The Role of Pancreatic Preproglucagon in Regulating Local Inflammation in Mice

Highlights

What are the main findings?

• Obesity increases the satiety and inflammatory responses to exogenous LPS and increases the production of pancreatic GLP-1.
• Pancreatic Gcg is necessary for restraining the local inflammatory environment in response to LPS.

What is the implication of the main finding?

• Pancreatic production of GLP-1 increases with inflammatory stress and is necessary for limiting local macrophage accumulation and inflammation.
• Expression of the GLP-1R on macrophages suggests that pancreatic production of the GLP-1R functions to regulate the local inflammatory environment.

Abstract

Data suggest that both pancreatic and intestinally produced glucagon-like peptide-1 (GLP-1) increases in response to inflammation. Here, we set out to determine the tissue-specific function of increased GLP-1 during inflammatory stimuli. Using our innovative mouse model of tissue-specific Gcg (the gene that encodes GLP-1) expression, we explored the function of GLP-1 under severe inflammatory conditions induced by lipopolysaccharide (LPS) administration in lean and obese mice. High-fat diet (HFD) increased the LPS-induced suppression of feeding and increased the plasma levels of pro-inflammatory cytokines and GLP-1. Both pancreatic and intestinal Gcg expression contribute to LPS-induced increases in GLP-1, but Gcg was not necessary for the glucoregulatory or suppressed feeding responses to LPS. While Gcg was not necessary for systemic cytokine increases with LPS in either chow- or HFD-fed mice, whole-body Gcg-null animals had increased macrophage accumulation and an increased expression of genes reflecting pro-inflammatory signaling in the pancreas. We then performed flow cytometry on the pancreas from mice expressing a fluorescent marker on the GLP-1 receptor (GLP-1R). In response to LPS, we found that pancreatic CD64+/CD11b+ macrophages expressed the GLP-1R. We conclude that under severe inflammatory conditions, pancreatic production of GLP-1 functions in an immunological rather than a metabolic role to directly regulate local macrophage accumulation.

1. Introduction

Meta-inflammation refers to the chronic low-grade inflammatory state that links obesity to metabolic comorbidities (e.g., type 2 diabetes mellitus) [1]. Understanding the factors that link metabolism and immune function are critical to gaining deeper insight into the physiology and pathophysiology of these systems. One potential link are the nutrient-responsive peptide glucagons like peptide-1 (GLP-1). GLP-1 is classically described as an incretin hormone, coded by the preproglucagon gene (Gcg). Gcg is expressed in the α-cells of the pancreas, the L-cells of the intestinal tract, and in the nucleus of the solitary tract (NTS) [2]. GLP-1 is secreted postprandially predominantly from the intestinal L-cells and functions to decrease blood glucose in an insulin-dependent manner [3,4,5,6] and to promote satiety via brain GLP-1 receptors (GLP-1R).

However, plasma GLP-1 also increases in response to inflammatory stress such as with cytokine or lipopolysaccharide (LPS) administration [7,8,9,10,11] and after gut barrier injury, suggesting that L-cells respond not only to nutrients but also pathogen stimuli [12].

Although a rise in plasma GLP-1 levels is commonly observed in both patients and preclinical models with severe inflammation, the precise function of this increase remains elusive.

While the primary product of Gcg within the pancreas is glucagon, under metabolic and inflammatory stress there is also increased pancreatic production of GLP-1 [8,13,14]. The physiological function of this increase in pancreatic GLP-1 during an immune response remains unknown, especially in obesity, where GLP-1 secretion is impaired. The purpose of this study was to determine if Gcg peptides are necessary for the metabolic or cytokine responses to severe inflammation and the impact of obesity on these responses. Here, we found that obesity increases inflammatory and plasma GLP-1 responses and leads to greater suppression of feeding without a corresponding decrease in energy expenditure in response to LPS administration. We then induced obesity in our tissue-specific Gcg expression mouse model to explore the function of intestinal vs. pancreatic Gcg-derived peptides in response to inflammation. We hypothesized that intestinal Gcg regulates the anorectic responses and that pancreatic Gcg regulates the glucoregulatory responses to inflammation.

2. Materials and Methods

Animal Care: Our study exclusively examined male mice. Eight-week old C57BL6/J male mice (The Jackson Laboratory) or genetically crossed mice (discussed below) were used for all studies and housed in the University of Michigan North Campus Research Complex animal facilities under controlled conditions (12:12 light-dark cycle, 50–60% humidity, and 25 °C) with ad libitum access to water, and normal chow (Chow; Lab Diet, 5L0D) or a high-fat diet (HFD; Research Diets D12492; 20% Kcal protein, 60% kcal fat, 20% kcal carbohydrate; 5.21 kcal/g) as indicated by each study. All mouse cages were provided nestlets and/or huts for enrichment. Intestinal or pancreatic reactivation of the endogenous Gcg gene was accomplished as described previously [15,16,17]. Briefly, Gcg-null mice were crossed with villin 1-Cre (Jax Laboratories, stock number 004586) and PDX1-Cre (Jax Laboratories, stock number 014647) mice, respectively. Offspring with Cre-specific reactivation of the Gcg gene, Gcg-null, and Cre littermate controls (Con) were generated. Glp1r-GFP reporter mice were previously validated and generated as reported [18]. Briefly, Glp1r^Cre/+ mice (Jax Laboratory; 029283) were crossed to Cre-dependent EGFP-L10a mice [19] to generate offspring with a GFP labeled Glp1r. Mice were split into LPS- vs. vehicle (Veh)-treated groups by matching body weight within diet or genotype depending on the study. During data collection, only the study coordinator was aware of the treatment group or genotype of the mice. The University of Michigan Institutional Animal Care and Use Committee approved all procedures.

