Comparison of RNAseq datasets from TCGA and the Cancer Cell Line Encyclopedia (CCLE) with The Genotype-Tissue Expression (GTEx) portal revealed that ~20–25% of PDACs exhibit increased expression of SLC16A1, the gene encoding MCT1, a mediator of 3BP activity. We, then, explored the activity of 3BP as a free drug in a panel of PDAC cell lines, including MIA PaCa-2, PSN-1, Panc 02.13, AsPC-1, BxPC-3, and CFPAC-1. Cells were treated with increasing concentrations of 3BP (0–220 µM) and assessed for cell death with real-time quantitative live-cell imaging. Five of the cell lines, MIA PaCa-2, PSN-1, Panc 02.13, AsPC-1, and BxPC-3, were sensitive to 3BP, with IC50s ranging from 24 to 40 µM (Figure 1A and B). In contrast, CFPAC-1 was highly resistant to 3BP, at concentrations of up to 220 µM. RNA expression levels of three cellular transporters (Cunningham et al., 2022; Barretina et al., 2012), SLC2A1 (GLUT1), SLC16A1 (MCT1), and SLC16A3 (MCT4), were examined to identify potential associations between the drug and cell viability (Figure 1B). The expression of MCT1 was much lower in the 3BP-resistant cell line CFPAC-1 than in the other five cell lines. However, no clear relationship was found between 3BP resistance and the expression of the other transporters. MCT1 immunohistochemical (IHC) analyses furthermore demonstrated that the expression of the MCT1 protein on the cell surface paralleled RNA levels in these cell lines (Figure 1C).

To further demonstrate the specificity of the drug, we genetically inactivated MCT1 in cells and compared the viability of inactivated cells with their parental counterparts in the presence of 3BP. A similar strategy was used to establish MCT1 as essential for 3BP-mediated cell death in KBM7 cells, a unique near-haploid cell line derived from a leukemia (Birsoy et al., 2013). We deleted the gene encoding MCT1, SLC16A1, in the 3BP-sensitive cell line MIA PaCa-2 that expresses high levels of MCT1 (Figure 1A and B). Next-generation sequencing confirmed knockout of MCT1 in six clones (Figure 2A and B), and the results were validated with IHC (Figure 2C and D). Cells from the six clones were then mixed at equal ratios to create an MCT1 KO cell population (KO cells), thereby avoiding potential variables often associated with the evaluation of a single clone (Torrance et al., 2001; Lee et al., 2019).

Next, we introduced nuclear-restricted green and red fluorescent proteins into the KO and parental cells, respectively, to enable simultaneous visualization in co-culture. KO cells grew at a rate indistinguishable from the parental cells under normal culture conditions (Figure 2E). However, in the presence of 50 µM 3BP, the parental cells (Figure 2G), were almost completely eliminated (IC50 24 uM) while the KO cells continued to proliferate (Figure 2F and Videos 1 and 2). These results indicated the specificity of 3BP for cells expressing MCT1.

