|Home | About | Journals | Submit | Contact Us | Français|
The endoplasmic reticulum (ER) resident PKR-like kinase (PERK) is necessary for Akt activation in response to ER stress. We demonstrate that PERK harbors intrinsic lipid kinase, favoring diacylglycerol (DAG) as a substrate and generating phosphatidic acid (PA). This activity of PERK correlates with activation of mTOR and phosphorylation of Akt on Ser473. PERK lipid kinase activity is regulated in a phosphatidylinositol 3-kinase (PI3K) p85α-dependent manner. Moreover, PERK activity is essential during adipocyte differentiation. Because PA and Akt regulate many cellular functions, including cellular survival, proliferation, migratory responses, and metabolic adaptation, our findings suggest that PERK has a more extensive role in insulin signaling, insulin resistance, obesity, and tumorigenesis than previously thought.
The cellular membrane phospholipid phosphatidylinositol (PtdIns) and its metabolites are critical signaling molecules. PtdIns is synthesized at the endoplasmic reticulum (ER) membrane (1) and can then be phosphorylated at the 3, 4, and 5 positions of the inositol ring, generating a variety of monophosphorylated metabolites. These metabolites serve as precursors for additional phosphorylation events that result in the generation of PtdInsP2 and PtdInsP3 (10, 42, 48). PtdIns(4,5)P2 is in turn hydrolyzed by phospholipase C (PLC), generating diacylglycerol (DAG) and inositol-1,4,5-triphosphate, resulting in the formation of additional molecules capable of intracellular signaling (38). PtdIns(3,4,5)P3 (PIP3) is generated by the PtdIns3-kinase (PI3K) superfamily of lipid kinases (16). PI3K activity and PIP3 production are regulated by growth factors and chemokines, leading to the activation of Akt, one of the key growth and survival pathways in the cell. Additionally, generation of phosphatidic acid via the mitogen-stimulated activation of phospholipase D (PLD) provides another signal promoting Akt activation due to the phosphatidic acid-dependent assembly of the mTORC2 complex (18, 49).
PI3K class IA is composed of a 110-kDa catalytic subunit (p110) and an 85-kDa adaptor/regulatory subunit (p85). Mammalian cells have three p110 isoforms (p110α, -β, and -δ), encoded by three separate genes and at least seven adaptor proteins that are generated through alternative splicing of transcripts encoded by three distinct genes, p85α, p85β, and p55γ. The p85 subunit has two Src homology 2 (SH2) domains that dock with phosphorylated tyrosine residues generated by activated tyrosine kinases (3, 11). The p85 SH2 domain mediates recruitment of the cytosolic PI3Ks to cellular membranes where their lipid substrates reside. The p110 subunit-binding site within p85 is located between the two SH2 domains (inter-SH2 domain) (13, 21).
The ER transmembrane serine/threonine kinase PERK, or EIF2AK3 (26, 45), is activated under conditions of physiological ER stress such as low carbon source (glucose deprivation), low oxygen (hypoxia), or increased synthetic demand in secretory tissues, as well as by chemical inducers of ER stress (tunicamycin, thapsigargin). PERK-dependent phosphorylation of eIF2α triggers decreased protein synthesis to alleviate ER client protein load while simultaneously increasing production of key protein substrates necessary for cell adaptation (26). PERK also maintains cellular redox homeostasis via direct phosphorylation of the Nrf2 transcription factor (12). Moreover, PERK has been linked to the activation of PI3K signaling and Akt during conditions of ER stress (25, 29, 31). PKR, another eIF2α kinase, can also stimulate Akt phosphorylation (31). Since these two kinases share eIF2α as their substrate, it was proposed that PI3K activation is dependent on the effect of PERK/PKR on translation attenuation.
Here, we present evidence that PERK possesses an intrinsic lipid kinase activity toward diacylglycerol (DAG), generating phosphatidic acid (PA) as a major product. Additional evidence is provided demonstrating that PERK lipid kinase activity is regulated in a PI3K regulatory subunit p85α-dependent manner. The lipid kinase function of PERK mediates mTOR, Akt, and Erk1/2 activation during ER stress. Critically, the activity of PERK is necessary for the activation of anabolic pathways downstream of Akt in a physiologically relevant setting.
