PERK (EIF2AK3) Regulates Proinsulin Trafficking and Quality Control in the Secretory Pathway

The following antibodies were used: golgin 97 (Abcam), GM130 (BD Biosciences), FK2 (Biomol), HA (Cell Signaling), V5 (Invitrogen), insulin (Linco), proinsulin (Beta Cell Biology Consortium and Hytest), proinsulin C-peptide (Millipore), c-myc 9E10, glucagon (Santa Cruz), ERGIC-53 (p58), green fluorescent protein, α-tubulin (Sigma), calnexin, and Erp72 (Stressgen). The CT-A antibody was a gift from the Lencer Lab, and the C8 proteasome subunit antibody (AbC8) was a gift from the Monaco Lab. The in situ cell death detection kit, TMR Red (Roche), was used to detect transferase-mediated dUTP nick-end labeling (TUNEL) cells.

Vesicular stomatitis virus G-protein (ts045-VSVG-GFP) in pcDNA3.1 was provided by Jennifer Lippincott-Schwartz (41). Human wild-type proinsulin and the human C(B19)A proinsulin mutant in pcDNA3.1 were from Peter Arvan (42). The pCMVshort-EYFP-ATF6α in pEYFO-C1 was provided by Kazutoshi Mori (43), pCGN-HA-S2P was obtained from Ron Prywes (18), and C terminus deleted 9E10-tagged mouse PERK (DNPerk) in pcDNA was provided by David Ron (44). Transduction of cells with DNPerk reduces phosphorylation of eIF2α to 26% normal (13) similar to the reduction seen in Perk KO pancreata (9). Wild-type and mutant proinsulin genes were subcloned into pIRESbleo3 with a V5 epitope at the C terminus. The wild-type proinsulin-KDEL construct was generated by inserting a KDEL ER-retention sequence 3′ to the V5 epitope-tag. A small-interfering RNA (siRNA) directed against human Perk mRNA coding region nucleotides 2,237–2,255 was used to knockdown PERK (45) in human-derived cells lines. Standard transfection protocols were followed (31,45). AD293 and HepG2 cells were cultured in high-glucose DMEM and 10% FBS at 37°C in 5% CO₂. A short-hairpin RNA directed against the rat Perk mRNA (shPerk) and scramble, randomized sequence (shScram) were from Fumihiko Urano (Program in Gene Function and Expression, University of Massachusetts). The shPerk and shScram are stably integrated into the genome of INS1 832/13 β-cell lines and under the inducible regulation of doxycycline.

Proinsulin synthesis was determined as previously described (8). For protein synthesis islets or cultured cells were labeled with S³⁵ labeled Met/Cys (500 μCi/ml) at 37°C for 30 min, precipitated with 10% trichloroacetic acid. Radioactivity was measured by scintillation counting and normalized to protein content.

Chymotryptic activity of the proteasome in cell lysates was determined using the 20S Proteasome Activity Assay Kit (Chemicon) (46). The cleavage of LLVY-AMC and release of fluorigenic AMC was measured (360/460 nm) to determine relative proteasome activity. Retrotranslocation assays were performed as previously described (21). Cholera toxin (CT) A (CTA) and CTA1 CT subunits were detected by Western blot.

For trafficking assays ts045 VSVG-GFP was transfected into cells. At 8 h posttransfection, cells were transferred to the restrictive temperature of 40°C for 16 h followed by a shift to the permissive temperature of 32°C for 15 min (unless otherwise specified). Cells were then harvested for immunofluorescence, subcellular fractionation, and EndoH assays (47). The VSVG protein was detected using an antibody to the GFP-tag.

The Perk KO mice strains (9) were congenic for C57BL/6J or 129SvEvTac or were of mixed background, and each strain exhibited the defects described herein. To increase the Perk dosage in the β-cells of Perk KO mice, a Perk transgene under the control of the rat insulin promoter (βPerk) was introduced (7). Mice bearing the Akita mutation in a C57BL/6J background were from The Jackson Laboratories. These mice were crossed into Perk^+/−, βPerk mice to generate the strains used. The Akita mutants (C57BL/6J) were crossed with Perk^+/− in the 129SvEvTac background to generate Ins2^+/Akita, Perk^+/− double heterozygotes or Ins2^+/Akita, Perk^+/+ mice that were F1 hybrids. All other strains were in a mixed genetic background. All animal studies were approved by the institutional animal care and use committee of Pennsylvania State University.