One of the major challenges of using 3BP as a chemotherapeutic agent is that it is rapidly inactivated by the serum proteins it alkylates before it leaves the circulation and enters cancer cells (Chang et al., 2007). We have previously shown that encapsulation of 3BP with a cyclodextrin can mitigate its toxicity, presumably by permitting the use of lower concentrations of the encapsulated drug compared to the free drug (Chapiro et al., 2014). Therefore, we focused on optimizing the microencapsulation procedure and then examined the serum stability of the drug. To optimize microencapsulation of 3BP in cyclodextrins, we chose the β-cyclodextrin scaffold rather than alpha or gamma cyclodextrins based on structural modeling (Chapiro et al., 2014). We first evaluated two types of β-cyclodextrin modifications, one with the hydroxyls substituted with succinyl groups and the other with the hydroxyls substituted with 2-hydroxypropyl groups. Second, for the modifications involving succinyl groups, we sought to determine the optimal number of hydroxyl substitutions. Finally, we investigated the optimal ratio of cyclodextrin to 3BP. To evaluate these various formulations, we first used size-exclusion chromatography (SEC) to ensure that the amount of free 3BP was minimized (Figure 3A). Second, we used cell toxicity as a measure of the stability of these formulations. Because the cytotoxicity of 3BP decreases upon interaction with serum proteins, residual cell toxicity following exposure to serum is directly related to its resistance to serum degradation. Therefore, the formulations were incubated with human serum at 37°C and aliquots were collected at various time points for assessment of cell toxicity in DLD-1 cells which have high expression of MCT1 and sensitivity to 3BP (IC50, 22.64 µM; Figure 1B). Among the β-cyclodextrin formulations tested, we found that succinyl-substituted cyclodextrins were capable of protecting 3BP in sera more efficiently than hydroxypropyl-substituted versions. Moreover, a succinyl-substituted cyclodextrin with an average of 3.4 succinyl groups per cyclodextrin, at a molar ratio of 1.2 β-cyclodextrin per 3BP, performed best (Figure 3A). This MicroEncapsulated formulation was named ‘ME3BP-7’. Although free 3BP lost 90% of its activity by 30 min of exposure to serum, ME3BP-7 lost <10% of its activity in 30 min (Figure 3B). Remarkably, even after 8 hr of incubation with serum, ME3BP-7 retained >70% of its activity (Figure 3B). Hydroxypropyl-substituted cyclodextrin-3BP complexes were more stable than free 3BP, losing just over 50% of its activity after 30 min of exposure to serum. However, these complexes were not nearly as stable as ME3BP-7 (Figure 3B).

Finally, parental MIA PaCa-2 (MCT1 WT) in contrast to MCT1 KO cells were sensitive to the encapsulation of 3BP within ME3BP-7, indicating that the compound retained specificity as well as cytotoxicity for cells expressing MCT1 (Figure 3C).

We also noticed that the morphology of the MIA PaCa-2 cells rapidly changed in response to exposure to 3BP or ME3BP-7. To determine the significance of this change in morphology, we performed washout experiments on co-cultures of WT and KO cells exposed to drugs. Co-cultures were exposed to 3BP or ME3BP-7 at various concentrations and time periods, and cell proliferation was followed for 48 hr after removal of the drug. Exposure to either 3BP or ME3BP-7 for as little as 30 min led to the loss of the majority of WT MIA PaCa-2 cells. In contrast, KO MIA PaCa-2 cells continued to proliferate under all concentrations of the drugs and time points tested (Figure 4A–D, Figure 4—figure supplements 1 and 2). In addition, gemcitabine or any of the FOLFIRINOX agents, including irinotecan, oxaliplatin (Video 3), and 5-FU, exhibited no differential effect on WT and KO cells, thus emphasizing the specificity of 3BP and ME3BP-7 for MCT1-expressing cells. Moreover, 3BP and ME3BP-7 caused cell death of MIA PaCa-2 cells, while the standard of care agents only decreased their proliferation (Figure 4, Figure 4—figure supplements 1 and 2).

To assess efficacy in vivo, we treated mouse models of PDAC with ME3BP-7. We first tested the feasibility of delivering ME3BP-7 systemically and compared the toxicity of free 3BP with that of ME3BP-7 in athymic nude mice. Escalating doses ranging from 8.0 to 33 mg/kg of free 3BP (as ME3BP-7) were used to identify a maximum tolerated dose of ME3BP-7. Equivalent amounts of free 3BP and ME3BP-7 were infused via vascular access buttons (VABs) (33 mg/kg of 3BP) in 200 µL of PBS every other day for 1 week. In preliminary experiments, two-thirds of mice in the free 3BP group became ill, showing irrecoverable weight loss, pallor, and decreased activity, and died after only two doses. No overt toxicity was evident even after 5 weeks of treatment with ME3BP-7 at the same dose (33 mg/kg) (Figure 5—figure supplement 1A).

