Asna1/TRC40 Controls β-Cell Function and Endoplasmic Reticulum Homeostasis by Ensuring Retrograde Transport

Secretory proteins (e.g., insulin) are inserted into the endoplasmic reticulum (ER) where they are posttranslationally modified, folded, and then trafficked further through the endomembrane system. If the protein load exceeds the protein folding capacity of the ER, unfolded and misfolded proteins accumulate within the ER, and ER stress develops, leading to activation of the unfolded protein response (UPR). During the development of type 2 diabetes (T2D), pancreatic β-cells initially compensate for insulin resistance successfully by increasing insulin biosynthesis and secretion. However, conditions that lead to sustained ER stress (i.e., prolonged and persistent insulin resistance and/or failure to reestablish proper ER homeostasis) are implicated in the deterioration of β-cell function and the development of overt diabetes (1–3). Thus, identification of key molecules and factors that ensure proper membrane trafficking and ER homeostasis, and thereby β-cell function and survival, is important to gaining insight into the etiology of T2D.

In yeast, the Guided Entry of Tail-anchored proteins (GET) pathway (i.e., the tail recognition complex [TRC] pathway equivalent) is associated with a broad range of phenotypes (4–9). The GET complex has been suggested to genetically associate with endomembrane trafficking pathways (10,11), and inactivation of the GET pathway results in ER stress and activation of the UPR (12). Mechanistic studies, primarily in cell-free systems, have suggested a role for Get3 and the mammalian homolog Asna1 (also known as TRC40) in delivering tail-anchored (TA) proteins for posttranslational insertion into the ER through the CAML/WRB receptor complex (13–16). In agreement with the proposed role for the GET/TRC pathway in membrane trafficking within the secretory pathway, key regulators of membrane-mediated transport and protein translocation (e.g., soluble NSF attachment protein receptors [SNAREs] such as Sec22b and Sed5 as well as Sec61β and RAMP4) have been proposed as protein clients for this pathway (17). Recently, Get3/Asna1 was also shown to function, under oxidative stress conditions, as a molecular chaperone that binds unfolded proteins to prevent their irreversible aggregation (18). In Caenorhabditis elegans, Asna1 function is required for larval growth and resistance to cisplatin, an oxidative stress–inducing anticancer drug (19,20).

A mechanistic role for Asna1 in mammalian cells, however, has not been functionally assessed in vivo because global inactivation of Asna1 in mice results in embryonic lethality (21). To explore a potential role for Asna1 in mammalian cells, we generated β-cell–specific Asna1 mutant mice, denoted Asna1^β−/− mice. Asna1^β−/− mice displayed pancreatic hypoinsulinemia, impaired insulin secretion, and early onset diabetes. β-Cells of Asna1^β−/− mice showed impaired retrograde transport, reduced insulin content, and ER stress. Moreover, we show that Retro-2–mediated pharmacological inhibition of retrograde transport per se in isolated islets and insulinoma cells leads to decreased insulin content and ER stress. Thus, in addition to identifying a role for Asna1 to ensure retrograde transport as well as insulin and ER homeostasis in β-cells, the findings provide independent evidence for a role for retrograde transport in regulating β-cell function.

A detailed description of the generation and genotyping of the conditional Asna1 allele is described in the Supplementary Data. Briefly, two loxP sites flanking exon 2 of Asna1 were inserted by a recombination strategy essentially as previously described (22). CRE recombinase–mediated deletion of the intervening exon 2 is predicted to result in translational termination after exon 1. ERAI mice (23) were provided by the RIKEN BioResource Center through the National BioResource Project of the MEXT, Japan. The animal studies were approved by the Institutional Animal Care and Use Committee of Umeå University and were conducted in accordance with the guidelines for the care and use of laboratory animals.

Glucose tolerance test (GTT) and glucose-stimulated insulin secretion (GSIS) were performed on overnight-fasted (15–17 h) and sedated (Hypnorm and Dormicum) mice after intraperitoneal injection of glucose 2 g/kg body weight. Area under the curve (AUC) was calculated according to the trapezoidal rule (Supplementary Data).

Western blot expression data were normalized using GAPDH, β-actin, or α-tubulin expression. For detailed information and antibodies, see Supplementary Data.

All quantitative RT-PCR (qRT-PCR) data are presented as fold expression relative to the control sample and calculated using the ΔΔCq method. TBP was used as an internal reference gene. For detailed information, see Supplementary Data.

