Initial evidence for regulated LLO destruction in intact cells—generation of Dol-P by ER stress
Transfer of the 14-sugar G3
glycan from LLO to nascent ER polypeptide yields Dol-P-P as a byproduct, which is efficiently recycled to Dol-P to seed a new round of LLO synthesis (Schenk et al., 2001
). As shown by others (Spiro et al., 1976
Schmitt and Elbein, 1979
Hubbard and Robbins, 1980
Grant and Lennarz, 1983
) and reproduced by us (Gao and Lehrman, 2002b
), this “dolichol cycle” is interrupted by inhibitors of protein synthesis such as cycloheximide (CHX), which traps the Dol-P in LLO molecules. It therefore seemed reasonable that physiological stimuli that slow translation would also hinder the dolichol cycle and diminish Dol-P availability. One such translation-attenuating activity is mediated by the ER-associated transmembrane kinase “PKR-like ER kinase” (PERK, also known as PEK [
Sood et al., 2000
]). PERK has a lumenal stress-sensing domain and a cytosolic Ser/Thr kinase domain. In response to ER stress, PERK phosphorylates the translation initiation factor eIF2α, which results in attenuated translation initiation for ER client polypeptides (). Thus, we devised experiments to test the hypothesis that PERK activation would trap a large portion of the Dol-P pool as LLO.
Dol-P levels in normal mouse embryonic fibroblasts (MEFs) with functional PERK were consequently evaluated after treatment with either of two well-known ER stressors: dithiothreitol (DTT), which reduces disulfide bonds of ER proteins; and thapsigargin (TG), which blocks ER calcium channel ATPase activity, resulting in calcium loss. As for the M6P-dependent reaction discussed in the Introduction
, proper measurement of Dol-P in the dolichol cycle requires retention of fragile ER structure by gentle permeabilization with SLO. This procedure is unlike those employing homogenization and isolation of microsomal membranes, which obscure this pool of Dol-P (Gao and Lehrman, 2002b
). Thus, SLO treatment permits reliable study of the dolichol cycle with radioactive nucleotide sugar donors for Dol-P-dependent enzymes.
presents the reactions and reagents used in this analysis.
FIGURE 2: ER stress causes release of Dol-P from a preexisting LLO pool. (A) Influences of ER stressors and other factors on the dolichol cycle, and assay of Dol-P with nucleotide-[3H]sugars. Reactions in black occur in intact cells, and those in gray are in SLO-permeabilized (more ...)
In the first step of LLO formation, UDP-GlcNAc donates GlcNAc-1-P for Dol-P-dependent synthesis of GlcNAc-P-P-Dol (Lehrman, 1991
). Treatment with a three–amino acid OT acceptor peptide (AP), resulting in transfer of the LLO-associated glycan to the peptide and generation of Dol-P, increased GlcNAc-P-P-Dol synthesis in PERK+/+
samples to ~260% of control levels (). Although PERK-dependent translation attenuation was expected to strongly decrease GlcNAc-P-P-Dol synthesis by depleting Dol-P, only a mild loss of GlcNAc-P-P-Dol was measured with DTT (~60% of control levels). More surprisingly, an increase (~160% of control levels) in activity was detected with TG. Together, these data suggested that two opposing effects were in play: the anticipated suppression of Dol-P levels by ER stress due to PERK-dependent translation attenuation (), and offsetting production of Dol-P by an unknown effect of ER stress. Testing this hypothesis with PERK−/−
MEFs () revealed that treatments with DTT or TG each substantially increased Dol-P (~250% of unstressed controls). This represented a large fraction of the Dol-P in PERK−/−
MEFs that could be released from LLO by discharge with AP (~370% of controls). When mixed, the effects of DTT and AP were not additive, suggesting they acted upon the same pools of LLO and Dol-P (), and DTT was not behaving as a nonspecific enzyme stimulator.
These results with PERK−/− MEFs were not limited to the GlcNAc-1-P transferase itself, because similar effects were detected with GDP-mannose used to evaluate Dol-P-dependent mannose-P-dolichol synthesis (product was increased to ~180% of unstressed controls by ER stress [
]). In sharp contrast to ER stressors, CHX reduced GlcNAc-P-P-Dol synthesis to ~45% of unstressed controls, demonstrating that the dolichol cycle had the expected sensitivity to translation rate (). The ability of CHX to trap Dol-P as LLO was countered by acceptor peptide, which released the trapped Dol-P. As in
, the effects of the DTT and AP treatments shown in
were not additive.
