Increased Number of Islet-Associated Macrophages in Type 2 Diabetes

Specific human sample information is available in Table 1. Patients with pancreatitis, lymphoma, and systemic infection and who were on immunosuppressive therapy were excluded from analysis. Pancreata were procured for histology and islet isolation according to regulations and good practice rules applied at that time in Switzerland. Briefly, consent was considered obtained if the potential donor carried an official organ Swisstransplant (Swiss national organ sharing agency) donor card, on which individual reservations about procurement of specific organs or tissues are explicitly mentioned. For brain-dead potential donors not carrying an organ donor card, consent was obtained orally from the closest relatives and specifically mentioning the use of the pancreas for islet isolation or histology. Use of pancreatic tissue was approved by the cantonal ethical committee, number StV 29-2006.

Human tissue samples were fixed in formalin, 4-μm sections were cut, and immunohistochemistry was performed on an automated stainer (Ventana Benchmark; Ventana, Tucson, AZ) after protease 1 (Ventana) pretreatment. Sections were incubated with an anti-CD68, anti-CD163, or anti–HLA-2 antibody (mouse anti-human CD68, clone PG-M1, 1:50, Dako, Glostrup, Denmark; mouse anti-human CD163, clone 163C01/10D6, 1:100, NeoMarkers/Lab Vision, Newmarket Suffolk, U.K.; and mouse anti-human HLA class 2 [DP+DQ+DR], clone IQU9, 1:50, Novocastra Laboratories, Newcastle, U.K.), followed by a biotinylated secondary antibody (Ventana). Staining was visualized with the I-view DAB detection kit (Ventana). Sections were costained with an anti-insulin antibody (polyclonal guinea pig anti-insulin, 1:500, Dako), followed by a prediluted secondary antibody and chromogenically detected via the Ventana alkaline phosphatase Fast Red Kit. Counterstaining was done with hematoxylin. Islet-associated granulocytes were identified morphologically (hematoxylin and eosin staining) and using an anti-human myeloperoxidase antibody (rabbit polyclonal, 1:15,000; Dako). The CD68 antibody was used as previously shown (14) and controlled by mouse IgG antibody staining. For transferase-mediated dUTP nick-end labeling (TUNEL) detection, sections were permeabilized with proteinase K (20 μg/ml) and endogenous peroxidases blocked with 3% H₂O₂. Sections were incubated with working-strength TdT enzyme (ApopTag kit S7100; Millipore, Zug, Switzerland). After rinsing sections with a stop buffer, the slides were covered by anti–digoxigenin peroxidase conjugate, rinsed, and then incubated with rabbit anti-sheep horseradish peroxidase antibody (1:80). Detection was performed with DAB (Ventana), and hematoxylin was used for counterstaining.

Mouse pancreatic cryosections were incubated with an anti-Cd11b primary antibody (BD Pharmingen, Basel, Switzerland; 1:167), isotype rat IgG2B (Serotec, Düsseldorf, Germany), anti-insulin antibody (Dako), and Dapi to identify nuclei. We have described this Cd11b antibody previously (15); mouse spleen served as a positive tissue control. Primary antibodies were visualized using Strep-Cy3 secondary and fluorescein isothiocyanate (FITC) secondary antibodies (Jackson Immunoresearch, Newmarket, U.K.) and images captured with an Axioplan2 imaging system (Zeiss, Feldbach, Switzerland). Additionally, some sections were visualized with 3-amino-9-ethylcarbazole (AEC) substrate and counterstained with hematoxylin.

Wistar and GK rat cryosections were incubated with mouse anti–rat-IA (major histocompatibility complex [MHC] II; Serotec; 1:300) and ED1 mouse anti-rat CD68 (Serotec; 1:100), followed by incubation with goat anti-mouse secondary (Caltag, Cergy, France) and visualized with AEC substrate. For each series of pancreas sections, one slide was stained only with the second antibody as a control for endogenous peroxidase activity and nonspecific antibody binding, as described previously (16).