Lipopolysaccharide (LPS) Dosing: At 8 weeks of age, mice either remained on chow or were switched to a HFD for 9–16 weeks depending on the endpoint of interest to assess the interaction between diet-induced obesity and metabolic response to severe inflammation. To determine whether a HFD alters the GLP-1 response to LPS compared to chow, we analyzed the plasma levels of GLP-1 and glucose before and after LPS administration. To do this, mice were fasted 4 h after the onset of light phase. Mice were then administered LPS at a dose (100 ng/g BW) previously demonstrated to increase plasma levels of GLP-1 [10,11] (n = 4/diet) or vehicle (n = 4/diet) via IP injection. Based on our preliminary work and the literature [8,10,11], this dose is also sufficient to induce inflammatory responses while limiting fatalities. Tail blood was collected in EDTA-coated tubes at 2 and 4 h after LPS. Blood glucose was measured using a glucometer (AccuChek, Roche Pharmaceuticals, San Francisco, CA, USA) at baseline, and plasma GLP-1 (see kit information below) and blood glucose were measured 2 and 4 h after LPS. Due to the increase in body weight, the HFD-fed mice received a higher absolute dose of LPS. Therefore, the above experiment was repeated under the same dietary and experimental conditions except with a flat dose of LPS (3.2 µg delivered in 0.22 mL) (Chow-LPS, n = 8; HFD-LPS, n = 9) or vehicle treatment (Chow-Veh, n = 6; HFD-Veh, n = 9). The flat dose was calculated based on the average dose a chow-fed mouse would receive at 100 ng/g LPS dose.

Food Intake: On the day of the experiment, chow- or HFD-fed mice were fasted for 4 h prior to lights out. Immediately before lights out, LPS (100 ng/g) or vehicle (0.9% NaCl) was administered via IP injection. Food hoppers were weighed and food intake was calculated 2, 4, and 24 h after injections similar to our previous studies examining food intake in mice [18,20].

Energy Homeostasis: In order to understand the impact of a HFD on changes in energy expenditure during severe inflammatory conditions, we utilized indirect calorimetry cages. A cohort of mice was generated under the same chow vs. HFD dietary conditions as described above. Mice were housed in PhenoMaster chambers (TSE Systems, Bad Homburg, Germany) for 5 days to assess food intake, energy expenditure, and respiratory exchange ratio. Mice acclimated to the chambers and 22 °C for 2 days prior to the experiment day. Prior to lights out, mice received a flat LPS dose (4 µg delivered in 0.13 mL) (n = 6/diet) or vehicle (n = 6/diet) via IP injection. This dose was calculated from 100 ng multiplied by the average body weight of the whole cohort of mice. Mice remained in the metabolic chambers for 3 more days to monitor recovery from LPS. Just before and immediately after removal from the metabolic cages, body composition (fat and fat free mass) were assessed using NMR in conscious unrestrained mice (Echo MRI, Houston, TX, USA) as we have done previously [15,16,18].

Time Course of Glucose, Glucoregulatory Hormones, and Cytokine Responses to LPS: Because previous experiments were limited by time points and blood volume that could be taken via tail nick, we designed a study to examine the time course responses of the plasma levels of metabolic hormones and cytokines in response to LPS. A cohort of male C57Bl/6J animals was generated under the same chow vs. HFD dietary conditions described above. At a time 3–5 days before the experiment, under general anesthesia (isoflurane), mice had catheters placed in the carotid artery (MJC-02, SAI Infusion Technologies, Haryana, India) and jugular vein (BTPU-014, Instech Laboratories Inc. Plymouth Meeting, PA, USA; MRE037, Braintree Scientific Inc., Braintree, MA, USA) as we have done previously [15]. After recovery from surgery, mice regained at least 90% of their pre-surgery body weight by study day. On the experiment day mice were fasted 4 h after the onset of the light phase. To minimize handling of the mice, LPS (4 µg delivered in 0.13 mL) or vehicle was administered via the jugular vein. Blood samples, about 200 μL, were collected from the carotid artery in EDTA-coated tubes containing aprotinin and dipeptidyl peptidase-4 (DPP-4) inhibitor at baseline, 30, 60, 120 and 180 min post-LPS infusion for assessment of plasma GLP-1, insulin, glucagon, IL-6, TNF-alpha, and IL-1β (see kit information below). Additional measures of blood glucose were taken every 30 min throughout the experiment. To maintain blood volume due to the frequent sampling, heparin-washed red blood cells from donor mice were administered via the jugular vein throughout the three-hour experiment at a rate of 4.4 µL/minute for a total of 800 µL. During the experiment, we observed stroke-like symptoms in 5 mice and all data from these mice were excluded from analysis. Following the experiment, mice (Chow-Veh, n = 6; Chow-LPS, n = 9; HFD-Veh, n = 8; HFD-LPS, n = 8) were sacrificed via pentobarbital sodium (Fatal-Plus solution) administered via the jugular vein. Tissues were collected for qPCR analysis.

