GM-CSF Administration Improves Defects in Innate Immunity and Sepsis Survival in Obese Diabetic Mice

This article is featured in In This Issue, p.627

Despite progress in antibiotic therapy, ventilator management, resuscitative strategies, and blood glucose control, sepsis remains the leading cause of death in the intensive care units (ICU), impacting over one million people annually (1). Additionally, sepsis is the leading cause of organ injury after trauma and surgical procedures (2, 3). Although there have been advancements in the clinical understanding, disease definition, and immune pathophysiology that have led to improvements in 30-d mortality, long-term sepsis mortality remains in the 60–80% range (4).

Individuals who are obese and have type 2 diabetes (T2D) have increased recurrent, chronic, and nosocomial infections that generate hospital readmissions and worsen patient long-term morbidity and mortality from sepsis (5, 6). These outcomes are crucial to explore, given the rising prevalence of obesity and T2D. Worldwide, obesity has nearly tripled since 1975 (7), and T2D has nearly quadrupled since 1980 (8). An estimated 117 million people in the United States are obese (9). An estimated 29.1 million people in the United States have diabetes of all types, with T2D comprising well over 90% of the total diabetic population (10). Patients with T2D experience far greater rates of infection, organ failure (11, 12), and mortality during sepsis (13–21), although a few studies have shown equivocal effects (11, 12, 22). Cell dysfunction in the innate immune system has been incriminated as one of the leading factors underlying disease progression and appearance of complications (23–25). One of the hallmarks of obesity is increased accumulation of immune cells, especially macrophages, within the adipose tissue (26, 27), which contributes to a chronic inflammatory state (28). Neutrophils in individuals with T2D show defects in almost all functions, including migration to inflammatory sites, release of lytic proteases, phagocytosis, production of reactive oxygen species (ROS), and apoptosis (29).

Animal models of sepsis demonstrate that obese, diabetic mice have much lower survival rates compared with nondiabetic mice (30, 31). We have shown in sepsis that neutrophils are essential for eradication of bacteria, prevention of infectious complications, and sepsis survival (32). Although many investigators report that neutrophil chemotaxis (33), phagocytosis (34), and ROS generation (35) are reduced in diabetic states, the contribution these deficits make to infection eradication and mortality in sepsis is unclear. We propose that obesity and T2D create an underlying functional immune deficiency, which directly impairs neutrophil and monocyte capabilities, resulting in decreased bacteria clearance, infection persistence, and increased sepsis mortality.

Additionally, because both short- and long-term sepsis mortality have been attributed to innate immune dysfunction (36), we hypothesize that improving the innate immune response during sepsis will result in improvements in sepsis mortality. There has been considerable emphasis on neutrophil and monocyte functional augmentation during sepsis (36). GM-CSF, a cytokine that enhances neutrophil and monocyte production and function, has shown therapeutic promise to improve outcomes in human lung injury trials (37). Based on the known GM-CSF effects, including amplification of neutrophil and monocyte production, function, and immune surveillance, we hypothesized that GM-CSF administration during infectious sepsis would increase survival by improving innate immune function in obese, diabetic mice.

All experiments were approved by the Institutional Animal Care and Use Committee at the University of Michigan. C57BL/6 and C57BL/6J diet-induced obese (DIO) male mice were purchased from The Jackson Laboratory between 24 and 26 wk of age. All mice were maintained at the University of Michigan and were studied between 28 and 32 wk of age. DIO mice were fed a high-fat diet (60 kcal% fat) compared with C57BL/6 mice, which were fed a normal-feed diet (13 kcal% fat).

For induction of abdominal sepsis, mice underwent cecal ligation and puncture (CLP), which includes ligation of the cecum and a double colon puncture (38, 39). For all survival experiments, a 20-gauge needle was used, leading to a mortality rate of 10–20%. Control mice underwent a sham laparotomy in each experiment. All survival experiments were repeated multiple times with n = 20 mice per experimental group. Mouse weights were collected during survival experiments. All mice were weighed presepsis induction and then daily after sepsis initiation until the conclusion of the experiment. Spleen weights were measured in C57BL/6 and DIO naive mice (n = 15 mice per experimental group) and C57BL/6 and DIO mice at serial time points between day 1 and 14 post-CLP (n = 3 mice per experimental group at each time point). Experimental groups of both C57BL/6 and DIO mice who underwent CLP were also treated with Recombinant Mouse GM-CSF (carrier-free) (BioLegend). GM-CSF (100–200 ng) were placed into 100 μl of PBS (Thermo Fisher Scientific) with 0.4% Albumin Solution (Sigma-Aldrich) and administered s.c. every 12 h for 7 d starting 6 h after CLP.