Mouse embryo fibroblasts (MEFs) were grown in Dulbecco's modified Eagle's medium (DMEM) (high glucose formulation) with 4 mM l-glutamine, 10% (vol/vol) fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin, and 55 mM β-mercaptoethanol. H1299 cells were grown in RPMI 1640 medium with 10% (vol/vol) fetal bovine serum. PERKLoxP/LoxP MEFs were derived as previously described (7) and immortalized using 3T9 protocol. Transient expression of plasmids encoding Myc-PERK and Myc-K618APERK, HA-p85α and FLAG-p110α, was achieved using Lipofectamine Plus (Invitrogen). For MEF in vitro differentiation, cells were grown to confluence in 60-cm2 dishes and then treated with adipocyte differentiation cocktail as described previously (8). Small molecule inhibitors used were as follows: FIPI (Cayman Chemical); rapamycin (Calbiochem); PIK-294, IC-87114, and MK-2206 (Selleck Chem Company); and wortmannin (Millipore). Compounds were added directly to cell culture media at the concentrations indicated.
PERK deletion mutants were generated by site-directed mutagenesis with the QuikChange mutagenesis kit (Stratagene) according to the manufacturer's instructions and confirmed by sequencing.
Cells were lysed in EBC buffer (50 mM Tris, pH 8.0, 120 mM NaCl, 0.5% NP-40) supplemented with protease and phosphatase inhibitors. The following antibodies were used: PERK (Rockland Immunochemicals); phospho-eIF2α, phospho-Ser473 Akt, Akt, eIF4E, phospho-S6 ribosomal protein (Ser235/236), S6 ribosomal protein, phospho-p44/42 MAPK (Thr202/Tyr204 of Erk1 and Thr185/Tyr187 of Erk2), and p44/42 MAPK (Erk1/2), (Cell Signaling); CHOP (Santa Cruz Biotechnology); p110α (BD Biosciences); and p85 and IRS1 (Millipore).
PERK was isolated from cells by immunoprecipitation, or recombinant GST-tagged ΔN-PERK or ΔN-K618A PERK was purified from bacteria, using glutathione (GSH)-Sepharose. Commercial recombinant PERK was obtained from Invitrogen. Recombinant PKR was purchased from Millipore. Recombinant p38α was obtained from Upstate. Recombinant GSK3β and PKA C-α were purchased from Cell Signaling. Recombinant DGK was purchased from Sigma. 1,2-Dilauroyl-sn-glycerol (DAGC12), 1,2-dimyristoyl-sn-glycerol (DAGC14), 1,2-dipalmitoyl-sn-glycerol (DAG16), and 1,2-distearoyl-sn-glycerol (DAG18) were from Echelon Biosciences. A lipid kinase assay was performed as described previously (50). 1,2-Dihexanoyl-sn-glycero-3-phosphate (06:0 PA) was purchased from Avanti Polar Lipids.
In vitro transcription and translation were performed with expression plasmid-encoding p85α using a coupled transcription and translation system (Promega) according to the manufacturer's instructions.
Lipids on cultured cells were visualized after cells were fixed for 2 min in 3.7% formaldehyde, washed with deionized water, and stained with Oil Red O solution.
Total cellular phosphatidic acid content was determined using either the total phosphatidic acid kit (Cayman Chemical) or fluorescent monitors of PA in living cells (36). In brief, cells were transfected with a PA biosensor expression construct (wild-type [WT] biosensor; Pii-DK) or control plasmid with mutated PA binding domain (mutant [MUT] biosensor; Pii-DK-9A). Twenty-four hours after transfection, cells were serum-starved (0.1% fetal bovine serum [FBS]) for 16 h and then treated with 500 nM thapsigargin. Cyan fluorescent protein (CFP) (excitation [ex] of 440 nm; emission [em] of 480 nm) and fluorescence resonance energy transfer (FRET) (ex of 440 nm; em of 550 nm) values were obtained using fluorescence microscopy. FRET data represent results from three independent experiments where data were acquired from a minimum of three cells. For data acquisition, a Leica AF6000LX system equipped with a Hamamatsu ORCA-R2 charge-coupled-device (CCD) camera using a 20× objective at 37°C was used. FRET/CFP ratios, the corresponding fold change in PA production, and P values were calculated using Sigma Stat software from results of a minimum of three independent experiments.