Since DTT could not supplement the effects of AP, ER stress did not appear to cause new Dol-P synthesis. Rather, the results were best explained by ER stress liberating Dol-P from preformed LLO, as did AP, by cleavage within the pyrophosphate linkage (accompanied if necessary by either pyrophosphatase or dolichol kinase action [
Schenk et al., 2001
]). As shown in
, active PERK obscured this observation. Consequently, for clarity, most of our subsequent biochemical analysis of this unexpected LLO degradation was performed with PERK−/−
MEFs, although key results were repeated with PERK+/+
ER stress increases M6P and decreases steady-state LLO levels
No signaling pathways are known that explain LLO cleavage and Dol-P generation in ER-stressed MEFs indicated in the schematic in
. As reviewed in the Introduction, we previously demonstrated that M6P can stimulate release of free glycans from LLO in permeabilized cells (, inset), and therefore speculated this mechanism might be utilized by cells under the type of ER stress induced by DTT and TG. This hypothesis makes several testable predictions: 1) ER stress should elevate M6P to concentrations sufficient to induce LLO cleavage; 2) elevated M6P levels caused by ER stress should decrease LLOs and increase the products of their cleavage—namely Dol-P () and free glycans; and 3) the process should be regulated by an authentic ER stress–sensing molecule. We thus set about testing these predictions systematically, beginning with an assessment of M6P responsiveness and levels in MEFs.
We used fluorophore-assisted carbohydrate electrophoresis (FACE) as a general method for profiling steady-state levels of intracellular saccharides and associated molecules (e.g., LLOs). FACE takes advantage of fluorophore conjugation to carbohydrates, with differing purification schemes and fluorophore reactivities useful in profiling different carbohydrate species. FACE also has the advantage of not requiring glucose deprivation of cells, which is often used during metabolic labeling studies and which is capable of perturbing ER function and saccharide analysis (Gao and Lehrman, 2002a
). Using FACE, we first asked whether M6P specifically stimulated release of G3
in SLO-treated MEFs, a permeabilized cell type that had not been previously evaluated for this reaction.
As shown in
, 50 μM M6P, but not G6P, promoted G3
release from LLO in SLO-permeabilized MEFs, indicating these cells contained the components needed to respond to M6P. We previously reported that several hexose-Ps in mammalian cells were mobilized by ER stress, presumably to aid glycoconjugate synthesis and ATP production (Doerrler and Lehrman, 1999
Gill et al., 2002
). To determine specifically whether M6P increased in ER-stressed MEFs, we coupled cellular pools of hexose-Ps to the FACE fluorophore 2-aminoacridone (AMAC). As shown in
, glucose-1-P (G1P), G6P, fructose-6-P (F6P), and M6P (as well as several unidentified species) were all increased by DTT and TG. When corrected for cellular volume, M6P concentrations in PERK−/−
MEFs increased about fourfold, from 60 μM to 250 μM (). Notably, G1P is a direct product and the other hexose-Ps are indirect products of glycogen breakdown by glycogen phosphorylase (GP).
FIGURE 3: ER stress elevates M6P. (A) SLO-permeabilized normal (MPI+/+) MEFs were incubated 1 h at 37°C with nucleotide sugar LLO precursors in the absence or presence of 100 μM G6P, 100 μM M6P, and/or 200 μM M6Po. Released free (more ...)
These data showed that MEFs contain the M6P-responsive component(s) needed for LLO cleavage, and that ER stress can increase M6P in intact MEFs in the required concentration range. To directly assess whether the elevated M6P might have caused LLO cleavage, and therefore provide an explanation for the generation of Dol-P ( and
) in intact stressed MEFs, we used a variation of the FACE technique in which the glycan portions of LLOs were released from the lipid carrier with mild acid, and then labeled with the fluorophore 7-amino-1,3-naphthalenedisulfonic acid (ANDS;
Gao and Lehrman, 2006
). Interestingly, both DTT and TG stresses resulted in a statistically significant suppression of G3
-P-P-Dol levels in PERK−/−
MEFs (). No truncated LLOs were generated (). Thus, there was no degradation within the LLO glycan moiety, further supporting the hypothesis that ER stress activated LLO cleavage within the pyrophosphate linkage.