Flow cytometry samples were acquired and analyzed using a LSR II flow cytometer (BD Biosciences) and FlowJo software. Abs used in flow cytometry analysis were Rat Anti-Mouse CD16/Cd32, BV421 Rat Anti-Mouse CD11b, APC Anti-Mouse Ly6G, FITC Hamster Anti-Mouse CD11c, BV421 Rat Anti-Mouse F4/80 (all from BD Biosciences), and APC Anti-Mouse CD206 (BioLegend).

C57BL/6 and DIO mice underwent CLP. On day 1, 3, 5, 7, and 14 post-CLP, 1 ml PBS (Thomas Fisher Scientific) was injected i.p., the abdomen manually massaged and then opened, and all of the peritoneal fluid collected (n = 3 mice per experimental group at each time point were analyzed). BD Cell Viability Kit (BD Biosciences) was used to determine the live bacteria in the abdominal fluid sample. Analysis was completed using BD LSR II flow cytometer and FlowJo Software. First, bacteria were gated from cells, and then live bacteria were gated from dead bacteria. The live bacteria concentration was subsequently determined.

C57BL/6 and DIO mice underwent CLP or sham procedure ^+/− recombinant GM-CSF administration. On day 1, 3, 5, 7, and 14 post-CLP, blood, bone marrow from one femur per mouse, spleen, peritoneal fluid, and abdominal adipose tissue were harvested (n = 3 mice per experimental group at each time point were analyzed). Additionally, n = 15 naive C57BL/6 and DIO mice were analyzed. Single-cell suspensions were created using a 70-μm pore size cell strainer (Falcon). Erythrocytes were lysed with a standard ammonium chloride solution. Neutrophils and monocytes were isolated using Histopaque-1119 and Histopaque-1077 and labeled for surface markers anti-Ly6G and anti-CD11b. Cell pellets were suspended in fixed volumes. Analysis was completed using a BD LSR II flow cytometer and FlowJo Software.

C57BL/6 and DIO mice underwent CLP or sham procedure ^+/− recombinant GM-CSF administration. Using a cardiac puncture collection method, blood from each mouse was collected on day 1, 3, 5, 7, and 14 post-CLP (n = 3 mice per experimental group at each time point were analyzed). Additionally, n = 15 naive C57BL/6 and DIO mice were analyzed. Erythrocytes were lysed with a standard ammonium chloride solution, and then the blood cells were labeled for surface markers anti-Ly6G and anti-CD11b. Phagocytic ability was evaluated using pHrodo Green Escherichia coli BioParticles Conjugate for flow cytometry (Thermo Fisher Scientific) using a BD LSR II flow cytometer and FlowJo software.

C57BL/6 and DIO mice underwent CLP or sham procedure. Bone marrow from both femurs from each mouse were collected on day 1, 3, 5, 7, and 14 post-CLP (n = 3 mice per experimental group at each time point were analyzed). Additionally, n = 15 naive C57BL/6 and DIO mice were analyzed. Neutrophils and monocytes were isolated using Histopaque-1119 and Histopaque-1077 and labeled for surface markers anti-Ly6G and anti-CD11b. ROS production was determined using dihydrorhodamine 123 (Invitrogen). Twenty microliters (5 μg/ml) of working dihydrorhodamine solution was added to cell suspensions in 180 μl PBS and incubated at 37°C for 5 min. Cell suspensions were stimulated with PMA at 37°C and rhodamine fluorescence. Analysis was completed using a BD LSR II flow cytometer and FlowJo Software. First, live cells were gated from dead cells, then neutrophil and monocytes were isolated, and finally, mean fluorescence intensity (MFI) was determined.

Bone marrow was harvested from one femur per mouse. Neutrophils (CD11b^high, Ly6G^high) were enriched by negative selection bone marrow using Miltenyi Biotec MicroBeads and a QuadroMACS magnet. Nonneutrophils are incubated with a mixture of biotinylated Abs and Anti-Biotin MicroBeads for depletion. Neutrophil RNA was extracted using TRIzol reagent (Invitrogen). The RNA was reverse-transcribed using QuantiTect Reverse Transcription Kit (QIAGEN). Real-time PCR was performed with an Applied Biosystems 7500 Real-Time PCR System continuous fluorescence detector. Primers were designed using Primer-BLAST (40). A species-specific multiplex real-time PCR array (RT² Profiler PCR Arrays; QIAGEN) was used to detect the expression patterns of 84 genes relevant to phagocytosis signaling in naive C57BL/6 and DIO mice (n = 3 mice per experimental group) as well as in C57BL/6 and DIO on day 3 post-CLP (n = 3 mice per experimental group were analyzed). This identified multiple gene differences between C57BL/6 and DIO mice. Two specific genes identified were then validated by quantitative PCR (qPCR).

