Physical Activity Decreases Inflammation and Delays the Development of Obesity-Associated Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is deadly, with a 5-year survival rate of 13%, due to lack of early detection, prevention, and treatment approaches (1, 2). Consequently, patients with PDAC are diagnosed at advanced stages of disease when surgical resection is often not possible (1, 2).

Obesity adversely influences mortality and increases PDAC risk (3) and high body mass index (BMI) is associated with earlier age of onset, decreased survival, and therapy resistance of PDAC (4). This is concerning because 30% of the world’s population is overweight or obese and rates are increasing (5). We and others have shown that obesity (genetic or diet-induced) accelerates tumor progression in mouse models and is associated with increased systemic levels of inflammatory mediators, like IL1ß, IL6, TNFα, and lipocalin 2 (LCN2), and promote tumor growth (6–9). As obesity is a modifiable risk factor for PDAC, identifying effective lifestyle changes for obesity management, or obesity-related signaling pathways, could help minimize the development and progression of PDAC (10).

Lifestyle interventions that reduce obesity, like diet alterations and increased activity, improve the outcomes of several diseases (11). Moderate-to-vigorous physical activity (PA) reduces the risk of some cancers and chronic diseases (10, 12, 13). PA decreases obesity-associated inflammation by stimulating the release of beneficial cytokines, myokines, and/or adipokines (14). PA reduces PDAC risk (15), but clinical studies involving PA focus on increasing treatment efficacy and quality of life and do not evaluate the effects of PA or diet-induced weight loss (DI-WL) as a preventive strategy for PDAC (10). Clinical studies that use PA to prevent PDAC are important, but challenging with the low incidence rate of PDAC (10, 16). In mouse models, we demonstrated that diet-induced obesity (DIO) accelerated pancreatic inflammation, fibrosis, pancreatic intraepithelial neoplasia (PanIN) lesions, and PDAC in a KRAS-dependent manner (17). Therefore, implementing lifestyle interventions to decrease obesity through PA and/or DI-WL could prevent obesity-associated PDAC.

This study investigates changes in systemic levels of cytokines before and after PA in overweight or obese subjects or mice with obesity and PDAC. We assess whether increased PA, DI-WL, or both, alone or combined with chemotherapy, could prevent tumor development and progression, tumor growth, and inflammation in preclinical models of obesity-associated PDAC and whether increasing the expression of Il15 (upregulated by PA) in the adipose tissue of mice delays tumor growth.

Animal studies adhere to the Animal Research: Reporting of In Vivo Experiments guidelines and were compliant with regulatory and ethical guidelines for animal studies approved by the Institutional Animal Care and Use Committees at The Ohio State University and The University of Texas MD Anderson Cancer Center. The Institutional Review Board of The University of Texas MD Anderson Cancer Center approved the human study and the study was conducted in accordance with US Common Rule.

Obese and overweight subjects were recruited (n = 12) in the spring and fall of 2014 for a pilot study in partnership with the Power in Motion community organization that provides a 10-week running clinic for new and intermediate runners preparing to complete a 5K race (http://powerinmotion.org). Inclusion criteria included subjects >18 years old, able to walk 30 minutes continuously, read and write English, provide written informed consent, and have not participated in a 5K race (or other endurance athletic event) within the past 2 years. There were no exclusion criteria. Subjects participated in a 10-week training course that included weekly talks and small group runs adapted to the participants’ fitness levels. Blood samples and body composition using anthropometric measures and dual-energy X-ray absorptiometry were compared before and within 2 weeks after the program.

LSL-Kras^G12D mice (18) were bred with the Ela-Cre^ErT mice (RRID:IMSR_JAX:025736; ref. 19) Kras^G12D/CRE mice (Kras^G12D; mixed background) and littermates (control) and given tamoxifen (TAM) orally for 3 days, as described (17). Body weight (BW) was measured weekly. Blood collection was performed via submandibular bleeding, and glucose was measured using Contour Next glucometer 7308 (Ascensia Diabetes Care).

