Distinction Between Interleukin-1–Induced Necrosis and Apoptosis of Islet Cells

Pancreatic islets were isolated from adult male Wistar rats, as described previously (26). Freshly isolated islets were dissociated in calcium-free medium and purified into single β-cells (>92% β-cells) and α-cells (>75% α-cells) by autofluorescence-activated sorting (26). The purified β-cell preparation contains <5% α-cells, <5% PP cells, <5% δ-cells, and <5% nonendocrine islet cells and contaminating exocrine cells; the α-cell preparation contains 10% β-cells, 5% PP cells, <5% δ-cells, and <5% nonendocrine islet cells and contaminating exocrine cells (26). Isolated single cells were cultured in polylysine-coated microtiter cups (3 × 10³, 24 × 10³, or 48 × 10³ cells/cup in 200 μl medium) (27). Isolated β-cells were also reaggregated in a rotary shaking incubator (28) before further culture in polylysine-coated multiwells (15 × 10³ cells/well in 1 ml medium); isolated islets were cultured in the same conditions (15 islets/well in 1 ml medium). Culture was conducted in serum-free medium that was previously found to maintain the viability and glucose responsiveness of isolated rat and human β-cells (27,29,30). This medium consists of Ham’s F10 (Gibco Laboratories, Grand Island, NY) supplemented with 2 mmol/l l-glutamine, 50 μmol/l 3-isobutyl-1-methylxanthine, 0.075 g/l penicillin, 0.1 g/l streptomycin, 10 mmol/l glucose, and 10 g/l charcoal-treated bovine serum albumin (factor V, radioimmunoassay grade; Sigma, St. Louis, MO). In experiments where cytotoxicity on rat β-cells, α-cells, and/or islets was compared, this medium was supplemented with glucagon (10⁻⁸ mol/l, Sigma) and recombinant human growth hormone (0.5 ng/ml, Amersham Pharmacia Biotech, Buckinghamshire, U.K.) because these hormonal supplements were beneficial for the survival of isolated α-cells.

Human β-cell preparations were made available by a multicenter program on islet transplantation in diabetes (31). Donor pancreases were procured by European transplantation centers through the Eurotransplant Foundation (Leiden, the Netherlands). The methods for isolation and culture of human pancreatic islet cells have been previously described (30,31). After culture for 4–19 days, these preparations were dissociated in calcium-free medium containing trypsin and then sorted to be enriched in single β-cells (purity 77 ± 4%) (26). The non–β-cell fraction is a mixture of variable proportions of endocrine islet cells and nonendocrine duct cells (31). The purified cells were cultured in polylysine-coated microtiter cups in the same medium as described for rat β-cells, with 7.5 mmol/l glucose.

After 16-h preculture, the cell preparations were exposed (on day 0) for 3, 6, and 9 days to one or two of the following cytokines: recombinant human IL-1β (200 U/ng; Genzyme, Cambridge, MA; and 38 U/ng, a gift from Dr. C.W. Reynolds, National Cancer Institute, Bethesda, MD), recombinant murine IFN-γ (for experiments with rat cells; 10 U/ng; Holland Biotechnology, Leiden, the Netherlands), and recombinant human IFN-γ (for experiments with human cells; 48 U/ng; Genzyme). Concentrations (50 U/ml IL-1β and 1,000 U/ml IFN-γ) were selected from previous studies (14,17,23); the effects were also tested in the presence of 1 mmol/l of the inducible nitric oxide synthase (iNOS) inhibitor l-methylarginine (Sigma). Medium was replaced every 3 days. Medium nitrites were measured by the Griess reaction (32).

The effect of NO was assessed on both rat β-cells and human islet cells by exposing them for 15 min to the NO donor GEA 3162 [1,2,3,4-oxatriazolium,5-amino-3-(3,4-dichlorophenyl), chloride; Alexis, Nottingham, U.K.] before washing and culturing for 1 or 6 days in the same medium as above.