At 6, 12, and 18 h and 1, 3, 5, 7, and 14 d after sham procedure or CLP, whole blood was harvested by cardiac puncture collection method and immediately centrifuged at 1500 × g for 15 min. The plasma supernatant was isolated and stored at −80°C until the time of analysis. The plasma samples were analyzed for cytokines using Luminex technology with Mouse Cytokine Magnetic 20-Plex Panel (Life Technologies).

GraphPad Prism software was used for statistical analyses. Data were evaluated for normality distribution. Data not normally distributed were analyzed by nonparametric testing. Differences among groups in flow cytometric analyses were evaluated by ANOVA for multiple groups and by the Student t test for two groups. Post hoc comparisons were performed using Tukey tests. Statistical analysis for survival was performed using the Mantel–Cox test and Gehan–Breslow–Wilcoxon test. In all cases, significance was designated at the 95% confidence level using a two-tailed test.

To replicate humans with obesity and T2D, we used 28–32-wk-old male DIO mice. The DIO mouse model is characterized by marked obesity, hyperglycemia, hyperinsulinemia, insulin resistance, and glucose intolerance (41). This represents both the genetic and environmental risk factors for the development of obesity and T2D in humans. When critically ill, these patients experience intensified sepsis mortality (10–12). Because multiple studies have shown that obesity and T2D increase sepsis mortality (42, 43), we sought to determine if there was a survival difference between DIO and C57BL/6J (lean) mice after an episode of polymicrobial sepsis. We used our established CLP model (44) and looked for a survival difference between these two cohorts undergoing sublethal CLP (LD_10–20) or sham procedure. Following CLP, death occurs predominantly within the first 3 d (44). We found a significant survival difference between lean and DIO mice after CLP sepsis; lean mice had a 20% mortality rate compared with a 60% morality rate in the DIO mice (Fig. 1A).

The poor DIO survival was associated with significantly more weight loss after CLP (Fig. 1B). Following sepsis initiation, both lean and DIO mice have rapid weight loss. However, by day 7, the lean mice, which reach a maximum of 17% body weight loss, had already begun to regain weight. By day 28, they were almost back to their baseline weight. Conversely, the DIO mice lost significantly more weight over a longer period of time. They lost a maximum of 29% of their body weight by day 18, at which point they started to slowly regain weight. However, by day 28, they still had an 18% weight loss. Additionally, we observed substantial weight loss in DIO mice who had undergone a sham procedure. Similar to the CLP experimental cohorts, the DIO sham mice also experienced a rapid weight loss after laparotomy, losing 16% of their body weight by day 7. Then they slowly started to regain weight. At day 28, they remained at a 12% total weight loss. In sharp contrast, the lean sham mice experienced less than a 5% weight loss throughout the entire experiment. These data suggested that DIO septic mice had limited physiologic reserve. They were unable to handle even a small traumatic injury, and their subsequent recovery was much lengthier compared with that in lean mice. Given these observed findings, we sought to understand the pathophysiology behind these differences.

The innate immune system, mainly neutrophils and monocytes, is essential for successful host eradication of bacterial pathogens (45–47). When an organism is exposed to a pathogen, the bone marrow provides a large reserve of neutrophils that are released into the peripheral circulation, which then traverse to sites of inflammation (48). Previous studies have demonstrated that survival after sepsis depends on neutrophil-mediated microbe eradication (45, 46). Given the survival differences between lean and DIO mice after CLP, we sought to determine if there were differences in bacterial clearance. At serial time points from day 1 to 14 after sepsis induction, the number of live bacteria in the mouse peritoneal cavity was quantified. A dramatic difference in the total number of live bacteria in the peritoneal cavity was observed between lean and DIO CLP animals, with significant differences at days 5 and 7 after CLP (Fig. 2B). Although most deaths following sepsis happen within the first 3 d following CLP, the bacterial loads at these early time points only trended toward being significant. The most significant differences in bacterial loads between lean and DIO mice following sepsis were at later time points, days 5 and 7 following CLP, at which point all of the septic mice will survive. It was interesting to us that bacterial load did not correlate with morality. However, the DIO mice were unable to effectively contain and clear bacteria from the peritoneal cavity. This discovery led us to investigate why DIO septic mice were unable to clear bacteria as effectively as the lean mice cohort.