We then generated orthotopic xenografts with Panc 02.13, a patient-derived PDAC cell line with high expression of MCT1 (RNA expression of 137 transcripts per million) (Figure 1B), engineered to express firefly luciferase. Tumor-bearing animals were randomized into three treatment arms on the basis of intravital multiphoton imaging (IVIS), and treatment was initiated a day later (Figure 5A). Guided by our dose-finding toxicity studies, no treatment arm with free 3BP was included in this study as none of the animals would survive the course of the therapy. Over the course of 4 weeks of treatment (dosed Mondays, Wednesdays, and Fridays), no discernible adverse effects or weight loss was observed in the treated animals (Figure 5—figure supplement 1A). Animals in both treatment cohorts showed near-complete abrogation of tumor growth (p<0.01, one-way ANOVA, Figure 5B and C). Finally, the weight of residual tumors after 4 weeks of ME3BP-7 was significantly decreased in treated animals relative to controls (p=0.01, Mann-Whitney U test, Figure 5D).

We next tested the efficacy of ME3BP-7 in an orthotopic xenograft model generated with patient-derived PDAC TM01212 tumor pieces (Figure 6A). Orthotopic TM0212 xenografts expressed moderately high levels of both MCT1 RNA (37 transcripts per million) and protein (Figure 6B). Histological evaluation revealed the development of duct-like structures within these xenografts (Figure 6B). More importantly, all control animals developed metastases to the lung and liver, mimicking the behavior of human PDACs (Figure 6F–H). However, only 3/7 mice treated with ME3BP-7 had any distant metastases, and the number and size of metastases in the treated vs. untreated mice were significantly decreased in treated animals relative to controls (p<0.01, Mann-Whitney U test, Figure 6H). Again, histological evaluation of major organs did not reveal any pathological changes of significance as a result of potential toxicity in the treated mice (Figure 6—figure supplement 1 and Supplementary file 1).

We evaluated the efficacy of ME3BP-7 in a second PDAC patient-derived xenograft model, TM01098 (Figure 6—figure supplement 2A). Although these tumor cells expressed high levels of MCT1 RNA (96 transcripts per million), protein expression exhibited focal staining (Figure 6—figure supplement 2B), which was in contrast to the uniform expression in Panc 02.13 (Figure 1C) and TM01212 xenografts (Figure 6B). Despite focal MCT1 protein expression, significantly slower growth of TM01098 xenografts occurred in animals treated with ME3BP-7 compared to controls, based on ultrasound (US) (p<0.01, Mann-Whitney U test, Figure 6—figure supplement 2D) and tumor weight (p=0.0028, Figure 6—figure supplement 2E).

To further validate the mRNA expression data in human tumors in publicly available databases (Cancer Genome Atlas Research Network, 2017; Chen et al., 2019b, Figure 7—figure supplement 1), immunohistochemistry was performed on a total of 95 PDACs on tissue arrays. Of the 95 samples analyzed, 56% (n = 53) were positive for MCT1 and 44% (n = 42) were negative. The staining intensity of the 53 MCT1 positive pancreatic cancer tissues was scored as follows: 58% (n = 31), low intensity; 30% (n = 16), intermediate intensity; and 12% (n = 6) high intensity (Figure 7A and C).

None of the 51 normal tissues recorded in the GTEx database (https://gtexportal.org/home/gene/SLC16A1) expressed average levels of SLC16A1 RNA greater than the approximate threshold required for sensitivity to 3BP in cancer cell lines (~50 transcripts per million, Figure 7—figure supplement 2). The highest expression levels in normal tissues were found in the colon, testis, and uterus (median 37, 38, and 32 transcripts per million, respectively, among 142–376 tissue samples assessed for each tissue type) (Uhlén et al., 2015). The RNA levels were consistent with the immunohistochemistry; staining was low or absent in normal pancreas, brain, or ovary while readily observed in testis, colon, and uterine tissue (Figure 7—figure supplement 3). On the basis of this information, we carefully examined ME3BP-7-treated mice at necropsy and found no discernible damage after 4 weeks of treatment in either the colon or uterus, while the pancreatic cancers showed extensive necrosis (Supplementary file 1).