Pancreatic resection samples (three control subjects and four type 2 diabetic patients) and sorted human non–β-cells plated on extracellular matrix (ECM) were analyzed for IL-8 expression. Human glioblastoma sections were used as a positive control for IL-8 staining as described previously (17). Sections and sorted non–β-cells were incubated with a rabbit anti–IL-8 primary antibody (ab16223; Abcam, Cambridge, U.K.; 1:50) or isotype control (rabbit IgG; R&D Systems, Abingdon, U.K.). Antibody specificity was tested using recombinant IL-8 protein (Abcam ab6931) to block binding. IL-8 was visualized using AEC, Cy-3 anti-rabbit, or Alexa Fluor 488 donkey anti-rabbit IgG secondary antibodies (Molecular Probes, Eugene, OR). Sections were further incubated with guinea pig anti-insulin or guinea pig anti-glucagon (Dako; 1:50), followed by FITC secondary antibodies. Non–β-cells were incubated with the above glucagon antibody and rhodamine-conjugated goat anti-guinea pig (Jackson, Suffolk, U.K.) secondary antibody, and the nuclei were labeled with Hoechst 33342 (Sigma, Buchs, Switzerland). Further, the same IL-8 antibody was used in Western blotting of human islet samples. Recombinant IL-8 (R&D Systems) was used as a positive control for Western blotting.

An average of 43 ± 17 islets from nondiabetic (n = 7) and 35 ± 12 islets from diabetic (n = 9) pancreatic sections were blindly scored for CD68-positive cells around the periphery and/or within islets by two investigators (A.P. and X.G.). CD163 and HLA-2 was used to confirm macrophage identity in resection samples. To evaluate TUNEL-positive cells localized to CD68-positive infiltrated islets, 190 islets from three diabetic patients (resection samples) showing strong infiltration were evaluated in serial sections stained for TUNEL and CD68. Postmortem interval time for autopsy samples ranged from 7 to 24 h, and archive time for all samples ranged from 24 to 99 months. Pancreatic sections from the corpus and tail of the pancreas were examined.

Islet-associated Cd11b-positive cells were scored by a single investigator (J.A.E.) blinded to the conditions. Only Cd11b-positive cells around the periphery of pancreatic islets or within islets were scored. For each animal in the study (n = 3–7), four to eight pancreatic sections cut at 60- to 80-μm intervals were scored. Data are expressed as Cd11b-positive cells/islet, where “islet” refers to a cross-sectionally detected islet defined by size (small islet: 1–5 cross-sectional cells; medium islet: 5–20 cross-sectional cells; and large islet: 20–50 cross-sectional cells). A total of 100–200 islets per treatment group were scored. Islet area was measured by assessing the area of insulin-immunopositive cells, traced manually, and computed using analySIS 3.1 software (Soft Imaging System, Münster, Germany). TUNEL-positive cells were analyzed in 20.6 ± 0.4 islets and 25.0 ± 5.2 islets/animal in 8- and 16-week standard diet–and high-fat–fed animals (8 weeks, n = 5 and 16 weeks, n = 6; In Situ Cell Detection Kit, AP, Roche, Basel, Switzerland). Islet-associated CD68-positive and MHC II–positive cells in Wistar and GK rats were scored in islets from six to nine different animals.

Male C57BL/6J mice (Harlan, Horst, Netherlands) were used for all mouse islet experiments. In some cases, animals were fed a hypercaloric (high-fat) diet (Research Diets, New Brunswick, NJ). The high-fat diet contained 58, 26, and 16% calories from fat, carbohydrate, and protein, respectively, and a total of 5.6 kcal/g, whereas the standard diet (Provimi Kliba, Kaiseraugst, Switzerland) contained 29, 39, and 32% calories from fat, carbohydrate, and protein, respectively, and a total of 2.8 kcal/g. For assessment of Cd11b-positive cells around islets, animals were started on a high-fat diet at age 3–4 weeks. For ex vivo determination of islet cytokines and chemokines, animals were started on a high-fat diet at age 8 weeks. For glucose tolerance testing, mice were injected intraperitoneally with 2 mg/g body wt glucose (intraperitoneal glucose tolerance test) and blood glucose concentration measured with a Freestyle glucometer (Abbott, Baar, Switzerland).

