Short Term Exposure of Beta Cells to Low Concentrations of Interleukin-1β Improves Insulin Secretion through Focal Adhesion and Actin Remodeling and Regulation of Gene Expression*

Background: Molecular mechanisms involved in beneficial effects of IL-1β on beta cell function are poorly understood. Results: Short exposure of beta cells to low concentrations of IL-1β enhanced stimulated insulin secretion through focal adhesion, actin remodeling, and transcription. Conclusion: The beneficial effects of IL-1β occur at multiple molecular levels. Significance: IL-1β has opposing effects on beta cells depending on time of exposure and concentration. Type 2 diabetes involves defective insulin secretion with islet inflammation governed in part by IL-1β. Prolonged exposure of islets to high concentrations of IL-1β (>24 h, 20 ng/ml) impairs beta cell function and survival. Conversely, exposure to lower concentrations of IL-1β for >24 h improves these same parameters. The impact on insulin secretion of shorter exposure times to IL-1β and the underlying molecular mechanisms are poorly understood and were the focus of this study. Treatment of rat primary beta cells, as well as rat or human whole islets, with 0.1 ng/ml IL-1β for 2 h increased glucose-stimulated (but not basal) insulin secretion, whereas 20 ng/ml was without effect. Similar differential effects of IL-1β depending on concentration were observed after 15 min of KCl stimulation but were prevented by diazoxide. Studies on sorted rat beta cells indicated that the enhancement of stimulated secretion by 0.1 ng/ml IL-1β was mediated by the NF-κB pathway and c-JUN/JNK pathway acting in parallel to elicit focal adhesion remodeling and the phosphorylation of paxillin independently of upstream regulation by focal adhesion kinase. Because the beneficial effect of IL-1β was dependent in part upon transcription, gene expression was analyzed by RNAseq. There were 18 genes regulated uniquely by 0.1 but not 20 ng/ml IL-1β, which are mostly involved in transcription and apoptosis. These results indicate that 2 h of exposure of beta cells to a low but not a high concentration of IL-1β enhances glucose-stimulated insulin secretion through focal adhesion and actin remodeling, as well as modulation of gene expression.

Type 1 and type 2 diabetes are characterized by absolute or relative beta cell failure respectively. Although long considered diseases with disparate etiology, current thinking suggests that both major forms of diabetes may be caused in part by the presence of inflammatory mediators in the islet that trigger a final common pathway of beta cell loss of function and/or death (1,2). The IL family of cytokines consists of 11 members closely linked to the innate immune response and known to mediate autoinflammatory diseases (3). IL-1␤ has been suggested as a key inflammatory cytokine in both type 1 and type 2 diabetes and is considered a promising target for treatment of both diseases (4,5). Although pilot studies in type 2 diabetes proved successful (6,7), such was not the case for type 1 diabetes (8). Although invading immune competent cells are most likely the major source of IL-1␤ within islets during disease progression, intriguingly in isolated human islets beta cells themselves produce IL-1␤ (9), as well as islet vascular endothelial cells (10). Beta cell IL-1␤ production is elevated by IL-1␤ itself, and more interestingly by high concentrations of certain fatty acids or glucose (11,12), even if the latter observation and its suggested relevance to the pathophysiology and possible treatment of type 2 diabetes has been contested (1). Furthermore, beta cells were shown to contain IL-1␤ protein in diabetic animals and most significantly in humans with type 2 diabetes (13). Taken together, all these data suggest a deleterious role for elevated islet IL-1␤ production, leading to beta cell death in type 1 diabetes and impaired function with only limited cell death in type 2 diabetes. It is indeed well established that in vitro culture of islets or beta cell lines with high concentrations of IL-1␤ (5-20 ng/ml) for long periods of time (Ͼ 24 h) induces apoptosis and necrosis through multiple pathways including endoplasmic reticulum and metabolic stress (14 -18). However, it should not be forgotten that inflammation is initially a repair response to an insult, and it has indeed been known for several decades that IL-1␤ can also exert beneficial effects on beta cells when used at low concentrations (0.01-0.1 ng/ml). Such concentrations improve insulin biosynthesis and secretion and increase beta cell proliferation and survival after up to 2 days of exposure (12,19,20). We have also shown that IL-1␤ production by beta cells underlies the beneficial effects of culture on extracellular matrix and suggested that activation of the canonical NF-B pathway could be involved in both the positive and negative actions of IL-1␤ on beta cells following exposure to low or high concentrations, respectively, for many hours or days (21). However, the precise mechanism(s) through which IL-1␤ improves insulin secretion and the possible effects of much shorter times of exposure remain poorly characterized following a series of early studies dating to the 1980s and 1990s (22)(23)(24)(25)(26)(27)(28)(29).