Islet isolation and insulin secretion were performed essentially as previously described (24). For islet insulin secretion experiments, five equally sized islets were incubated in CMRL-1066 (#21530; Gibco) supplemented with 10% FBS (#10500; Gibco) at 37°C for 2 h. The islets were equilibrated in ubiquitin (UB) buffer (2.8 mmol/L glucose, 0.1% BSA) at 37°C for 1 h and then transferred to UB buffer containing either 2.8 mmol/L glucose, 16.8 mmol/L glucose, or 30 mmol/L KCl and incubated at 37°C for an additional 1 h (Supplementary Data). For Retro-2 treatment experiments, MIN6 cells were passaged 1:3, cultured for 48 h, and exposed to Retro-2 for 24 h, and islets were cultured for 48 h after isolation and exposed to Retro-2 for an additional 48 h. UB buffer (10×) was prepared as follows: NaCl 14.6 g, KCl 880 mg, CaCl2 Å∼ H₂O 376 mg, MgCl₂ Å∼ 6H₂O 488 mg, and HEPES 11.9 g was dissolved in 200 mL H₂O. Upon dilution, pH was set at 7.35 and 0.1% BSA was added (ICN #105033, fatty acid free).

Isolated islets were first incubated for 1 h in CMRL-1066 supplemented with 10% FCS. For the COPI-independent Golgi-to-ER retrograde transport assay, islets were then transferred to media containing brefeldin A (BFA) 0.5 μg/mL, and islets were removed and fixed after 0, 2.5, 5, 10, 20, and 40 min. For the Golgi anterograde transport assay, islets that had been incubated 1 h in CMRL-1066 supplemented with 10% FCS were transferred to media with BFA and incubated for an additional 1 h. Islets were then washed and incubated in medium without BFA, removed, and fixed after 0, 30, 45, 60, and 120 min.

All numerical data are presented as mean ± SEM. All statistical analyses were performed by heteroscedastic two-tailed Student t test. P < 0.05 was considered statistically significant.

To elucidate the functional role of Asna1 in vivo, we generated β-cell–specific deletion of Asna1 in mice by breeding Asna1^flox/flox mice (Supplementary Fig. 1A–H) with Ins1^+/Cre mice (i.e., mice where the gene encoding the CRE recombinase was inserted in one of the Ins1 alleles [25,26]). Asna1 gene expression was progressively reduced between 2 and 4 weeks in islets of ∼4-week-old Asna1^β−/− mice (Supplementary Fig. 2A), and consequently, Asna1 protein levels were decreased in Asna1^β−/− islets at 4 weeks (Supplementary Fig. 2B). The residual Asna1 expression at 4 weeks likely predominantly reflects expression in non-β islet cells, although we cannot exclude a potential minor contribution from a few β-cells that had not yet fully deleted both Asna1 alleles at this stage. Nonfasted glucose levels of Asna1^β−/− mice were already mildly increased from ∼2 weeks of age, and Asna1^β−/− mice progressed to overt diabetes between 6 and 10 weeks of age (Fig. 1A). The increase in nonfasting hyperglycemia and development of diabetes was more pronounced in Asna1^β−/− males (Fig. 1A) than in Asna1^β−/− females (Supplementary Fig. 3). These findings show a requirement for Asna1 in the maintenance of β-cell function and glucose homeostasis. Subsequent experiments were performed on 3- to 4-week-old mice (i.e., before the onset of overt diabetes) using a mix of males and females.

To assess how Asna1 ensures β-cell function, we analyzed glucose tolerance and GSIS in response to exogenous, intraperitoneal administration of glucose (i.e., GTT) in Asna1^β−/− mice. The targeted insertion of Cre into one of the Ins1 alleles, Ins1^+/Cre; global inactivation of one Asna1 flox allele, Asna1^flox/−; or β-cell conditional inactivation of one Asna1 flox allele, Asna1^β+/−, affected neither glucose tolerance nor GSIS (Supplementary Fig. 4A–F). Thus, Asna1^β+/−, Asna1^+/flox, Asna1^flox/flox, and Asna1^β+/− mice, collectively denoted Asna1^βctrl, were all used as controls. In contrast to Asna1^β+/− littermates, 3-week-old Asna1^β−/− mice exhibited impaired glucose tolerance (Fig. 1B) as well as reduced insulin secretion in response to both glucose (Fig. 1C) and the insulin secretagogue arginine (Fig. 1D). Of note, glucose tolerance and insulin secretion were affected in both males and females at this stage (Supplementary Fig. 5A–F). Together, these data show that loss of Asna1 in β-cells leads to diabetes due to impaired β-cell function.