Many PERK-dependent biological effects result from reduced translation following direct phosphorylation of eIF2α (26, 30, 43). It was therefore important to establish whether there was a role for eIF2α in PERK-mediated activation of Akt following ER stress. For these experiments, we used tunicamycin, an inhibitor of N-linked glycosylation and a commonly used inducer of protein misfolding in the endoplasmic reticulum. As previously noted, induction of CHOP (C/EBP homologous protein) was ablated in eIF2αS51A knock-in cells (Fig. 1a) (43). In contrast, the ER stress-dependent increases in Akt phosphorylation on Ser473 as well as mTOR-dependent ribosomal S6 phosphorylation were preserved in eIF2αS51A knock-in cells (Fig. 1a), demonstrating that neither was a feature of altered eIF2α-dependent translation initiation. Because the PI3K p110α catalytic subunit is responsible for the majority of class IA PI3K activity in fibroblasts, p110α−/− cells were utilized to determine its contribution to ER stress-dependent Akt phosphorylation (57). Akt phosphorylation on Ser473 was still induced in p110α−/− cells (data not shown), suggesting that it is a p110-independent activity. However, because fibroblasts also express p110β, we used pharmacological inhibition as a mechanism to ablate all p110-related PI3K activity. PERK-proficient fibroblasts were treated with one of three distinct chemical inhibitors of the catalytic p110 subunit of PI3K: wortmannin, PIK-294, and IC-87114. Treatment of cells with these inhibitors did not prevent ER stress-induced AKT phosphorylation (Fig. 1b), whereas they did inhibit serum-induced Akt phosphorylation (data not shown). Finally, we also used a p110α dominant negative allele as an independent method to reduce endogenous total PI3K function. Expression of dominant negative p110α in H1299 cells did not significantly diminish Akt phosphorylation in response to ER stress (Fig. 1c; compare lanes 2 and 6). Increased phosphorylation of 4E-BP1, a substrate of mTOR, was also observed in PERK+/+ but not PERK−/− MEFs, revealing PERK-dependent activation of mTOR signaling (Fig. 1d; third panel). This was paralleled by reduced affinity of 4E-BP1 for the 7-methyl-GTP-Sepharose-bound complex of the mRNA Cap-binding proteins (Fig. 1d, upper panel).
We next interrogated the mechanism whereby PERK regulates mTOR. Surprisingly, we observed that PERK immunopurified from either p110+/+ or p110α−/− cells treated with thapsigargin, which induces protein misfolding due to calcium depletion in the ER as a consequence of SERCA pump inhibition, exhibited an ability to phosphorylate lipids present in a preparation of phosphatidylinositol (PI) (Fig. 2a). For direct assessment, we performed an in vitro lipid kinase assay utilizing the recombinant catalytic domain of PERK and PI. The catalytic domain of PERK, but not kinase-dead PERK-K618A, generated a phosphorylated lipid product in vitro (Fig. 2b). No significant lipid kinase activity was detected when other recombinant active kinases (p38, GSK3β, PKA Cα) were tested (Fig. 2c). To identify the phospholipid product, in vitro lipid kinase assays were carried out using recombinant PERK and various phosphoinositide substrates. Products were resolved by thin-layer chromatography (TLC) (Fig. 2d) and subsequently excised and subjected to high-performance liquid chromatography (HPLC) for identification (Fig. 2e). PERK-dependent phosphorylation of phosphoinositides resulted in the production of phosphatidic acid (PA), as evidenced by the mobility of this metabolite on TLC plates (Fig. 2d) and the elution profile of deacylated product on HPLC (Fig. 2e). We hypothesized that PERK was preferentially phosphorylating a small pool of diacylglycerol (DAG) present in PI preparation. Consistent with this hypothesis, PERK phosphorylated purified DAG in vitro, demonstrating a preference for shorter-fatty-acid side chains (Fig. 2f). The identity of the PERK-generated metabolite as PA was also confirmed by using recombinant diacylglycerol kinase (DGK) (Fig. 2g) as a control. Thus, the serine/threonine kinase PERK is a bifunctional enzyme possessing both protein kinase and lipid kinase activity toward DAG generating PA as a product in vitro.