The percentages of viable, apoptotic, and necrotic cells were assessed by a previously reported assay (25). After 3, 6, and/or 9 days of cytokine treatment or after 1 or 6 days of culture after exposure to the NO donor, the cells were incubated for 15 min with the DNA-binding dyes Hoechst 33342 (HO 342) (20 μg/ml; Calbiochem-Novabiochem, La Jolla, CA) and propidium iodide (PI) (10 μg/ml; Sigma). The HO 342 compound is known to cross the plasma membrane of all cells, whether they are damaged or not, causing a blue fluorescence of their nuclei. The polar PI only penetrates cells with damaged membranes and leads to a red nuclear fluorescence. Examination in an inverted fluorescence microscope with ultraviolet excitation at 340–380 nm allows counting of the percentage of viable cells (intact nuclei with blue HO 342 fluorescence), the percentage of necrotic cells (intact nuclei with yellow fluorescence due to binding of both HO 342 and PI), and the percentage of apoptotic cells (fragmented nuclei with either blue [HO 342 binding, early apoptosis] or yellow [HO 342 plus PI binding, late apoptosis] fluorescence). In each cup or well, a minimum of 500 cells were counted. The necrosis index was calculated as described earlier (14,23,33):

The apoptosis index was obtained by the same formula, using the percentage apoptotic cells instead of the percentage necrotic cells. The control condition (i.e., culture without cytotoxic agents) with 3 × 10³ cells/200 μl maintained the survival of single rat β- and α-cells: 82 ± 7 and 74 ± 2% living β-cells and 79 ± 8 and 67 ± 4% living α-cells on days 0 and 9, respectively (P > 0.05). On the other hand, the percentage of living human islet cells decreased from 85 ± 3% on day 0 to 66 ± 6% on day 9 (P < 0.05). Survival of rat β-cells was maintained at high cell densities: 76 ± 2 and 75 ± 4% living β-cells on day 9 at 24 and 48 × 10³ cells/200 μl, respectively (paired t test, P > 0.05 vs. culture with 3 × 10³ cells/200 μl). Addition of l-methylarginine to the culture medium did not influence the survival of the cells in control conditions (data not shown). The percentage of apoptotic cells in control cultures of rat β-cells at the end of a 9-day culture period was 7 ± 1%; in rat α-cell cultures and human β-cell cultures, we found 19 ± 3 and 20 ± 5% apoptotic cells, respectively, after 9 days of culture. Cell death in control conditions mainly occurs during the first day of culture (15–21% of all cells), which suggests that it involves cells that were damaged during the isolation procedure and/or nonendocrine contaminating cells that did not survive under the selected culture conditions. Experimental conditions are introduced after this first day and tested for a period of a maximum of 9 days, during which cell death in the control conditions is only 1% per day. These control values confirm the ability of our serum-free culture medium to maintain survival of endocrine islet cells during 9 days of culture (27).

Cell death in aggregated cells and islets was evaluated with the same fluorescence methods. However, precise quantification of the number of living, apoptotic, and necrotic cells was not possible in this compact tissue. This technical obstacle cannot be overcome by dissociating the aggregates or islets before counting because such a step disrupts cells with damaged membranes and in itself causes cell death (34). We can thus only identify conditions with marked apoptosis and/or necrosis in these preparations.

Data represent means ± SE of n independent experiments. Each experiment is conducted with cells obtained from different rat or human pancreases. Statistical significance of differences was determined using the Student’s paired t test.

