KPC we examined whether Hilpda exerts tumor-promoting properties in pancreatic cancer xenografts. WT and KO cells were injected subcutaneously into the backs of nude mice (11 mice per genotype) and tumor sizes were measured with calipers. The results showed that loss of Hilpda significantly decreased the growth rate of KPC tumors, suggesting that Hilpda can positively regulate tumor growth, independently of lipolytic control (Fig. 4A and B). At the completion of the xenograft growth, we excised the tumors and confirmed their Hilpda genotype by western blot analysis (Fig. 4C) and measured their triglyceride content (P<0.05) (Fig. 4D). The WT tumors contained two times more triglycerides than the KOs, indicating that tumor microenvironmental conditions, such as hypoxia, stimulate Hilpda-dependent lipid storage.