We next assessed PERK-dependent generation of PA in vivo. We detected a significant increase in intracellular levels of total PA in PERK+/+ cells treated with thapsigargin but not in PERK−/− cells (Fig. 3a). PA production in response to thapsigargin was retained in eIF2αS51A knock-in fibroblasts (Fig. 3b), consistent with this being a translation-independent activity of PERK. To address the role of endogenous DGK in the generation of PA during an ER stress response, we treated cells concurrently with thapsigargin and a pharmacological inhibitor of DGK R59949; no significant inhibition of ER stress-dependent PA production was noted (Fig. 3c), suggesting that DGK does not contribute significantly under these conditions. We also addressed the potential contribution of PLD to stress-dependent PA production. Treatment of cells with the PLD inhibitor, FIPI, failed to reduce stress-dependent PA generation (Fig. 3c). Finally, we also evaluated stress-dependent PA generation following inhibition of PI3K activity (wortmannin), mTORC1 (rapamycin), and Akt (MK-2206). Consistent with the ability of PERK to directly generate PA, inhibition of these kinases failed to significantly inhibit ER stress-dependent PA induction (Fig. 3d).
As an independent test of PA production, we utilized a FRET biosensor (Pii-DK) (36) to measure PA production in live cells (Fig. 3e). Consistent with PA production, the FRET signal (as represented by FRET/CFP) decreased in a time-dependent manner in PERK+/+ cells treated with thapsigargin; in contrast, high FRET levels were sustained in PERK+/+ cells expressing a mutant biosensor (Pii-DK-9A); a high FRET level was also maintained with the wild-type or mutant biosensors in PERK−/− cells (Fig. 3e and f). A significant increase in PA production was measured in PERK+/+ cells but not in PERK−/− cells in response to thapsigargin treatment (Fig. 3g). Thus, PERK lipid kinase activity can directly generate PA in vitro and contribute to PA production in vivo. However, because more than one product migrated on a TLC plate when PI was used as a substrate in an in vitro assay (Fig. 2a and b), we cannot exclude at this point that PERK initially mediates PI phosphorylation followed by lipase-like cleavage of PIP2 generating DAG substrate.
We next determined whether PI3K class IA p85 regulatory subunits contribute to PERK-dependent Akt activation. Indeed, ER stress-dependent Akt phosphorylation was significantly reduced in p85α/β double-knockout (DKO) MEFs (Fig. 4a and and5a)5a) (9). This result suggests a potential functional regulatory relationship between PERK and p85α. The most direct scenario would be one mediated by direct association. Indeed, purified GST-tagged catalytic domain of PERK and K618A PERK associated with in vitro transcribed and translated p85α (Fig. 4b). Supporting this in vitro interaction, PERK was also recovered in p85 immune complexes collected from 293T cells expressing exogenous PERK, p85α, and p110α (Fig. 4c and d). This interaction was diminished when the C terminus of PERK was deleted (Fig. 4c). Finally, endogenous p85 exhibited stress-dependent association with PERK in wild-type murine fibroblasts (Fig. 4e).
To establish the functional relationship between p85 and PERK, we utilized p85α/p85β double-knockout fibroblasts or those reconstituted with wild-type p85α. While ER stress-dependent activation of Akt was greatly diminished in knockout cells, reconstitution with p85 restored Akt activation (Fig. 5a). Finally, in an in vitro lipid kinase assay, addition of purified p85α stimulated PERK-dependent lipid kinase activity (Fig. 5b) revealing a regulatory role for p85.