FIGURE 4: LLO destruction in ER-stressed MEFs. (A) LLO cleavage is expected to elevate levels of Dol-P/Dol-P-P and lumenal free glycans, the latter digested by lumenal glycosidases. (B) FACE measurements of G3M9Gn2-P-P-Dol for DTT- (black) or TG- (gray) treated (more ...)
ER stress causes the appearance of LLO-derived free glycans
Using FACE, we next measured levels of free glycans predicted by hydrolysis of G3
-P-P-Dol (). Conjugation of the ANDS fluorophore requires a reducing end on the oligosaccharide, which could result from cleavage between the reducing terminal GlcNAc and the pyrophosphate linkage, or cleavage within the pyrophosphate linkage followed by dephosphorylation of the resulting phospho-oligosaccharide. After DTT or TG treatments, free glycan pools in PERK−/−
MEFs were altered significantly () with an approximately twofold increase of the total pool (presented below in
). Although G3
itself was not apparent in free glycan pools, it could be preserved by inclusion of the glucosidase inhibitor castanospermine (Figure S1), consistent with rapid digestion by intracellular glycosidases. However, castanospermine was not used routinely in our experiments because it causes ER stress (Doerrler and Lehrman, 1999
FIGURE 5: ER stress generates LLO-derived lumenal free glycans. (A) Free glycans in PERK−/− MEFs after treatments with DTT, TG, or TN. The positions of M5Gn2 and G3M9Gn2 standards on FACE gels are shown. (B) PERK−/− MEFs were left (more ...)
FIGURE 6: Glycogen phosphorylase inhibitors block ER stress–induced hexose-P mobilization and free glycan increases. (A) Scheme linking glycogenolysis and extracellular mannose to hexose-Ps and GDP-mannose. Three GP inhibitors are indicated. Determinations (more ...)
Significantly, free glycans released by ER stress were sensitive to diagnostic enzymatic digestions for LLO-type structures. Glycans that migrated more slowly than the G6 standard (), in the range expected for M6Gn2 to G3M9Gn2, were susceptible to endoglycosidase H—an enzyme that senses the presence of a specific α1,3-linked mannosyl residue on mature LLOs and many of the larger LLO intermediates. Moreover, all of the glycans were sensitive to jack bean α-mannosidase. This enzyme cleaves exposed α-linked mannosyl residues expected on LLO-derived glycans. The observation of a resistant fragment in the G5–G6 range is consistent with the presence of blocking glucosyl residues, which are also a feature of some LLO-derived glycans. Tunicamycin (TN), an inhibitor of LLO synthesis as well as an inducer of ER stress, did not increase free glycans ().
These data were highly consistent with destruction of LLO by ER stress, yielding Dol-P and free glycans (). However, a possibility remained that the free glycan increases were actually caused by ERAD of nascent glycoproteins destabilized by ER stressors (). This could be especially severe in PERK−/−
MEFs, which are unable to limit new glycoprotein synthesis in response to ER stress. To distinguish between LLO cleavage and ERAD, we analyzed the subcellular compartmentalization of the free glycans. Glycans increased by ERAD should be cytosolic, while LLO-derived glycans should be lumenal (Chantret and Moore, 2008
). These populations can be distinguished as retained (lumenal) and diffusible (cytosolic) glycan pools after permeabilization with SLO (Moore et al., 1995
As expected, the two types of free glycan pools in MEFs were distinctly different (). Significantly, both DTT and TG increased lumenal free glycan pools, while there were no effects on cytosolic pools. This showed that ER stress caused LLO cleavage, but without a free glycan contribution by ERAD. Further, z-VAK-fmk (an inhibitor of cytosolic N-
Misaghi et al., 2004
]), did not prevent the ER stress–dependent increase of free glycans. Experiments presented with inhibitors of glycogen phosphorylase also address ERAD.
In summary, ER stress–induced increases of M6P in intact cells correspond with loss of G3M9Gn2-P-P-Dol and appearance of LLO-derived Dol-P and lumenal free glycans. While fastidious LLO biosynthesis is critical for proper N-linked glycosylation, our results suggested M6P might have a paradoxical function in impeding LLO usage. The experiments that follow, which were intended to further elucidate M6P's role, evaluated the metabolic origin of M6P, its necessity for LLO cleavage, and its signaling pathway.