We designed a new formulation of the cytotoxic agent 3BP to be resistant to serum degradation and systemically administered. Modified cyclodextrins were investigated for their efficiency of encapsulating 3BP, and the activity of the encapsulated drug was analyzed after exposure to serum. We used CRISPR-mediated technologies to knock out the cell membrane receptor for 3BP, MCT1, in pancreatic cancer cell lines in vitro. We also modified pancreatic cancer cell lines to express fluorescent or bioluminescent markers for the purposes of tracking cellular responses to ME3BP-7 as well as other drugs commonly used in the clinic for treating PDAC patients. Tumor cells or tumor fragments were orthotopically implanted into the pancreas of mice for in vivo experiments. Implantation was confirmed with luciferase expression in cell lines or US, followed by stratification and randomization into treatment and control groups. The MCT1-dependent killing potential of ME3BP-7 was measured with time-lapse recordings obtained with an IncuCyte Live Cell-Analysis System (Essen Bioscience; Ann Arbor, MI, USA). Necropsies were performed for tumor resection and organ evaluation at the end of the experiment. Sample sizes for animal experiments were selected based on previous experience with the animal models but were not predetermined by power analysis. No animals were excluded from the study due to illness unless indicated. The number of replicates in each experiment is noted in the figure legends.

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Surojit Sur (ssur1@jhmi.edu). Isogenic knockout and engineered cell lines are available through The Genetic Resources Core Facility at Johns Hopkins School of Medicine (jhbiobank@jhmi.edu, MD, USA).

Drug batches were generated as previously described. In brief, a solid sample of 3BP (1 mmol, 167 mg) was added in small portions to a solution of the corresponding cyclodextrin (1.0–1.2 mmol) in 25 mL of deionized (DI) water under constant agitation. After complete addition over 15–20 min, the samples were further agitated for 1–4 hr at 25°C. Finally, the samples were flash-frozen in liquid nitrogen or dry ice/acetone and lyophilized overnight. The lyophilized reagent was diluted in PBS 15 and 30 min prior to in vitro and in vivo experiments, respectively.

SEC was performed with a Shodex-OH Pak at an elution rate of 1 mL/min of PBS under isocratic conditions and monitored at 220 nm. The samples included (i) free 3BP (1 mg/mL); (ii) succinyl-β-CD without 3BP (20 mg/mL); (iii) a mixture of 10 µL of free 3BP at 1 mg/mL and 10 µL ME3BP-7 at 10 mg/mL; and (iv) ME3BP-7 (10 mg/mL).

Human PDAC cell lines, including MIA PaCa-2, PSN-1, Panc 02.13, AsPC-1, BxPC-3, and CFPAC-1, were obtained from the American Type Cell Culture (ATCC) (Manassas, VA, USA), authenticated by STR and tested for mycoplasma. Cell lines were maintained at 37°C in a humidified 5% CO₂ atmosphere in T75 tissue culture flasks containing 14 mL of the medium recommended by ATCC (DMEM for MIA PaCa-1 parental and MCT1 KO; RPMI 1640 for Panc 02.13, AsPC-1, BxPC-3, and DLD-1; and IMDM for CFPAC-1). All media were supplemented with 10% fetal bovine serum (Mediatech, Inc; Manassas, VA, USA) and 1 % penicillin-streptomycin. Cells were passaged at the ATCC recommended ratio every 4–5 days with trypsinization and resuspended in fresh medium in a new flask. Only early passaged cells harvested at 70–85% confluence were used for in vitro and in vivo experiments.

MIA PaCa-2 or Panc 02.13 cells were transduced with a CMV-Firefly luciferase lentivirus carrying a puromycin-selectable marker (Cellomics Tech; Halethorpe, MD, USA) according to the manufacturer’s instructions. Parental and MCT1 KO MIA PaCa-2 cells were transduced with Lentiviral NucLight Red and Green vectors (Sartorius), respectively, carrying a puromycin-selectable marker. All transduced cells were selected in growth medium containing puromycin (4 μg/mL) for 96 hr.