We have previously shown that acute glucose stimulation of beta cells induces focal adhesion (FA) 2 and acto-myosin IIA remodeling necessary for insulin granule recruitment at the basal membrane and secretion (30,31). Focal adhesion kinase (PTK2) and paxillin (PXN) are two focal adhesion-associated proteins that function in transmitting signals downstream of integrins. PTK2 is a nonreceptor tyrosine kinase that upon integrin ligation is activated to autophosphorylate Tyr 397 , which increases its activity necessary for the binding and the phosphorylation of PXN in particular, which is involved in many cellular processes (32). Glucose was seen to induce phosphorylation of PTK2 and PXN, necessary for downstream activation of ERK, and induced their recruitment to newly formed protrusions at the basal membrane of beta cells. Interestingly, IL-1␤ is known to induce actin remodeling and FA maturation linked to integrin activation in fibroblasts (33,34), and FA signaling can also impact IL-1R expression, cluster these receptors at FAs, and induce ERK activation in other tissues (35). Taking all these earlier observations into consideration, we have now studied the impact of short term (2 h or less) exposure of primary beta cells to IL-1␤ with a special focus on possible beneficial effects of low concentrations of the cytokine on beta cell function and the underlying molecular mechanisms including specifically FA and actin cytoskeleton remodeling.
Cells and Culture Conditions-Rat islet isolation, beta cell sorting by FACS and monolayer culture on extracellular matrix derived from rat bladder carcinoma 804G cells (804G-ECM) were performed as previously described (36). Human islets were generously provided by the islet transplantation groups at the University of Geneva, Switzerland, through Juvenile Diabetes Research Foundation Award 31-2008-416 (European Consortium for Islet Transplantation Islet for Basic Research Program). Human islets isolated from cadaveric donor pancreas were maintained in CMRL-1066 medium containing 5 mM glucose, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM GlutaMAX, 250 g/ml gentamycin, 10 mM Hepes, and 10% FCS (Invitrogen). Fifty human islet equivalents were plated on 804G-ECM per condition and were cultured for 24 h to adhere before changing the medium for the insulin secretion experiments.
Insulin Secretion-Beta cells or islets were washed, preincubated, and incubated for insulin secretion assays, and insulin was measured by radioimmunoassay as previously described (36). In brief, cells were preincubated 2 h in Krebs-Ringer bicarbonate buffer at 2.8 mM glucose followed by a further 1 h at 2.8 mM to measure basal insulin secretion and a further 15 min to 1 h (as indicated) at 16.7 mM glucose or for 15 min with 30 mM KCl to measure stimulated secretion. IL-1␤ was added to the incubations as indicated for each specific experimental condition. Cellular insulin was extracted at the end of the experimental period, and secretion is expressed as a function of total insulin (cell content ϩ basal ϩ stimulated secretion).
SDS-PAGE and Western Blotting-Protein samples were prepared and immunoblots were analyzed as described (37). Western blots were quantified by densitometry and band density of phosphoproteins normalized to that of the corresponding total protein and/or to total actin as indicated in the figure Legends.
Statistical Analysis-Statistical significance for differences between experimental conditions was determined using Graph-Pad Prim 5 software by one-or two-way ANOVA as indicated or Student's t test for unpaired groups for comparison of two conditions. p values less than 0.05 were considered significant.
Immunofluorescence and Confocal Microscopy-Immunofluorescence was performed as previously described (37). Basal membranes or the central plane of cells were observed by confocal microscopy using a Zeiss LSM510 Meta microscope with a 63ϫ oil immersion lens, and images were acquired and processed using LSM510 software (Carl Zeiss) and ImageJ (National Institutes of Health).
RNA Extraction-Rat primary sorted beta cells were detached from culture dishes using QIAzol and centrifuged; the supernatant was removed, and the cell pellet was disrupted in RLT buffer (RNeasy; Qiagen). Total RNA was prepared according to the standard RNeasy protocol.
Preparation, Sequencing, and Mapping-Total RNA libraries were constructed following customary Illumina TruSeq protocols for next generation sequencing. Poly(A)-selected mRNA was purified, size-fractioned, and subsequently converted to single-stranded cDNA by random hexamer priming. Following second strand synthesis, double-stranded cDNAs were blunt end-fragmented and indexed using adapter ligation, after which they were amplified and sequenced according to protocol. RNA libraries were 49-bp paired-end sequenced with one or a maximum of two samples per HiSeq 2000 lane. Standard quality checks for material degradation (Bioanalyzer; Agilent Technologies, Wokingham, UK) and concentration (Qubit; Invitrogen) were done before and after library construction, ensuring that samples were suitable for sequencing. Paired end reads were mapped to the rat genome (version r5 of Rattus norvegicus) employing the rna-pipeline from GemTools, version 1.6.2. 3 The GEM mapping suite (38) was used first to map and subsequently to split map all reads that did not map entirely. The mapping pipeline and settings can be found online. For gene and transcript quantifications, we used Cufflinks software (version 2.1.1 (39,40)) with the default parameters, using Ensembl gene annotation Rnor_5.0.73 as a reference annotation. Gene expression values were expressed as values of reads per kilobase of transcript per million mapped reads (RPKM). The number of uniquely mapped reads counts was extracted with htseqcount script (htseq).
Differential Expression Analysis and Statistical Significance-Differential expression analysis and statistical significance were evaluated with GraphPad Prim 5 software by one-way ANOVA when more than two experimental conditions were compared. All the three different conditions were grouped by day of experiment and tested for differential expression: control versus low concentration of IL-1␤, control versus high concentration of IL-1␤, and low concentration versus high concentration of IL-1␤. p values less than 0.05 were considered significant.