A 9-day exposure of single rat or human β-cells (3 × 10³ cells/200 μl) to IL-1β up to 50 U/ml was not cytotoxic (Fig. 1). Combination of IL-1β (50 U/ml) plus IFN-γ (1,000 U/ml) induced apoptosis in both preparations, starting after 3 days (apoptosis index 10 ± 1%) and progressively increasing with culture duration (apoptosis index 33 ± 2% after 9 days; Fig. 1). It also caused necrosis in a small percentage of rat β-cells but not in human islet cells; this effect did not increase with the exposure time (14 ± 3 and 13 ± 4% after 6 and 9 days culture, respectively; P < 0.05 vs. control culture without cytokines, paired t test). Culture with only IFN-γ did not exert any cytotoxic effect (Fig. 1). After 9 days of culture, the percentage of dead cells in the IL-1β plus IFN-γ condition was 69 ± 2 vs. 28 ± 5% in the control for rat β-cells and 56 ± 6% in the IL-1β plus IFN-γ condition vs. 34 ± 6% in the control for human β-cells. All further analysis is carried out on the calculated necrosis and apoptosis index, which facilitates comparisons of the experimental conditions.

The presence of l-methylarginine completely prevented cytokine-induced necrosis of rat β-cells (Fig. 1) but exerted only a partial and transient protection against apoptosis: apoptosis index 4 ± 1, 13 ± 4, and 28 ± 7% after 3, 6, and 9 days of culture, respectively, with IL-1β plus IFN-γ (P < 0.05, <0.05, and >0.05 vs. 10 ± 1, 25 ± 3, and 33 ± 2, respectively, in the absence of l-methylarginine; Fig. 1). In human β-cell preparations, l-methylarginine did not reduce the apoptosis index after 9 days of culture with the cytokine combination (14). The protective effect of l-methylarginine against cytotoxicity induced by exposure of single rat β-cells to IL-1β plus IFN-γ was associated with its prevention of NO accumulation in the medium (medium nitrites in cultures of single rat β-cells exposed to IL-1β plus IFN-γ were 1.3 ± 0.3 and 5.5 ± 0.7 pmol/μl medium in the presence and absence of l-methylarginine). Nitrite levels were only measured during the first 72 h because their values are difficult to interpret at later time points in view of the increasing number of dead cells (Fig. 1).

These data indicate that both rat and human β-cells can undergo apoptosis after prolonged exposure to IL-1β plus IFN-γ but not to the single cytokines, suggesting involvement of two pathways. This process does not require production of NO but seems accelerated by it (see data on rat β-cells). The cytokine combination can also cause necrosis of rat β-cells, but this effect is completely NO-dependent; this might explain why no necrosis was noticed in human β-cell preparations (Fig. 1) in which the cytokines failed to induce detectable NO levels under the present experimental conditions (<0.5 pmol/μl after 3 days of culture).

The combination of IL-1β plus IFN-γ was at least equally potent in inducing apoptosis of rat α-cells and rat β-cells but failed to cause necrosis of these cells (apoptosis index in rat α-cells 50 ± 6% and necrosis index 10 ± 3%, P < 0.01 and >0.05, respectively, vs. control condition without cytokines, n = 4). At variance with β-cells, the α-cells also underwent apoptosis when cultured with only IFN-γ (apoptosis index 25 ± 2% and necrosis index 6 ± 3%, P < 0.01 and >0.05, respectively, vs. control condition without cytokines, n = 4). Exposure to IL-1β alone was not cytotoxic for rat α-cells (apoptosis and necrosis index <5%, n = 4). The cytokine-induced apoptosis of α-cells was not prevented by l-methylarginine (apoptosis index 41 ± 5% after exposure to IFN-γ in the presence of l-methylarginine and 52 ± 5% after exposure to IL-1β plus IFN-γ in the presence of l-methylarginine, P < 0.01 vs. control condition without cytokines, n = 4). In none of the α-cell cultures did medium nitrite rise to detectable levels, suggesting independence of NO production (<0.5 pmol/μl after 3 days of culture). This result also indicates that none of the cell types present in this fraction produce any detectable NO levels. Some of these cell types contaminate the purified β-cell preparation, but at lower percentages than in the α-cell preparation; they therefore do not contribute to the NO levels that are measured in the β-cell preparation.