Islets were isolated using a modified histopaque-1077 separation method (11). Islets were cultured in RPMI for 6–12 h at 37°C in 5% CO₂ prior to experimental processing.

For studies of in vivo glucose-stimulated insulin secretion (GSIS), serum was collected after a 16-h fast and 20 min after the intraperitoneal injection of 2 mg glucose/g body wt. Serum insulin concentrations were determined by immunoassay (Meso Scale Discovery). For total pancreatic insulin measurement, pancreata were sonicated in 1 ml of acid ethanol (95% ethanol:10.2N HCl in a 50:1 ratio), incubated overnight at 4°C, centrifuged and the supernatant assayed for insulin.

Statistical analysis was performed using Student t test, and P < 0.05 was accepted as significant. Error bars represent the SE of the mean.

In wild-type β-cells (Fig. 1,A and B), proinsulin was found to be localized in a perinuclear pattern within the Golgi (Fig. 1,C), as previously reported (5). A substantial fraction (ca 30–40%) of PERK-deficient β-cells exhibit an extremely distorted cellular morphology (Fig. 1,D–F), which we denote as an impacted-ER phenotype characterized by expanded ER and massive accumulation of proinsulin (Fig. 1,F, arrows) colocalizing with ER markers (11). Quantitative analysis of the signal intensity in wild-type, knockout nonimpacted-, and knockout impacted-ER β-cells revealed that insulin was not significantly different (Fig. 1 G). Proinsulin levels were not significantly different between wild-type and Perk KO nonimpacted-ER β-cells; however, the proinsulin level in the Perk KO impacted-ER β-cells is more than sevenfold higher compared with wild-type or Perk KO nonimpacted-ER β-cells. Developmental studies by us have suggested that the impacted-ER β-cells in Perk KO mice are derived from nonimpacted-ER β-cells (11).

To determine whether uncontrolled synthesis caused overaccumulation of proinsulin in the Perk KO impacted-ER β-cell, we examined protein or proinsulin synthesis rates in PERK-deficient cells. We found that global protein synthesis rates in INS1-832/13 (13) and AD293-expressing DNPerk, Perk KO islets (Fig. 1,H), and shPerk-expressing INS1-832/13 β-cells (Fig. 1,I) were not significantly different from wild-type controls. In addition, proinsulin synthesis is actually somewhat reduced in shPerk INS1-832/13 β-cells (Fig. 1,I). Therefore, the abnormal accumulation of proinsulin in the ER is not caused by uncontrolled or derepressed proinsulin synthesis and is not correlated with an increase in general protein synthesis. Moreover, insulin and proinsulin content in the Perk KO nonimpacted β-cells are not elevated (Fig. 1 G) as would be expected if proinsulin synthesis was derepressed.

Abnormal accumulation of proinsulin in the ER of PERK-deficient β-cells could be caused by defects in ER-Golgi trafficking, and to address this question we assessed the trafficking of the temperature-sensitive variant of the ts045 VSVG-GFP (14) in cells in which PERK activity was suppressed by a dominant-negative mutation of Perk (DNPerk). We found that suppressing PERK activity in HepG2 (Fig. 2,A–D), INS-1 832/13 (Fig. 2,E–H) and AD293 (Fig. 2,I–R, supplementary Fig. S1E, available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1064/DC1) cells blocked anterograde movement of ts045 VSVG-GFP from the ER to the Golgi compared with vector-transfected control cells (supplementary Fig. S1A–D). This trafficking defect was also confirmed using the EndoH-resistance assay in AD293 cells (Fig. 2 S), which showed an almost complete block in anterograde trafficking of VSVG following a shift to the permissive temperature.