Functional Annotation Analysis-Gene functional analyses were performed using The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7. The functional annotation chart tool was used to analyze genes of interest (e.g. low concentration of IL-1␤ specific) in the context of the whole rat genome. The DAVID default annotation categories were used, and enrichment results were sorted by significance of p values. The Kyoto Encyclopedia of Genes and Genomes database was used to determine molecular network between genes of interest.

Low but Not High Dose of IL-1␤ for 2 h Increased Glucosestimulated Insulin Secretion-
To determine the effect of IL-1␤ on insulin secretion, we treated sorted rat primary beta cells plated on extracellular matrix from 804G cells, with different cytokine concentrations (0.1, 2, and 20 ng/ml) for 2 h. Treatment with 0.1 and 2 ng/ml IL-1␤ increased glucose-stimulated insulin secretion (GSIS) by 68.58 Ϯ 16.1% (p ϭ 0.0024) and 60.75 Ϯ 12.6% (p ϭ 0.0061), respectively, whereas 20 ng/ml had no significant effect (Fig. 1A) and neither did a lower dose of 0.01 ng/ml (data not shown). Moreover, basal secretion and insulin content (data not shown) was unchanged between all the different conditions. Similar results were obtained in rat whole islets using only the lowest and highest dose of IL-1␤ (Fig.  1B). As a proof of concept, we performed the same experiments on human islets with very similar results (Fig. 1C): 0.1 ng/ml IL-1␤ increased significantly GSIS by 74.99% Ϯ 20.54% (p ϭ 0.0014), whereas 20 ng/ml was without effect. Further control studies using sorted rat beta cells indicated that 0.1 ng/ml IL-1␤ was without effect on GSIS if added to high glucose for only 15 min (Fig. 1D) or 1 h (Fig. 1E), and neither was there any effect of this concentration of IL-1␤ on basal secretion over 2 h (Fig. 1F).
Low Dose of IL-1␤ Affected FA Morphology and Actin Organization-Because it has been observed that IL-1␤ is involved in actin (33) and FA remodeling (34) in other cell types, we hypothesized that IL-1␤ treatment of beta cells could similarly have an effect on these key events previously implicated in insulin secretion (30,31). Confocal immunofluorescence studies of the basal membrane of glucose-stimulated rat primary beta cells revealed that exposure to 0.1 and 2 ng/ml IL-1␤ increased the percentage of actin and PXN-containing FAs with a length superior or equal to 2.5 m, with an average FA length of 2.82 Ϯ 0.16 m (0.1 ng/ml IL-1␤) and 2.83 Ϯ 0.14 m (2 ng/ml IL-1␤) compared with 1.91 Ϯ 0.08 m for the control condition (Fig. 2, A and B). No significant effect on FA length was observed with 20 ng/ml IL-1␤ treatment (Fig. 2, A and B), and there was no significant impact on FA number by any concentration of IL-1␤ (Fig. 2D). In the basal condition (2.8 mM glucose), no effect of 0.1 ng/ml IL-1␤ was observed on FA morphology or number (Fig. 2E); nevertheless actin organization was modified with the formation of small protrusions and numerous thin filaments as previously observed in fibroblasts (33). As demonstrated in Fig. 1, the lowest dose of IL-1␤ enhanced insulin secretion in a glucose-dependent manner that correlates with these results obtained on FA remodeling. High glucose is thus necessary for IL-1␤-induced FA remodeling; however, glucose stimulation seems to be not necessary for IL-1␤-induced actin cytoskeleton remodeling.
Low Dose of IL-1␤ Induced FA Remodeling Independently of PTK2-To gain insight into a possible impact of IL-1␤ on FA protein regulation, we next studied the phosphorylation status of PTK2 (Tyr 397 ) and two of its downstream targets, PXN (Tyr 118 ) and ERK1/2, that are both known to be phosphorylated in response to short term glucose stimulation and necessary for GSIS (31). Beta cells were treated for 1 h with low glucose followed by 15 min in 16.7 mM glucose in the continued presence or absence of 0.1 or 20 ng/ml IL-1␤. There was no change in total protein levels; however, the low dose of IL-1␤ induced an increase of phospho-PXN and -ERK1/2, whereas 20 ng/ml IL-1␤ did not modify glucose-induced phosphorylation of either proteins. Surprisingly, glucose-induced PTK2 phosphorylation was not significantly affected by 0.1 ng/ml IL-1␤, suggesting a potential regulation of the FA signaling pathway independently of PTK2 (Fig. 3A).