Increased PA generation can directly promote mTORC1 and mTORC2 complex assembly and Akt Ser473 phosphorylation (49). PA also contributes to Ras activation via recruitment of the nucleotide exchange factor Son of Sevenless (SOS) to membranes (35, 56) and through recruitment of cRaf-1 to the plasma membrane followed by its interaction with Ras (39); both events contribute to activation of Erk1/2. We thus hypothesized that PERK-dependent generation of PA should link the unfolded protein response (UPR) with the induction of pathways downstream of Ras, such as MAPK, Akt, and mTOR. Indeed, we observed PERK-dependent activation of Akt (25, 29), as well as increased phosphorylation of S6, an event dependent upon mTOR, and Erk1/2 phosphorylation (Fig. 6a). Activation of mTOR and Ras pathways in response to ER stress was lost in PERK−/− MEFs; importantly, this could be restored by addition of PA to the growth medium in the presence or absence of tunicamycin (Fig. 6a). These data reveal that PA generation is sufficient to trigger signaling downstream of Ras in cells experiencing a robust ER stress response. Thus, PERK-dependent generation of PA is necessary and sufficient for mTOR-Akt and Ras-MEK-Erk1/2 signaling following initiation of ER stress.
To further test the impact of PERK lipid kinase activity on growth factor receptor signaling, we measured Akt activation in response to insulin stimulation in the context of PERK wild-type and PERK-deficient cellular environment. Indeed, insulin-dependent Akt phosphorylation was attenuated in PERK−/− cells (Fig. 6b). In contrast, PERK−/− cells remained responsive to serum stimulation (Fig. 6c), demonstrating that while PERK function contributes to insulin signaling, it is not required for generalized growth factor signaling. Assessment of PERK lipid kinase activity in response to insulin treatment revealed an insulin-dependent 2-fold increase in PERK-lipid kinase activity (Fig. 6d). We also detected recruitment of p85α in PERK immune complexes upon insulin treatment (Fig. 6e). Finally, thapsigargin pretreatment of PERK+/+ cells enhanced the amplitude of PERK-dependent Akt phosphorylation elicited by insulin treatment (Fig. 6f). Thus, PERK-dependent generation of PA promotes mitogenic signaling in the setting of ER stress.
To query the physiological significance of PERK lipid kinase activity, we utilized an in vitro adipocyte differentiation assay. We previously reported that differentiation of MEFs transduced with Myc-SREBP1 was attenuated in the absence of PERK (8). We measured levels of Akt phosphorylation in PERK wild-type and PERK-null MEFs during in vitro differentiation. Akt phosphorylation was severely attenuated in the absence of PERK (Fig. 7a). This correlated with reduced accumulation of lipid droplets in PERK-null MEFs as evidenced by the Oil Red O staining (Fig. 7b). We isolated p85α, PERK, and the insulin receptor substrate 1 (IRS1) by immunoprecipitation from wild-type MEFs at different stages of differentiation (Fig. 7c). Recruitment of the p85α into both IRS1 and PERK immune complexes was observed by day 7 of differentiation (Fig. 7c). p110α was detected in p85α and IRS1 but not in PERK immune complexes even though total levels of p85α were similar in PERK and IRS1 immune complexes (Fig. 7c). Finally, we measured PERK lipid kinase activity and detected an increase in the ability of immunopurified PERK to phosphorylate phosphatidylinositol in vitro on day 7 and day 10 of differentiation (Fig. 7d), consistent with recruitment of p85α into PERK immune complexes at this time (Fig. 7c). The increase in PERK activity does not reflect changes in PERK expression, as PERK levels remain constant through the first 7 days of differentiation followed by a modest decline at day 10 (8). These data provide proof of principle for the importance of PERK lipid kinase activity in a physiologically relevant model.
