Insulin secretory granules control autophagy in pancreatic β cells

See allHide authors and affiliations

Science  20 Feb 2015:
Vol. 347, Issue 6224, pp. 878-882
DOI: 10.1126/science.aaa2628

Too hungry to eat, too hungry not to eat

Pancreatic beta cells, the source of insulin in response to food, employ an unusual mechanism to adapt to nutrient depletion. Goginashvili et al. found that starvation of beta cells induced selective degradation of newly formed insulin granules through their fusion with lysosomes, the cell's garbage disposal units (see the Perspective by Rutter). The nutrient sensor mTOR is recruited to these lysosomes, leading to its local activation and the suppression of autophagy—a process by which cells “eat” their own constituents. Protein kinase D, a major regulator of insulin granule biogenesis, controls this granule degradation in response to nutrient availability. Thus, unlike most other cells, autophagy is not the strategy of choice in beta cells to adapt to starvation.

Science, this issue p. 878; see also p. 826


Pancreatic β cells lower insulin release in response to nutrient depletion. The question of whether starved β cells induce macroautophagy, a predominant mechanism maintaining energy homeostasis, remains poorly explored. We found that, in contrast to many mammalian cells, macroautophagy in pancreatic β cells was suppressed upon starvation. Instead, starved β cells induced lysosomal degradation of nascent secretory insulin granules, which was controlled by protein kinase D (PKD), a key player in secretory granule biogenesis. Starvation-induced nascent granule degradation triggered lysosomal recruitment and activation of mechanistic target of rapamycin that suppressed macroautophagy. Switching from macroautophagy to insulin granule degradation was important to keep insulin secretion low upon fasting. Thus, β cells use a PKD-dependent mechanism to adapt to nutrient availability and couple autophagy flux to secretory function.

Upon feeding, the pancreatic β cell catabolizes nutrients to secrete insulin. Conversely, insulin secretion decreases upon fasting (1). A shortage of nutrients also induces macroautophagy (hereafter referred to as “autophagy”). During autophagy, cellular components are sequestered into autophagosomes and degraded upon targeting to lysosomes (autolysosomes). Resulting catabolites maintain cells in a metabolically active state and ensure cell survival (2). However, catabolism of nutrients could also trigger insulin release, which should be avoided during fasting. Autophagy might thus be a bad strategy for a β cell to cope with nutrient deprivation. Both induction and lack of induction of autophagy have been reported in fasted β cells (3, 4). We thus wanted to assess the importance of autophagy in fasted β cells.

Microtubule-associated protein 1 light chain 3 B (LC3B) incorporates into membranes of autophagosomes (5). Unexpectedly, INS1 cells (a rat insulinoma–derived β cell line) exogenously expressing LC3B tagged with green fluorescent protein (GFP) (LC3B-GFP) deprived of serum and amino acids or glucose (Glc), but not of serum alone, decreased LC3B-GFP puncta (fig. S1, A to C). Conversely, starved human embryonic kidney (HEK) 293 cells increased LC3B-GFP puncta (fig. S1, D and E) (6). INS1 cells endogenously expressing LC3B-GFP (INS1LC3B-GFPendo cells) (fig. S2, A to C) treated without or with bafilomycin A1 (BafA1), which blocks autolysosomal degradation of LC3B (7), decreased LC3B-GFP puncta upon starvation (Fig. 1A and fig. S2, D to F). Tandem RFP-GFP–tagged LC3 (ptfLC3; RFP, red fluorescent protein) allows for discrimination of yellow fluorescent autophagosomes and red fluorescent autolysosomes (8). Multiple RFP-GFP puncta were converted into RFP-only puncta upon the onset of starvation. A reduction in RFP puncta and no reappearance of RFP-GFP puncta were observed over time (fig. S3, A and B). Correlative light and electron microscopy (CLEM) confirmed their autophagic origin (Fig. 1B). Moreover, lipidated autophagosomal LC3B (LC3B-II) decreased in starved INS1 cells in the absence and presence of BafA1 (Fig. 1C). In contrast, starvation increased LC3B-II in HEK293 cells (fig. S4). p62 binds polyubiquitinylated substrates for targeting to autophagosomes (9). p62 was moderately increased during starvation, independent of BafA1 (Fig. 1C). p62 puncta clustered and increased in size (fig. S5A). Accordingly, p62/LC3B-GFP colocalization was decreased, indicating suppressed autophagy-dependent clearance of p62. ATG16L1 binds to preautophagosomes (10). ATG16L1- and ATG16L1/LC3B-GFP–positive puncta moderately decreased upon starvation (fig. S5B). Quantitative electron microscopy (QEM) confirmed reduced autophagic compartments in β cells of starved murine primary islets (fig. S6). Moreover, in LC3B-GFP–expressing mice (11), autophagosomes decreased in β cells upon fasting (Fig. 1D and fig. S7). What about other secretory cells, in which nutrients do not act as secretagogues? In contrast to β cells, primary murine plasma cells induced autophagy upon starvation, whereas secretion of immunoglobulin G was reduced (fig. S8, A and B). Thus, nutrient-sensing β cells appear to use a distinct mechanism to overcome shortage of nutrients.