Pancreatic tumor cells derived from an LSL-Kras^G12D, LSL-Trp53^−/−, and Pdx1-CRE (KPC) genetically engineered mouse model (GEMM) expressing firefly luciferase (KPC-LUC) were utilized in the PDAC orthotopic model (7). Viable cells were counted and mixed in PBS with 20% Matrigel (BD Biosciences). C57BL/6J mice (The Jackson Laboratories; RRID:IMSR_JAX:000664) were orthotopically injected with KPC-LUC cells (0.25 × 10⁶ cells/mouse) as described (7) and given buprenorphine subcutaneously every 12 hours for 48 hours (Covetrus). Subcutaneous injections of D-luciferin (1.5 mg/mouse; Caliper Life Sciences) were given to visualize tumor growth weekly using the In Vivo Imaging System (IVIS; Caliper Life Sciences; RRID:SCR_018621) and measured using the Living Image software (RRID:SCR_014247; ref. 7). Body composition (fat and lean mass) was measured using an EchoMRI system (EchoMRI LLC; RRID:SCR_017104). Food consumption was measured weekly by weighing the food placed in a cage and subtracting the remaining food.

PA mice had access to a low-profile running wheel, whereas the control groups had access to a locked low-profile running wheel for mice (4 hours/day, 5 days/week; Med Associates Inc.; RRID:SCR_024879). All mice were housed individually to allow for activity tracking and provided with in-cage enrichment to minimize stress. As mice are nocturnal, they were placed in an altered light cycle environment (noon–midnight dark cycle), a week before starting interventions to adjust and maximize PA performance. A wireless USB hub with Wheel Manager 2.03.00 acquisition software (Med Associates Inc.; RRID:SCR_024880) tracked distance ran.

A gonadal white adipose tissue (gWAT) recombinant adeno-associated vector (rAAV) was generated as described (20) containing dual expression cassettes that restrict off-target transduction in the liver. The first cassette consists of the cytomegalovirus enhancer and chicken β-actin promoter, the woodchuck posttranscriptional regulatory element and the bovine growth hormone polyA tail flanked by AAV2 inverted terminal repeats as described (21). The second cassette encodes a microRNA driven by the albumin basic promoter to target the woodchuck posttranscriptional regulatory element (20). Mouse Il15 complementary DNA (GenBank: NM_001254747.4) was subcloned into a multiple cloning site downstream of chicken β-actin promoter as described (22). The sequence was confirmed, and an empty or green fluorescent protein vector was used as control. rAAV vectors were packaged into a Rec2 capsid. AAV viruses were purified by iodixanol gradient centrifugation as described (20, 23). Mice received rAAV/empty, rAAV/GFP, rAAV/IL15, or rAAV/IL15/IL15 receptor agonist (ra) vectors (2 × 10¹⁰ vg per mouse via intraperitoneal injection in 100 μL AAV dilution buffer). After 3 to 4 weeks KPC-LUC cells were orthotopically implanted as described (7).

DIO was achieved using a high-fat diet (HFD) in which 60% of the energy was derived from fat (DIO 58Y1 van Heek Series; Test Diet) as described (7, 17). A lower-fat control diet was used for the DI-WL interventions, in which 10% of the energy was derived from fat (DIO 58Y2; Test Diet or Pico Lab Rodent Diet 20 (5053); Lab Diet; ref. 17). For some Kras^G12Dand control mice, TAM was administered and induced DIO at the same time (TAM pre-DIO) or induced DIO prior to TAM administration (TAM post-DIO).

Hematoxylin and eosin sections were evaluated and scored by a board-certified veterinary pathologist (S.E. Knoblaugh) blinded to genotype and intervention. PanIN lesions were classified and graded according to standard criteria (24). Pancreata were also scored for the most severe and frequent lesions per section using a scoring scheme adapted from a grading scheme used for lesions in transgenic adenocarcinoma of the mouse prostate (25). Fibrosis and inflammation were assessed on a scale of 0 to 4. The adjusted lesion scores and the fibrosis and inflammation scores generate the total pathology score. (Supplementary Table S1)

Mouse splenocytes were isolated, cryopreserved, thawed, and resuspended in 5% FBS in PBS. Cells (5 × 10⁵/mL⁻¹ × 10⁶/mL) were incubated at 4°C with fluorochrome-labeled antibodies and appropriate isotype controls (Supplementary Table S2), fixed (1% formalin), and measured using a Fortessa (BD Biosciences) or an Attune (Life Technologies) flow cytometer and analyzed using FlowJo software (BD Biosciences; RRID:SCR_008520).

Statistical analyses were performed using Prism 10 (GraphPad Software; RRID:SCR_002798) or JMP Pro 17 (JMP Statistical Discovery; RRID:SCR_014242). A paired Wilcoxon test was used for data from the same subjects before and after an intervention. Unpaired t tests were used for comparing one variable between two groups or Mann–Whitney tests if non-normally distributed. A repeated-measures mixed between-within ANOVA was used to assess groups over time. A One-way ANOVA test was used for comparing more than two groups or a Kruskal–Wallis test for non-normally distributed data and two-way ANOVA (regular or repeated measures) was used when comparing multiple factors with a log transformation for non-normally distributed data. Data are expressed as the mean ± standard error of the mean unless otherwise noted in the figure legend. Significance: ^∗, P ≤ 0.05; ^∗∗, P ≤ 0.01; ^∗∗∗, P ≤ 0.001; ^∗∗∗∗, P ≤ 0.0001.