To investigate further whether the potentiation of GSIS by 0.1 ng/ml IL-1␤ was independent of PTK2, we performed experiments using compound Y15, which specifically inhibits PTK2-Tyr 397 autophosphorylation and phosphorylation of its downstream targets, PXN and ERK1/2, in the glucose-stimulated condition in beta cells (31). First, Y15 alone decreased the number of phospho-PXN-containing FAs at the basal membrane of rat primary beta cells following glucose-stimulation, as previously shown (31), with a morphology similar to the 2.8 mM glucose condition (basal). Interestingly, despite pretreatment with Y15, low dose IL-1␤ still induced FA remodeling with a significant increase of FA length compared with Y15 alone or control condition (Fig. 3B). Finally, we investigated the effect of inhibition of PTK2 in presence of IL-1␤ on insulin secretion and protein phosphorylation, using the standard conditions of 1 h at 2.8 mM followed by 1 h at 16.7 mM glucose, in the continued presence of IL-1␤ and/or Y15. As shown in Fig. 3C, Y15 alone inhibited GSIS by 32.49 Ϯ 7.4% (p ϭ 0.038) in confirmation of our previous study (31). Although 0.1 ng/ml IL-1␤ was still able to enhance GSIS in combination with Y15 ( Fig. 3C), the absolute amount of insulin secreted was comparable to the control condition (without cytokine or inhibitor). We next examined whether IL-1␤-induced phosphorylation of PXN and ERK required upstream PTK2 activation. Beta cells treated with Y15 displayed significantly decreased glucose-induced phosphorylation of PXN, ERK1/2, and PTK2 as shown previously (31). The effect of Y15 on IL-1␤-induced phosphorylation of PXN and ERK (Fig. 3D) was quite similar to that seen for GSIS, but there was no significant effect of IL-1␤ on PTK2 phosphorylation either in the presence or absence of Y15. Taken together, these data indicate that IL-1␤ induced-FA remodeling and potentiation of GSIS are independent of PTK2.
IL-1␤ Enhances Insulin Secretion Stimulated in Response to Elevated Cytosolic Ca 2ϩ -GSIS is triggered by an increase in beta cell cytosolic free Ca 2ϩ concentration, which increases the number of granules undergoing exocytosis. This rise in cytosolic Ca 2ϩ stems from glucose metabolism, closure of ATPsensitive potassium channels, depolarization, opening of voltage-gated calcium channels, and calcium influx (41). KCl causes depolarization, thereby increasing cytosolic Ca 2ϩ to trigger insulin secretion (42). To explore whether KCl-induced insulin secretion could also be potentiated by IL-1␤ treatment, rat beta cells were stimulated for 15 min with 30 mM KCl at the end of 2-h treatment with 0.1 or 20 ng/ml IL-1␤. Only the low dose of IL-1␤ significantly increased KCl-induced insulin secretion ( Fig. 4A) just as observed for enhancement of GSIS (Fig. 1A). The shift of equilibrium potential of K ϩ (E k ) to less negative values after adding 30 mM KCl caused a sustained membrane depolarization of the beta cell (43). A low dose of IL-1␤ could modify this membrane potential value, which triggered the increase of intracellular Ca 2ϩ level and insulin secretion. To assess whether the IL-1␤ effect on insulin secretion requires Ca 2ϩ entry, beta cells were treated with diazoxide, known to block GSIS by opening ATP-sensitive potassium channels and thereby preventing glucose from depolarizing the beta cells membrane and the downstream events leading to increased cytosolic Ca 2ϩ and exocytosis (43,44). Diazoxide abolished GSIS in our pretreated cells even in the presence of a low dose of IL-1␤ (Fig. 4B). Diazoxide also affected glucose-induced FA remodeling and blunted glucose-in-duced newly formed protrusions containing phospho-PXN with a morphology similar to the basal condition and significantly decreased FA length. IL-1␤ treatment was unable to prevent this effect of diazoxide on FA remodeling (Fig. 4C). Regarding these results, IL-1␤ enhancement of stimulated insulin secretion thus arises through mechanisms that could depend on increased cytosolic Ca 2ϩ , leading to FA and actin remodeling and not be due to nonphysiological cell death or lysis-related event.
Rapid Relocalization of NF-B to the Nucleus Induced by IL-1␤ Is Necessary for FA Remodeling and Insulin Secretion but Is Not Dependent on Actin Remodeling-The NF-B pathway is involved in the cytotoxic effects of IL-1␤ toward beta cells (2) but also in its beneficial effects (18,21). We therefore wished to determine whether the canonical NF-B pathway, with nuclear translocation of the p65 subunit (45), was activated by short times of exposure of beta cells to low concentrations of IL-1␤. To this end, the cellular localization of p65 subunit of NF-B in rat primary beta cells was observed by immunofluorescence after exposure to 0.1 ng/ml IL-1␤. As shown in Fig. 5A, in glucose-stimulated conditions, only 15 min of treatment with IL-1␤ was necessary to induce NF-B relocalization to the nucleus and was maintained after longer treatment up to 2 h (see also Fig. 5C). It was also apparent that IL-1␤-induced NF-B relocalization was independent of glucose stimulation (data not shown).
IL-1␤-induced actin reorganization has been shown to be implicated in NF-B relocalization to the nucleus in other cell types (46 -48). To test this, we treated primary beta cells with two agents that disrupt actin remodeling, jasplakinolide, which stabilizes actin filaments (49) and induces an increase of GSIS in mouse islets (50), and latrunculin B, which is known to depolymerize actin microfilaments and also increase GSIS from both cultured MIN6 and isolated rat islet cells (51). Both jasplakinolide and latrunculin B treatment led to actin disruption and also FA remodeling. However, IL-1␤ still induced NF-B nuclear localization despite such perturbations to the actin cytoskeleton, confirming that actin remodeling is not an upstream regulator of IL-1␤-induced NF-B nuclear localization in beta cells under the present experimental conditions (data not shown).