ER stress–dependent M6P originates from glycogen and is necessary for LLO cleavage
We considered hypotheses for both intracellular and extracellular sources of the ER stress–generated hexose-Ps. Since ER stress does not appear to stimulate hexose transport or hexokinase activities (Doerrler and Lehrman, 1999
Gill et al., 2002
), it was unlikely that intracellular hexose-Ps were increased by these mechanisms. In contrast, ER stress did correlate with decreased glycogen content (Gill et al., 2002
), and it also correlated with increases in metabolites of glycogenolysis ( and
), suggesting that glycogen may be an intracellular source of these sugars.
The hypothesis that ER stress triggered glycogenolysis was tested by treating ER-stressed MEFs with a highly efficacious inhibitor of GP, the indole-2-carboxamide CP-91149 (;
Martin et al., 1998
). As shown in
, CP-91149 efficiently blocked the appearance of all DTT- and TG-inducible G1P, the immediate product of glycogenolysis. Equally as important, blocking glycogenolysis with CP-91149 also prevented the increases of G6P, F6P, and M6P (), all of which can be formed from G1P (). To address the possibility that a nonglycogen source of sugars was blocked by an unexpected off-target activity of CP-91149, we evaluated two additional GP inhibitors having different mechanisms than CP91149. While CP-91149 interacts with a novel GP site (Oikonomakos et al., 2000
Rath et al., 2000
), caffeine inhibits the known purine-binding site (Johnson, 1992
), and 1,4-dideoxy-1,4-imino-d
-arabinitol (DAB) targets the catalytic site (Oikonomakos et al., 2006
). Like CP-91149, these two GP inhibitors impaired DTT- and TG-dependent hexose-P mobilization (unpublished data). From these experiments and earlier results (Gill et al., 2002
), we conclude that glycogen is the source of ER stress–dependent hexose-Ps.
If LLO cleavage is caused by M6P produced via ER stress–dependent glycogenolysis, then GP inhibitors should prevent the formation of LLO-derived free glycans. As shown in
, this was indeed the case: CP-91149, caffeine, and DAB each hindered DTT- and TG-dependent free glycan increases. These results also further ruled out a contribution by ERAD (which should have been insensitive to these inhibitors), as well as LLO cleavage by a nonspecific effect of ER stress on the labile pyrophosphate bond. Collectively, these results link ER stress to glycogenolysis, elevated M6P, and LLO cleavage.
M6P introduced in the absence of ER stress is sufficient to cause LLO cleavage
To test the sufficiency of M6P for LLO cleavage by an independent method, we introduced it into cells at appropriate concentrations, but in the absence of ER stress. As shown in
, a 60-min incubation of PERK−/− MEFs in medium supplemented with 10 mM d-mannose increased intracellular M6P by three- to sixfold, to ~180 μM (). Using analyses similar to those described in the preceding sections for ER-stressed cells, we found that these conditions stimulated LLO cleavage: G3M9Gn2-P-P-Dol decreased to 82 ± 2% (n = 2) of untreated levels, and significantly, lumenal free glycans increased (). There was no effect of incubation with 10 mM d-galactose, the 2- and 4-epimer of d-mannose. LLO cleavage was therefore not explained by nonspecific osmotic factors.
FIGURE 7: Elevation of M6P in the absence of ER stress releases free glycans. (A–D) PERK−/− MEFs were incubated with (A, B) variable concentrations of mannose for 60 min; (C) 10 mM mannose for variable times; or (D) 10 mM mannose for 60 (more ...)
Mannose treatment itself did not cause ER stress, because stress-associated hexose-6-Ps other than M6P (i.e., G6P and F6P) were not elevated ( and
). Mannose treatment of MEFs did not increase GRP78/BiP mRNA, and neither brief nor extended mannose treatments increased splicing of XBP1 mRNA (Figure S2), both sensitive markers for ER stress signaling. Mannose treatment generated free glycans even during CHX treatment, to prevent synthesis of newly synthesized glycoproteins, which might be ERAD substrates and hence glycan sources (unpublished data). Therefore, we find no evidence that the free glycans generated by mannose treatment were due to inadvertent ER stress or ERAD.