Real-time Quantitative PCR (qPCR): To examine systemic cytokine expression, necropsied livers were rapidly removed and frozen for later analysis. After homogenization in Trizol reagent, tissue RNA was extracted using a Pure Link RNA mini kit (Invitrogen, Carlsbad, CA, USA). cDNA was synthesized (iScript cDNA synthesis kit, BioRad, Hercules, CA, USA), and qPCR was performed using a TaqMan 7900 Sequence Detection System with a TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA). Relative mRNA expression of hepatic Il6 (Mm00446190_m1), Tnfa (Mm00443258_m1), IL-10 (Mm01288386_m1) and Ccl2 (Mm00441242_m1) were calculated relative to β-actin (Mm02619580_g1) samples using the ΔΔCT method.

Hormone Assays and Cytokine Measurements: The plasma levels of total GLP-1 and IL-6, TNF-alpha, and IL-1β (Mesoscale Discovery, Rockville, MD, USA), insulin (Crystal Chem Inc., Elk Grove Village, IL, USA) and glucagon (Mercodia, Uppsala, Sweden) were measured according to the manufacturer’s instructions using commercially available kits.

Tissue Macrophage Accumulation: On experiment day, mice (n = 3–4/group) were administered LPS (100 ng/g) or vehicle treatment via IP injection. A total of 24 h after administration, mice were euthanized by CO₂ asphyxiation and the pancreas was immediately removed and placed in 10% formalin overnight. Tissues were processed for subsequent paraffin embedding and cut into 4 μm sections for immunohistochemistry staining (VECTASTAIN Elite ABC-HRP Kit, Peroxidase (Standard) (PK-6100), Vector Laboratories, Newark, CA, USA). The sections were incubated overnight with a F4/80 antibody (1:250; Cell Signaling Technology, Danvers, MA, USA) as a marker of general macrophage accumulation and counterstained hematoxylin. The slides were scanned using a Leica Aperio AT2^® scanner (Leica Biosystems, Nussloch, Germany) and quantified using ImageJ version 1.54k as the number of F4/80 cells per mm² of pancreas area of the total, islet, or acinar area.

Macrophage Characterization via Flow Cytometry: Mouse pancreata (Con-Veh n = 5, Con-LPS n = 5, Gcg-null-Veh n = 5, Gcg-null-LPS n = 5, HFD-Con-LPS n = 5, HFD-Con-Veh n = 5, HFD Gcg-null n = 4, Gcg-null-LPS n = 4, wild-type n = 1, Glp1r-GFP-LPS n = 2) were assessed for macrophage characterization using flow cytometry similar to previous protocols [21]. Pancreata were excised immediately after euthanasia and then mechanically dissociated with scissors in sterile PBS. Pancreata were washed and pelleted under centrifugation, resuspended in 1 mg/mL Collagenase V (Sigma Aldrich, St. Louis, MO, USA), digested for 15 min at 37 °C, quenched with RPMI + 10% FBS, then filtered through 40 μm filters (Corning Life Sciences Products, Corning, NY, USA). The cells underwent RBC lysis, were washed in PBS, then blocked and stained with CD45-PerCP (BioLegend, San Diego, CA, USA) for total immune cells, CD64-PE (BD Biosciences, San Diego, CA, USA) for macrophages, and CD11b-APC (Biolegend, San Diego, CA, USA) for a specific subpopulation of macrophages at 1:100 dilution in 100% FBS. Separately, bone marrow was extracted (Wild-type = 1, Glp1r-GFP-LPS n = 2), homogenized to a single cell suspension and stained with the following antibodies: CD45-eFluor450 (Thermo Fisher Scientific, Carlsbad, CA, USA) and CD11b-APC (Invitrogen/eBioscience) [22]. The cells were then washed in FACS buffer (1% FBS, HBSS), and run on a ZE4 Analyzer flow cytometer (BioRad Laboratories, Hercules, CA, USA). While we had a validated F4/80 IHC antibody, we tested both F4/80 and CD64 to identify macrophage populations with flow cytometry, and the CD64 population was more distinct for flow cytometry of the pancreas.

Pancreas RNA Isolation and Sequencing: Mice were euthanized, the common bile duct was clamped, and RNAlater was injected towards the pancreas for subsequent RNA sequencing similar to previous RNA isolation protocols [15]. Pancreata were then removed, placed overnight in RNAlater at 4 °C, and RNA was extracted the following day. Poly A mRNA sequencing was performed by the University of Colorado Cancer Center Genomics Core via the Illumina NovaSeq X platform. Raw reads were trimmed using cutadapt (version 4.8, python version 3.10.14) [23] and aligned to the mouse GRCm38 genome using STAR (version 2.7.11b) [24]. Gene expression was quantified with featureCounts (version 2.0.6) [25] and differential expression analysis was conducted using DESeq2 (version 1.44.0) [26] with R version 4.4.1. Analysis can be replicated using code available on github [27] with docker images for preprocessing [28] (version 1) and differential expression and plotting [29] (version 1). Pathway analyses were performed via enrichr [30] using the MSigDB Hallmark 2020 database [31].