To explore whether NF-B could conversely be an upstream regulator of actin and FA remodeling, we used the compound   MARCH 6, 2015 • VOLUME 290 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 6659  previously shown to inhibit NF-B, GSIS, cell spreading, and actin cytoskeleton remodeling in rat primary beta cells (21). As shown in Fig. 5 (B and C), inhibition of the NF-B pathway by BAY 11-7082 prevented NF-B relocalization to the nucleus and altered the morphology of the beta cells by a complete disruption of actin and FA remodeling (Fig. 5C) and also decreased GSIS (Fig. 5B) despite the presence of IL-1␤. Indeed, inhibition of the NF-B pathway blunted PXN phosphorylation and its recruitment toward FAs, and IL-1␤ was unable to rescue glucose-induced FA and actin remodeling in the presence of this inhibitor (Fig. 5D). NF-B is thus an important regulator of beta cell function, and IL-1␤ treatment cannot prevent the multiple changes induced by BAY 11-7082.

Beneficial Effects of Interleukin 1␤ on Beta Cell Function
Potentiation of Insulin Secretion and FA Remodeling Induced by 2-h Exposure to 0.1 ng/ml IL-1␤ Is Mediated in Part through the Regulation of Gene Transcription-IL-1␤ is known to regulate gene transcription in many cell types (52). The rapid nuclear localization of the transcription factor NF-B seen after only 15 min of 0.1 ng/ml IL-1␤ treatment (Fig. 5A), with nuclear localization maintained at 2 h, suggested that modulation of transcription could contribute to the enhancement of GSIS seen after that time of incubation with the cytokine. To study this further, we treated rat primary beta cells with actinomycin D (ActD), an inhibitor of transcription (53), in the presence or absence of IL-1␤. ActD alone had no significant effect on GSIS (Fig. 6A). IL-1␤ was still able to increase GSIS in the presence of ActD (compare ActD versus ActD ϩ IL-1␤; Fig. 6A), but GSIS was nonetheless less than with IL-1␤ alone. This intermediary result suggests that IL-1␤ treatment only regulates GSIS partially through gene transcription. Interestingly, ActD completely prevented the IL-1␤ FA remodeling (Fig. 6B), suggesting that this arm of IL-1␤ action is dependent on gene transcription and is required for full enhancement of GSIS.
Because the IL-1␤ enhancement of GSIS was only partially mediated through gene transcription, we next tested the effect of cycloheximide, an inhibitor of translation. The results were quite similar to those obtained with ActD, with inhibition of translation preventing only partially the enhancement of GSIS induced by low dose IL-1␤ (Fig. 6C).
Transcriptome Profiling Revealed Differential Gene Expression Depending on the IL-1␤ Concentration-According to our results, low and high concentrations of IL-1␤ induce differential effects on beta cell function, and the lowest concentration regulates GSIS in part through gene transcription. We therefore went on to compare gene expression by RNAseq in rat primary beta cells exposed to 0 (control), 0.1 and 20 ng/ml IL-1␤ for 1 h in low glucose (2.8 mM) followed by 1 h at high glucose (16.7 mM), corresponding to the standard conditions used to study beta cell function (three independent experi-ments). First, clustering by the Pearson correlation distance method demonstrated that patterns of gene expression changes did not cluster between each experimental condition but only between independent experiments regardless of the condition (data not shown). This result was to be expected, given very low variation in expression induced by 2 h of treatment with either concentration of IL-1␤ for the vast majority of genes in the face of interexperiment variability typical of primary cells. One-way ANOVA statistical comparison between each condition, taking into account interexperiment variability, was used to identify genes regulated by IL-1␤. First, we classified genes with the most important increase or decrease in expression induced by both concentrations of IL-1␤ (Table 1). Cxcl1 expression was increased 75-and 68-fold by 2-h treatment with low and high dose IL-1␤, respectively. This was the largest increase in expression observed and is consistent with our previous study in rat beta cells showing that Cxcl1 expression is regulated in part by IL-1␤ in an autocrine manner and is NF-B-dependent (12). We also observed major increases in expression of Lif, Tnf, Egr4, Ccl20, Cxcl2, Atf3, and two unknown genes irrespective of the IL-1␤ concentration. These genes, except Egr4, are involved in immune and inflammatory processes (DAVID functional annotation analysis) and may set the stage for the deleterious effects of IL-1␤ expected after longer treatment (1) but not manifest after just 2 h. The four genes most down-regulated by both concentrations of IL-1␤ were Adamts15, Dnaaf2, Sox12, and Mllt11 (Table 1). None of these genes has been reported to be involved in beta cell function.