As seen in
, 10 mM mannose treatment increased both M6P and M1P. However, M1P was not increased by ER stress, and none of the other ER stress-induced metabolites of glycogenolysis (G1P, G6P, and F6P) were increased by mannose treatment. Thus M6P was the only hexose-P increased under all conditions causing LLO cleavage. It is unclear why elevated M6P appeared to drive formation of M1P (a single enzymatic step;
) after mannose incubation; but not with ER stress; perhaps ER stress also promotes reactions that consume M1P. Coupled with the strong specificity for M6P in permeabilized cell experiments (;
Gao et al., 2005
), we conclude that increased M6P is sufficient to drive LLO cleavage in intact cells, even in the absence of ER stress.
Inhibition by a structural analogue reveals a defined M6P target
Taken together, our data surprisingly implicate M6P as a signaling molecule. As such, it would be predicted to have a defined target, and be regulated by other signaling components. To understand these biological properties in greater detail, we adopted two approaches. First, we probed for the presence of an M6P interaction site in LLO cleavage reactions. Second, we sought to identify the signal transducer that triggers glycogenolysis (and hence M6P production) in response to ER stress. Some aspects of the M6P interaction site could already be deduced by our earlier permeabilized-cell experiments. Our findings that both M1P and G6P (the 2-epimer of M6P) failed to promote LLO cleavage suggested that the orientation of the 2-hydroxyl and the position of the phosphate in M6P are important for the M6P interaction and M6P-dependent signaling.
To further explore the requirements for M6P's interaction with a target site, we synthesized the nonhydrolyzable analogue mannose-6-phosphonate (M6Po; Figure S3A; compound 11 in
Belakhov et al. [2004
]). The M6Po preparation was free of detectable contaminating M6P (Figure S3B), and M6Po alone did not cause LLO cleavage with SLO-permeabilized cells (Figure S3C). However, inclusion of M6Po partially inhibited LLO cleavage caused by M6P, while inclusion of G6P or M1P had no effect ( and S3D). The esterifying oxygen atom of M6P is therefore important for LLO cleavage, but not for binding to the M6P target site, allowing M6Po to antagonize M6P. It remains to be determined whether the esterifying oxygen mediates an activation step at the target site, and/or is involved in a requirement for M6P hydrolysis.
Authentic UPR signaling mobilizes hexose phosphates
To identify signaling components that might regulate M6P, we focused upon the abilities of DTT and TG to promote glycogenolysis, and considered two possibilities. First, as known ER stressors, DTT and TG might have triggered the unfolded protein response (UPR), an elaborate set of signaling events that occur in response to the presence of excess misfolded protein in the ER lumen (Schröder and Kaufman, 2005
Ron and Walter, 2007
). Alternatively, these agents may have unexpectedly altered hexose metabolic pathways by their effects on redox potential and calcium homeostasis. If the UPR was involved, hexose-P mobilization should be diminished by stable overexpression of GRP78/BiP (Dorner et al., 1992
) to counteract dissociation of this chaperone from lumenal stress-sensing domains of UPR-signaling proteins and attenuate their activities ().
FIGURE 8: UPR signaling by IRE1α mobilizes hexose-Ps. (A) UPR-dependent actions of chemicals that cause protein misfolding can be distinguished from other possible effects of these chemicals by overexpression of GRP78/BiP. Under normal conditions, sufficient (more ...)
As shown in Figure S4A, overexpression of GRP78/BiP caused the expected reduction of UPR-dependent XBP1 splicing by either DTT or TG. Since the GRP78/BiP overexpressing cells were derived from tumor-like CHO-K1 cells, they were expected to have enhanced glycolysis (Levine and Puzio-Kuter, 2010
), which could complicate hexose-P analysis; most of the F6P from glycogenolysis would likely be catabolized for ATP production, leaving little for conversion to M6P. We therefore focused on G6P, the intermediate preceding F6P, as an alternative readout for glycogenolysis (). As shown in
, DTT and TG clearly elevated G6P in unmodified CHO-K1 cells, while G6P production was suppressed by overexpressing GRP78/BiP. Thus, one or more UPR-signaling proteins were responsible for mobilization of hexose-Ps.
Hexose-6-P mobilization is fast (detectable within 10 min of DTT application [
Gill et al., 2002
]), arguing against transcriptional control. This focused our attention on the UPR transmembrane kinases PERK and inositol-requiring enzyme 1 (IRE1), which employ rapid autophosphorylation mechanisms and can be evaluated directly with MEF knockout lines. Knockout of PERK did not affect the abilities of DTT or TG to mobilize G6P or M6P ( and
). However, the response was blocked by knockout of IRE1α (). Three nucleotide sugars (UDP-GlcNAc, GDP-mannose, and UDP-glucose) synthesized from hexose-Ps also increased with ER stress, but also required IRE1α (Figure S4B). These results implicated IRE1α as an essential UPR component involved in glycogenolysis and hexose-P generation.