When single rat β-cells were plated at higher densities than the control condition (3 × 10³ cells/200 μl), they became subject to IL-1β–induced necrosis but not to apoptosis (Table 1). At 8- and 16-fold higher cell densities, the single cytokine caused a necrosis index of 27 ± 12 and 45 ± 7% after 9 days of culture (Table 1). This cytotoxic effect was not yet apparent after 3 days of culture (necrosis index <5% for the 16-fold higher cell density) but was clear after 6 days (necrosis index 17 ± 6% for the 16-fold higher cell density, P < 0.05 vs. control condition with 3 × 10³ cells/200 μl). At the end of the first 3 days, medium nitrite concentration was two- to threefold higher at the higher cell densities. In the presence of l-methylarginine, medium nitrites did not rise above control values, and no necrosis was induced by IL-1β. When medium nitrite accumulation over the first 3 days was expressed as a function of the respective cell numbers, higher values were found for the low–cell density condition, indicating that the absence of necrosis in this preparation is not due to lower NO production rates (Table 1). During subsequent days, the nitrite accumulation in the medium markedly decreased in the high cell density conditions (data not shown), which is attributed to the increasing proportion of dead cells.

In reaggregated β-cells or in intact islet tissue, a precise quantification of the percentage of necrotic and apoptotic cells is difficult if not impossible. The currently used fluorescence method can be used to identify outspoken conditions of necrosis or apoptosis in such preparations. It was thus noticed that IL-1β alone caused massive necrosis in reaggregated β-cells; this effect did not occur in the presence of l-methylarginine (Fig. 2). Similar observations were made in cultured islets; electron micrographs indicated that necrosis affected both α- and β-cells (data not shown). The appearance of necrosis in β-cell aggregates and in islets occurred at an earlier time point (after 3 days of culture) than in high-density monolayers (see above). The medium nitrite accumulation during the first 3 days was lower than that in the condition of single β-cells (Table 2). When this nitrite accumulation was expressed as a function of the number of β-cells or of the DNA content, values were also lower than those in the single β-cell condition (Table 2).

A 15-min exposure of rat and human islet cells to the NO donor GEA 3162 caused necrosis within 24 h (Table 3). Further culture for 5 days did not increase the percentage of necrotic cells. In the three tested cell preparations, the transition from marginal to massive cytotoxicity occurred in a narrow concentration range, i.e., between 25 and 100 μmol/l (Table 3). Within this narrow range, rat α-cells appeared more sensitive than rat β-cells (Table 3). In none of the conditions did the NO donor induce apoptosis.

Numerous studies have reported that proinflammatory cytokines can be toxic for islet β-cells (2–4). Using different techniques, several laboratories have noticed that cytokines can induce apoptosis in islet cells (14,35–38). A microscopic fluorescence assay allows one to quantify the number of apoptotic cells during culture and to determine the relative importance of apoptosis versus necrosis (25). We have used this method to investigate the effects of IL-1β, which has been a major component in most previous studies. It was found that IL-1β can kill β-cells by necrosis as well as by apoptosis. However, induction of one or the other form of cell death depends on the presence of additional environmental factors.

Single rat β-cells were not killed by IL-1β when cultured at low plating densities (±90 cells/mm²) but underwent necrosis from day 3 on at high densities (±1,500 cells/mm²). Absence of IL-1–induced cytotoxicity in dispersed β-cell preparations (11,19,20, and present data) is thus not caused by an insensitivity to IL-1β, for example, as a result of the dissociation procedure or of the cellular attachment to dishes. It is not attributable to a lower NO production, as judged from the nitrite formation per 10³ β-cells, but rather is related to a higher distance between β-cells. Dilution of released NO may avoid toxic NO levels at neighboring cells. The role of NO-dependent cell necrosis indeed increases with plating density and is also high in cultures of reaggregated β-cells and of intact islets in which dilution is probably minimal. Our observations clarify the discrepancy that was previously noticed for the IL-1 effects in different in vitro models. They indicate that static cultures of compact islet cells can lead to toxic intercellular NO concentrations that cause necrosis. The question now is whether these NO levels can be reached in vivo, for example, when the flow in the islet interstitium is severely impaired.