We have also identified using stringent one-way ANOVA statistics genes that are differentially regulated by the low dose of IL-1␤ and that could thus conceivably be implicated in the enhancement of GSIS by 0.1 ng/ml IL-1␤. We separated these genes into two groups ( Table 2). Group A genes are significantly up-or down-regulated to a significantly greater extent by 0.1 ng/ml than by 20 ng/ml IL-1␤ (p Ͻ 0.05 for control versus 0.1 ng/ml IL-1␤; 20 ng/ml IL-1␤ versus 0.1 ng/ml IL-1␤; control versus 20 ng/ml IL-1␤). There were five genes in this group, all of which were up-regulated by both concentrations of IL-1␤, but to a greater extent by the lower dose. Group B genes are significantly up-or down-regulated only by 0.1 ng/ml but not 20 ng/ml IL-1␤ (p Ͻ 0.05 for control versus 0.1 ng/ml IL-1␤; 20 ng/ml IL-1␤ versus 0.1 ng/ml IL-1␤ but not for control versus 20 ng/ml IL-1␤). There were 13 genes in this group, with just 2 up-regulated uniquely by 0.1 ng/ml IL-1␤ and 11 down-regulated selectively by the low dose of cytokine. Regarding the characterized genes, any clear relation and common cellular processes between all these genes were known in the literature. Atf4, Jun, Neurod1, and Mecom are related to gene transcription regulation processes, and Jun, Neurod1, Pcsk6, and Timp1 are related to the mechanisms involved in apoptosis regulation (Table 3). Kyoto Encyclope-  dia of Genes and Genomes pathway analysis revealed that MAPK pathway is linked to Atf4, Jun, and Mecom gene regulation. Interestingly, JUN is also linked to focal adhesion pathway (Table 4).
We have also documented changes in the expression of genes known to be regulated by IL-1␤ in other contexts/conditions or involved in the secretory function of the beta cell. Surprisingly, IL-1␤, IL-1r1, and IL-1r2 were not regulated by either the low or high dose of IL-1␤ after just 2 h of exposure, and neither were Pdx1, Glut2, and Vamp2 (Table 1). However, Cdc42 expression was increased, and Rab3a and Rhoa were decreased in the presence of both IL-1␤ concentrations (Table 1). Interestingly, two important genes involved in insulin granule trafficking, Snap25 and Scg2 (secretogranin II), were down-regulated only with the

RPKM values corresponding to gene expression
The RPKM values corresponding to gene expression were obtained from deep RNA sequencing of rat primary beta cells exposed to 0, 0.1, and 20 ng/ml IL-1␤ for 1 h at 2.8 mM glucose followed by 1 h at 16.7 mM glucose. The data are presented as means from three independent experiments with samples from each sequenced in triplicate. Genes are grouped in different cellular processes categories. p values were calculated using one-way ANOVA. CTL, control. high dose ( Table 1), suggesting that 2 h of treatment with 20 ng/ml IL-1␤ may already allow for the start of deleterious changes that will become manifest after longer times of exposure. Inspection of the expression of genes known to be regu-lated by NF-B showed up-regulation by both concentrations of IL-1␤ for some but only by 20 ng/ml for others (Table 1). Concerning genes involved in insulin secretion and related to FAs, Pxn, Vcn, and Tln2 gene expression did not change, RPKM values corresponding to gene expression obtained from deep RNA sequencing of rat primary beta cells exposed to 0, 0.1, and 20 ng/ml IL-1␤ for 1 h at 2.8 mM glucose followed by 1 h at 16.7 mM glucose. The data are presented as means from three independent experiments with samples from each sequenced in triplicate. Genes are grouped for statistically significant changes in expression for group A (control versus 0.1 ng/ml IL-1␤ and 0.1 ng/ml versus 20 ng/ml and control versus 20 ng/ml) and group B (control versus 0.1 ng/ml IL-1␤ and 0.1 ng/ml versus 20 ng/ml but NOT control versus 20 ng/ml). CTL, control.

TABLE 3 Gene ontology classification with David functional annotation analysis
whereas Itgb1 and Itga5/Icam were down-regulated and upregulated, respectively, by both concentrations of IL-1␤. Furthermore, there was a slight decrease of Ptk2 only with the high concentration, but the total protein levels remained unchanged (Fig. 3A), suggesting that sustained exposure to this high concentration of IL-1␤ may decrease PTK2 protein levels and impair FA remodeling (Table 1). For the interested reader, the complete data set providing RPKM values for all genes recorded in primary rat beta cells under the three experimental conditions is provided in supplemental Table S1. The Effects of IL-1␤ on FA Remodeling and GSIS Are Mediated in Part by the c-JUN/JNK Pathway-Transcriptomic analysis demonstrated a differential regulation of Jun gene expression depending of the IL-1␤ concentration (Table 2). Interestingly, there were similar changes at the protein level with a 8.69 Ϯ 0.77-and 6.28 Ϯ 0.59-fold increase of c-JUN protein levels by 0.1 and 20 ng/ml of IL-1␤, respectively, compared with the nontreated condition, with a significant difference between the low and high dose of IL-1␤ (p ϭ 0.0403) (Fig.  7A). Given these results and our previous observation of ERK activation by 0.1 ng/ml IL-1␤ (Fig. 3A), we examined the impact of pharmacological inhibition of these two key signaling pathways using the JNKII and the ERK inhibitor PD98059. In a first set of Western blot experiments (data not shown), both inhibitors were shown to be effective: PD98059 blunted IL-1␤-induced ERK phosphorylation, and JNKII decreased IL-1␤-induced c-JUN phosphorylation, known to be a downstream target of JNK. Unfortunately, we were not able to confirm directly the efficiency of the JNKII inhibitor using anti-phospho-JNK antibodies (from Cell Signaling Technology catalog nos. 9251 and 9255) that both gave too many unspecific bands to allow for any firm conclusion; therefore we used c-JUN phosphorylation as a surrogate marker to successfully validate the effectiveness of the JNKII inhibitor (data not shown).