To elucidate the mechanism underlying β-cell failure in Asna1^β−/− mice, we next determined pancreatic islet area and β-cell number. Pancreatic islet area and β-cell number were both normal (Fig. 2A and B), whereas pancreatic insulin content was reduced by ∼60% in Asna1^β−/− mice (Fig. 2C), suggesting that Asna1^β−/− mice develop diabetes as a consequence of reduced amounts of insulin. In mice, insulin is encoded by two highly homologous genes Ins1 and Ins2, and as expected, Ins1 expression was nearly significantly (P = 0.055, n = 4) reduced by ∼50% in islets of mice carrying the Ins1^Cre knockin allele (Supplementary Fig. 6A). However, because of the low expression of Ins1 relative to Ins2, the reduced Ins1 expression did not in itself affect total insulin (Ins1 + Ins2) expression or insulin protein content in Ins1^+/Cre mice (Supplementary Fig. 6A and B). Islet proinsulin and insulin content, however, was reduced by ∼70% in Asna1^β−/− islets compared with Asna1^β+/− islets (Fig. 2D) without a corresponding difference in total insulin (Ins1 + Ins2) mRNA levels (Supplementary Fig. 7), thus providing evidence for a posttranscriptional reduction in insulin biogenesis in β-cells of Asna1^β−/− mice. In contrast, pancreatic content of the hormone islet amyloid polypeptide, which is stored and cosecreted with insulin, was unaffected (Supplementary Fig. 7B). No significant difference in islet hormone content was observed when comparing males and females (Supplementary Fig. 7C).

The pancreatic proinsulin/insulin ratio was normal in Asna1^β−/− islets (Fig. 2E), and inhibition of protein degradation using proteasomal and lysosomal inhibitors had similar effects on proinsulin and insulin content in Asna1^β−/− and Asna1^β+/− islets (Supplementary Fig. 8A and B). These results argue against an increased rate of (pro)insulin degradation as the underlying mechanism for the reduced insulin content in Asna1^β−/− β-cells. In agreement with the perturbed glucose tolerance and impaired glucose- and arginine-induced insulin secretion observed in vivo, isolated Asna1^β−/− islets secreted less insulin in response to both glucose and the membrane depolarizer KCl (Fig. 2F). Normalization of the amount of secreted insulin-to-total insulin content, to adjust for the reduced insulin content in Asna1^β−/− islets, however, revealed that the relative insulin secretion potential of Asna1^β−/− islets was similar to that of control islets (Fig. 2G). Of note, plasma levels of insulin and proinsulin showed an increased proinsulin/insulin ratio in Asna1^β−/− mice (Supplementary Fig. 8C), suggesting that proinsulin secretion is relatively greater than insulin secretion in Asna1^β−/− mice.

The expression of genes involved in glucose uptake and metabolism (i.e., Glut2 and Gck) was reduced by ∼70% and ∼30%, respectively (Supplementary Fig. 9A). The expression of both Glut2 and Gck however, has been observed to be negatively affected by hyperglycemia (27), leaving open the possibility that the reduced expression of these genes in Asna1^β−/− islets may be secondary to the modest hyperglycemia observed at the stage of islet isolation (i.e., 3–4 weeks of age). The expression of genes involved in membrane depolarization (Sur1 and Kir6.2) and insulin exocytosis (Rab3a, Rab3b, Rab27, Snap25, Syt7, Stx1a, Stx4a, Vamp2, and Vamp3) was essentially normal with the exception of Rab3b expression, which was reduced by ∼40% (Supplementary Fig. 9B). Taken together, these findings suggest that Asna1^β−/− mice develop insulin insufficiency and diabetes largely as a consequence of impaired insulin biogenesis, although it is possible that the reduced expression of Glut2, Gck, and Rab3b may contribute to the impaired GSIS observed in vivo.