All data and analytic methods are included in the article, and supplementary material is available upon request from the corresponding author.

We assessed whether a PA intervention in overweight and obese subjects could reduce inflammation by measuring serum levels of cytokines of unfit individuals before and after a 10-week voluntary running program. After the intervention, subjects exhibited decreased levels of several proinflammatory cytokines and growth factors including, IFNγ (P = 0.001), IL12p70 (P = 0.031), IL17 (P = 0.005), IL1ra (P = 0.031), IL1β (P = 0.016), IL6 (P = 0.004), IL8 (P = 0.002), monocyte chemoattractant protein 1 (MCP1; P = 0.001), macrophage inflammatory protein 1α (MIP1α; P = 0.002), TNFα (P = 0.012), VEGF (P = 0.027), and lipocalin 2 (LCN2; P = 0.021) and a decreasing trend in eotaxin (P = 0.052) compared with before the intervention (Fig. 1; Supplementary Table S3). Moreover, the waist-to-hip ratio (P = 0.027) and waist circumference (P = 0.004), more accurate indicators of health than BMI (26), were reduced, but not BMI (P = 0.187) after the intervention (Table 1). We observed an increasing trend in the percentage of body fat after the intervention that, with changes in waist-to-hip ratios, could suggest that the intervention promoted fat storage redistribution. These results confirm that increased PA can decrease systemic inflammation and modulate body composition in overweight and obese subjects.

To determine whether voluntary PA could reduce inflammation and delay the development of obesity-associated PDAC, obesity was induced in Kras^G12D and control mice (TAM pre-DIO) and randomized into a control (DIO) or PA (DIO + PA) intervention group (Fig. 2A). Mice on DIO + PA ran similar distances (∼4–5 km/day; Fig. 2B) and experienced attenuated weight gain compared with the DIO intervention (P = 0.047; Fig. 2C). At the end of the study, the Kras^G12D mice that underwent the DIO + PA intervention gained less weight (P = 0.020) compared with the DIO control (Fig. 2D). The pancreas weight and the nonfasting glucose levels remained similar between the groups (Fig. 2E; Supplementary Fig. S1A); however, the DIO + PA Kras^G12D mice displayed lower levels of pancreatic inflammation (P = 0.001), fibrosis (P = 0.001), and total pathology score (P = 0.027) compared with the DIO control group (Fig. 2F–I). There was a negative correlation between the total pathology score and the average distance ran (P = 0.033; Fig. 2J) and a positive correlation between the total pathology score and BW change (Supplementary Fig. S1B), indicating that advanced cancer was related to decreased activity and increased BW gain. Within the Kras^G12D mice, the DIO + PA group had fewer PanIN-1 lesions than the DIO control group (P = 0.002; Fig. 2K). Notably, 43% of the mice in the DIO developed PDAC, whereas none of the mice in the DIO + PA developed PDAC (Fig. 2L). IHC showed that the pancreas of the DIO + PA mice had less fibrosis (αSMA, P = 0.029) and cell proliferation (Ki-67, P = 0.029) and a trend toward less macrophage infiltration (F4/80, P > 0.05) compared with the DIO control group (Supplementary Fig. S1C–S1F). Kras^G12D DIO mice showed a decrease in circulating levels of IL22 (P = 0.024) and IL12p70 (P = 0.012) over time and a trend toward increased IL6 (Fig. 2M). Kras^G12D mice in the DIO + PA group had decreased circulating levels of IFNγ (P = 0.012), MIP2 (P = 0.039), the epithelial-derived neutrophil-activating (ENA) peptide 78 (P = 0.024), IL18 (P = 0.039), MIP1β (P = 0.012), a decreasing trend for IL22 (P = 0.09), and increased levels of circulating IL6 (P = 0.012), and LCN2 (P = 0.039) over time (Fig. 2M). Overall, these results demonstrate that PA delays PDAC development in a DIO GEMM and dampens systemic inflammation during DIO and PDAC.