Although GSIS was blunted by PD98059 on its own, 0.1 ng/ml IL-1␤ was still able to enhance GSIS in the presence of this ERK inhibitor (Fig. 7B). Interestingly, the presence of JNKII inhibitor alone had no effect on GSIS; however, it prevented the low dose IL-1␤-induced enhancement of GSIS (Fig. 7B). This suggests that IL-1␤ mediates its effect on insulin secretion in part through the c-JUN/JNK pathway. We performed immunofluorescence and Western blot studies to determine whether this pathway could also mediate IL-1␤-induced FA remodeling. As shown in Fig. 7C, JNK inhibition blunted glucose and IL-1␤induced actin and PXN recruitment toward protrusions. By contrast, ERK inhibition blunted glucose-induced FA remodeling, but a low dose of IL-1␤ restored the recruitment of phospho-PXN to protrusions, as well as actin remodeling, to form a thick network as observed in the control condition. Additional data concerning PXN phosphorylation showed that both ERK and JNK inhibition decreased significantly IL-1␤-induced PNX phosphorylation in the high glucose condition, however, to a higher extent in presence of JNKII (Fig. 7D). c-JUN/JNK pathway therefore seems to be an important mediator of the beneficial effect of IL-1␤ on GSIS, acting through the regulation of FA and actin remodeling circuits.

DISCUSSION
Here we report a selective potentiation of glucose-stimulated insulin secretion when rat primary beta cells or rat and human whole islets were exposed for just 2 h to a low (0.1 and 2 ng/ml) but not a high (20 ng/ml) concentration of IL-1␤. At the lowest effective concentration studied, 0.1 ng/ml, IL-1␤ enhanced GSIS and increased FA length and phosphorylation of key proteins involved in secretion including PXN (paxillin) and ERK in a NF-B-dependent manner and acted also through the c-JUN/ JNK pathway. We have also observed that cytosolic Ca 2ϩ influx mediates the IL-1␤ effect. This same low concentration of IL-1␤ also selectively altered the expression of 18 genes after 2-h treatment of primary rat beta cells. Our results using sorted primary beta cells further indicate for the first time that IL-1␤ has a direct beneficial effect on beta cell function, without the need for other islet cell types. We thus observed a similar enhancement of GSIS after 2 h of treatment with the lower

TABLE 4 Kyoto Encyclopedia of Genes and Genomes pathway analysis realized with David functional annotation analysis
The table represents a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis realized with David functional annotation analysis of our 18 genes specially regulated by the low concentrations of IL-1␤. Only five are classified because the others are not related to any of the known pathway in KEGG. Neurogenic differentiation 1 rno04950: maturity onset diabetes of the young concentrations of IL-1␤ in sorted rat primary beta cells and in rat or human whole islets. A dual effect of IL-1␤ on beta cell function and/or survival has been demonstrated previously and seems to be largely dependent on NF-B (9). Deleterious effects of high concentrations of the cytokine over several days on insulin secretion, cell proliferation and survival, and inflammation within islets have been carefully characterized (9). However, the molecular mechanisms involved in the beneficial effects on beta cells of lower concentrations of IL-1␤ maintenance are not so clearly under-stood, even if a role for the FAS/FLIP pathway has been postulated in increased beta cell proliferation evoked by very low concentrations of IL-1␤ over several days (12,19,20). In one very early study (24), rat islets were exposed for up to 2 h to a low concentration of IL-1␤, and the enhancement of GSIS with no change in basal insulin secretion was similar to the present data. The authors attributed this to an increase in glucose oxidation and suggested that mitochondria were a prime target for IL-1␤ in the beta cell. Increased metabolic flux through the oxidative pathway would be expected to lead to increased cyto- solic Ca 2ϩ and in turn exocytosis, whereas another group postulated a role for phosphoinositide hydrolysis in increased exocytosis in response to IL-1␤ (29). Consistent with our data, the former hypothesis concerning the significance of cytosolic Ca 2ϩ in the observed beneficial IL-1␤ effect was confirmed in the present study and is necessary to allow for IL-1␤ potentiation of GSIS. Despite much early interest and speculation on the mechanism(s) underlying the enhancement of GSIS by low concentrations of IL-1␤ (27), there has subsequently been much less attention paid to such beneficial effects than to the cytotoxic effects of prolonged exposure of islets to high concentrations of this cytokine and implications for diabetes therapy.