During septic episodes, the hematopoietic system responds to the invading pathogen by increasing production and release of innate immune cells, including neutrophils and monocytes (48). This “emergency granulopoiesis” leads to an increase in and release of cells from the bone marrow to fight infection (49). The bone marrow myeloid compartment consists of a heterogeneous population of cells that includes neutrophils, both mature and immature monocytes, macrophages, dendritic cells, and erythroid phenotypes (50). Neutrophils are an essential component of the innate immune system and are necessary for bacterial eradication and sepsis survival (45, 46). Monocytes/macrophages are also integral to innate immune function. These cells can alter their phenotype and function depending on cues from the microenvironment (47), leading to either proinflammatory macrophages (M1) or anti-inflammatory macrophages (M2) (51). An efficient M1 response is necessary to eradicate microbial invaders (47). Studies have shown that the majority cell released form the bone marrow during septic episodes consists of mature neutrophils, whereas monocytes and lymphocytes make up a minority (48, 50). Our investigations demonstrated that, at baseline, naive DIO mice had significantly more neutrophils and monocytes in the bone marrow than lean mice. However, at days 3 and 7 after sublethal CLP, DIO mice had a significant decrease in the number of monocytes in the bone marrow compared with lean mice. There was no difference in the number of neutrophils (Fig. 3A). In addition, we and others have demonstrated that the spleen can serve as a site of extramedullary hematopoiesis after acute infections (38, 52). Accordingly, we investigated if there were differences in cellular counts in the spleen. There were no differences in the number of neutrophils or monocytes in the spleen between naive DIO and lean mice. However, there was a significant decrease in the number of both splenic neutrophils and monocytes after CLP (Fig. 3B). This also correlated with a significant decrease in splenic mass in DIO mice after CLP (Fig. 3C). These findings indicate dysfunctional emergency granulopoiesis after acute infection in DIO mice. If DIO mice are unable to respond to an acute infection with ample emergency granulopoiesis, they will be unable to eradicate the infection, or promote wound healing and adequate tissue regeneration, leading to poorer outcomes.

In addition to an emergency granulopoiesis response, there must also be successful migration of the neutrophils and monocytes to the site of infection to clear the invading pathogen. We sought to evaluate if there was a difference in the migration ability of neutrophils and monocytes by looking at the number of neutrophils and monocytes in the peritoneal cavity post-CLP sepsis in lean and in DIO mice. On days 1 and 3 post-CLP, peritoneal fluid was collected, and the neutrophils and monocytes were isolated. At both days, DIO mice had significantly more neutrophils and M1 monocytes compared with lean mice (Fig. 4A). So, despite DIO mice having decreased hematopoiesis compared with lean mice and lower overall cell counts in the bone marrow and spleen after CLP, they actually had more neutrophils and M1 monocytes at the site of infection. It appears that there were no deficiencies in the ability of these cells to home to sites of infection. However, as described above, the DIO mice also had increased numbers of live bacteria in the peritoneal cavity. These findings suggest that although adequate numbers of neutrophils and monocytes were recruited to the peritoneal cavity, their function was inadequate to clear bacteria. We next chose to investigate the functional ability of neutrophils and monocytes in lean and DIO mice. Additionally, we looked at the M1 to M2 monocyte ratio (Fig. 4B, 4C). Compared to lean mice, DIO mice had an increased M1/M2 ratio at day 3. In sepsis, an ongoing debate persists as to whether inflammatory/anti-inflammatory processes or innate/adaptive immune dysfunction are more detrimental to overall survival (53). However, our data were extended by genomic studies on tissue samples from septic and severely injured trauma patients (54). This identified an enduring and simultaneous inflammatory and anti-inflammatory state, driven by a dysfunctional innate and suppressed adaptive immunity, which culminates in persistent organ injury (55), inflammation, and patient death (56, 57). The inability of DIO mice to mount an appropriate M2 response likely contributes to the worsened outcomes observed.

Once primed by bacterial products and endogenous mediators, neutrophils and monocytes contain invading organisms. To effectively eliminate the pathogen, the immune cells have to be able to both phagocytize the bacteria as well as kill it for processing after engulfment. One mechanism by which a host phagocytic cell kills a pathogen is by releasing ROS into the phagosome (58). Neutrophils and monocytes produce ROS through NADPH oxidase (59, 60).

We set out to investigate both phagocytosis and ROS generation. First, we looked to see if there are any differences in phagocytic activity in PMNs and monocytes from lean and DIO naive (nonseptic) mice. We did not observe any differences in either neutrophils or monocytes in phagocytic ability (Fig. 5B). We then looked at cells from lean and DIO mice after CLP. At time points from 1 to 14 d after sepsis induction, we isolated neutrophils and monocytes from the blood and evaluated phagocytic ability using flow cytometry. DIO neutrophils had a significant decrease in phagocytic ability at all time points after CLP compared with lean mice. DIO monocytes had a significant decrease in phagocytic ability on days 1, 3, and 14 post-CLP, whereas the other days trended toward a significant decrease (Fig. 5C).