To understand the interaction between the timing of obesity, cancer induction, and weight loss, we induced mutant Kras^G12D expression either at the same time that mice were started on a DIO (Kras^G12D pre-DIO) or after 8 weeks of DIO (Kras^G12D post-DIO) and then mice underwent a PA intervention while on DIO (Fig. 3A). On average, mice ran similar km/day (Fig. 3B) and the Kras^G12D mice with DIO + PA tended to gain less weight (Fig. 3C). PA had an overall effect in preventing weight gain (P = 0.008), particularly in the Kras^G12D pre-DIO (P = 0.036; Fig. 3D). Moreover, Kras^G12D pre-DIO mice in the PA intervention gained less fat mass (P = 0.009) and a trend in Kras^G12D post-DIO (P = 0.093; Fig. 3E), but not lean mass, than in control mice (Supplementary Fig. S2A). Food consumption was similar between all groups, suggesting voluntary PA does not affect appetite (Supplementary Fig. S2B).

Mice on the PA intervention exhibited no difference in pancreas weight compared with mice on DIO (Fig. 3F). The timing of mutant Kras^G12D expression relative to DIO and the PA intervention did not affect pancreatic fibrosis (Fig. 3G and H), inflammation (Fig. 3I), or total pathology score (Fig. 3J). Mice in the Kras^G12D post-DIO group had more PanIN-1 lesions than mice in the Kras^G12D pre-DIO group (P < 0.001; Fig. 3K). Additionally, Kras^G12D pre-DIO + PA mice, but not the Kras^G12D post-DIO + PA mice, trended toward fewer PanIN-1 lesions compared with the DIO-only Kras^G12D mice, and there were no differences in the number of PanIN-2 and PanIN-3 lesions (Fig. 3K). Kras^G12D pre-DIO mice also displayed less PanIN lesions than Kras^G12D post-DIO mice (Fig. 3K). Several mice in the Kras^G12D pre-DIO + PA had no cancer or PanIN lesions; however, this was not observed in the Kras^G12D post-DIO + PA mice (Fig. 3L). Notably, there was no correlation between total pathology score and average distance ran (data not shown), or between total pathology score and BW change (Supplementary Fig. S2C and S2D), regardless of the timing of DIO and Kras^G12D expression. These data suggest that the effects of obesity on PDAC prior to Kras^G12D expression are harder to prevent.

As hyperinsulinemia can promote PDAC growth and PA is known to increase insulin sensitivity (9, 27), we assessed changes in glucose and insulin levels over time. Both glucose and insulin generally increased after DIO regardless of the timing of Kras^G12D expression. However, serum glucose decreased after PA only in the Kras^G12D pre-DIO + PA mice (P = 0.002; Supplementary Fig. S2E–S2H). Interestingly, glucose and insulin decreased at an endpoint in the Kras^G12D post-DIO mice after mice were left on DIO (glucose, P < 0.001; insulin, P = 0.006; Supplementary Fig. S2F and S2H), which suggests that the metabolism of mice that are obese prior to Kras^G12D expression (post-DIO) is different and should be explored further.

We explored additional methods to decrease obesity by evaluating whether a combination of PA and diet modulation could further delay obesity-associated PDAC. Kras^G12D and control mice were induced with TAM for Kras^G12D activation prior to DIO and randomized to the interventions. One group continued the DIO, a second group was switched to a DI-WL intervention, and a third group was switched to a DI-WL + PA intervention (Fig. 4A). All mice in the DI-WL + PA group ran similar distances daily (∼4.7 km/day; Fig. 4B). Mice switched to the DI-WL or the combined DI-WL + PA experienced weight loss (P < 0.001; Fig. 4C and D). The control mice on both interventions (DI-WL and DI-WL + PA) experienced lower nonfasting glucose levels compared with the DIO group (DI-WL, P = 0.038; DI-WL + PA, P = 0.003; Fig. 4E). However, only the Kras^G12D mice on the DI-WL + PA intervention had lower nonfasting glucose levels compared with the DIO group (P = 0.035; Fig. 4E). Pancreas weights were reduced in the Kras^G12D mice on the DI-WL (P = 0.002) and the DI-WL + PA (P = 0.013) interventions compared with the DIO control. In contrast, pancreas weight remained unchanged in the control mice after the interventions (Fig. 4F). These results suggest that reducing dietary fat intake promotes weight loss, whereas PA synergistically enhances the effects of diet modulation in a DIO PDAC GEMM.