Gene ID
We also show for the first time that 2-h treatment with the lowest but not high concentrations of IL-1␤ improved insulin secretion, as well as inducing an increase of FA length and phosphorylation of PXN and ERK independently of PTK2. Studies in other cell types have reported that IL-1␤ induces FA maturation through ␤-integrin activation necessary for downstream activation of ERK (34,54). Interestingly, we have observed that IL-1␤ did not change glucose-induced phosphorylation of PTK2, known to be a major upstream regulator of PXN and important for insulin secretion (31). Tyrosine phosphorylation of PXN is observed following integrin-dependent cell adhesion to extracellular matrix and triggers different cellular functions. However, PTK2 is not the only kinase responsible for PXN phosphorylation and activation (55). In fibroblasts, in the absence of PTK2, Src induces phosphorylation of PXN (Y118) and controls migration (56) that in some respects depends upon the same cellular events as exocytosis. PXN can be also directly phosphorylated by JNK leading to modification of FA size, distribution, and rate of assembly in fibroblasts (57).
Unfortunately, despite the use of deep RNA sequencing of purified beta cells and our best attempts to restrict our scrutiny to just those genes that were differentially regulated by the lowest but not the highest concentration of IL-1␤, it was not possible to see any obvious link between altered gene expression and enhanced GSIS. Transcriptome profiling in this experimental setting is quite challenging. First, there was the expected interexperimental variability inherent to studies on primary cells, in the face of small changes in the expression of most genes regulated uniquely by 0.1 ng/ml IL-1␤, as anticipated following just 2 h of incubation with the cytokine. Indeed, of the 15 genes regulated in this fashion (group B; Table 1) and expressed in the control situation with an RPKM of Ͼ0.1, no gene was up-or down-regulated by more than a factor of 2. Furthermore, IL-1␤ regulates gene expression through NF-B, a transcription factor that displays highly characteristic oscillations that in turn govern selective gene activation (58,59). Time is thus of the essence, with no steady state expected; capturing just one snapshot of gene expression at 2 h offers only a static representation of a plastic process. This already complicated scenario is further impacted by the very complex dose-and time-response relationships for IL-1␤ actions on the beta cell and by cross-talk between the NF-B and other signaling pathways, as demonstrated for example in muscle cells (60). A complete transcriptome analysis would have required multiple time points and additional IL-1␤ concentrations that would have been way beyond the scope of the present study. A final impor-tant consideration is that changes in mRNA levels, especially in a relatively short time frame, provide little insight into changes in the levels of the corresponding proteins driving cellular phenotype.
Despite these serious limitations, we have determined that c-Jun is significantly more up-regulated by low than high concentrations of IL-1␤ at both the mRNA and protein level. c-Jun is linked to the FA pathway using Kyoto Encyclopedia of Genes and Genomes pathway analysis (Table 4). Indeed, it has been shown that JNK, an upstream regulator of the protein c-JUN/ AP-1, phosphorylates and activates PXN and promotes its recruitment toward FAs to induce cell migration (61,62). Furthermore, an interesting study showed an important role of c-JUN in Src kinase protein regulation, which can phosphorylate directly PXN to induce maturation of FAs necessary for cell migration (63). Thus, the JNK/c-JUN pathway could regulate FA remodeling in a noninflammatory context through the direct phosphorylation and activation of PXN. Because we have observed no regulation of PTK2 phosphorylation by IL-1␤, the regulation of PXN through c-JUN/JNK underlies IL-1␤-induced FA remodeling independent of PTK2. Indeed, we have confirmed that the c-JUN/JNK pathway mediated the beneficial effect of IL-1␤ on insulin secretion through the regulation of actin and FA remodeling. In support of these results, 0.1 ng/ml IL-1␤ differentially regulated the expression of three genes coding for the transcription factors Jun, Atf4, and Mecom that are linked to the MAPK pathway composed in part by JNK, ERK, and p38 (Tables 3 and 4). Note, however, that Mecom was expressed at vanishingly low levels even in the control situation and even lower following 2 h with 0.2 ng/ml (RPKM 0.077 and 0.022, respectively). Our results support a complex role for ERK in IL-1␤-induced enhancement of GSIS; although ERK activation and presumably localization to FAs seem to be important for normal GSIS, there would seem to be parallel signaling pathways for enhancement of GSIS by IL-1␤ involving ERK but independent of FA remodeling.
Pcsk6 (PACE4) was down-regulated in the low IL-1␤ condition by ϳ30% with no significant change at 20 ng/ml. This gene encodes a member of the subtilisin-like proprotein convertase family, which is a heparin-binding protein localized previously to the ECM of HEK293 cells (64) and shown to mediate cancer cell migration through the regulation of matrix metalloproteinase proteins (65). Nothing is known about a potential role of this protein in beta cell function, but it might be linked to ECM proteins and FA protein regulation. Intriguingly it has been known for many years to be expressed most abundantly in neuroendocrine cells including the anterior pituitary (66).
In conclusion, this study sheds important new light on the beneficial short term effects on beta cell function of low concentrations of IL-1␤. We confirm the enhancement of GSIS first observed in the late 1980s and implicate both FA remodeling and transcriptional regulation as key modulators of this process. This must be taken into consideration when attempting to blockade IL-1␤ in the treatment of diabetes. Leaving aside the expected immunological consequences of severe abrogation of IL-1␤ signaling, this may further compromise insulin secretion in patients with severely damaged or destroyed beta cells.