Fig. 1 Nutrient depletion suppresses autophagy in β cells.

(A) LC3B-GFP puncta in INS1LC3B-GFPendo cells under growing culture (GC) and no amino acids and fetal calf serum (AA/FCS) for 1 hour in the absence and presence of BafA1 (10 nM). DAPI, 4′,6-diamidino-2-phenylindole. (B) CLEM of ptfLC3-expressing INS1 cells under GC and no AA/FCS for 2 hours. Regions of interest (ROI) are indicated with labeled dashed squares. Yellow and black asterisks indicate autophagosomes and autolysosomes, respectively. EM, electron microscopy. (C) Immunoblot of LC3B and p62 using soluble and insoluble fractions of lysates of INS1 cells under GC and no AA/FCS for 1.5 hours, nontreated or treated with 1 nM BafA1 for the last hour of incubation. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as a loading control. (D) Immunofluorescence (IF) of LC3B-GFP puncta (white arrowheads) and insulin (red) in β cells in islets of fed and fasted LC3B-GFP–expressing mice. The nuclei were stained with DAPI.

Targeting of secretory granules (SGs) to lysosomes might be an alternative strategy (12). Lysosomal Lamp1 and SG protein Phogrin colocalized close to the Golgi upon starvation in INS1 cells. Lysosomal inhibitors (LIs) increased their colocalization (fig. S9, A to C). QEM, immunogold labeling, and CLEM revealed abundant granule-containing lysosomes (GCLs) (fig. S10, A to D). We next used density gradients to purify GCLs (see supplementary materials and methods). Upon starvation, Phogrin- and lysosomal Lamp2-positive signals shifted to heavier fractions containing GCLs. LC3B was almost undetectable in shifted fractions, suggesting independence of GCLs from autophagy (Fig. 2A). Accordingly, starvation did not increase LC3B-GFP/Phogrin colocalization (fig. S11). Moreover, inactivation of autophagy did not change the amount of GCLs in starved cells (fig. S12, A and B).

Fig. 2 Nutrient depletion induces SINGD in β cells.

(A) Immunoblot of Lamp2, Phogrin, and LC3B using lysates of indicated fractions from INS1 cells under GC and no AA/FCS for 30 min. EM of GCLs in shifted fractions is indicated by red dashed boxes. (B) Immunoblot of proinsulin using lysates of INS1 cells treated or not with LIs under GC and no AA/FCS for 6 hours. GAPDH was served a loading control. (C) EM of Golgi areas in primary murine islets under GC and no AA/FCS for 2 hours. Yellow asterisks indicate GCLs. ROI are indicated with dashed squares. (D) IF of (Pro)insulin and Lamp2 (left) and (Pro)insulin and LC3B-GFP (right) in β cells in islets of fed and fasted LC3B-GFP expressing mice. The nuclei were stained with DAPI.

Secretory granules are generated at the Golgi. Proinsulin, a marker for nascent SGs (13), markedly decreased upon starvation (Fig. 2B). Its reduction was partially restored by LIs. Abundant GCLs were confirmed ex vivo by QEM of starved primary murine islets (Fig. 2C and fig. S13A). In fasted GFP-LC3B–expressing mice, Lamp2 and proinsulin/insulin [(Pro)insulin] colocalization increased, whereas there was no increase in LC3B-GFP/(Pro)insulin colocalization (Fig. 2D and fig. S13, B and C). Thus, β cells employ starvation-induced nascent granule degradation (SINGD) instead of autophagy during fasting.