After engulfing the pathogen, the phagocyte then needs to destroy it. Hence, we examined if there were differences in ROS generation between the lean and DIO cohorts. There was no difference in the ability to generate cytosolic ROS between lean and DIO naive mice (Fig. 6B). However, there were differences after CLP. At serial time points from 1 to 14 d after CLP, neutrophils and monocytes were isolated from the bone marrow of both femurs, and ROS generation was assessed between 0 and 30 min. DIO neutrophils had significantly decreased ROS generation on days 3, 7, and 14 post-CLP, whereas DIO monocytes have decreased generation at all time points after sepsis (Fig. 6C). There was one time point on both day 1 and 7 post-CLP in which DIO monocytes showed greater generation of ROS compared with lean mice; however, these observations were at single time points and did not persist over time. We conclude that obesity and diabetes compromised phagocytosis and ROS generation in both neutrophils and monocytes from DIO septic mice as compared with cells from septic lean mice. These findings illustrate broad functional defects in DIO neutrophils and monocytes, which directly account for the decreased, defective peritoneal bacterial clearance leading to decreased survival.

To define mechanisms underlying the neutrophil functional deficiencies observed in septic DIO mice, we incorporated multiplex genomic technology. We used a species-specific multiplex real-time PCR array to detect the expression patterns of 84 genes relevant to phagocytosis signaling. We were able to identify multiple genomic differences between neutrophils in lean and DIO naive mice (data not shown). The most prominent of these gene changes were Axl and Mertk, which were significantly underexpressed in naive nonseptic DIO mice compared with lean mice (data not shown). Axl and Mertk belong to the TAM family of receptor tyrosine kinases, which have pivotal roles in innate immunity (61). TAM receptor kinases inhibit inflammation in dendritic cells and macrophages, promote phagocytosis of apoptotic cells and membranous organelles, and stimulate the maturation of NK cells (61). TAM receptor signaling is also required to protect immune cells from apoptosis and is important in increasing cell migration, aggregation, and growth (62). In TAM triple knockout mice, the ability to phagocytose apoptotic cells by APCs is lost (61). Additionally, mice with Mer inhibited show an inability to clear apoptotic cells (63). Given that both Axl and Mertk transcripts were deceased in naive nonseptic DIO mice neutrophils using a multiplex array, we then validated these results by qPCR. We identified decreases in Axl transcripts in naive DIO mice and Mertk transcripts in DIO mice 3 d post-CLP (Fig. 7). Downregulation of these two genes may contribute to the decreased functional ability that we observe in the DIO neutrophils.

In addition to optimal cellular function, effective bacterial clearance requires appropriate cytokine production to enhance host immune responses and mitigate tissue damage. Using a multiplex approach, we compared 23 blood cytokine levels involved in immune cell signaling and inflammation prior to and at serial time points during sepsis. Although there were differences in production across many cytokines tested, we found several cytokine differences that may explain the inability of DIO mice to eradicate bacteria (Fig. 8). Although obesity and T2D are thought to produce an inflammatory state, the lean mice produced significantly more TNF-α and MIP-1α (CCL3) compared with the DIO animals. The DIO mice produced significantly more IL-6 and IFN-γ compared with lean controls. Both lean and DIO mice produced very little GM-CSF early after sepsis. In addition to the lack of inflammatory cytokine production, DIO mice also produced significantly less IL-10, a compensatory anti-inflammatory cytokine. These data suggest why DIO neutrophils and monocytes may perform bacterial eradication in a suboptimal manner after sepsis. The lack of TNF-α and MIP-1α production in DIO mice may negatively impact neutrophil and monocyte recruitment and activation. The elevated DIO IFN-γ production can downregulate IFN-responsive genes such as Mertk. Decreased transcription of these genes may result in suboptimal phagocytosis and subsequent bacterial eradication, processing, and Ag presentation, leading to an inadequate innate and adaptive immune response. Lastly, with decreased levels of IL-10, the DIO mice demonstrated an imbalance between proinflammatory and anti-inflammatory pathways, indicating that inflammation resolution in DIO mice versus lean mice with sepsis is suboptimal for effective immune responses and monocyte and neutrophil cellular function.