On a systemic level, Get3/Asna1 has been associated with maintenance of ER homeostasis (10,12,28), thus constituting a likely intersection point for Asna1 TA-targeting activity and β-cell function. The expression and localization of markers for the ER-Golgi intermediate compartment (ERGIC53), cis Golgi (Gm130), trans Golgi network (TGN46), proinsulin vesicles, endosome (EEA1), and lysosome (Lamp1) compartments all appeared normal in Asna1^β−/− islets (Fig. 3A), whereas the expression of the ER stress-response chaperones BiP (Grp78) and Grp94, as judged by KDEL immunostaining, appeared more intense (Fig. 3A). Moreover, strong nuclear ATF4 expression was observed (Fig. 3B) and IRE1α activity, monitored using ERAI reporter mice (23) on an Asna1^β−/− background, was enhanced (Fig. 3C). Additionally, transmission electron microscopy revealed dilated ER cisterna in a subset of Asna1^β−/− β-cells at 5–6 weeks of age (Supplementary Fig. 9C). The expression of UPR genes, including BiP (Hspa5, grp78), Grp94 (Hsp90b1), DnaJc3, Ero1lb, Erp29, Pdia4 (Erp72), Edem2, HRD, Herpud1, Sel1l, Atf3, Atf4, Chop10 (Ddit3), and Trib3, were, with the exception of Ero1lb, all increased in Asna1^β−/− islets (Fig. 3D). Together, these findings, particularly the activation of all three branches of the UPR, suggest that loss of Asna1 in β-cells leads to ER stress.

The normal pancreatic proinsulin/insulin ratio (Fig. 2E) together with the normal localization of proinsulin to vesicular structures adjacent to the trans Golgi network (TGN) in Asna1^β−/− β- cells (Fig. 3A) suggest that proinsulin folding, transport, and processing are largely unaffected in Asna1^β−/− β-cells. Moreover, treatment of islets with the chemical chaperones 4-PBA and TUDCA, which are known to improve ER folding capacity (29,30), failed to normalize islets proinsulin and insulin content (Supplementary Fig. 10A and B) and to alleviate ER stress (Supplementary Fig. 10C and D) in Asna1^β−/− islets. Taken together, these results argue against impaired folding of proinsulin as the underlying cause for ER stress in Asna1^β−/− islets.

Get3/Asna1 has been suggested to genetically interact with the retromer complex, the COG complex, and the COPI coatomer complex, implying a potential role for Get3/Asna1 in retrograde transport (11). Retrograde transport between the Golgi and the ER are mediated by both COPI coatomer-coated vesicles and Rab6-dependent membrane tubules (31). To assess retrograde transport in Asna1^β−/− islets, we first made use of BFA, which 1) inhibits COPI-dependent traffic by blocking the assembly of the COPI vesicle coat and 2) collapses the Golgi into the ER through the alternative Rab6-dependent route (32). Similar to HeLa cells with impaired Rab6 activity (32,33), Asna1^β−/− islets showed a delayed and incomplete collapse of the Golgi compartment when exposed to BFA (Fig. 4A and Supplementary Fig. 11A and B). These data suggest that the Rab6-dependent retrograde transport route between the Golgi and ER is impaired in Asna1^β−/− β-cells. In contrast, rebuilding of the Golgi after BFA washout appeared largely unaltered, suggesting that anterograde transport was normal in Asna1^β−/− islets (Fig. 4B).

We next assayed plasma membrane (PM)-to-TGN retrograde transport by monitoring the uptake and transport of fluorescently labeled Ricin toxin (594-RiTx), which is transported through endosomes, TGN, and cis Golgi into the ER after endocytosis (34). In dispersed β-cells from control islets, 594-RiTx accumulated in the TGN after 90 min (Fig. 4C). In contrast, accumulation of 594-RiTx in the TGN was significantly reduced by 20% in Asna1^β−/− β-cells (Fig. 4C and D), providing evidence that PM-to-TGN retrograde transport is impaired in Asna1^β−/− β-cells. These results demonstrate that Asna1 is required for sustained retrograde transport to both the TGN and the ER.

Vesicle transport and membrane recycling among the PM, endosome, and TGN critically depends on TA-SNAREs, such as syntaxin (Stx) 5, Stx6, and Vamp3 (35,36). Thus, we reasoned that Asna1 might regulate retrograde transport by ensuring proper localization of Stx5, Stx6, and/or Vamp3 to the cis Golgi, TGN, and endosomes, respectively. Consistent with this notion, Stx5, Stx6, and Vamp3 proteins were barely detectable in Asna1^β−/− β-cells, whereas the localization of the Golgi TA protein giantin was unaltered (Fig. 5A). The reduced Golgi localization of Stx5 and Stx6 was already observed in ∼1-week-old Asna1^β−/− mice (Supplementary Fig. 12), whereas the loss of Vamp3 was observed first after 3 weeks of age (Fig. 5A), suggesting that the loss of Vamp3 may be secondary to impaired retrograde transport and/or deterioration of β-cell function.