To independently assess the requirement for IRE1α, we used RNA interference. Suppression of IRE1α in normal MEFs with four pooled small interfering RNAs (siRNAs) caused 68–95% loss of IRE1α protein () and inhibited hexose-P mobilization (). This result was extended with two individual siRNAs from the pool found most responsible for IRE1α suppression. Knockdown with either siRNA (designated siRNA1 and siRNA2, which reduced IRE1α protein to 48% or 38% of the buffer control, respectively) also prevented hexose-P mobilization. By comparison, a pool of four nontargeting siRNAs had no effect (). Taken together, knockout and knockdown experiments showed that IRE1α is necessary for hexose-P mobilization by the UPR.
Signaling by the IRE1α kinase domain triggers glycogenolysis
IRE1α is a single-pass transmembrane protein composed of a lumenal stress-sensing domain, and a cytosolic unit with a Ser/Thr kinase domain followed by an RNase domain. Both of the cytosolic domains are catalytically functional. In metazoans, IRE1α exerts most of its homeostatic effects by activating transcriptional programs via the ability of the RNase domain to regulate splicing of mRNA encoding the transcription factor XBP1 (Schröder and Kaufman, 2005
Ron and Walter, 2007
). However, hexose-P mobilization and nucleotide sugar increases still occurred in ER-stressed XBP1−/−
MEFs, although the TG responses were somewhat attenuated compared with DTT responses for reasons which remain unclear ( and S4C). The absence of a requirement for XBP1 suggested two potential mechanisms for increasing hexose-Ps. As shown previously, it was possible that the IRE1α RNase domain caused degradation of a preexisting regulatory mRNA (Hollien and Weissman, 2006
Hetz and Glimcher, 2009
). Such a mechanism might have degraded an mRNA involved in limiting glycogenolysis, causing elevation of hexose-P levels. Alternatively, IRE1α might have sent a signal directly from its kinase domain to stimulate glycogenolysis, without participation by the RNase domain.
To determine whether the entire cytosolic unit (kinase plus RNase domains) or the isolated kinase domain of IRE1α could activate glycogenolysis, we employed strategies used previously for expressing soluble fusions of the cytosolic units of yeast IRE (Aragon et al., 2009
) and mammalian IRE1α (Back et al., 2006
) to FKBP12-derived modules that bind the dimerizer AP20187 (). By dimerizing the fusion proteins, AP20187 promotes direct interaction of the attached domains and consequent activation of downstream events, including those possibly requiring transautophosphorylation. Rather than transient transfection (Back et al., 2006
), we prepared stable CHO-K1 transfectants to avoid contributions of hexose-Ps from untransfected cells in transiently transfected pools (). Thus, as for
, we limited our analysis of these CHO-K1–derived lines to G6P.
FIGURE 9: Inducible activation of the IRE1α kinase domain signals hexose-P mobilization. (A) AP20187 was used to dimerize the entire cytosolic unit of IRE1α (RNase + kinase domain) or the isolated kinase domain. (B) Immunoblots (anti-HA antibody) (more ...)
In cells expressing the empty vector, AP20187 was inert as expected. However, in cells expressing the entire IRE1α cytosolic unit (kinase and RNase), AP20187 increased G6P (). This showed that IRE1α signaling is sufficient to mobilize hexose-P. As anticipated, cells expressing the kinase-only fusion protein failed to stimulate XBP1 splicing with AP20187 (Figure S5A). Yet, G6P in these cells was mobilized within 20 min of AP20187 addition at levels equivalent to those achieved with DTT-induced stress (). Preliminary experiments suggest this activity requires kinase function, because it was abrogated by a Lys599Ala kinase-dead mutation, although the mutant protein was stably expressed. The effects of IRE1α activation were transient, suggesting a process for disengagement of signaling; after an initial increase, we detected a recovery phase within 1 h (). To verify that glycogenolysis was involved, we tested the GP inhibitor CP-91149. As shown in
, CP-91149 blocked the AP20187-dependent increase of G6P in cells expressing the AP20187-responsive kinase domain. An extensive analysis of potential off-target effects of CP-91149 and AP20187 on reactions relevant to this study was performed (Figures S5B–F). Although some unanticipated effects of CP-91149 were identified, they did not affect the final outcomes of the experiments reported here. Moreover, there was no unexpected interference between CP-91149 and AP20187.