Cells (3–12 × 10³) were seeded onto the wells of 96-well plates and exposed to drug 24 hr after plating. For MIA PaCa-2 parental and MCT1 KO co-culture experiments, cells were suspended in medium at equal concentration and seeded at 12,000/genotype onto wells. Cells in experiments were imaged with time-lapse fluorography using the IncuCyte Live Cell-Analysis System (Essen Bioscience; Ann Arbor, MI, USA). For green- and red-labeled cells, a processing definition for the IncuCyte was created, and the green or red fluorescence channels were used for analysis. All in vitro experiments were conducted at least in duplicate. Green and red object counts were used to track parental and MCT1 KO MIA PaCa-2 experiments, while confluence was used to track unlabeled cells. Cell viabilities were recorded as the fraction of cells that survived treatment at the indicated drug molarities relative to the control wells. Dose-response curves were fit using Prism 9 software.

Free 3BP, ME3BP-7, and HPCD-3BP were incubated at various concentrations (12.5 μM, 25 μM, 50 μM, 100 μM, and 200 μM) with 90 μL of human serum (Sigma, Cat No. H3667; St. Louis, MO, USA) at 37°C for up to 8 hr. Aliquots were collected at 30 min, 1 hr, 2 hr, 4 hr, and 8 hr and stored at –80°C until further analysis. For the cell toxicity assay used to assess the amount of biologically active 3BP at each time point, parental DLD-1 (MCT1, 387 TPM) cells were plated at 35–40% confluence and exposed to 10 µL of the collected drug sample diluted with 190 µL of medium. Cell death was measured by imaging cells after 72 hr in culture with the IncuCyte Live Cell-Analysis System (Essen Bioscience). Residual drug activity was estimated by comparison of cell death induced by drug samples incubated with human serum for different periods of time to the cell death induced by an unincubated drug sample to derive the % drug activity.

Cells were seeded onto 96-well plates at a density of 24,000 cells per well (12,000 parental MIA PaCa-2 red and 12,000 MCT1 KO MIA PaCa-2 green). After 24 hr, cells were treated with vehicle (complete DMEM) or a serial dilution of a drug (bromopyruvic acid, Sigma-Aldrich; gemcitabine HCl, irinotecan, oxaliplatin, and 5-fluorouracil, Selleck Chemicals; Houston, TX, USA). Each treatment condition was conducted in triplicate unless otherwise stated. The drugs were either removed from the plates at 30 min or 2 hr, or left on cells. Wells were gently washed with 200 µL of PBS twice and replaced with fresh medium. Residual cells were assessed with IncuCyte Live Cell-Analysis System. Cell viabilities were reported as the percentage of cells remaining after treatment at the indicated drug molarities relative to the average of three control wells without drug.

The Alt-R CRISPR system (Integrated DNA Technologies [IDT]; Coralville, IA, USA) was used to delete SLC16A1, the gene encoding MCT1 protein, in the DLD-1 and MIA PaCa-2 cell lines. The gRNA sequence was designed with CHOPCHOP v3 (Labun et al., 2019). Alt-R CRISPR Cas9 crRNAs (ACCATGCCATTCAGGCTAGT, IDT; SEQ ID NO:1) and Alt-R CRISPR-Cas9 tracrRNA (1072532, IDT) were resuspended in Nuclease-Free Duplex Buffer (IDT) at a concentration of 100 μM. The crRNAs and tracrRNA were mixed in a 1:1 molar ratio and denatured for 5 min at 95°C, followed by slow cooling to room temperature for duplexing. Cas9 Nuclease (1081059, IDT) was then added at a 1.2:1 molar ratio and allowed to stand at room temperature for 15 min. Forty pmoles of the Cas9 ribonucleoprotein containing tracrRNA/MCT1 crRNA duplex was mixed with 2 × 10⁵ cells in 20 μL of OptiMEM (31985088, Thermo Fisher Scientific; Waltham, MA, USA), loaded into a 0.1 cm cuvette (1652089, Bio-Rad; Hercules, CA, USA), and electroporated at 120 V for 16 ms using an ECM 2001 (Harvard Apparatus; Holliston, MA, USA). Cells were immediately transferred to complete growth medium and cultured for 1 week. Upon reaching confluence, cells were plated at a density of 0.5–2 cells per well in 96-well plates and cultured for 3 weeks. Single colonies were transferred into 2 replica 96-well plates. Genomic DNA was harvested from one of the plates using the Quick-DNA 96 Kit (Zymo Research; Orange, CA, USA) and PCR-amplified using Q5 Hot Start High-Fidelity 2X Master Mix (New England BioLabs; Ipswich, MA, USA). SafeSeqS (Kinde et al., 2011) was used to confirm the mutation status of selected clones. Single MIA PaCa-2 clones deleted for SLC16A were collected from the matched replica plate, and six clones were pooled at equal ratios for further experiments.