Lysosome-derived amino acids induce translocation of mechanistic target of rapamycin (mTOR) complex 1 (mTORC1) to lysosomal membranes, mTORC1 activation (14, 15), and subsequent suppression of autophagy (1618). The mTOR inhibitors rapamycin and torin 1 increased the number of LC3B-GFP puncta in starved INS1 cells (Fig. 3A). Moreover, starvation localized mTOR to Phogrin/Lamp1-positive puncta close to the Golgi (Fig. 3B). mTORC1 suppresses starvation-induced autophagy through phosphorylation of Unc-51–like kinase 1 (ULK1) (19). Indeed, phospho-ULK1 remained high upon starvation and was abolished by rapamycin, whereas mTORC1-mediated phosphorylation of S6K1 (20) diminished during starvation (Fig. 3C). Accordingly, S6K1 phosphorylation was shown to require higher mTOR activity than ULK1 phosphorylation (21). Moreover, starvation led to formation of large phospho-ULK1 puncta that colocalized with Lamp1/Phogrin (fig. S14, A and B). Thus, nutrient depletion leads to induction of SINGD that locally activates mTOR to suppress autophagy.

Fig. 3 SINGD suppresses autophagy in a mTOR-dependent manner to prevent insulin release.

(A) IF of LC3B-GFP puncta (white arrowheads) in INS1 cells under GC and no AA/FCS for 2 hours treated or not with 100 nM rapamycin or 250 nM torin 1, as indicated. Quantification of LC3B-GFP puncta per cell (error bars denote mean ± SEM). **P < 0.01. (B) IF of mTOR in INS1 cells coexpressing Phogrin-GFP and Lamp1-RFP under GC and no AA/FCS for 2 hours. White arrowheads in ROI (dashed squares) indicate colocalization of Phogrin-GFP with mTOR and Phogrin-GFP with Lamp1-RFP. A Golgi marker (Giantin) was used. (C) Immunoblot of indicated proteins using lysates of INS1 cells treated or not with 100 nM rapamycin under GC, no AA/FCS, or no Glc/FCS for 2 hours. GAPDH served as a loading control. DMSO, dimethyl sulfoxide. (D) Insulin in supernatants of human islets treated as indicated. Insulin concentrations are expressed as a percentage of insulin upon 16.7 mM stimulatory Glc (mean ± SEM). ***P < 0.001.

Fasting suppresses both insulin secretion and autophagy. Would insulin secretion increase if autophagy is induced upon starvation? Keeping autophagy high by rapamycin precluded suppression of insulin secretion in INS1 cells (fig. S15). We next used tat-beclin1 (22) to specifically trigger autophagy in β cells of murine primary islets upon nonstimulatory Glc (fig. S16, A and B). Tat-beclin1 increased insulin secretion to levels comparable with those observed upon stimulatory Glc (fig. S16C). Potassium channel closure that evokes calcium influx and exocytosis of SGs (1) was required, as the potassium channel activator diazoxide abolished this effect (fig. S16D). Insulin secretion was also potentiated by tat-beclin1 in murine islets at stimulatory Glc, which was blocked by diazoxide (fig. S16, E and F). Tat-beclin1 also increased insulin secretion from ex vivo fasted human islets (Fig. 3D and fig. S17, A and B). Thus, SINGD-mediated suppression of autophagy is important to keep insulin secretion low during fasting.

Protein kinase D (PKD) controls insulin SG biogenesis at the Golgi (23, 24). Thus, PKD inactivation could have an effect on the turnover of nascent SGs. Indeed, inhibition of PKD decreased the amount of proinsulin (Fig. 4A). Proinsulin biosynthesis was unchanged in PKD1-depleted cells (fig. S18, A and B). However, accumulation of newly formed insulin was decreased, suggesting enhanced degradation of de novo synthesized insulin (fig. S18C). QEM and immunogold analyses revealed abundant GCLs in the Golgi area (Fig. 4B and fig. S19, A and B), which was further corroborated by subcellular fractionation (fig. S19C). PKD inhibition increased Phogrin/Lamp1 colocalization, which was more prominent upon LI treatment (fig S19, D to F). Phogrin/LC3B-GFP colocalization remained unchanged (fig. S20). mTOR largely colocalized with Lamp1 (Fig. 4C), and phospho-ULK1 was increased (Fig. 4D) in PKD1-depleted cells. PKD1 knockdown decreased accumulation of LC3B-II in the presence of BafA1 (fig. S21A). Moreover, PKD inhibition decreased LC3B-GFP puncta in absence and presence of BafA1 (fig. S21B).

Fig. 4 PKD controls SINGD.