Characteristics of the GK rat maintained in the colony at the Paris 7 University have been described previously (18). This animal model was developed by inbreeding Wistar rats with mild hyperglycemia. Male GK rats are normoglycemic before weaning (1 month), with hyperglycemia, hypercholesterolemia, and hypertriglyceridemia developing shortly after weaning, followed by insulin resistance at 2 months. Male db/db and db/+ littermate controls were purchased from The Jackson Laboratory (Bar Harbor, ME). Guidelines for the use and care of laboratory animals at the University of Zurich were followed.

Human islets were isolated from pancreata of 14 organ donors at the University of Geneva Medical Center and the University of Illinois at Chicago. All human islet preparations were stained with dithizone (>80% purity of islets) and insulin (40–50% β-cells/islet) to monitor purity and were handpicked and plated by a single investigator (J.A.E.) to maintain consistency. Human non–β-cells were isolated using a method adapted from Gmyr et al. (19) and Ichii et al. (20) (G. Parnaud, unpublished observations). Mouse islets and nonendocrine pancreatic tissue were isolated from C57BL/6J mice by collagenase digestion and handpicking of islets. After collagenase digestion, nonendocrine tissue was recovered as a pellet and islets removed by handpicking as reported (21). Human islets were cultured in CMRL-1066 medium containing 5.5 mmol/l glucose, 100 units/ml penicillin, 100 μg/ml streptomycin, and 10% FCS (Invitrogen, Basel, Switzerland). Mouse islets and nonendocrine tissue were cultured in RPMI-1640 medium containing 11 mmol/l glucose, 100 units/ml penicillin, 100 μg/ml streptomycin, 40 μg/ml gentamicin, and 10% FCS (hereafter referred to as islet media). Islets were cultured on ECM-coated plates (at 20 islets/plate) derived from bovine corneal endothelial cells (Novamed, Jerusalem, Israel) as previously described (22). In experiments using ECM dishes, islets and nonendocrine tissue were left in islet media for 48 h to adhere and spread before initiation of experiments. INS-1 cells were kindly donated by Dr. S.A. Hinke (Brussels Free University VUB, Brussels, Belgium) and MIN-6 cells by Dr. P. Halban (University of Geneva Medical Center, Geneva, Switzerland) and cells cultured as previously described (23). MIN-6 and INS-1 cells were seeded at 5 × 10⁵ cells/well, and control conditions included 25 mmol/l and 11 mmol/l glucose media, respectively.

In some experiments, islets were treated with 33 mmol/l glucose and/or 0.5 mmol/l palmitate (Sigma). Palmitic acid was dissolved at 10 mmol/l in RPMI-1640 medium containing 11% fatty acid–free BSA (Sigma) under an N₂ atmosphere, shaken overnight at 55°C, sonicated for 15 min, and filtrated under sterile conditions. For control incubations, 11% BSA was prepared as described above. Before use, the effective free fatty acid concentrations were controlled with a commercially available kit (Wako, Neuss, Germany). In some experiments, 500 nmol/l staurosporine or 0.1 and 1 mmol/l streptozotocin were added to mouse islets for 48 h to induce cell death. Cell death was confirmed by TUNEL (Roche).

For acute insulin release in response to glucose, islets were washed and incubated in Krebs-Ringer buffer containing 2.8 or 16.7 mmol/l glucose and 0.5% BSA for 1 h. Islet insulin was extracted with 0.18 mol/l HCl in 70% ethanol for determination of insulin content. Secreted insulin and insulin content was assayed by radioimmunoassay (CIS Biointernational, Gif-sur-Yvette, France).

Conditioned media and serum cytokines and chemokines were assayed using human, mouse, and rat Luminex kits. In some islet experiments, cytokine/chemokine release was normalized to total islet protein, extracted using lysis buffer, and measured using a bicinchoninic acid assay (Pierce, Rockford, IL).