Kras^G12D mice in the DI-WL + PA intervention showed a decreasing trend of pancreatic inflammation (P = 0.054), fibrosis (P = 0.068), and total pathology score (P = 0.066) compared with the DIO group (Fig. 4G–J). The DIO group had fewer PanIN-1 lesions than the DI-WL (P < 0.05) or the combined DI-WL + PA (P < 0.01) intervention groups (Fig. 4K); however, 100% of the mice in the DIO group developed PDAC, whereas only 67% in the DI-WL and 40% in the DI-WL + PA groups developed PDAC (Fig. 4L). To understand whether these beneficial effects were more associated with a change in weight loss rather than the PA, we correlated total pathology score and weight change, which was not significant (Supplementary Fig. S3A). These findings illustrate that weight loss from a combination of dietary and PA interventions can delay tumor progression in a DIO PDAC GEMM.

To determine whether diet and PA interventions could be preventive strategies to delay PDAC growth after obesity, C57BL/6J mice were exposed to DIO and then randomized to either remain with DIO, or to interventions for PA (DIO + PA), DI-WL, or a combined DI-WL and PA (DI-WL + PA). After the interventions, KPC-LUC cells were implanted orthotopically while the mice continued the dietary interventions without PA (Fig. 5A). Mice in the DIO + PA and DI-WL + PA interventions ran similar distances (Fig. 5B), which resulted in reduced weight (P = 0.002) and fat mass (P < 0.001) before PDAC cell implantation compared with the mice in the DIO control group (Fig. 5C and D). Furthermore, the DI-WL group also experienced weight loss (P < 0.001) and reduced fat mass (P < 0.001) compared with the DIO group (Fig. 5C and D). Therefore, a DI-WL intervention drives weight loss, but additional PA can still delay weight gain in the context of obesity.

At study completion, BWs were decreased in the DI-WL (P = 0.005) and DI-WL + PA (P = 0.001) groups compared with the DIO control (Fig. 5E), whereas the DIO + PA group had similar BW changes to the HFD control. Tumor weights and weekly tumor growth remained similar across all groups (Fig. 5F and G). Interestingly, nonfasting glucose increased in all the groups that underwent either DI-WL or PA, with the DIO group displaying the lowest levels of nonfasting glucose (Supplementary Fig. S3B). Therefore, neither PA nor a DI-WL delayed tumor growth in a DIO orthotopic mouse model of PDAC.

To understand how the interventions affect immune cells and inflammation, we measured splenic immune cell populations to evaluate the effects of PA and diet on the immune system of the DIO orthotopic model. There were no differences between the percentages of myeloid-derived suppressor cells (MDSC), F480⁺ cells, NK cells, CD8⁺, and CD4⁺ T cells between groups (Supplementary Fig. S3C), which aligned with the lack of changes in tumor growth. In contrast, when nonobese tumor-bearing mice were treated with gemcitabine in combination with PA (Supplementary Fig. S4A), splenic CD8⁺ T cells were elevated compared with the saline group without PA (P = 0.002), and the saline group with PA (P = 0.001; Supplementary Fig. S4B). We did not observe any changes in BW (Supplementary Fig. S4C) but did observe reduced body fat composition during the PA intervention, particularly in those that underwent PA (P < 0.001; Supplementary Fig. S4D). Although we saw a delay in tumor growth 7 days posttumor implantation due to PA compared with no PA in this model (P = 0.032; Supplementary Fig. S4E), we did not observe any additional benefits of combined gemcitabine and PA intervention on tumor growth (Supplementary Fig. S4F and

To understand how PA impacts inflammatory signaling in the adipose tissue, we performed a transcriptomic analysis of the gWAT from the Kras^G12D and control mice from Fig. 2. PA downregulated genes involved in several immune-related pathways associated with a proinflammatory state (Fig. 6A and B; Supplementary Table S4). Furthermore, we observed an upregulation of the IL15/IL15ra anti-inflammatory axis due to PA, which we verified by quantitative RT-PCR (P = 0.003; Fig. 6C and D).

To explore whether adipose-derived IL15 could be used as a therapeutic target in obesity-associated PDAC, we upregulated the expression of Il15/Il15ra in the gWAT of obese C56BL/6J mice and performed orthotopic implantation of PDAC cells to measure tumor growth (Supplementary Fig. S5A). gWAT Il15 expression was verified via RT-qPCR after the intervention (P < 0.001, Supplementary Fig. S5B); however, IL15 levels were mostly undetectable in the serum (Supplementary Fig. S5C). BW and pancreas tumor weight were similar between the Il15/Il15ra overexpression and control mice (Supplementary Fig. S5D and S5E), suggesting that Il15 gWAT overexpression was not sufficient to exert beneficial changes in PDAC in the context of obesity.