To examine whether Golgi components were redistributed into the ER in PERK-suppressed cells as a consequence of the trafficking defect, we carried out subcellular fractionation (SCF) analysis. For this purpose, AD293 cells cotransfected with ts045 VSVG-GFP and DNPerk were incubated overnight at the restrictive temperature followed by a shift to the permissive temperature for 15 min. Equivalent numbers of cells were then harvested for SCF. In the empty-vector controls, VSVG was successfully trafficked from the ER-enriched fractions (fractions 7–9) to the ER-Golgi intermediate compartment (ERGIC) and partially into the Golgi (Fig. 3,A). In contrast, in the DNPerk cotransfected cells the VSVG cargo was retained within the ER-enriched fractions (fractions 7–9) (Fig. 3,B). In these cells, the marker for the ERGIC compartment was redistributed into the ER-enriched fractions, while the Golgi marker, golgin, showed very low levels of expression relative to the ER marker calnexin (Fig. 3 B).

Immunohistochemical (IHC) analysis revealed an impacted-ER phenotype due to abnormal ER retention of misfolded ts045 VSVG-GFP in AD293 cells, in which PERK activity was suppressed by DNPerk (supplementary Fig. S1E, arrowheads) (Fig. 2,C, G, and K) or by Perk siRNA–mediated knockdown (Fig. 2,M and P). In addition, the Golgi marker GM130 appeared to be dispersed throughout the cytoplasm (Fig. 2,B, F, J, N, and Q). Similarly, in Perk KO impacted-ER β-cells, GM130 was no longer condensed into perinuclear structures but instead was redistributed throughout the cytoplasm (Fig. 3,C–E). The localization of the ERGIC marker protein ERGIC-53 was examined to determine whether defects in the intermediate compartment were also present in Perk KO impacted-ER β-cells. We found that in wild-type β-cells the ERGIC compartment forms a previously unidentified structure that encircles a large concentration of proinsulin in wild-type β-cells (Fig. 3,F–H). However, in Perk-deficient impacted-ER β-cells, the ERGIC compartment was dispersed and was no longer in juxtaposition to proinsulin (Fig. 3,I–K, box). In addition, the level of proinsulin was reduced in the ERGIC compartment in the nonimpacted-ER β-cells (Fig. 3 I–K, arrows).

ATF6 positively regulates ERp72 and GRP78/BiP (15) and also represses insulin gene transcription (16) and therefore may be the cause of induction these two ER chaperones (13) in PERK-deficient β-cells as well as the reduced insulin gene expression that we reported previously (11). ATF6 is an ER resident protein that upon induction is trafficked to the Golgi where it is processed by the site 1 and 2 proteases (S1P and S2P, respectively) to generate the mature transcriptionally active form of ATF6 (17). We observed a constitutive reduction in full-length ATF6 and induction of the cleaved nuclear form (Fig. 4). In addition, S2P was found to be abnormally redistributed into the ER of PERK-suppressed cells (supplementary Fig. S2). These findings are consistent with previous reports that redistribution of Golgi proteases to the ER caused by a pharmacological block in trafficking can result in constitutive processing of ATF6 (18). Associated with the redistribution of Golgi components, we found substantial changes in the lipid composition of the ER and Golgi (supplementary Fig. S3A and B). Because overexpression of DNPERK protein may effect lipid composition and other associated defects in the ER for reasons other than suppressing PERK activity, these results need to be confirmed by ablating Perk via siRNA methods.

A large fraction of all newly synthesized cargo proteins in the ER fail to attain their native conformation (19), and these proteins undergo ERAD, which is comprised of three steps: ER-cytosolic retrotranslocation, ubiquitylation, and, finally, proteasome-mediated degradation (20). Retrotranslocation from the ER to the cytoplasm was examined by monitoring the fate of CT subunits targeted to the ER (21) (Fig. 5,A). Suppression of PERK in AD293 cells substantially reduced the amount of retrotranslocated CTA1 CT subunit, suggesting that Perk is necessary for retrotranslocation activity. ERp72, a protein disulfide isomerase-like chaperone, is localized in the ER and is a negative regulator of retrotranslocation (21). Suppression of PERK activity in AD293 cells led to a 3.3-fold induction in ERp72 protein levels (Fig. 5,B), and this was correlated with the localization of ERp72 to large balloon-like structures in impacted-ER β-cells (Fig. 1 D and E, arrowheads) seen in Perk KO mice (13).