Statistics: The number of mice studied per treatment and genotype are indicated within each experiment and were determined by a power analysis completed using variance from previous data. All data are expressed as mean ± SEM. The statistical procedures were performed as indicated in the figure legends. Data were normally distributed based on the Shapiro–Wilk test and were analyzed via standard parametric statistics including two-way or three-way ANOVAs with repeated measures when required or t-tests where appropriate. In some cases, there are missing time points for a given endpoint. Those data were analyzed via a mixed-effects model. If significant interactions were detected between independent variables, then Tukey’s post hoc analysis was performed to determine where the significant differences lie. Statistical analyses were performed using GraphPad Prism 10, and statistical significance was accepted when p < 0.05.

3. Results

3.1. HFD-Fed Mice Have a Prolonged Anorectic Response to LPS Compared to Chow-Fed Mice

Baseline body weight was greater in HFD- vs. chow-fed mice but was similar within each diet group treated with LPS vs. Veh (Figure 1A). Consistent with the literature, LPS-treated mice had a significant decrease in body weight (Figure 1B) for 4 days following administration. In addition, there was a significant decrease in lean but not fat mass 4 days after LPS, indicating that most of the body weight lost was from muscle mass (Figure 1C,D). To further examine the interaction between energy homeostasis, immune responses, and diet, we placed lean and obese mice in metabolic chambers and administered LPS. A typical response to LPS is a decrease in energy expenditure to prioritize immune cell function [32]. We found a significantly lower energy expenditure in the Chow-LPS vs. Veh group during the Day 1 dark cycle and then no significant difference between these groups during the first light cycle or thereafter (Figure 1E,F), indicating that the chow-fed mice were quickly recovering from the LPS injection. However, the HFD-LPS group had similar energy expenditure relative to their Veh-treated control after LPS (Figure 1E,F), implying that HFD induces a dysregulation of energy prioritization. Because the HFD-fed mice have increased body weight, we chose to analyze absolute energy expenditure. However, we saw similar results when energy expenditure was expressed relative to body weight as well (Supplementary Figure S1).

Under immune activation, competing metabolic demands of immune cells vs. physiological functions must be balanced [32]. The respiratory exchange ratio (RER) is an indicator of whole-body fuel utilization and responses in the RER over time after LPS are depicted in Figure 1G. The average RER (Figure 1H) for Day 1 was significantly higher in the Chow-Veh group compared to all others. On Days 2 and 3 both Veh- and LPS-treated chow groups were significantly higher compared to the HFD-fed groups. These RER data suggest that the chow-fed mice shift from carbohydrate to fat oxidation when exposed to LPS (100% carbohydrate oxidation indicated by a RER = 1) (Figure 1G). Fuel oxidation in Veh-treated mice is already primarily from fatty acids over carbohydrates in the HFD-fed mice (100% fatty acid oxidation indicated by a RER = 0.7) at baseline inducing a floor effect such that LPS treatment cannot further decrease the RER. By Day 2 after LPS, the RER in the chow-fed mice was similar to baseline, indicating that fuel utilization processes had recovered from the LPS administration. Conversely, the HFD-LPS mice dipped below 0.7 on the RER graph, implying the mice are oxidizing fatty acids and/or ketones. By day 3 after LPS treatment, both LPS-treated dietary groups had similar RERs to their respective control groups.

As expected, LPS-treated mice decreased their food intake within hours of administration (Figure 1I) and this suppression of feeding lasted for 3 days in both chow- vs. HFD-fed mice with the reduction being significantly greater in HFD-fed mice (Figure 1J). Specifically, one day after LPS, the HFD mice had a significantly decreased food intake compared to the chow-fed LPS group (Figure 1I). As the mice recovered on Days 2 and 3, the HFD-LPS mice continued to have decreased cumulative food intake compared to chow-fed mice (Figure 1J). Together these data suggest that HFD-fed mice have dysregulated energy expenditure and increased sickness-induced anorexia in response to severe inflammation.

3.2. Impact of LPS on Plasma GLP-1 in HFD- Versus Chow-Fed Mice

LPS is a well-established tool to induce inflammation and also increases plasma GLP-1 levels in an IL-6 dependent manner [10,33], but the impact of obesity on GLP-1 responses to inflammation is less clear. Because we saw greater suppression of feeding in obese mice, we speculated the plasma GLP-1 response may be higher. We analyzed the plasma levels of GLP-1 and glucose before and after LPS administration in chow- and HFD-fed mice. As expected, 8 weeks of HFD significantly increased body weight (mean ± SEM: Chow-Veh 30.2 ± 1.11 g, Chow-LPS 29.5 ± 0.95 g, HFD-Veh 48.4 ± 2.16 g, HFD-LPS 47.9 ± 2.30 g, p < 0.05) and baseline glucose levels (Figure 2A,B) compared to chow-fed mice. Two hours after LPS treatment, both chow- and HFD-fed mice had decreased blood glucose levels compared to baseline, but HFD animals’ were still significantly higher compared to chow-fed mice (Figure 2A,B). Four hours after LPS treatment, there were no longer statistical differences in glucose levels between the LPS-treated chow and HFD animals indicating a greater relative fall from baseline in HFD-fed mice. Total GLP-1 (Figure 2C) levels were increased with LPS treatment and were significantly greater in HFD-fed compared to chow-fed mice 2 h after LPS. Consistent with previous work, the LPS dose was calculated based on body weight [10,11]. However, because HFD-fed mice weigh more, their absolute LPS dose was higher, which, in and of itself, could increase plasma GLP-1 levels in the obese animals. To determine if the higher GLP-1 was a dose rather than diet effect, we administered a flat dose (3.2 μg) of LPS to a second cohort of chow- vs. HFD-fed mice. Again, a HFD significantly increased body weight (mean ± SEM: Chow-Veh 32.43 ± 0.66 g, Chow-LPS 32.07 ± 0.32 g, HFD-Veh 45.66 ± 1.86 g, HFD-LPS 45.43 ± 1.90 g; p < 0.05). Similar to the body weight-relative LPS dose, glucose (Figure 2B) levels in chow- and HFD-fed decreased over 4 h after LPS and a greater relative decrease was observed in HFD-fed mice. Total plasma GLP-1 (Figure 2D) was also significantly increased in the HFD- compared to chow-fed mice 2 h after LPS. Taken together, these data indicate that a HFD lowers glucose and increases GLP-1 responses to LPS.