(A) Immunoblot of proinsulin using lysates of INS1 cells treated with CID755673 for the indicated times. GAPDH served as a loading control. (B) EM of Golgi areas in nonsilenced (NS) and PKD1-depleted (shPKD1) INS1 cells. Yellow asterisks indicate GCLs. ROI are indicated with dashed squares. (C) IF of mTOR and Lamp1 in nonsilenced and PKD1-depleted INS1 cells. White arrowheads indicate colocalization of mTOR with Lamp1. Nuclei were stained with DAPI. (D) Immunoblot of indicated proteins using lysates of nonsilenced and PKD1-depleted INS1 cells. GAPDH served as a loading control. (E) (Top) EM of Golgi areas of fasted β cells in primary islets of p38δ+/+ and p38δ−/− mice. Yellow asterisks indicate GCLs. (Bottom) EM of cytoplasm of fasted β cells in primary islets of p38δ+/+ and p38δ−/− mice. The yellow asterisk indicates an autophagosome. (F) Model linking SINGD, secretion, and autophagy. PM, plasma membrane.

If PKD controls SINGD, its activity should be regulated by nutrients. PKD activity at the Golgi decreased upon starvation (fig. S22, A and B). Loss of the p38δ kinase leads to activation of PKD in β cells (23). GCLs were markedly decreased in β cells of ex vivo fasted p38δ knockout (p38δ−/−) islets (Fig. 4E and fig. S23A). Accordingly, (Pro)insulin/Lamp2 colocalized to a lesser extent in β cells of fasted p38δ−/− mice (fig. S23, C and D), suggesting that high PKD activity prevents SINGD. In contrast, autophagic compartments were increased in fasted p38δ−/− β cells, indicating that SINGD-dependent suppression of autophagy was diminished (Fig. 4E and fig. S23B). Thus, PKD is a major regulator of SINGD and autophagy in β cells.

Altogether, inactivation of PKD evokes SINGD, localized activation of mTOR, and suppression of autophagy, which is critical to prevent insulin release during fasting. On the other hand, PKD activity is required for SG biogenesis (Fig. 4F). Because the nascent SGs are preferentially secreted (25), SINGD-mediated suppression of autophagy may represent an optimal strategy to counteract insulin secretion, at the same time providing sufficient nutrients. Because SINGD-mediated suppression of autophagy depends on abundant nascent SGs, their depletion will probably derepress autophagy without increasing insulin release. Timing when depletion occurs may vary substantially depending on the models and protocols used and may thus explain previously described inconsistencies (3). Accordingly, prolonged fasting of mice can induce autophagy in β cells (26). Our model could also explain the regulation of autophagy despite constitutively high adenosine monophosphate–activated protein kinase activity in β cells (27).

Our data show that triggering autophagy results in increased secretion of insulin. Although this should be avoided during fasting, it may be beneficial when insulin demands are high; for example, after a meal or in diabetes (28). The positive correlation between autophagy and insulin secretion may suggest an involvement of autophagy in postprandial insulin release, probably going beyond the widely established housekeeping role of autophagy.

Supplementary Materials

Materials and Methods

Figs. S1 to S23

References (2936)

References and Notes

  1. Acknowledgments: We thank H. de F. Magliarelli, H. Gehart, O. Sumara, G. Sumara, N. Djouder, E. Hafen, R. J. Loewith, J. Klumperman, E. Polishchuk, R. Polishchuk, and all current members of the Ricci laboratory for critical scientific inputs. We thank A. Hausser for providing G-PKDrep-live, G. Rutter for providing Phogrin-GFP, J. Hutton (deceased) and H. Davidson for an antibody against Phogrin, and J. Klumperman for an antibody against Lamp1. We thank M. Koch, N. Messaddeq, and C. Ruhlmann for support in imaging; N. Mizushima for providing GFP-LC3 transgenic mice; and P. Halban and J.-I. Miyazaki for sharing the MIN6B1 cell line. This work was supported by a European Research Council (ERC) starting grant (ERC-2011-StG, 281271-STRESSMETABOL) and a European Foundation for the Study of Diabetes/Lilly European Diabetes Research Programme grant. A.G. and Z.Z. were supported by the ERC grant, M.M. was supported by a fellowship of the Deutsche Forschungsgemeinschaft, A.P. was supported by an IGBMC Ph.D. fellowship, K.K. was supported by a fellowship (Bourse Régionale de Recherche from region Alsace and INSERM), I.S. was supported by the Action Thématique et Incitative sur Programme (ATIP)–Avenir program, L.C. and C. Spiegelhalter were supported by a Telethon Grant, and N.S. and Y.S. were supported by EMBL internal funding. A.G. and R.R. have applied for a patent that protects autophagy-inducing molecules to increase insulin release.
View Abstract

Navigate This Article