DLD-1 parental and isogenic DLD-1-MCT1 KO cells were used as the positive and negative controls for testing the specificity of five commercially available MCT1 antibodies. Cells were trypsinized, washed in media to inactivate the trypsin, pelleted at 400×g for 10 min, fixed in 10% formalin, and embedded in paraffin. Sections (4 µm) were cut from the blocks and incubated with different dilutions of primary and species-appropriate secondary antibodies. The anti-MCT1 mouse antibody from Santa Cruz Biotechnology (SC-365501, Lot number D2319; Dallas, TX, USA) was chosen for immunostaining with a Ventana Discovery Ultra autostainer (Roche Diagnostics; Indianapolis, IN, USA) at the Oncology Tissue Services Core of Johns Hopkins University School of Medicine. Briefly, following deparaffinization and rehydration of sections, epitope retrieval was performed in Ventana Ultra CC1 buffer (catalog# 6414575001, Roche Diagnostics) at 96°C for 64 min. Antibodies were diluted in antibody dilution buffer (catalog# 5280524001, Roche Diagnostics; anti-MCT1 antibody, 1:2000 for cell pellets, 1:200 for patient-derived xenografts, and 1:100 for other tissues or tissue microarrays) and incubated with the slides at 36°C for 60 min. Following standard washing in the Ventana auto-stainer, bound antibodies were detected with an anti-mouse HQ detection system (catalog# 7017936001 with 7017782001, Roche Diagnostics) and a Chromomap DAB IHC detection kit (catalog # 5266645001, Roche Diagnostics). The slides were then counterstained with Mayer’s hematoxylin, dehydrated, and mounted in Toluene mounting medium (MER 7720, Mercedes Scientific; Lakewood Ranch, FL, USA).

IHC for MCT1 was performed on two commercially available FFPE tissue microarrays (BC001130 and HPanA150CS03; US Biomax, Inc; Rockville, MD, USA) for a total of 100 cases of pancreatic cancer. BC001130 included 20 cases of pancreatic carcinoma (triplicate cores per case), and HPanA150CS03 included 80 cases with adjacent normal tissue. Two independent reviewers, including a pathologist, scored the intensity and pattern of MCT1 staining for tumor and normal tissues. Any differences in grading were reviewed and resolved. Upon evaluation, 5 samples were considered to be of poor quality and disregarded. Of the 95 samples analyzed, 56% (n = 53) were positive for MCT1 and 44% (n = 42) were negative. The staining intensity of the 53 MCT1 positive pancreatic cancer tissues was scored as follows: 58% (n = 31), low intensity; 30% (n = 16), intermediate intensity; and 12% (n = 6) high intensity.

Escalating doses ranging from 8.0 to 33 mg/kg of free 3BP (as ME3BP-7) were infused via the tail vein of healthy nude mice to determine the maximum tolerated dose. Repeated tail vein injections of free 3BP caused significant scarring of the tail, limiting the use for sustained injections of the free drug, while injections of ME3BP-7 exhibited minimal sclerosis. For optimal comparison with the free drug, implantable VABs were used to provide direct access to the jugular vein for accurate administration of both drugs. Crl:NU(NCr)-Foxn1^nu (athymic nude mice; Charles River (490); Wilmington, MA, USA; Figure 5) were used for experiments performed with pancreatic tumor cell lines. More severely, immunodeficient mice were used for patient-derived xenografts which did not uniformly grow in athymic nude mice. In addition, different types of severely immunocompromised mice were used for the patient-derived xenografts due to limited availability of animals during the COVID pandemic: NOD.Cg-Prkdc Il2rg /SzJ ((NSG), the Jackson Laboratory (005557); Bar Harbor, ME, USA; Figure 6); and NOD-Prkdc^em26Cd52Il2rg^em26Cd22/NjuCrl (NCG) (Charles River (572)); Figure 5—figure supplement 1.