The ER can be viewed as a sensor of metabolic status in the cell. Imbalance between available intracellular resources relative to the functional demand of the ER results in protein misfolding and activation of the adaptive response pathway referred to as the unfolded protein response (UPR) and sometimes as the integrated stress response (ISR) due to the contribution of cytosolic signal transducers that regulate common downstream pathways. Activation of the UPR is mediated by three major proximal sensor molecules: PERK (26, 45), inositol-requiring enzyme 1 (Ire 1α/β) (47, 51), and transmembrane transcription factor ATF6 (27, 33, 52). The transcriptional programs activated by Ire1, ATF6, and PERK either adjust ER functional capacity and alleviate ER client protein load or trigger apoptosis if nutrient/energy availability is severely compromised. Accumulating evidence demonstrates that UPR signaling molecules regulate both ER functional capacity and cellular metabolic pathways. PERK (8) and Xbp1 (32, 46) have been implicated in anabolic regulation of lipid synthesis. The intersection with lipid synthesis likely reflects a need for ER membrane expansion necessary for adaptation. Another connection between UPR and cellular metabolic networks is provided by the PERK-dependent activation of the Akt pathway (25, 29, 31). Akt is the primary regulator of anabolic metabolism in the cell via its effects on glucose transport (14, 53), capture by hexokinase, and utilization (15, 23, 40).
Initial work suggested that ER stress-dependent Akt activation depends on phosphorylation of eIF2α and PI3K catalytic function (31). However, our data demonstrate that Akt phosphorylation is independent of eIF2α phosphorylation and p110 catalytic function. The discrepancy likely arises from use of LY294002 (29), which at the dose used also inhibits mTOR activity and, thus, Akt Ser-473 phosphorylation. Our data support a model where Akt activation is mediated by PERK-dependent generation of PA which can, in turn, either directly promote assembly of the mTORC2 complex (49) or function through SOS-Ras activation (35, 56).
Our data demonstrate that PERK can generate PA via direct phosphorylation of DAG and that PERK is required for PA generation in cells exposed to ER stress-inducing agents. Canonically, PA is generated via pathways wherein phospholipase D (PLD) utilizes phosphatidylcholine (PC) as a substrate (20) or where DAG kinases (DGKs) directly phosphorylate cellular DAG (34). The use of inhibitors that directly target each molecule failed to inhibit ER stress-dependent PA generation, suggesting that these molecules do not significantly contribute to PA generation under these conditions.
A question that arises from PERK-dependent generation of PA concerns the source of the DAG substrate. As an ER transmembrane protein kinase, PERK localization should facilitate access to phosphoinositides. Indeed, the ER is a major compartment of lipid and phospholipid generation in cells (17). DAG can be generated via phospholipase C-mediated hydrolysis of phosphoinositides. Environmental growth factors can trigger the utilization of PC at least in part via activation of protein kinase C (PKC) (4). There is little direct evidence to support an active role for PKC in ER stress signaling. However, PKCδ activation has been observed in steatohepatitis-associated ER stress (24). While we cannot rule out this mechanism, there are few data to support ER stress-dependent activation of PKC in other systems. Finally, studies utilizing labeled glucose suggest that the majority of DAG comes from de novo synthesis from glycolytic precursors and thus ultimately requires the action of phosphatidic acid phosphatase, which utilizes PA as a substrate (41). As this pathway would oppose PA accumulation, it seems unlikely to play a major role during ER stress.
Akt activation promotes anabolic metabolism, which may seem counterintuitive with regard to cell adaptation to stress, as the cell needs to be reprogrammed to preserve energy and halt anabolic reactions. However, increased activity of Akt and the subsequent increase in intracellular glucose availability would help to generate additional energy via glycolysis and/or oxidative phosphorylation as well as maximize glucose-derived cellular biosynthetic precursors necessary to synthesize ER membrane lipids, thereby accelerating recovery from stress. Critically, the accumulation of long-chain fatty acids and PA at the ER acts as a signal for increased association of the phosphatidate phosphatase 1 with the ER membrane, thereby increasing synthesis of triacylglycerols (TAG) (28).