Total mouse islet RNA was extracted as described (22) and reverse transcribed using random hexamers. Commercially available mouse primers to 18S rRNA, IL-6, chemokine KC, G-CSF, and MIP-1α were purchased and assayed according to the manufacturer's protocol using the ABI 7000 system (Applied Biosystems, Foster City, CA). Changes in mRNA expression were calculated using difference of Ct (cycle threshold) values.

To evaluate monocyte and neutrophil migration, peripheral blood mononuclear cells and granulocytes were isolated from a single healthy male donor using Histopaque per the manufacturer's protocol (Sigma). Migration was tested using Transwell membranes by loading a mix of 1 × 10⁶ peripheral blood mononuclear cells and 5 × 10⁵ granulocytes into the upper chamber and human islet medium or human islet supernatant into the lower chamber. Experiments were carried out in X-Vivo 15 medium (Cambrex, Verviers, Belgium), in which islet culture medium and islet supernatants were diluted 10 times. Human islet supernatants treated without (untreated) and with 33 mmol/l glucose and 0.5 mmol/l palmitate (treated) for 48 h were used. Migration was allowed to proceed for 4 h at 37°C before evaluation of total cells migrated by flow cytometry (FACScan; BD Biosciences). Identification of migrated monocytes and neutrophils was achieved using FITC-conjugated anti-CD14 and -CD15 monoclonal antibodies, respectively (BD Biosciences). Appropriate isotype controls were used to ensure antibody specificity. IL-8 was neutralized by addition of an IL-8 antibody or normal goat isotype control (30 μg/ml; R&D Systems) to the lower chamber and preincubation for 30 min with conditioned media before the addition of cells to the upper chamber. In control experiments, the IL-8 antibody was found to block recombinant IL-8–induced migration.

Islets from four separate human islet isolations were fixed by immersion in a fixation solution containing 2.5% paraformaldehyde, 0.1% glutaraldehyde, and 0.01% picric acid for 4 h. Thereafter, specimens were dehydrated and embedded routinely in LR White (Polysciences, Warrington, PA). Ultrathin sections were cut at 90 nm and transferred onto nickel grids (mesh size 100). Sections were incubated with a mouse glucagon antiserum (G-2654; Sigma, St. Louis, MO; 1:100), followed by biotinylated anti-mouse IgG (Amersham International, Dübendorf, Switzerland) and a streptavidin gold 5-nm complex (Amersham). IL-8 was visualized using rabbit IL-8 antiserum (Abcam; 1:50) followed by biotinylated goat anti-rabbit IgG (Bioscience, Emmenbrücke, Switzerland) and a streptavidin gold 15-nm complex (Amersham). Sections were examined with a Philips CM 100 electron microscope and digitally analyzed with a Gatan Bioscan Digital Micrograph (Gatan, Pleasanton, CA).

Data are expressed as means ± SE, with the number of individual experiments presented in the figure legends. All data were tested for normality and analyzed using the nonlinear regression analysis program PRISM (GraphPad, San Diego, CA). Significance was tested using the Student's t test and ANOVA with Bonferonni's or Dunnett's post hoc test for multiple comparison analysis. Significance was set at P < 0.05.

We investigated whether type 2 diabetic islets display immune cell infiltration. With respect to human samples, we observed increased numbers of islet-associated macrophages (based on CD68, CD163, and HLA-2 immunolabeling; CD163 and HLA-2 not shown) in human type 2 diabetic tissue from both autopsy and resection samples (Table 1 and Fig. 1A and B). Islets with increased numbers of macrophages (more than three CD68-positive cells per islet) were observed more frequently in type 2 diabetic samples than in nondiabetic controls (Fig. 1A) (maximum of 19 CD68-positive cells/islet in type 2 diabetic samples). In contrast to islets of control subjects with exclusive perivascular location of CD68-positive macrophages, the affected islets showed intraislet invasion (Fig. 1B). Type 2 diabetic islets, characterized by increased CD68-positive cells, did not display increased TUNEL-positive cells, and we did not observe macrophages in the vicinity of apoptotic β-cells (Fig. 1B). Numbers of TUNEL-positive cells were 0.023 ± 0.011 versus 0.046 ± 0.004 TUNEL-positive cells/islet for nondiabetic (n = 4) versus diabetic (n = 3) resection samples analyzed, respectively. Numbers of CD68-positive cells/islet did not correlate with postmortem interval of autopsy samples (r² = 0.15, P = 0.34, n = 8) or tissue archive time (r² = 0.00066, P = 0.93, n = 15). Cases with intraislet invasion were associated with decreased insulin immunoreactivity and amyloid deposits. These macrophages were positive for HLA-2 and CD163 (data not shown). We saw no differences in pancreatic islet or exocrine-associated granulocytes or CD3-positive T-cells in diabetic versus nondiabetic samples, while some CD3 T-cells were observed in the peri-islet region of all samples.