Nonetheless, as previous studies show that exercise-induced IL15/IL15ra axis promotes antitumor immunity of PDAC in the nonobese setting (28), we aimed to test if the adipose-specific expression of Il15 could still cause beneficial changes in PDAC growth in the nonobese mice (Fig. 6E). In this case, tumor growth was delayed in the mice with adipose tissue-specific Il15 upregulation compared with control mice (P = 0.032; Fig. 6F and G). Expression of Il15 in the gWAT of the mice was verified by quantitative RT-PCR (P = 0.046; Fig. 6H). These results suggest that PA modulation of the IL15/IL15ra anti-inflammatory pathway in the adipose tissue could contribute to the benefits of PA for the treatment of PDAC but less effective in the context of obesity.

Here, we show that a voluntary running-based PA intervention reduces systemic inflammation in overweight and obese subjects and in obesity-associated PDAC in mice. We also demonstrate that PDAC mice with DIO on a PA intervention had less weight gain, systemic inflammation, tumor progression, and a decrease in adipose proinflammatory signals. However, these benefits were not as evident when obesity preceded pancreas Kras^G12D expression. Combining PA with DI-WL–delayed obesity-associated PDAC progression, but neither PA alone nor combined with DI-WL or chemotherapy prevented PDAC growth in obese or nonobese orthotopic PDAC models. We show that PA increases the expression of IL15ra in the adipose tissue and that increasing adipose-specific Il15/Il15ra expression could be a strategy to delay PDAC in the nonobese setting. These findings suggest that weight loss via PA and dietary changes reduce inflammation and delay obesity-associated PDAC development or progression.

The benefits of PA weight loss interventions in lowering the risk of several types of cancer and chronic diseases, including PDAC, are known (15, 29). However, modeling this in rodents is challenging as the impact of weight loss interventions on cancer outcomes can depend on the degree of metabolic dysregulation due to the length of DIO and mouse strain utilized (30–32). Our results in the obese GEMM with the Kras^G12D induction post-DIO, or in the obese orthotopic PDAC model in which obesity preceded PDAC, demonstrated that PA was insufficient to improve tumor outcomes compared with Kras^G12D induction pre-DIO. Here, the DIO priming may have a stronger effect on PDAC progression than the timing of Kras expression or PA because Kras post-DIO mice had a higher frequency of PDAC compared with Kras pre-DIO mice. We observe differences in the response to DIO between the Kras pre-DIO groups (Figs. 2 and 4), and the Kras pre- and post-DIO experiment (Fig. 3). These differences may be related to a variation in the background strain of Kras mice used in Fig. 3, a limitation to this study. However, strain-related differences in DIO responses have been reported and proper controls were utilized (33). Similarly, in breast cancer, DI-WL was insufficient to reverse obesity-associated inflammation and did not have any effect on tumor growth (34). Likewise, a patient-derived subcutaneous model of PDAC did not differ in the tumor growth of mice after a posttumor implantation treadmill intervention (35). The lack of tumor initiation steps in the orthotopic and subcutaneous models eliminates any effect that the interventions have on the early stages of the disease. Therefore, PA, alone or with DI-WL, may be more effective in delaying early tumor development rather than tumor growth.

Glucose and insulin control is important because patients with new-onset diabetes have a higher risk of developing PDAC and can develop hyperglycemia prior to diagnosis (36). PA can improve systemic glycemic control in patients with PDAC after tumor resection (37). Hyperinsulinemia, a condition associated with obesity and diabetes, can promote PDAC tumor initiation (9, 27), increasing the relevance of PA to regulate insulin in this high-risk PDAC population. We demonstrated that combining PA and DI-WL in obesity-associated PDAC GEMMs lowered blood glucose; however, the effect is reduced when obesity precedes PDAC. In the obese orthotopic PDAC model, glucose levels increased with the PA and diet interventions, emphasizing the variations in outcomes based on the intervention and model utilized. This highlights the potential for PA to mediate metabolic changes in obesity-associated PDAC.