3.3. HFD-Fed Mice Have Increased Inflammatory Responses to LPS Compared to Chow-Fed Mice

We next sought to determine a more detailed time course of plasma GLP-1 as well as insulin, glucagon, and circulating cytokines in chow- vs. HFD-fed mice after LPS. Mice underwent vascular catheterization and after recovery from surgery, LPS (4 μg; flat dose) was administered via the jugular vein while blood draws were taken from the carotid artery. This method is beneficial as it allows for a higher volume and frequency of blood draws, and it is less stressful for the mice compared to repeated tail bleeding. We maintained blood volume with a constant red blood cell infusion through the jugular vein. As in Figure 1, the LPS-treated mice had significant decreases in blood glucose levels (Figure 3A) and increases in plasma GLP-1 (Figure 3B) over time compared to Veh-treated mice and the degree of these changes was magnified (Diet × Treatment p = 0.02) in mice fed a HFD. Interestingly, although insulin was higher in HFD- vs. chow-fed mice, there were no differences in insulin with LPS treatment nor did the increase in insulin with a HFD correlate with the increase in GLP-1 with LPS treatment (Figure 3C). The LPS-treated mice had significantly increased glucagon (Figure 3D) levels compared to Veh-treated mice at 120 and 180 min after LPS and there was no additional impact of diet on these responses.

The pro-inflammatory cytokines IL-6 (Figure 3E) and TNF-α (Figure 3F) were greater in the HFD-LPS group compared to the Chow-LPS group while IL-1β (Figure 3G) increased with LPS similarly between the two dietary groups. Because the liver is an important link between the gut and systemic circulation, we evaluated hepatic expression of IL-6 and TNFα, as the plasma levels of both cytokines were significantly greater in the HFD-fed mice. We also assessed hepatic expression of IL-10 and MCP-1, as these cytokines are produced by differing cell populations including macrophages in the liver. The LPS-treated mice had a significant increase in liver cytokine expression of IL-6 (Figure 3H), TNF-α (Figure 3I), IL-10 (Figure 3J), and MCP-1 (Figure 3K) compared to Veh-treatment, but there were no significant interactions between diet and LPS. Thus, plasma, but not hepatic cytokine expression in response to LPS was elevated with HFD.

3.4. The Role of Gcg in the Physiological Responses to LPS in Chow- and HFD-Fed Mice

HFD-fed mice had greater suppression of feeding and greater reductions in blood glucose compared to chow-fed animals in response to LPS. We also found greater levels of plasma GLP-1, but not glucagon, in response to LPS in HFD vs. chow animals. Next, we wanted to determine if the greater suppression of feeding and glucose was due to the greater increases in GLP-1 with LPS. Further, while intestinal GLP-1 has been found to regulate local inflammatory processes [34], pancreatic production of GLP-1 has also been suggested to be increased with increases in inflammatory stress [13,14,35]. To examine the function of these two sources of GLP-1, we utilized a Cre-LoxP strategy targeted to Gcg expression, the gene that encodes both GLP-1 and glucagon. To do this, a Cre mouse targeted to Pdx1 or the Villin promoter was crossed to a mouse with a floxed transcriptional blocking cassette inserted into the Gcg gene (Gcg-null). The resulting mice are designated as GcgRA^ΔPanc for pancreatic reactivated or GcgRA^ΔInt for intestinally reactivated Gcg (Figure 4A). PDX1-Cre and Vil-Cre litter-mate controls were utilized and are indicated by Con in figures. The Gcg-null mice served as whole-body KO controls. These mice have previously been extensively validated and phenotyped [15,16,17].

There was no significant impact of genotype on to the ability of LPS to suppress feeding (Figure 4B,C), demonstrating that Gcg is not necessary, nor is pancreatic or intestinal Gcg sufficient, for the anorectic effect of LPS. There was no significant change in glucose in response to vehicle (Figure 4D) over time or between genotypes. While glucose was significantly lower at 240 min in response to LPS, (Figure 4E) there was no significant difference in this decrease between the genotypes. At 240 min after LPS (Figure 4F), total GLP-1 was increased in the chow-fed control and GcgRA^ΔPanc mice and, as expected, was undetected in the Gcg-null mice. The GcgRA^ΔInt mice had no significant increase in total GLP-1 levels in response to LPS. This indicates that the pancreas is a significant contributor to plasma GLP-1 responses during LPS-induced inflammation. There was no significant impact of genotype or LPS on insulin levels (Figure 4G). Together these data suggest that GLP-1 does not function to increase insulin during severe inflammation and is not necessary for the glucose response to LPS. Glucagon (Figure 4H) was significantly higher in response to LPS in control and GcgRA^ΔPanc mice and as expected due to the genetic manipulation was undetectable in both the Gcg-null and GcgRA^ΔInt mice. Several cytokines were significantly increased in response to LPS in both the plasma (Figure 4I–K) and liver (Figure 4L,M) but there was no additional impact of genotype. Because genotype did not impact hepatic expression of cytokines in the control, Gcg-null or GcgRA^ΔPanc mice, this endpoint was not assessed GcgRA^ΔInt groups.