The cellular function of PA may be dependent upon the structure of the PA molecule generated. With regard to de novo synthesis from glycolytic precursors, the newly generated PA is associated with decreased mTORC2 activity. This is achieved through PA-dependent dissociation of the mTORC2-specific component Rictor (55). Thus, these data are in direct contrast with our results and those of other groups (2, 18, 49) demonstrating that PA stimulates mTOR function. A resolution to this seeming paradox likely stems from functionally and structurally distinct PA. Our data reveal that PERK favors short-chain DAG as a substrate (C12), while the chain length of PA molecules generated from the biosynthetic pools is unknown (55). The possibility that structurally distinct PA molecules could have opposing effects on cellular signaling is an exciting notion that will require additional investigation.
Although lipid synthesis and storage in adipocytes are attenuated in p110α-null MEFs, deletion of p110 was also accompanied by decreased levels of two major adipogenic transcription factors, PPARγ and C/EBPα, as well as p85α (57). Because p85α deficiency results in reduced PERK-dependent lipid kinase activity, the reduced lipid production during adipogenesis in p110α-null MEFs could also reflect reduced PERK function. Indeed, a significant increase in PERK lipid kinase activity is apparent upon p85α recruitment to PERK. Our data support a model wherein activation of the lipogenic program in adipocytes is coordinately regulated by PERK and PI3K p110α for the optimal induction of lipid synthesis.
PERK lipid kinase function is not only essential for engaging mitogenic and prosurvival pathways during a bona fide ER stress response; it also synergizes with the insulin receptor signaling. Since PERK is not generally required for growth factor (e.g., serum)-dependent activation of mitogenic pathways, it is not as yet clear how PERK synergizes with insulin. One possibility is that in addition to engaging Ras, insulin also triggers ER stress, which engages PERK, and together these signals contribute to maximum activation of relevant pathways.
In addition to connecting with PERK, monomeric p85α and p85β subunits regulate nuclear transport of the ATF6/Ire1-regulated transcription factor Xbp1 (37, 54), and this mechanism appears to be downstream of insulin receptor signaling. Critically, this novel regulatory mechanism is lost in ob/ob mice, leading to insulin resistance that could be rescued by adenoviral delivery of p85. Thus, p85-dependent regulation of at least two UPR transducers may provide a mechanism for increasing insulin sensitivity with regard to PERK and preventing development of insulin resistance due to the unresolved stress in the case of Xbp1.
Finally, the third major UPR transducer, ATF6α, was directly implicated in transcriptional upregulation of the mTOR activator Ras homolog enriched in brain (RheB), which correlated with increased levels of phospho-S6 ribosomal protein in ATF6α-positive cells, consistent with higher mTOR activity (44). Thus, the UPR/ISR/PI3K/mTOR signaling pathways appear to be hyper-connected and may cooperate to promote survival of the cell under conditions of stress.
The PI3K-mTOR pathway represents an attractive pharmacological target in cancer. We have observed that while insulin- and PI3K p110α-dependent activation of Akt is impaired in response to growth factors, its activation is essentially unaffected in response to ER stress. This may have significant implications for the development of successful treatment protocols. For instance, increased expression of ER chaperone and the ER stress marker BiP was observed in human breast tumors and cancer-derived cell lines (19, 22), suggesting that the UPR and PERK play active roles in promoting tumorigenesis. Given recent work demonstrating a role for PERK in tumor progression (5–7), it is tempting to speculate that PERK may directly contribute to tumorigenesis by promoting PA-dependent regulation of mTOR, thereby triggering Akt activation. Critically, PERK lipid kinase activity may also influence the efficacy of strategies that target PI3K/mTOR.
We thank Lucia Rameh for assistance in conduction of HPLC characterization of lipid metabolites, Craig B. Thompson for insightful comments, Lewis Cantley, Jean Zhao, Morris Birnbaum, Randal Kaufman, and Etsuko Kiyokawa for sharing cell lines and reagents, and Margarita Romero for outstanding technical assistance.
This work was supported by National Institutes of Health grants F32CA1238252 (E.B.-M.) and P01 CA104838 and a Leukemia & Lymphoma Scholar award (J.A.D.).
Published ahead of print 9 April 2012
E.B.-M. and D.P. contributed equally to the manuscript.