To investigate the onset of increased islet-associated immune cells, we fed C57BL/6J mice a standard or high-fat diet and evaluated islet-associated CD11b-positive cells (a marker for macrophages in addition to dendritic and other myeloid cell lineages) after 4, 8, and 16 weeks. After 8 weeks of high-fat feeding, mice displayed glucose intolerance and continued to do so until the experiment was terminated at 16 weeks (Fig. 1C). In this model, Cd11b-positive cells were observed mostly at the periphery of islets (Fig. 1D); the spleen served as a positive control for Cd11b staining (Fig. 1E). Already after 8 weeks of high-fat feeding, we detected a doubling in the number of islet-associated CD11b-positive cells exclusively around large islets in high-fat–fed animals versus standard diet controls of the same age. At 16 weeks, there was a trend toward increased CD11b-positive cells around medium-sized islets as well. Large islets with more than three CD11b-positive cells were observed more frequently in high-fat sections than in standard diet controls (10 ± 6%, n = 4 vs. 48 ± 15% of total islets with more than three CD11b-positive cells, n = 5, in 8-week standard diet–and high-fat–fed samples, respectively; P < 0.05). Islet area of those large islets evaluated at 8 weeks was not significantly different in high-fat–versus standard diet–fed animals (28,720 ± 8,930 μm², n = 4 vs. 27,440 ± 4,210 μm², n = 5). Thus, the increase in islet-associated CD11b-positive cells around large islets was not due to a difference in islet size between standard diet–and high-fat–fed animals and was detected as an early event following high-fat feeding. Analysis of islets in 8- and 16-week standard diet–fed animals and in 8-week high-fat–fed animals revealed no TUNEL-positive cells. In 16-week high-fat–fed animals, 0.031 TUNEL-positive cells per islet (2 TUNEL-positive cells/65 islets analyzed from n = 3 animals with highest number of Cd11b-positive cells/islet) were detected.

The GK rat is a rodent model of spontaneous type 2 diabetes established by inbreeding Wistar rats selected from the upper limit of a normal distribution for glucose tolerance (18). We analyzed the presence of islet-associated macrophages in 1-month-old (weaning; normogylcemic) and 2-month-old male Wistar and GK rats (1 month after chronic mild hyperglycemia; fasting glycemia: 6.3 ± 0.2 vs. 11.3 ± 0.6 mmol/l, n = 10, respectively; P < 0.05) using two different antibodies, ED-1 (CD68) and anti-MHC class 2. There was no difference in macrophages associated with islets at 1 month of age (not shown). While few macrophages were present around Wistar islets, GK islets were characterized by pronounced macrophage infiltration at 2 months of age (Fig. 2). The mean islet area analyzed for CD68-positive and MHC-2–positive cells was identical in both strains for these comparisons (Fig. 2B and D). Finally, we investigated 8- to 9-week-old db/db mice and littermates (glucose intolerance develops between 4 and 8 weeks of age) for islet-associated Cd11b-positive cells. Fasting glycemia was 4.0 ± 0.2 and 8.3 ± 0.7 mmol/l for db/+ and db/db mice, respectively (n = 10, P < 0.05). Compared with db/+ littermates, db/db mouse islets were characterized by increased peri-islet CD11b-positive cell infiltration (2.6 ± 1.7%, n = 4 vs. 24.8 ± 4.9% of total islets with more than three CD11b-positive cells, n = 5; P < 0.05; 1.6 ± 0.1 vs. 2.4 ± 0.3 Cd11b-positive cells/islet; P < 0.05, n = 5 for db/+ and db/db, respectively).