The protein levels of the major 35-kDa Golgi Stx5 isoform and Stx6 were largely unaltered in Asna1^β−/− islets (Fig. 5B), although there was a tendency for the minor 42-kDa Stx5 ER isoform (37) to be slightly reduced. Together, these results show that Stx5 and Stx6 are not degraded but likely become mislocalized or redistributed from their respective Golgi compartment in Asna1^β−/− β-cells. Stx5 was successfully coimmunoprecipitated when coexpressed with Myc-tagged Asna1 in insulinoma cell lines (Fig. 5C), leaving open the possibility that analogous to Get3/Asna1-Sed5 interaction (28), Asna1 mediates membrane insertion of Stx5 in β-cells. The localization of other proposed Asna1 targets, such as the ER resident proteins Sec61β and Sec22b (Fig. 5D) as well as the post-ER TA-SNAREs Vamp2 and Stx1a, which are implicated in insulin vesicle exocytosis (Fig. 5E), revealed a normal localization and expression, arguing against a critical role for Asna1 in the biogenesis of these TA proteins in β-cells. Taken together, these results provide evidence of a requirement for Asna1 function in ensuring Golgi localization of Stx5 and Stx6 in β-cells.

To test whether the activation of the UPR and/or reduced insulin content observed in Asna1^β−/− β-cells might be secondary to impaired retrograde transport, we used the small molecule inhibitor Retro-2 that inhibits retrograde transport from early endosomes (EEs) to the TGN (38). Of note, analogous to the phenotype of Asna1^−/− β-cells, exposure of isolated wild-type islets to Retro-2 resulted in not only reduced Golgi localization of Stx5 but also, albeit less pronounced, reduced Stx6 Golgi localization (Fig. 6A). However, as observed in Asna1^−/− β-cells, the levels of the 35-kDa isoform of Stx5 were largely unaltered, whereas the less abundant 42-kDa expression tended to be reduced in Retro-2–treated islets (Supplementary Fig. 13A). Thus, similar to that previously observed in HeLa cells (38), exposure of islets to Retro-2 results in a redistribution of Stx5 from the Golgi compartment. Moreover, and like Asna1^−/−β-cells, COPI-independent Golgi-to-ER retrograde transport was impaired in Retro-2–treated islets (Supplementary Fig. 13C–E). Because ex vivo cultivation of isolated islets in itself provokes ER stress, the functional effects of Retro-2 on UPR were assessed on day 2 after isolation (i.e., at a stage when the acute activation of ER stress genes is somewhat dampened) (Supplementary Fig. 14). Retro-2–exposed islets displayed reduced insulin content by ∼30% and enhanced expression of UPR genes, indicating ER stress (Fig. 6B and C). Additionally, insulin expression was significantly reduced in Retro-2–exposed islets (Fig. 6D), thus likely contributing to the observed reduction in insulin content in these islets. The subcellular location of Stx5 was unaffected by the ER stress/UPR activators tunicamycin and thapsigargin (Supplementary Fig. 15), providing evidence that the mislocalization of Stx5 is not secondary to ER stress.

To independently assess the effects of Retro-2 on insulin biogenesis and UPR activation in β-cells, obviating potential confounding effects of ER stress induced by isolation and ex vivo cultivation of islets, we next exposed MIN6 cells to Retro-2. Exposure of MIN6 cells to Retro-2 also resulted in reduced Golgi localization, but not expression, of Stx5 (Fig. 6E and Supplementary Fig. 13B), reduced insulin content, and enhanced expression of UPR genes (Fig. 6F and G). However, although insulin content was reduced by ∼30% in Retro-2–treated MIN6 cells, insulin expression was not significantly reduced (Fig. 6H). Taken together, these data show that inhibition of retrograde transport in β-cells results in ER stress and impaired insulin biogenesis. Additionally, these findings provide evidence that the reduced insulin content and ER stress observed in Asna1-deficient β-cells are a consequence of impaired PM/EE-to-TGN retrograde transport.