To examine the fate of proinsulin in the ER, independent of anterograde trafficking, we expressed proinsulin containing an ER retention signal (Ins⁺-KDEL) (supplementary Fig. S4A and B). Following cotransfection of Ins⁺-KDEL with DNPerk into AD293 cells, we noted multiple high–molecular weight forms that we identified as ubiquitylated isoforms of proinsulin (Fig. 5,C). We also found a highly significant 3.6-fold increase in the accumulation of Ins⁺-KDEL in the PERK-suppressed cells compared with the empty-vector cotransfected controls (Fig. 5,D), while proinsulin and steady-state protein synthesis levels in these cells showed no significant changes (Fig. 1,H–I; supplementary Fig. S4C). This suggests that the abnormal accumulation of proinsulin can occur independent of both the anterograde trafficking defect and runaway protein synthesis. PERK-suppressed cells also showed a significant accumulation of the ubiquitylated proinsulin isoforms within the ER (supplementary Fig. S4D). This was seen when AD293 cells were cotransfected with DNPerk and Ins⁺-KDEL (55.3% increase) (Fig. 5 D, arrow) or wild-type proinsulin (62% increase) (supplementary Fig. S4E, arrow), suggesting that they are inefficiently degraded and retained within the ER.

An important step in proteasomal assembly occurs at the ER membrane through the action of proteasome maturation protein (POMP) to form the 20S core complex (22). We found that the amount of the 20S proteasome precursor was greatly diminished in the impacted-ER β-cells of Perk KO mice (Fig. 6,A–D), suggesting that assembly or synthesis of the proteasome may be negatively impacted. Additionally, proteasomal activity in islets isolated from neonatal Perk KO mice was found to be significantly repressed by 39% (Fig. 6,E) compared with wild-type littermates, although no reduction in proteasomal activity was found in PERK-suppressed AD293 cells. Moreover, both an acute ablation of Perk in INS1 (832/13) β-cells (13), as well as a treatment of these cells with the nonspecific proteasomal inhibitor MG132 (Fig. 6 F and G), phenocopies the accumulation of proinsulin in the ER that is indistinguishable from the impacted-ER β-cells of Perk KO mice and DNPerk-832/13 cells.

The Akita-Ins2 mutation (C96Y) in mice leads to the development of severe hyperglycemia by the postnatal age of 4–6 weeks (23), and the same mutation in the single human insulin gene results in permanent neonatal diabetes (24). Unexpectedly, we found that β-cell mass was expanded in Ins2^+/Akita heterozygous mice at the time these mice are rapidly progressing to frank diabetes (Table 1). However, proinsulin and insulin levels within the β-cells are dramatically reduced in both Ins2^+/Akita heterozygotes and Ins2^Akita/Akita homozygotes compared with robust expression of glucagon in neighboring α-cells (supplementary Fig. S5). We crossed the Ins2^+/Akita heterozygotes with mice that were also heterozygous for Perk and found that the double heterozygous (Ins2^+/Akita; Perk^+/−) mice exhibited a substantial delay in developing diabetes compared with the Ins2^+/Akita; Perk^+/+ mice (Fig. 7,A). Increasing Perk gene dosage, specifically in the β-cells of the Akita heterozygotes, by genetically introducing the wild-type Perk gene under the control of a truncated rat Ins2 promoter (βPerk) (7), had the opposite effect by further accelerating the progression to overt diabetes (Fig. 7,A). The more rapid progression to diabetes seen in the Akita homozygous mice was also delayed by reducing Perk dosage (Fig. 7,B). Total pancreatic insulin content was inversely correlated with serum glucose as a function of Perk dosage in the Ins2^+/Akita mice (Fig. 7,C compared with Fig. 7,A). Increasing Perk gene dosage in the β-cells of wild-type Ins2^+/+ mice (βPerk) also decreased pancreatic insulin content (Fig. 7,D) and in vivo GSIS (Fig. 7,E). In vivo GSIS and basal serum insulin was repressed in the Ins2^+/Akita mice independent of Perk genotype (Fig. 7,E). However, no significant increase was seen in apparent β-cell death in the Akita mice as assessed by TUNEL analysis (Table 1), nor was the proapoptotic Chop (GADD153) gene expression elevated. Even more surprisingly, β-cell mass and islet size was significantly expanded in Ins2^+/Akita mice (Table 1) during the time that these mice are rapidly progressing to frank diabetes.