We hypothesized that pancreatic islets secrete factors that may attract macrophages under pathological conditions. After screening human islet, mouse islet, and cell line preparations, we concentrated on the regulation of IL-6, IL-8, G-CSF, interferon-inducible protein 10 (IP-10), MIP-1α, MCP-1, and chemokine KC in the rodents (all known to be elevated in type 2 diabetic and obese subjects [1,2,24]). After 48 h of treatment with 33 mmol/l glucose or 0.5 mmol/l palmitate (in combination or separately), we evaluated glucose-stimulated insulin secretion to ensure β-cell dysfunction (Fig. 3A and B). As seen in Fig. 3C–E, mouse and human islets released profoundly more IL-6, IL-8, chemokine KC, and G-CSF in response to elevated glucose and palmitate in combination after 48 h of treatment, while palmitate alone stimulated only some of these factors. Note that rodents do not express IL-8, and chemokine KC is thought to be its functional homolog in rodents. Further, IP-10 showed a trend toward regulation in human islets, with no change in mouse islets (Fig. 3F). MIP-1α release also was significantly increased in human islet preparations in response to a diabetic milieu (Fig. 3G) but was undetectable in mouse islet supernatants (data not shown). Finally, MCP-1 remained unaffected in human islets (Fig. 3H; undetectable in mouse islets).

The specificity of this islet inflammatory response to elevated glucose and palmitate was tested by comparison of endocrine and nonendocrine tissue, by analysis of β-cell lines, and by the induction of cell death. While glucolipotoxic stress increased IL-6, chemokine KC, and G-CSF release from islets, these factors were not significantly increased in an equal quantity of nonendocrine tissue (Fig. 4A–D). Further, to support our claim that these factors are islet cell derived, both MIN-6 and INS-1 cells responded to elevated glucose and palmitate by releasing increased amounts of chemokine KC, G-CSF, and MIP-1α (Fig. 4E–H). Finally, to rule out that an unspecific stimulation by the apoptotic/necrotic process induced by glucose and palmitate was responsible for the increase in cytokine/chemokine release, we tested the effect of a 48-h treatment with 500 nmol/l staurosporine and 0.1 and 1 mmol/l streptozotocin on mouse islets. Islet cell death induced by either agent did not increase cytokine/chemokine release (Fig. 4I–K).

To examine whether the nutrient effects on the above cytokine/chemokines are mediated at the transcriptional level, we isolated mouse islet RNA after a 48-h treatment under glucolipotoxic conditions. In contrast to the response seen at the protein level, the IL-6 transcript was downregulated by a diabetic milieu, while KC and G-CSF paralleled their protein response. Further, mouse MIP-1α also was strongly upregulated at the mRNA level (Fig. 5).

We also tested the hypothesis that chemokine KC and G-CSF may exert direct effects on islet function. Both factors were initially tested at 1–100 ng/ml, with maximal effects seen at 100 ng/ml. When added at 100 ng/ml for 4 days, both factors had a mild effect on β-cell apoptosis (control: 0.33 ± 0.11; 100 ng/ml KC: 0.85 ± 0.35; 100 ng/ml G-CSF: 0.41 ± 0.03 TUNEL β-cells/islet; P > 0.05, n = 5) and a minimal effect on glucose-stimulated insulin secretion (control: 2.9 ± 0.2-fold; 100 ng/ml KC: 2.3 ± 0.3-fold insulin secretion; P < 0.05, n = 4). Thus, we hypothesized that these factors were more important in mediating indirect effects on islets rather than having direct effects on β-cells themselves.