Given the observed findings described above, we wanted to investigate whether GM-CSF would improve outcomes in the DIO mice cohorts. GM-CSF is a product of cells activated during inflammatory conditions (64). It enhances neutrophil and monocyte production and function and has shown therapeutic promise to improve outcomes in human lung injury trials (37). Therefore, we sought to determine if we could improve DIO mouse survival by administering exogenous GM-CSF after sepsis induction. One hundred to two hundred nanograms of GM-CSF were given s.c. twice daily for 7 d starting 6 h after CLP. Limited studies exist on GM-CSF in murine sepsis models. We based our dosing on a study focusing on GM-CSF therapy in mice with burns (65). This study observed that a lower dose, 10 ng, given twice a day in 8–10-wk-old mice resulted in significantly improved survival, whereas higher doses of GM-CSF did not show the same results. As we used 28–32-wk-old mice, which weigh at least double that of 8–10-wk-old mice, we used a higher dose in our experiments to account for this weight difference. We saw significantly improved survival in the DIO mice treated with GM-CSF compared with DIO mice treated with placebo. We did not observe a similar survival benefit in the lean mice treated with GM-CSF (Fig. 9A). Similar to the burn murine study, we did not see the same improvement when a higher dose of GM-CSF is used (data not shown), suggesting that the dose of GM-CSF was critical for improved outcomes in mice. Despite a survival improvement, GM-CSF did not reduce the weight loss observed in the lean or DIO CLP cohorts (Fig. 9B). We then looked at peritoneal bacterial counts 1 d after CLP in DIO mice treated with GM-CSF versus those treated with placebo. We did not observe any difference in bacterial peritoneal counts after GM-CSF administration in DIO mice (Fig. 10A). We then looked at neutrophil and monocyte cell counts after GM-CSF administration. We did not find any differences in the number of neutrophils or monocytes in the bone marrow or spleen after GM-CSF administration 3 d after CLP sepsis (data not shown). Finally, we evaluated neutrophil and monocyte cellular function. Monocytes from DIO mice treated with GM-CSF had no difference in phagocytic ability 1 d after CLP sepsis initiation; however, we found that DIO mice treated with GM-CSF had significantly increased phagocytic ability in monocytes 3 d after CLP sepsis initiation (Fig. 10B). There were no differences in neutrophil phagocytic ability (data not shown). We also found that ROS generation of both neutrophils and monocytes in the blood were improved both 1 and 3 d after CLP sepsis in the DIO mice treated with GM-CSF (Fig. 10C, 10D). These results suggest that GM-CSF may be a useful therapeutic agent to improve both neutrophil and monocyte cellular function, leading to improved sepsis survival in obese, diabetic mice.

Individuals with obesity and T2D experience far greater rates of infection, organ failure (11, 12), and mortality from sepsis (13–21). These worsened outcomes are crucial to explore given the rising prevalence of obesity and T2D worldwide. Cell dysfunction in the innate immune system has been incriminated as one of the leading factors underlying sepsis progression and complication generation (23–25). One of the hallmarks of obesity includes increased accumulation of leukocytes, especially M1 macrophages, within adipose tissue (26, 27), which contributes to a chronic inflammatory state that has been shown to be deleterious in patients with trauma and/or sepsis (28). Neutrophils in individuals with T2D show defects in almost all functions, including migration to inflammatory sites, release of lytic proteases, phagocytosis, production of ROS, and apoptosis (29).

Survival from sepsis requires the mobilization of functional neutrophils to sites of infection (32). Successful neutrophil infiltration promotes effective monocyte influx that orchestrates and optimizes innate and adaptive immune interplay. Although a few researchers have shown that obese, diabetic mice have defective neutrophil function and poor survival after sepsis (30, 31), there are few studies that focus on neutrophil dysregulation and the effects on innate immune responses.

Our results reveal defects in the release of neutrophils and monocytes in obese, diabetic mice during sepsis. The hematopoietic system is responsible for increasing production and releasing innate immune cells, including neutrophils and monocytes, during an acute infection (48). One aspect observed during chronic inflammatory states, such as obesity, diabetes, and infection (27), is the appearance of extramedullary hematopoiesis, in which bone marrow–derived cells home to, proliferate, and mature in extramedullary organs, such as the spleen and liver (66). Previous studies have shown that extramedullary hematopoiesis can help to overcome the lack of myeloid cellularity during these chronic inflammatory states (38, 52, 66). We show that DIO mice have both decreased intra- and extramedullary hematopoiesis during sepsis. Not only is the bone marrow unable to meet the increased need for more innate immune cells in the circulation, but the spleen, a site of extramedullary hematopoiesis, is unable to increase myeloid production of innate cells to a level that will overcome the myeloid cellular deficit that immediately follows a septic insult. This deficiency does not exist while in the naive state, as DIO mice had increased myeloid cells in the bone marrow compared with lean mice. However, once DIO mice are faced with an infectious insult, the proliferation deficiency becomes apparent as a failure to release adequate numbers into the blood. Our results build off of information known from prior studies. For instance, Kanakadurga, Singer, and colleagues (67) demonstrated that diet-induced obesity promoted dysfunctional myelopoiesis in hematopoietic stem cells. In addition, Gallagher and coauthors used DIO mice to show that repressive histone methylation was decreased at the IL-12 promoter gene in bone marrow progenitors and was passed down to macrophages appearing in wound sites. These macrophages were primed in peripheral tissues and subsequently negatively impacted diabetic wound repair (68).