This study shows a critical role for Asna1 in ensuring β-cell function. Loss of Asna1 in β-cells of mice results in pancreatic hypoinsulinemia, impaired insulin secretion, and early onset diabetes. Additionally, β-cells of Asna1^β−/− mice showed impaired PM-to-TGN as well as Golgi-to-ER retrograde transport, ER stress, and mislocalization of Stx5 and Stx6. Of note, we also show that inhibition of retrograde transport at the level of EE-to-TGN in isolated islet and insulinoma cells results in impaired Golgi-to-ER retrograde transport, decreased insulin content, and ER stress. Thus, the findings provide evidence that Asna1 is required in β-cells to ensure retrograde transport, which in turn appears to be essential for ER homeostasis and proinsulin biogenesis. Additionally, the perturbed Golgi-to-ER retrograde transport in Retro-2–treated primary islets suggests that the impairment of this step in Asna1^β−/− β-cells likely is secondary to the inhibition of retrograde transport at the level of EE-to-TGN.

The primary cause of diabetes in Asna1^β−/− mice appears to be insufficient production of insulin. The current data do not support a role for Asna1 in the posttranslational targeting of insulin itself, such as has been described for other short secretory proteins (39), because we did not observe an increase or alteration in pancreatic proinsulin levels. Although the activation of the UPR indicates that ER homeostasis is perturbed in the absence of Asna1 function, we found no evidence of impaired proinsulin folding or anterograde transport through the secretory pathway. Moreover, vesicular proinsulin localization appears unaltered in Asna1^β−/− β-cells, pancreatic islet amyloid polypeptide content is unaffected, and insulin maturation as well as exocytosis seem largely unaffected. Thus, activation of the UPR is likely not provoked by insufficient insulin folding capacity. Of note, we observed an increased plasma proinsulin/insulin ratio, suggesting that secretion of newly synthesized proinsulin is increased and raising the possibility that Asna1-dependent EE-to-TGN retrograde transport normally counteracts leakage or premature secretion of proinsulin through the endosomes and the constitutive-like pathway (40).

Instead, we favor the idea that impaired retrograde transport in Asna1^β−/− mice provokes reduced insulin biogenesis as a negative consequence of UPR activation on protein translation (41). Additionally, there was a tendency, albeit nonsignificant, for insulin expression levels to be reduced in Asna1^β−/− mice, which would also be consistent with a negative effect of UPR activation (e.g., through IRE1α/XBP1 and Trib3 induction, on insulin transcription [42,43], or as a consequence of IRE1α/XBP1–mediated degradation of insulin mRNA [44,45]). Moreover, Retro-2–mediated impairment of retrograde transport resulted in robust activation of UPR genes and decreased insulin protein content in both isolated islets and MIN6 cells and were accompanied, albeit to a different extent, by suppression of insulin transcription. Thus, although we cannot exclude that Retro-2 has additional UPR-independent effects on insulin biogenesis, the reduced insulin content observed in Retro-2–treated islets and MIN6 cells appears to be secondary to UPR-mediated impairment of insulin biogenesis at both the transcriptional and the translational level. These findings also provide strong evidence that impairment of retrograde transport in β-cells by Retro-2 in vitro and ex vivo and due to loss of Asna1 function in vivo leads to activation of the UPR that in turn negatively affects insulin biogenesis, which, in vivo, results in insulin insufficiency and the development of diabetes (Supplementary Fig. 16).

Previous studies, primarily in cell-free systems, have outlined a role for Get3/Asna1/Asna1 in the targeting of TA proteins to the ER membrane receptor Get1/Get2 (CAML/WRB in mammals), thus facilitating their insertion into the ER membrane (15,16). The current data indicate that at least in β-cells, Golgi localization of the TA-SNARE proteins Stx5 and Stx6 depend on Asna1 function. Total levels of Stx5 and Stx6 mRNA and protein were unaltered in Asna1^β−/− β-cells, suggesting that Stx5 and Stx6 become redistributed and/or mislocalized in the absence of Asna1 activity. In contrast, the localization of the TA proteins Sec61β, Sec22b, giantin, Stx1, and Vamp2 was unaffected, suggesting that other chaperone systems, such as the Hsc70/Hsp40 pair (17), compensate for the targeting of these TA proteins in β-cells lacking Asna1 function. In agreement with such a notion, the yeast TA proteins Sbh1 and Sbh2 (i.e., the homologs of Sec61β) and Scs2 and Ysy6 all retain a certain level of ER localization in yeast GET1/GET2 mutants (28). Consistent with a potential direct role for Asna1 in the targeting of Stx5 to the ER of β-cells, and like Get3/Asna1-Sed5 interactions, we found that Asna1 physically interacts with Stx5 in insulinoma cells upon coexpression of these proteins. We cannot, however, exclude the possibility that Stx5 and Stx6 are appropriately targeted in the absence of Asna1 activity but become mislocalized or redistributed to other cellular compartments as a consequence of impaired retrograde transport.