The diabetes seen in PERK-deficient humans and mice was speculated to be caused by dysfunctions in the ER stress response, including derepressed protein synthesis (8,9). However, we found that protein synthesis and proinsulin synthesis is not derepressed in Perk-deficient islets or other cell types that we have examined (12,13), and the majority of β-cells in Perk KO mice have no increase in proinsulin and have a significant reduction in insulin content. Therefore, the observed accumulation of proinsulin in the PERK-deficient β-cells, which exhibit the impacted-ER phenotype, is not caused by runaway protein synthesis. Rather, we found severe ablation of ER-Golgi anterograde trafficking, disruption of the Golgi complex, reduced retrotranslocation of proteins out of the ER, and a decrease in proteasomal activity that together provide an explanation for the accumulation of proinsulin in the ER and insulin insufficiency.

Possible PERK-dependent factors that impact multiple ER functions are the chaperone/folding proteins GRP78/BiP, ERp72, and ERO1β. We found that BiP and ERp72 are substantially increased upon acute ablation of PERK in 832/13 β-cells (13) and exhibited enhanced expression in the impacted-ER β-cells in mouse islets, whereas ERO1β is reduced in islets (11,13). Modulating the expression of these ER chaperones can have a multiplicity of effects because these proteins are essential for folding, anterograde trafficking, and quality control (21,25,26). Increasing the expression of BiP may result in enhanced retention of client proteins in the ER by impeding anterograde trafficking and secretion (25). Elevated expression of ERp72 also increases ER retention and impedes the retrotranslocation step in ERAD (21). Genetic ablation of ERO1β has recently been shown to negatively impact glucose homeostasis (27). We speculate that PERK may regulate the competitive balance among ER chaperones that promotes either anterograde trafficking or ERAD of client proteins.

Two findings support the hypothesis that PERK primarily regulates ERAD. First, by tagging proinsulin with an ER retention signal we reduced the possibility of anterograde trafficking. Yet, we found that proinsulin still accumulated at much higher levels in the ER when Perk was impaired, suggesting that accumulation of proinsulin in Perk KO β-cells is caused by a failure in ERAD rather than anterograde trafficking. Second, the impact of altering Perk gene dosage in the Akita mouse suggests that Perk promotes ERAD. We found that decreasing Perk gene dosage by half ameliorates the progression of the Akita mouse to overt diabetes, whereas increasing Perk gene dosage in the β-cells of the βPerk transgenic mouse hastens progression to diabetes. Initially, it was proposed that increased β-cell death was the cause of the dominant Akita phenotype (28,29). However, we found substantially higher β-cell mass in Akita mice compared with wild-type at the onset of diabetes, and we failed to see an increase in TUNEL-positive β-cells. Although β-cell mass is not reduced at the onset of diabetes in Akita mice, insulin secretion is severely reduced (23). We speculate that the observed increase in β-cell mass is due to a compensatory response to reduced insulin secretion. Recent studies (30,31) have suggested that the Ins^Akita allele acts dominantly to enhance degradation of both the Akita and wild-type proinsulin via the ERAD pathway, and we propose that Perk acts to modulate the level of wild-type proinsulin degradation stimulated by the Ins^Akita mutation. Under this hypothesis, reducing Perk gene dosage decreases ERAD, thus sparing the degradation of wild-type proinsulin; whereas, increasing PERK dosage enhances the degradation of wild-type proinsulin, thus reducing secreted insulin even further.

Two ERAD processes were found to be impaired in Perk-deficient β-cells, retrotranslocation and proteasomal degradation, which raises the question if they are interconnected. Although the proteasome is localized in the cytoplasm, the 20S precursor is assembled at the ER membrane (22). ER dysfunctions or stress can impede proteasomal degradation of proteins (32), and mutations in proteins that regulate retrotranslocation can also impair general proteasomal activity (33). In addition to reduced general proteasomal activity, the 20S proteasome precursor was greatly diminished in the impacted-ER β-cells of Perk KO mice (Fig. 6 E), suggesting that assembly or synthesis of the proteasome may be negatively impacted. Inhibition of proteasomal activity by MG132 also phenocopied the accumulation of proinsulin in the ER observed in Perk-deficient β-cells, consistent with the hypothesis that Perk may positively regulate ERAD.