Additionally, we observed that DIO mice have decreased bacterial clearance ability compared with cells from lean mice. After proliferating in the bone marrow or extramedullary hematopoietic site, the immune cells must then traffic to the site of infection to fight the invading pathogen. In our sepsis model, this was the peritoneal cavity. Because our results illustrated an overall reduction in the number of neutrophils and monocytes in the bone marrow and spleen, we initially hypothesized that diminished production was the underlying mechanism for this decreased bacterial clearance. However, our results showed that decreased production was only part of the overall dysfunction found in obese, diabetic mice. We looked at the overall number of neutrophils and monocytes in the peritoneal cavity after sepsis induction and unexpectedly found an overall increase in DIO mice compared with their lean cohorts. Despite a decrease in the overall number of cells produced in the bone marrow, DIO mice actually had more neutrophils and monocytes in the peritoneal cavity. We hypothesized that this was secondary to decreased functional capacity of these cells. If the recruited cells were not functioning properly, more cells would continue to be recruited to the site of inflammation. An alternative hypothesis was that this increased amount of M1 macrophages was secondary to local migration of these cells into the peritoneal cavity from the abdominal fat in the DIO mice in which they had become resident cells.

Monocytes/macrophages accumulate in intra-abdominal adipose tissue in obese individuals and formed crown-like structures around dead adipocytes (69). Traditionally, adipose tissue macrophage accumulation was thought to be a consequence of peripheral monocyte migration under inflammatory conditions; however, there has been increasing evidence that local proliferation of macrophages accounts for this buildup (69). One study revealed that resident macrophage proliferation contributed to adipose tissue accumulation during the early stages of obesity, whereas migrated monocytes contributed to adipose tissue macrophage accumulation during the late stages of obesity (69). It is possible that some of these locally proliferated monocytes were able to migrate out of the abdominal adipose tissue to inflammatory sites after exposure to the appropriate extracellular milieu. However, the question remains why the peritoneal M1 phenotypic monocytes were unable to participate in infection eradication. During inflammatory states, M1 monocytes typically coordinate the immune system processes for infection eradication. With so many M1 monocytes recruited to the peritoneal of the DIO mouse, it was surprising to find significantly elevated bacterial counts. Moreover, the increased M1:M2 macrophage ratio in the DIO peritoneal cavity did not provide the necessary anti-inflammatory response needed to resolve inflammation.

Given the decreased production of leukocytes in the bone marrow of septic mice but increased cellular counts in the peritoneal cavity, our results revealed defects in the functional ability of both neutrophils and monocytes in obese, diabetic mice during sepsis. Decreased functional ability resulted in decreased bacterial clearance. Once again, we did not find a difference between naive lean and DIO mice; however, after sepsis, we found a widespread decrease in the functional ability of these innate immune cells. Using multiplex genomic technology, we than identified multiple genomic differences between neutrophils in lean and DIO naive and CLP-treated mice, including the downregulation of the genes Axl and Mertk. Axl and Mertk belong to the TAM family (Tyro-3, Axl, and Mertk) of receptor tyrosine kinases (61), which promotes phagocytosis of apoptotic cells and enhances numbers of membranous organelles (61). Such responses protect immune cells from apoptosis (62). Additionally, these genes seemed to be important in increasing immune cell migration, aggregation, and cell growth (62). Downregulation of these genes may have contributed to decreased functional ability of recruited neutrophils.