For orthotopic models, cells or small pieces of xenografts were implanted into the pancreas of each mouse in protocols modified from previously described techniques (Erstad et al., 2018). Once under anesthesia, the skin over the left abdomen was shaved and sterilized. A horizontal 1 cm incision was made over the left upper quadrant of the abdomen, right below the ribs. The underlying peritoneum was incised, and the spleen and pancreas were located. The pancreatic tissue was gently extruded from the abdominal cavity for good visualization. Either a cell aliquot was injected with a Hamilton syringe, or a piece of tumor was sutured to the tail of the pancreas with a 7-0 polypropylene suture avoiding major vessels. The implantation of the tumor fragment was standardized by size, time from extraction to implantation, and implantation technique in the tail of the pancreas. After implantation, the pancreas was gently returned to the abdominal cavity. The abdominal wall was closed with absorbable sutures, and the skin incision was closed using wound clips. Wound clips were removed once the incision site had healed, usually ~10 days after surgery.

Mice were euthanized upon termination of the experiment or when major weight loss or toxicity was observed according to JHU Animal Care and Use Committee standards. Tumors and organs were harvested, weighed, placed in 10% formalin, processed, fixed in paraffin, and sectioned. Standard H&E staining was performed on sections, and an expert comparative pathologist reviewed all presented data.

Luminescence quantification was performed using the IVIS imaging system and Living Image software (PerkinElmer; Waltham, MA, USA). Mice were anesthetized at 37°C with inhaled isoflurane in an induction chamber for 5 min and received an intraperitoneal injection of luciferin (150 mL, RediJect D- Luciferin Ultra Bioluminescent Substrate, PerkinElmer, 770505). Bioluminescence images were taken 13 min after injection, and control fluorescence images were acquired at each screening to confirm satisfactory intraperitoneal luminescence.

A modification of an US imaging method for pancreas in mice was used (Ganapathy-Kanniappan et al., 2009). The left flanks of mice were shaved with a clipper. Mice were then injected with 2 mL of 0.9% sterile saline intraperitoneally to increase the contrast between intrabdominal organs, anesthetized with isoflurane in the induction chamber for 5 min, and placed in the lateral recumbent position with the left flank up on the imager over a heated pad. Continuous anesthesia was applied through a face cone. Pancreatic tumors were detected and measured through acquisition of a high-resolution US obtained with the VisualSonics Vevo2100 High-Resolution Ultrasound System. A 15 mm depth US window in B-mode acquisition was used on all mice. After applying gel over the area for US visualization, images were obtained in a standardized fashion. The spleen, liver, left kidney, and pancreas were identified first to confirm correct positioning. Tumors were hypoechoic (dark/gray), while the surrounding pancreas was hyperechoic (bright/white). The tumor sutures were the most hyperechoic and frequently marked the tumor center. Trans-axial US images of the tumor were obtained by placing the US probe parallel and distal to the rib cage with the notched side of the transducer to the left (pointing anteriorly). Multiple anterior and posterior images were taken in this plane spanning the pancreas. Longitudinal US images were then obtained by placing the US probe parallel to the mouse axis with the notch side pointing toward the head of the mouse. Videos were acquired across the tumor and at the point where the maximum tumor diameter was visualized, and multiple images were saved for measurement for the longitudinal and the trans-axial views. For mice with several discrete tumors (e.g. at the peritoneal wall), the same process was repeated for each tumor. After imaging, mice were cleaned and allowed to recover completely from anesthesia before returning to their cage. To obtain measurements, files were loaded with the US software study management function. Maximal longitudinal and transversal diameters were obtained for each tumor image. To calculate tumor volumes, the formula V=a*b²/2, where a is the largest and b is the smallest diameter within the 4 obtained measurements (2 trans-axial and 2 longitudinal) (Tang et al., 2012).

Data are presented as the mean ± SD unless otherwise specified. Statistical analyses were performed with the tests as indicated. A p-value of <0.05 was considered statistically significant unless otherwise indicated. All analyses and graph production were performed using Prism version 9.0 (GraphPad) or Microsoft Excel.