Perk deficiency also results in dramatic changes in organelle structure and ER membrane composition in a substantial fraction of the β-cells. The Golgi apparatus, which normally displays a compact perinuclear localization, is dispersed and no longer distinct from the ER. The activation of the transcription factor ATF6 is also dependent upon the maintenance of ER and Golgi integrity, and we found that that the processed nuclear form of ATF6 was constitutively expressed in PERK-deficient β-cells, which is most likely due to mislocalization of the Golgi-processing enzymes S1P and S2P into the ER. This is consistent with the role of ATF6 as a negative regulator of insulin gene transcription (16,34) and reduced insulin mRNA levels in PERK-suppressed β-cells in culture (13) and in neonatal mice (11).

We have previously reported that Perk-deficient mice have a low β-cell mass due to decreased levels of β-cell proliferation during the critical late-embryonic to early neonatal stages of life (11). Moreover, β-cell proliferation is also reduced in insulinomas of Perk-deficient mice (35). How might the defects in ER functions and organelle integrity impact β-cell proliferation? First, normal proteasomal activity is necessary for cell-cycle progression (36), and we found that proteasomal activity was reduced. Second, the integrity of the ER and Golgi are normally maintained at all stages of mitosis (37). The disruption in the normal subcellular organization of these two organelles along with the defect in proteasomal activity (36) likely causes the reduced proliferation of β-cells in Perk-deficient mice.

We speculate that the major function of PERK in the β-cells is to act as a metabolic sensor that regulates proinsulin degradation in concert with metabolic demand for insulin. PERK enzymatic activity is negatively regulated by GRP78/BiP, when bound to calcium (38). Disassociation of GRP78/BiP-Ca²⁺ from PERK results in dimerization, autophosphorylation, and activation of its catalytic activity to phosphorylate eIF2α (38,39). PERK activity can be activated by dynamic changes in ER calcium that occur during fasting and glucose-stimulated insulin secretion without ever causing an apparent ER stress response (9,40). Following a brief fasting period, the activation of PERK is increased, whereas treatment of islets in culture with high glucose represses PERK's activity. We suggest that the function of PERK is to regulate proinsulin turnover in response to reduced nutrient intake when the demand for insulin is reduced.

This work was supported by the National Institutes of Health Grants DK-062049 and GM-056957; the Pennsylvania Department of Health, Tobacco Settlement Funds (to D.R.C.); and a predoctoral fellowship from the American Heart Association (to S.G.).

No potential conflicts of interest relevant to this article were reported.

S.G., B.M., and D.C. wrote the manuscript and researched data.

We thank Kaori Iida, Rong Wang, Hanna Xu, Ajin Wang, Wenli Hu, and Siying Zhu for technical assistance. We thank Fumihiko Urano and Rajarshi Ghosh for the shPerk and shScramble 832/13 β-cell lines, Christopher Newgard for the INS1-832/13 cell line, Wayne I. Lencer (Harvard Medical School) for the CT-A antibody, J.J. Monaco (University of Cincinnati) for the C8 antibody, Jennifer Lippincott-Schwartz (National Institutes of Health) for the ts045 VSVG-GFP construct, Peter Arvan and Ming Liu (University of Michigan) for the (human) wild-type proinsulin and the (human) C(B19)A mutant proinsulin constructs, Kazutoshi Mori (Kyoto University) for the pCMVshort-EYFP-ATF6α construct, Ron Prywes (Columbia University) for the pCGN-HA-S2P construct, and David Ron (Skirball Institute) for the C terminus deleted 9E10 tagged mouse PERK (dominant negative) construct. We also acknowledge the assistance of the Melissa Rolls Lab (Penn State) with confocal microscopy and the Kouacou Konan Lab (Penn State) with subcellular fractionation experiments.