Given the greater increase in plasma GLP-1 with a HFD (Figure 1B,D), we next examined the impact of a HFD on LPS responses in these genetically manipulated animals. Again, there was no significant impact of genotype on the feeding (Figure 5A,B), glucose (Figure 5C,D), insulin (Figure 5F), or plasma cytokine or hepatic cytokine expression in responses to LPS (Figure 5G–J) in HFD-fed animals across genotypes. Plasma glucagon significantly increased with LPS treatment in both Con and GcgRA^ΔPanc. Together these data suggest that Gcg is not necessary for the glucose, feeding, or systemic cytokine responses to LPS, regardless of diet.

With a HFD, total GLP-1 responses to LPS were significantly higher in LPS- vs. Veh-treated mice in the Con and GcgRA^ΔPanc and this increase was greater in Con vs. GcgRA^ΔPanc (Figure 5K). There was no significant increase in plasma total GLP-1 levels in intestinally reactivated mice (Figure 5K). The pancreatic contribution to plasma GLP-1 trended towards being higher in the chow- vs. HFD-fed mice (chow-GcgRA^ΔPanc 364.40 ± 95.40, HFD-GcgRA^ΔPanc 196.71 ± 45.07, p = 0.08) while the intestinal contribution (chow-GcgRA^ΔInt 69.21 ± 21.35, HFD-GcgRA^ΔInt 181.54 ± 40.08, p = 0.01) was greater in the HFD- compared to chow-fed mice. These data imply that a HFD induces a shift in the organ contribution to plasma GLP-1 from the pancreas to the intestine. If contribution to plasma GLP-1 is a marker for the intestine vs. pancreatic responses to LPS, then these data suggest that overall, a HFD dampens the pancreatic responses (at least regarding endocrine function) to inflammation. In further support of this, in the HFD-fed mice, the Con-LPS group, but not the pancreatic reactivated animals, had a significant increase in glucagon (Figure 5L).

3.5. Gcg-Null Mice Have Increased Macrophage Accumulation in the Pancreas After LPS

To test whether Gcg had a role in regulating local inflammation in the pancreas, we determined pancreatic macrophage accumulation and characterization via two methods: flow cytometry and immunohistochemistry. Pancreatic immune cells sorted on a flow cytometer were gated on singlets, scatter, and CD45⁺ (leukocyte common antigen) cells (Figure 6A and Figure S2). The population of interest, CD64⁺CD11b⁺, which represents a specific subpopulation of macrophages, was quantified and analyzed (Figure 6A and Figure S2). The Gcg-null-LPS-treated mice had significantly increased macrophage accumulation compared to all other groups in the chow-fed mice (Figure 6A).

In a separate cohort of chow-fed mice, LPS was administered and pancreata slides were stained for F4/80 (Figure 6B) as a general marker for macrophages in the total pancreatic area, (Figure 6C), the islet area (Figure 6D), and the acinar area (Figure 6E), respectively. There was an increase in F4/80 staining in the total pancreas and acinar area in response to LPS, but this increase was similar across genotypes. The Gcg-null mice had a significantly higher degree of F4/80 staining in the islet area compared to all other groups. Together, these data imply that under severe inflammation, Gcg-null mice have more macrophage accumulation in the pancreas compared to controls and reactivation of Gcg in the pancreas or intestine is sufficient to normalize this response.

Additionally, we performed RNA sequencing of whole pancreas tissue 24 h after LPS administration in control vs. Gcg-null mice (n = 2–3 samples/group). Among the differentially expressed genes, 150 were downregulated in Gcg-nulls (including Gcg itself), and 98 genes were upregulated (Figure 6F). Pathway analysis revealed downregulation of xenobiotic and bile acid metabolism (

Table 1. Pathways regulated in Gcg-null mice.

Pathways upregulated in Gcg null mice
Pathway	p-Value	q-Value	Genes
Apoptosis	0.000001	0.000048	CCND2, KRT18, CCND1, TNFRSF12A, TAP1, CD14, CLU, IER3
TNFα signaling via NFĸB	0.000058	0.000731	DUSP4, SOCS3, CCND1, BCL3, TAP1, RELB, IER3
p53	0.000058	0.000731	PROCR, CCND2, EPHX1, TAP1, SDC1, PDGFA, IER3
Cholesterol homeostasis	0.00048	0.004556	ALCAM, TNFRSF12A, CLU, CXCL16
Angiogenesis	0.000724	0.005505	CCND2, SPP1, PDGFA
IL6/JAK/STAT3 signaling	0.000884	0.005597	SOC3, TNFRSF12A, LTB, CD14
IL2/STAT5 signaling	0.002982	0.012863	CCND2, ALCAM, SPP1, SLC39A8, LTB
Pathways downregulated in Gcg-null mice
Xenobiotic metabolism	0.000021	0.000597	VTN, GCKR, DDC, GSTO1, LCAT, RAP1GAP, GNMT, ALDH941
KRAS signaling Dn	0.000794	0.011515	GAMT, CYP39A1, NR4A2, CPA2, CPB1, CELSR2, TENT5C
Bile acid metabolism	0.001564	0.015120	CYP39A1, ACSL1, NR3C2, ALDH9A1, GNMT
Pancreata β-cells	0.003334	0.024171	G6PC2, Gcg, IAPP