In cancer, exercise induces immune cell-mediated effects like increased CD8⁺ T-cell infiltration and decreased MDSCs, suggesting a shift in the tumor microenvironment from immunosuppressive to immunocompetent, improving the effectiveness of immunotherapy (38). PA improves tumor vascularity and enhances chemotherapy outcomes in patients with PDAC; however, the effects of PA on immune cells are conflicting (35). Here, PA generally reduced pancreatic inflammation in obese mice, although differences in specific immune cells were not characterized. However, there were no PA-related differences in splenic immune cell populations. Rather, gemcitabine reduced orthotopic tumor size and increased the population of splenic CD8⁺ T cells, regardless of PA. Similarly, a subcutaneous PDAC mouse model with “low volume” continuous exercise had reduced tumor growth, but this effect was not seen when exercise was combined with chemotherapy and there was no immune system stimulation from the PA (28, 39). In contrast, increased tumor infiltration of CD8⁺ T cells, decreased MDSCs, and delayed tumor growth were observed in mice on a treadmill intervention with PDAC orthotopic tumors (28). This same study found increased CD8⁺ T cells in the tumors of patients with PDAC after a neoadjuvant exercise intervention (28). Furthermore, forced treadmill running versus voluntary wheel running have different impacts on stress, immune response, and treatment outcomes (40). Many PA interventions in the literature are not in the context of obesity, which may explain the lack of PA-associated differences in immune cell populations in our models.

In addition to a decrease in overall pancreatic inflammation, we observed a reduction in circulating proinflammatory cytokines like IL18, IFNγ, MIP2, and MIP1β, due to PA in PDAC mice. PA in overweight/obese humans resulted in reductions in similar proinflammatory cytokines, including IL6 and IFNγ, even though there was no weight loss after the interventions. After exercise, cytokines like IL6 are transiently increased and rapidly cleared through increased hepatic clearance (41), can promote insulin sensitivity, and stimulate anti-inflammatory cytokines like IL10 and IL1ra in the blood. Indeed, transiently elevated IL6 from aerobic exercise can reduce cancer cell growth (42). Here, IL6 was increased in both the DIO and the DIO + PA groups, suggesting that PA may not be sufficient to prevent the chronic elevation of IL6 due to the tumor. IFNγ was reduced by exercise in both humans and mice and involved in immune cell activation (43). Exercise induces an expansion of regulatory T cells in the muscle, limiting the overproduction of IFNγ by NK and T cells (44). IFNγ can have antitumor responses by increasing MHC class I molecules on tumor cells; however, IFNγ signals can also suppress immune activity, limiting the effectiveness of immunotherapies (43). Here, the decrease in cytokines in both humans and mice after PA is likely due to chronic increased clearance and promotion of anti-inflammatory signaling. However, we only show correlations between changes in cytokine levels from PA, but not weight loss, and a reduction in cancer development. Further research will be needed to establish causality between these factors and whether weight loss, like DI-WL, can also alter cytokines.

Apart from the effects of obesity, adipose tissue inflammation worsens cancer outcomes (45). Adipose tissue can contribute to tumorigenesis by activating the CXCL12-CXCR4/CXCR7 signaling axis in prostate cancer (46), increasing adipose fatty acid release in PDAC (47), and increasing adipose-derived proinflammatory cytokines (like IL6, MCP1, and RANTES) in breast cancer (48). We show that PA downregulated genes involved in immune-related proinflammatory pathways in the gWAT of Kras^G12D mice while also increasing the levels of IL15ra. Agreeing with the protective effect of adipose-derived IL15 in lean PDAC mice, a similar study found that exercise promoted the mobilization of IL15Rα⁺ CD8⁺ T cells to the tumor, improving PDAC outcomes in a nonobese mouse model (28). Notably, although PA induces IL15 signaling, it is not reflected in the serum IL15 levels after PA in humans or PDAC GEMMs (28). Serum IL15 is transiently upregulated postexercise, but not after chronic exercise, like the training interventions in this study (49). This means that the effects of chronic PA might be tissue-specific and other signaling effectors, like immune cells may reduce tumor growth after PA (28). Here, adipose-derived IL15 was not protective in an obese orthotopic model, suggesting that the beneficial role for targeting gWAT IL15 expression may only be effective in lean individuals as DIO may be priming PDAC beyond the effectiveness of gWAT-derived IL15. Future studies should focus on tumor initiation in autochthonous models. In humans, microdialysis perfusion of IL15 into the adipose tissue increased lipolysis in lean, but not in obese individuals (50). This suggests that the obesity-mediated roles of adipose signaling, both pro- and antitumorigenic, merit further investigation.

Overall, PA, alone or in combination with DI-WL, delays obesity-associated cancer initiation and progression, similar to the effects observed in nonobese models (28). These effects are related to reduced systemic and adipose tissue inflammation associated with increased PA. However, PA-related beneficial effects are ablated when obesity precedes PDAC initiation. Although additional studies are needed to better understand the mechanisms by which PA and DI-WL delay DIO PDAC progression, especially if it is affecting the tumor directly or indirectly via adipose tissue contributions and inflammation, this study provides strong evidence for the benefits of weight management interventions early in tumorigenesis. Therefore, lifestyle interventions to manage obesity or modulate weight loss-related pathways in combination with other strategies could prevent obesity-associated PDAC in high-risk individuals.