Exposure of MIN6 insulinoma cells and isolated islets to the small molecule inhibitor Retro-2 closely mimics key phenotypes of Asna1^β−/− β-cells, including 1) reduced insulin content, 2) activated UPR, and 3) redistributed Stx5 and Stx6. Although the molecular targets of Retro-2 are unknown and at what point the Retro-2 and Asna1 pathways intersect is unclear, the reduction of Stx5 in the cis Golgi and Stx6 in the TGN is likely to affect Asna1-dependent retrograde transport. Small interfering RNA–mediated knockdown of Stx5 and inhibition of Stx6 using blocking antibodies have both been shown to negatively affect EE-to-TGN recycling (35,36). However, two observations argue for a primary role for Stx5 in the context of retrograde transport and ER homeostasis. First, the effect of Retro-2 on Stx5 localization is acute and complete, whereas the effect on Stx6 localization is slower and less severe (Fig. 6A) (38), suggesting that Stx6 mislocalization may be secondary to the loss of Stx5 from the Golgi. Second, in yeast Get3/Asna1 mutants, Sed5 is mislocalized, and the Bip/grp78 ER chaperone homolog Kar2 is abnormally secreted (10,28), which may reflect an impaired retention or retrograde transport of Kar2 (10). Importantly, overexpression of the Stx5 homolog Sed5 in Get3/Asna1 mutants rescues the Kar2 (Bip) secretion phenotype (28). Taken together, these data are consistent with a chain of events where inactivation of Asna1/Get3, likely through mislocalization of Stx5/Sed5, perturbs retrograde transport and, thereby, ER homeostasis. In mammalian β-cells, impairment of retrograde transport appears to have the additional consequence of attenuating insulin biogenesis, thus leading to the development of diabetes.

Recently, an additional function distinct from its TA protein–targeting activity was described for Get3/Asna1(18). Under oxidative stress conditions and independent of ATP, Get3/Asna1 was shown to function as a molecular chaperone that binds unfolded proteins to prevent their irreversible aggregation. The proposed dual role for Get3/Asna1 in yeast is intriguing and implies a potential role for maintenance of both ER and redox homeostasis. Future studies are required to separate potential oxidative stress–induced chaperone activity of Asna1 from the TA protein–targeting function to fully understand how Asna1 ensures retrograde transport as well as insulin and ER homeostasis.

In conclusion, we show that Asna1 is critical for β-cell function. The results provide evidence that Asna1 plays a role in ensuring retrograde transport along a PM-to-TGN and a Golgi-to-ER route. Impairment of these functions in β-cells leads to ER stress, insulin insufficiency, and development of diabetes. The findings suggest that maintenance and/or restoration of retrograde transport in β-cells may be therapeutically relevant for T2D. The study was performed, however, using genetically modified mice and isolated mouse islets. Thus, although the data provide strong evidence for a role for Asna1 and retrograde transport in ensuring mouse β-cell function, a potential similar role for ASNA1 and/or retrograde transport for β-cell function in humans will require additional analyses.

Acknowledgments. The authors thank Elisabet Pålsson, Jurate Straseviciene, Fredrik Backlund, and Lisa Lundberg (Umeå Centre for Molecular Medicine) for technical assistance and members of the authors' laboratory for technical instructions, suggestions, and helpful discussions.

Funding. These studies were facilitated by support from the Strategic Research Program in Diabetes at Umeå University and supported by grants from the Swedish Research Council (521-2013-3215) and the Knut and Alice Wallenberg Foundation (KAW 2010.0033).

Duality of Interest. H.E. is a cofounder, shareholder, and consultant of the unlisted biotech company Betagenon AB. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.N. and V.S.P. contributed to the design and performance of experiments, data interpretation, discussion, and writing and editing of the manuscript. P.N. initiated the study, provided advice, and contributed to the discussion and review of the manuscript. H.E. initiated the study, designed and supervised the study, analyzed and interpreted the data, contributed to the discussion, and wrote the manuscript. H.E is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.