) including several metabolic and digestive enzymes (lecithin-cholesteron acyltransferase, Lcat; aldehyde dehydrogenase 9 family member A1, Aldh9a1; phenylalanine hydroxylase, Pah; amylase alpha 2B, Amy2b; amylase alpha 1, Amy1; and more). For the genes upregulated in Gcg-null mice after LPS, several were involved in innate immune cell infiltration and macrophage activation signaling, and the pathways for TNFα signaling via nuclear factor- κB (NF-κB) and IL-6 were enriched. Consistent with our cell sorting data showing increased CD64+/CD11b+ cells in Gcg-null pancreata after LPS, we found increased gene expression of a cluster of differentiation 14 (Cd14), Chemokine (C-X-C motif), ligand 16 (Cxcl16), clusterin (Clu), tumor necrosis factor receptor superfamily member 12A (Tnfrsf12a), and complement factor D (Cfd), altogether suggesting increased inflammatory macrophage signaling. Additionally, the upregulated C-C motif chemokine ligand 28 (Ccl28), lymphotoxin-β (Ltb), and C-reactive peptide (Crp) indicate a greater inflammatory response to LPS in Gcg-null pancreata.

3.6. Macrophages Isolated from the Pancreas Express GLP-1R

Whether GLP-1Rs are expressed on macrophages, and other immune cells for that matter, is unknown as commercially available GLP-1R antibodies are not validated and lack specificity [36]. To address this problem, we crossed a Cre-dependent eGFP-L10a reporter mouse to a Glp1r^ΔCre (Glp1r-GFP; Figure 7A) mouse. We administered LPS and pancreatic cells were stained for immune cells, acquired on the flow cytometer, and gated on singlets, scatter, CD45+, and CD64+ cells (Figure 7B). As expected, the wild-type mice showed no GFP signal (Figure 7C), whereas Glp1r-GFP mice showed a strong GFP signal (Figure 7C). Additionally, this population of GFP+ cells was positive for CD11b+, the subpopulation of macrophages we saw elevated in Gcg-null mice. Interestingly, immune cells from the pancreas, but not the in vivo bone marrow macrophages, contained these GLP-1R positive cells (Figure 7C). Although it is possible that circulating macrophages may express the GLP-1R only upon reaching the pancreas, we speculate that under the stress of inflammation, local, rather than circulating macrophages in the pancreas express GLP-1R and are likely directly regulated by locally produced GLP-1.

4. Discussion

Here we find that obese mice have increased markers of systemic inflammation, greater relative falls in glucose, and show distinct energy homeostatic responses to severe inflammation. These mice also have increased plasma levels of GLP-1 in response to LPS but this increase is not driving the increased suppression of feeding or greater falls in glucose also observed in these mice. Further, our data suggest that the increase in plasma GLP-1 with LPS stems from pancreatic vs. intestinal production in chow-fed mice. While our initial hypothesis was that pancreatic GLP-1 production under these circumstances would function to regulate glucose levels, instead our data suggest that pancreatic production of GLP-1 is necessary to restrain macrophage accumulation and inflammation in the pancreas.

High fat diet clearly impacted energy partitioning in response to LPS. LPS is known to decrease food intake, conserve energy expenditure, and alter fuel utilization away from carbohydrate and towards fat oxidation in lean animals [32]. Despite having similar reductions in body mass and composition, obese mice had greater suppression of feeding but had no significant decreases in energy expenditure in response to LPS. Thus, the similar LPS-induced decreases in body weight were due to a combination of decreased food intake and energy expenditure in chow-fed mice but were dominated by a decrease in feeding in HFD-fed mice. HFD-fed mice also took longer to recover feeding after LPS. Although it is unclear how these shifts in fuel metabolism, or lack thereof, impact the ability to defend against the stress of LPS administration, previous work found that the LPS-induced increases in fat oxidation supported immune function though a toll-like receptor 4 mechanism [37]. Indeed, our HFD-fed mice had higher levels of plasma cytokines in response to LPS. Other responses which may be linked to the altered fuel partitioning are the greater relative drops in glucose and the greater increases in plasma GLP-1.

LPS-induced increases in plasma GLP-1 are well-documented in chow-fed animals [10], and here we show that this increase is larger in obese mice. A HFD also caused greater LPS-induced suppression of feeding compared to mice fed a chow diet. Although GLP-1 regulates feeding and glucose levels, our data suggest that decreases in feeding and changes in glucose in response to LPS, whether the animals were on chow or a HFD, were not dependent on Gcg-derived peptides, including GLP-1. Previous work has demonstrated that IL-6 administration increases plasma GLP-1 [8] and that IL-6 knockout mice have no significant increase in plasma GLP-1 after LPS [10]. Thus, the increase in GLP-1 with LPS is IL-6 dependent. We speculate the greater increase in plasma IL-6 drove the greater increase in plasma GLP-1 in the HFD-fed mice.