To evaluate whether in vitro regulation of cytokine/chemokine release by a diabetic milieu could be relevant in vivo, C57BL/6J mice were subjected to high-fat diet feeding in order to investigate the islets ex vivo. After 4 weeks on a high-fat diet, there was no increase in ex vivo islet cytokine/chemokine release despite a slight impairment in islet function as assessed by glucose-stimulated insulin secretion (not shown). However, after 8 weeks of high-fat feeding, both an impairment in islet function and a doubling of the same cytokines/chemokines regulated by a diabetic milieu in vitro (IL-6, chemokine KC, and G-CSF) were observed compared with control islets (Fig. 6). Finally, circulating serum KC was significantly elevated in 8-week high-fat–fed animals versus controls (Fig. 6B, P < 0.05, n = 4). The present study and independent experiments in our laboratory have not found an increase in islet area as a result of 8 weeks of high-fat feeding (data not shown; n = 5), indicating that these effects are not secondary to an increase in islet mass.

Given those factors induced by a type 2 diabetic milieu in human islets, we hypothesized that the chemokine IL-8 may be responsible for the migration of monocytes toward islets in type 2 diabetes. IL-8 is known to attract both monocytes and neutrophils, and it was most strongly induced by glucolipotoxicty in human islets (Fig. 3D). Initially, we analyzed the localization of IL-8 in the human pancreas and in human isolated islets. We found IL-8 expression in human islets to be localized to glucagon-positive endocrine cells, suggesting that islet IL-8 is mainly α-cell derived (Fig. 7A, 1–15). Indeed, by electron microscopy on ultrathin serial sections of isolated human islets, IL-8 colocalized to glucagon-positive α-cell granules (Fig. 7B) but was not found in β- or δ-cells (not shown). The specificity of the antibody used for immunostaining was isotype controlled, tested by preabsorption with recombinant IL-8, tested on a positive control tissue (Fig. 7A), and confirmed to bind the 8-kDa IL-8 protein in isolated human islets by Western blot (Fig. 7C).

Next, we tested the hypothesis that those factors released by pancreatic islets exposed to a type 2 diabetic milieu may recruit leukocytes. Flow cytometry analysis of migrated leukocytes revealed that conditioned medium taken from human islets had no effect on T-cell, B-cell, and natural killer cell migration (data not shown). However, monocyte and neutrophil migration was clearly increased by conditioned medium taken from diabetic milieu–treated human islets (from islets treated with 33 mmol/l glucose and 0.5 mmol/l pamitate; “treated”) compared with conditioned medium from untreated human islets (“untreated”; Fig. 7D). Intriguingly, IL-8 neutralization completely reversed the increased monocyte and neutrophil migration induced by treatment of human islets with a diabetic milieu (Fig. 7D). When taken together with the human and mouse islet in vitro and ex vivo data, the concept that IL-8 (or possibly chemokine KC in the rodent) may contribute to the immune cell infiltration we observed in type 2 diabetes is supported.

To the best of our knowledge, macrophage infiltration of pancreatic islets has not been systematically investigated in type 2 diabetes. Our data support the conclusion that increased numbers of immune cells, specifically macrophages, are associated with pancreatic islets in type 2 diabetes. Further, increased numbers of immune cells were associated with islets of type 2 diabetes models, including the high-fat–fed C57BL/6J mouse, the GK rat, and the diabetic db/db mouse. Given the accumulation of immune cells around and within type 2 diabetic islets and our in vitro data highlighting the specificity of this response to endocrine cells, this inflammatory process likely is directed toward the endocrine pancreas. This is keeping in mind that in obese patients, other organs may display typical inflammatory characteristics (e.g., macrophages in adipose tissue [2,3] and in atherosclerotic plaques [25]).

Whether the presence of macrophages is causative to type 2 diabetes islet pathology requires further investigation. Hess et al. (26) have shown that bone marrow cells promote islet regeneration of damaged β-cells in a peri-islet fashion. Possibly, early infiltration of macrophages may be beneficial to islet function and plasticity. However, as the disease progresses, macrophages may play a role in accelerating pancreatic islet cell dysfunction and death. Macrophages also may be present after β-cell death, acting to phagocytose dead islet tissue. To this end, we did not detect apoptotic cells in the vicinity of infiltrating immune cells.