The reported cell death in IL-1–cultured rat islets (5–10) most likely corresponded with a process of necrosis. Nitrite measurements in collected medium allowed us to identify the NO dependency of this toxicity, but they are not a valid index for the occurrence of this form of cell death because they are not always indicative of the NO levels at the β-cells. Experiments with RINm5F insulinoma cells also indicated that NO production is necessary but not sufficient for a cytokine-induced cell destruction (39). We did not observe a correlation between medium nitrite levels and the occurrence of necrosis in single or aggregated cells. In fact, medium nitrites were lower in reaggregated β-cells and in islets: this probably reflects the functional deficiency that proceeds to necrosis.

Experiments with the NO donor GEA indicated that human β-cells exhibit a susceptibility to NO-induced necrosis comparable to that of rat β-cells. That IL-1β (alone or in combination with IFN-γ) did not induce necrosis in our human β-cell preparation can be attributed to the failure to induce NO production—at least to any detectable levels. We have recently reported that the combination of IL-1β plus IFN-γ can induce iNOS expression and NO production in pancreatic duct cells (40); this nonendocrine cell type is also identified in the purified human β-cell preparation but in numbers that are too low to produce any detectable NO (D.Pa., D.Pi., unpublished data). It is currently unknown whether in other circumstances duct cells can produce sufficient NO to cause necrosis of adjacent β-cells. Studies in freshly isolated human islet preparations have also shown that IL-1β plus IFN-γ can induce iNOS expression in human β-cells (41), which raises the possibility that the state of the β-cells or the nature of the surrounding cells determines their NO production when exposed to cytokines and, hence, their susceptibility to necrosis.

When IL-1β was combined with IFN-γ, both single rat and single human β-cells died by apoptosis starting after 3 days of exposure and progressively increasing with culture time; this process was somewhat delayed when NO production was suppressed by l-methylarginine but was essentially NO-independent. Exposure of β-cells to IL-1β might make them susceptible to apoptosis induction by IFN-γ. IFN-γ alone indeed causes apoptosis in islet α-cells but not in β-cells. We have recently reported that IL-1β alters the phenotype of rat β-cells, causing a reduction in their cell-specific properties (42). This alteration may well be accompanied by a loss in β-cell–specific protection against apoptosis (24), making the cells more susceptible to apoptosis inducers such as IFN-γ.

The present in vitro experiments do not provide information on the effectors and mode of β-cell death in type 1 diabetes. This objective is difficult to achieve with in vitro models. Our data nevertheless indicate that if IL-1β is released within the islets of the human pancreas, the cytokine will not necessarily cause the massive necrosis as is seen in cultured rat islets. To achieve this effect in human islets, local IL-1 levels will have to be chronically elevated and lead to toxic intercellular NO concentrations: this will probably require sufficient iNOS induction in β-cells (41) or neighboring non–β-cells (21,22,40) as well as a reduced flow of the interstitial fluid. If these conditions are not met, the cytokine may cause apoptosis of β-cells if these are simultaneously and chronically exposed to other cytokines such as IFN-γ. Alternatively, IL-1β may influence the functional state of the β-cells by altering their phenotype (42) and, hence, their responses to physiological and pathophysiological agents.

This work was supported by grants from Interuniversity Attraction Pole P4/21, the Belgian Fonds Wetenschappelijk Onderzoek—Vlaanderen (G.0376.97), and the Juvenile Diabetes Foundation (995004). A.H. was a research fellow (1993–1997) of the Belgian Fonds Wetenschappelijk Onderzoek—Vlaanderen; she is grateful to Miriam Marichal, head of the Pathology Department, and the other pathologists for taking over her clinical work during this period.

The authors thank their technical collaborators for preparing the rat and human β-cell preparations.