Administration of IL-6 twice daily for over a week in mice led to an increase in the protein content of pancreatic GLP-1 [8]. Our data are consistent with an increase in pancreatic GLP-1 production with inflammatory stress as only the chow-fed control and pancreatic reactivated mice, but not the Gcg-null or intestinal reactivated mice, had significant increases in plasma total GLP-1 levels. Interestingly, only the HFD control mice had a LPS-induced increase in plasma GLP-1. This lack of increase in the GcgRA^ΔPanc mice fed a HFD could be due to baseline increases in GLP-1 in the HFD-fed mice, or to the HFD-fed GcgRA^ΔPanc mice having dysregulated responses to diet and/or LPS due to lack of Gcg in either the gut or hindbrain.

One of the more striking findings was the greater LPS-induced pancreatic macrophage accumulation we observed with both IHC and flow cytometry in the Gcg-null animals. The flow cytometry data comparing Gcg-null versus control mice shows that after LPS, the Gcg-null-LPS group had significantly more CD64+CD11b+ cells compared to all other groups. Our RNA sequencing data suggested the pro-inflammatory state of the pancreas of Gcg-null mice after LPS, with upregulated genes within the TNFα signaling via NFκB and IL-6/JAK/STAT3 pathways. These data imply that the absence of pancreatic Gcg increases the inflammatory state of the pancreas and suggest that Gcg is necessary to prevent macrophage accumulation in the pancreas in states of systemic inflammation. This is similar to the action of intestinal GLP-1 receptor signaling that regulates local T-cell-driven inflammation [34].

At this point it is unknown whether pancreatic macrophage accumulation in response to LPS is due to infiltration from the circulation or to local macrophage expansion. This accumulation also seemed to be greater in the islets than the acinar tissue, but this requires further validation. When Gcg was “restored” in the intestine or pancreas, the IHC data suggest that the macrophage accumulation is normalized. This would suggest potential crosstalk between the gut and pancreas, presumably through GLP-1, in target organ responses to inflammation. Intestinal GLP-1 has demonstrated effects on local and cardiovascular inflammation supporting this possibility [34,38]. One of the limitations in trying to understand how GLP-1 could be anti-inflammatory is that commercially available GLP-1 receptor antibodies are inaccurate. Here, we crossed a GLP-1 receptor Cre mouse to a Cre-dependent GFP reporter mouse and used flow cytometry to sort for pancreatic immune cells. Interestingly, we found GFP+ (and presumably GLP-1R+) cells in the specific CD11c+ subpopulations of cells that are increased with LPS administration. Although loss-of-function studies are the logical next step, these studies are complicated when dealing with subpopulations of immune cells due to the lack of specificity of currently available Cre-mouse models. Regardless, these data offer the intriguing possibility that GLP-1 signaling regulates pancreatic immune cell response to inflammation.

GLP-1 agonists have been suggested to have anti-inflammatory effects in multiple tissues and experimental models [39,40]. However, the mechanism (or mechanisms) by which GLP-1 agonists contribute to an anti-inflammatory state remains unknown. Patients with sepsis have higher plasma GLP-1 levels but also have increased mortality [41], which is seemingly discordant from the pharmacological data which imply beneficial effects of GLP-1R agonism in inflammation. However, we also saw minimal differences in plasma cytokine responses to LPS between Gcg-null, GcgRA^ΔPanc, and GcgRA^ΔInt vs. control mice, regardless of diet. Thus, these data point to differences in the physiology vs. pharmacology of GLP-1 signaling in immune regulation. We speculate endogenously produced GLP-1 is important for regulating local tissue inflammatory processes whereas GLP-1 agonists can have a broader systemic impact.

It is important to note the limitations of these studies. We were unable to perform IHC and cell sorting of the pancreas in HFD-fed Gcg-null mice. Another limitation is that these studies were only conducted in male mice. Given the widespread sex differences in immune function [42], the resistance to dietary induced obesity in female mice [43], and potential sex differences in GLP-1 signaling [44], future studies powered to detect sex differences would be of interest. Studies conducted in mice have limitations in the translation to human physiology. Further limiting translation, some studies here were conducted in mice with constitutive deletion of Gcg which could lead to unknown developmental compensations. While GLP-1 has been demonstrated to be increased in patients with sepsis and chronic heart failure [41], two studies examining the impact of exogenous administration of LPS on humans conflict, with one study demonstrating an increase [45] and the other showing no change in the plasma levels of GLP-1 [46]. The timing of GLP-1 assessment or differing glycemic conditions could explain the conflicting results. Of course, it is also possible that the connection between immune function and GLP-1 is more closely linked in the mouse than in humans.

5. Conclusions

This study supports a body of work demonstrating that pancreatic GLP-1 production is increased under stressful conditions. In this case, inflammatory stress increases GLP-1 levels which regulate local pancreatic macrophage accumulation and inflammatory state. Together these data imply that the function of GLP-1 under inflammatory conditions differs from its postprandial functions. Whether this increase in GLP-1 is simply a marker of the body trying to defend itself against severe inflammation or whether it has an important regulatory role remains unclear. We predict that GLP-1 physiology is an interesting link between these pathological states and impaired immune responses.