L. Cao reports a patent for PCT/US2018/024305 pending, licensed, and with royalties paid from Zvelt Therapeutics; L. Cao is co-founder of Zvelt Therapeutics that licensed the patent of adipose targeting AAV vector system used to produce AAV vectors for this research. C.C. Coss reports grants from the NCI during the conduct of the study, as well as a patent for Methods and Compositions for the treatment of cancer cachexia pending and with royalties paid from Recursion Pharmaceuticals. K. Basen-Engquist reports grants from the NCI and Cancer Prevention and Research Institute of Texas outside the submitted work. Z. Cruz-Monserrate reports grants from UT MD Anderson Cancer Center, The Ohio State University Comprehensive Cancer Center, The National Center for Advancing Translational Sciences, NCI T32 Tumor Immunology Fellowship, MD Anderson Cancer Center Support Grant CA016672, OSUCCC P30 NCI, and Pelotonia Scholarship Program during the conduct of the study. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Any opinions, findings, and conclusions expressed in this material are those of the author(s) and do not necessarily reflect those of the Pelotonia Scholarship Program, The Ohio State University, and the NIH.

V. Pita-Grisanti: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. E. Velez-Bonet: Conceptualization, data curation, methodology. K. Chasser: Conceptualization, data curation, methodology. Z. Hurst: Data curation, methodology. A. Liette: Data curation. G. Vulic: Data curation. K. Dubay: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. A. Lahooti: Data curation, formal analysis, methodology, project administration, writing–review and editing. N. Badi: Data curation, formal analysis, methodology, project administration, writing–review and editing. O. Ueltschi: Data curation, formal analysis, project administration, writing–review and editing. K. Gumpper-Fedus: Data curation, formal analysis, methodology, writing–original draft, writing–review and editing. H.-Y. Hsueh: Formal analysis, writing–review and editing. I. Lahooti: Data curation, formal analysis, project administration, writing–review and editing. M. Chavez-Tomar: Formal analysis, writing–original draft, project administration, writing–review and editing. S. Terhorst: Data curation, project administration, writing–review and editing. S.E. Knoblaugh: Data curation, formal analysis, project administration, writing–review and editing. L. Cao: Data curation, formal analysis, project administration, writing–review and editing. W. Huang: Project administration. C.C. Coss: Project administration, writing–review and editing. T.A. Mace: Data curation, formal analysis, project administration, writing–review and editing. F. Choueiry: Data curation, formal analysis, project administration, writing–review and editing. A. Hinton: Formal analysis, writing–review and editing. S. Culp: Formal analysis, writing–review and editing. J.M. Mitchell: Data curation, formal analysis, project administration, writing–review and editing. R. Schmandt: Data curation, formal analysis, project administration, writing–review and editing. M. Onstad Grinsfelder: Data curation, formal analysis, writing–review and editing. K. Basen-Engquist: Data curation, formal analysis, project administration, writing–review and editing. Z. Cruz-Monserrate: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank The Ohio State University Comprehensive Cancer Center Small Animal Imaging Core Laboratory for their IVIS and echoMRI services, Comparative Pathology and Digital Imaging Shared Resource for IHC support and histopathology evaluation, Analytical Cytometry Shared Resource, and Genomics Shared Resource for their mRNA Affymetrix assay and sequencing support. Research reported in this publication was supported by Knowledge GAP MDACC (to Z. Cruz-Monserrate), Start-up funds from The Ohio State University Comprehensive Cancer Center (to Z. Cruz-Monserrate), Intramural Research Program Pelotonia Idea Award from The Ohio State University Comprehensive Cancer Center (to Z. Cruz-Monserrate), Pelotonia Scholarship Program (to M. Chavez-Tomar, K. Dubay, A. Lahooti, O. Ueltschi, and V. Pita-Grisanti), National Center for Advancing Translational Sciences TL1TR002735 (K. Gumpper-Fedus), and NCI T32 Tumor Immunology Fellowship T32CA090223 (to K. Chasser) and in part by The MD Anderson Cancer Center Support Grant CA016672 and OSUCCC P30 CA016058 NCI. The content in this article is solely the responsibility of the authors and does not necessarily represent the official views of the Pelotonia Scholarship Program, The Ohio State University, and the National Institutes of Health.