To further explore the molecular signals underlying increased numbers of macrophages associated with type 2 diabetic islets, we investigated the release of cytokines and chemokines from both human and mouse islets exposed to a type 2 diabetic milieu and ex vivo from high-fat–fed animals. In all cases, a type 2 diabetic milieu caused a pronounced increase in the release of IL-6, IL-8, chemokine KC (rodent islets only), G-CSF, and MIP-1α (human islets only). Further, in mouse islets, the increased release of KC and G-CSF could be blunted by treatment of islets with IL-1Ra, the endogenous receptor antagonist of IL-1β (n = 10, P < 0.05 [J.A.E., M.Y.D., unpublished data]). This suggests that β-cell production of IL-1β in diabetic islets (27,28) may be a key regulator of increased chemokine production. Further, IL-1β, or other effector mechanisms, may be at the origin of the observed increased rate of β-cell apoptosis, since the macrophages were not associated with apoptotic cells. Therefore, antagonism of IL-1 in patients with type 2 diabetes may protect the islets not only from the direct toxic effects of IL-1β but also from the consecutive inflammatory process (29).

While further investigation is required to test the true cellular origin of islet-derived chemokines both in vitro and in vivo, our data suggest that they are pancreatic islet cell derived. This may include an endocrine, endothelial, neuronal, and/or resident macrophage (or other immune cells) origin. However, in contrast to IL-8, G-CSF, and MIP-1α, nutrient regulation of IL-6 does not seem to occur at the mRNA level in islets, and its production level is similar in islets and nonislet pancreatic tissue. Our data are supported by a preliminary report that elevated palmitate can upregulate chemokine KC mRNA in addition to other chemokines (i.e., MCP-1, SDF-1) in MIN-6 β-cells (30). In addition, transcripts for chemokines in isolated β-cells and β-cell lines have been shown to be induced by cytokines mimicking the type 1 diabetic milieu (i.e., IL-1β, tumor necrosis factor α, IFNγ) (31–34). Chemokine KC also was strongly induced in these studies (33,34). Comparing these studies with our data shows that both the pattern of islet chemokines induced by glucose/palmitate versus IL-1β (or a cytokine cocktail) and the magnitude of effect may be different. However, these differences may simply be due to concentration-dependent effects of IL-1β.

Increased migration of both monocytes and neutrophils was induced by conditioned medium from islets exposed to a type 2 diabetic milieu and was completely reversed by IL-8 neutralization. Thus, IL-8 presents itself as an intriguing candidate for contributing toward inflammation in type 2 diabetic islets. Circulating IL-8 has been shown to be elevated in type 2 diabetic individuals (10,11), and IL-8 expression is elevated in the adipose tissue of obese insulin-resistant subjects (35). Further, hyperglycemia has been shown to increase aortic endothelial cell IL-8 secretion and thereby promote monocyte adhesion (13). Our levels of IL-8 release are very similar to those needed for monocyte adhesion to endothelia (12,13). Given that circulating IL-8 levels are very low, it is possible that islet-produced IL-8 may promote a concentration gradient leading to monocyte transmigration and infiltration.

In conclusion, we have detected the presence of increased numbers of macrophages in pancreatic islets from patients with type 2 diabetes. In fact, in high-fat–fed mice and GK rats, increased islet macrophages were detected early during disease progression. Further, elevated glucose and palmitate concentrations increased chemokine release from human and mouse pancreatic islets both in vitro and ex vivo. In particular, we localized IL-8 to the human α-cell and demonstrated the ability of a type 2 diabetic milieu to enhance immune cell chemotaxis, an effect regulated by islet-derived IL-8.

This work was supported by grants from the Swiss National Science Foundation (PP00B-68874/1), the European Foundation for the Study of Diabetes, and the University Research Priority Program “Integrative Human Physiology” at the University of Zürich. J.A.E. is supported by a Juvenile Diabetes Research Foundation postdoctoral fellowship.

We thank M. Borsig, I. Danneman, G. Seigfried-Kellenberger, E. Katz, and J. Coulaud for excellent technical assistance and W. Moritz for technical advice regarding immunohistochemistry.