Finally, we discovered that GM-CSF administration may serve as a therapeutic option in sepsis, reducing overall sepsis mortality in obese, diabetic mice. We found that GM-CSF improved the phagocytic functional ability of monocytes and ROS generation in neutrophils and monocytes after sepsis in DIO mice. GM-CSF is known to stimulate stem cells to produce neutrophils, monocytes, and macrophages (70). Recently, others have shown that in the immune suppressive phase of sepsis, ventilator-dependent patients treated with recombinant GM-CSF had fewer ventilator and ICU days (37, 71). Recombinant GM-CSF therapy in immune-suppressed pediatric patients with sepsis restored the TNF production in lymphocytes and reduced nosocomial infections (72). Furthermore, a metaanalysis of over 12 clinical studies involving either G-CSF or GM-CSF demonstrated that either therapy reduced the infection rate (73). Considering that 70–80% of patients who die of sepsis harbor bacteria or fail to resolve infections (54), GM-CSF alone or in combination with other immune modulatory agents may prove useful for infection eradication during sepsis and improved infection-free long-term survival after sepsis. This has the potential to have wide-ranging applicability, given the increasing population of individuals with obesity and diabetes. Intriguingly, the survival advantage after GM-CSF administration was not present in lean mice. The lean mice actually had worse survival after GM-CSF administration. We believe that this is because the functional defects present in the DIO mice after sepsis are not present in the lean mice. The lean mice have functioning monocytes and neutrophils; therefore, GM-CSF does not have the same therapeutic benefit in the lean mice. For GM-CSF administration to offer an advantage after sepsis, it seems as though functional defects in neutrophils and monocytes are a prerequisite. This is a vital concept to address because individualized precision medicine therapies seem to work best in well-defined patient populations. We need to understand the underlying innate immune cell functional defects associated with comorbidities, such as obesity and diabetes, so that we can provide directed therapeutic approaches that ameliorate the specific defects. Our DIO mice have functional defects in monocytes and, to a lesser degree, neutrophils. GM-CSF has the ability to improve these functional defects after just one dose. Therefore, using GM-CSF in individuals with these defects after sepsis has the ability to improve outcomes, but if these defects are not present, such as in lean mice, no improvements are observed in clinical outcomes. Nonetheless, much more research needs to be done to understand the specific monocyte and neutrophil functional defects induced by diabetes and obesity before expanding studies to clinical trials in septic humans.

Our study does have several limitations. First, all of our experiments are completed in male mice only. We acknowledge that all studies should be done in both sexes; however, currently, DIO mice from The Jackson Laboratory are only readily available as males. As all our experiments are completed on mice at least 28 wk of age, it is imperative that we are able to obtain these DIO mice at an older age to coincide with humans who have developed obesity and diabetes over time. To complete all of our experiments in female mice, we would need to obtain these mice at 6 wk of age and feed them a high-fat diet for at least 22 more weeks before they were acceptable for experimental study. Given the sheer number of mice that we use in our time series studies, this significant time delay would inhibit our ability to complete experiments in a timely manner. Additionally, male mice are more affected by diabetes than are female mice; therefore, male mice tend to be used much more frequently in diet-induced obesity models given the reproducibility of results (74).

Next, we use mice that are 28–32 wk of age to coincide with human patients who are in their early 40s. Although our mice still do not precisely represent human septic ICU populations, our 28–32-wk-old mice represent a more realistic human age correlation than do young mice <12 wk and old mice over 16 mo of age. Choosing mice around 30 wk of age was intentional because multiple studies have been published that show that both elderly mice and humans have increased mortality following sepsis and trauma (75–77). As we wanted to isolate the innate immune defects that were secondary to obesity and diabetes, we did not want our results to be affected by innate immune changes secondary to advanced age. We hypothesize that if elderly mice were used, survival would be much worse in the DIO mice just as the published literature shows in aged mice.

Lastly, as most deaths following CLP occur within the first 3 d following sepsis initiation, at all time points after 3 d, there is an inherent survival bias to all of the studies because we are effectively comparing mice that are going to survive the septic insult. This is an inherent bias of our study that is challenging if not impossible to eliminate. However, to increase the “survivor pool” from which we would analyze, we used both our 20-gauge needle CLP as well as a less severe CLP model to capture more surviving mice. Nevertheless, this method still has a survivor bias that is difficult to eliminate. However, we did our best to analyze a wide range of survivors in an attempt to identify and eliminate outliers. To do this, we would start a separate cohort for each specific time point that we analyzed and also started with a large enough cohort that more than three mice were surviving at the desired time point so that we could randomly select three mice to analyze.

In conclusion, we provide evidence that there are defects in both production and function of neutrophils and monocytes after infectious sepsis in obese, diabetic mice. These defects resulted in bacterial persistence and worsened sepsis mortality. We have found downregulation of the Axl and Mertk genes as possible mechanisms behind these observed defects. Finally, treatment with GM-CSF may be a potential therapeutic option to improve DIO sepsis survival through improvement in the functional ability of innate immune cells.

The authors have no financial conflicts of interest.