In summary, UPR signaling involving the IRE1α kinase domain activates glycogenolysis to elevate hexose-Ps. This allows IRE1α to regulate the ability of M6P to cause LLO destruction.
PERK stabilizes LLO pools despite ongoing LLO destruction
In general, the UPR plays an important role in maintaining ER homeostasis. Along these lines, the UPR can aid LLO assembly (Lehrman, 2006
). One mechanism involves a role for PERK in balancing translation rates to compensate when LLO flux is impaired (Shang et al., 2007
). Additionally, since ER stress increased hexose-Ps, which are precursors of UDP-GlcNAc, GDP-mannose, and UDP-glucose, the nucleotide sugar building blocks for LLOs, we measured these molecules directly. As shown in Figure S4B, all three nucleotide sugars were increased by the UPR in an IRE1α-dependent manner, suggesting yet another mechanism by which the UPR may assist LLO assembly. LLO destruction by the UPR therefore presents a serious paradox. To address this problem, we hypothesized that suppression of LLO levels by a destructive mechanism in normal cells under conventional ER stress might be compensated by PERK, which can reduce LLO consumption by limiting N-
glycoprotein synthesis ().
Like PERK−/− MEFs, PERK+/+ MEFs underwent LLO destruction in response to ER stress. Hexose-Ps, including M6P, were elevated ( and
) and free glycans, resulting from LLO cleavage, were formed (). The other LLO cleavage product, Dol-P, appeared to be generated as well, although PERK's translation-attenuating activity predictably reduced the amount of Dol-P detected (). Free glycan formation in PERK+/+ MEFs was also responsive to mannose treatment (). However, steady-state G3M9Gn2-P-P-Dol levels were unaffected by either DTT or TG stress in PERK+/+ MEFs (). In comparison, these stressors decreased LLO levels in PERK−/− MEFs by ~50%. This was attributable solely to differences in translation rate, because LLO levels in DTT-stressed PERK−/− MEFs were returned to normal by pharmacological rescue with the translation inhibitor CHX ().
FIGURE 10: HSV-1 infection mobilizes M6P, with destruction and depletion of LLOs. (A) Uninfected PERK−/− MEFs were treated with 2 mM DTT and/or 0.2 mM CHX for 1 h, as indicated, and G3M9Gn2-P-P-Dol was measured (n = 3, mean ± SEM). *, p < (more ...)
These results demonstrate that translation attenuation by ER stress–activated PERK can ameliorate the effect of LLO destruction by stabilizing steady-state G3M9Gn2-P-P-Dol levels. In this respect, the outcome of IRE1α/M6P signaling is antagonized by PERK. IRE1α/M6P signaling and LLO destruction themselves, however, are independent of PERK.
M6P signaling, LLO destruction, and LLO depletion during a pathogenic challenge with HSV-1
Since LLO destruction appeared benign in normal PERK-expressing cells under conventional ER stress with DTT or TG, we hypothesized that a biological context for LLO destruction might involve an abnormal condition, where loss of LLOs could be advantageous. One such condition could be infection with a pathogen required to synthesize N-
glycosylated components, a conjecture we chose to test with HSV-1, because this virus contains a number of envelope N-
glycoproteins (Compton and Courtney, 1984
). In addition, virally infected mammalian cells can attenuate translation to hinder synthesis of viral polypeptides and to initiate autophagy/xenophagy (Tallóczy et al., 2002
Orvedahl et al., 2007
), but HSV-1 has developed countermeasures to prevent translation attenuation, including that dependent upon PERK (Mulvey et al., 2007
). This increased the likelihood that any LLO destruction induced by HSV-1–associated ER stress might also suppress LLO levels, as we observed in PERK knockout cells stressed with DTT and TG ( and
Normal MEFs were mock-infected, infected with HSV-1 at saturating multiplicity, or TG-treated. Hexose-Ps, free glycans, and LLOs were then harvested. Shown in
, G6P and M6P increased similarly after TG treatment and 12 h postinfection (h.p.i.) with HSV-1, which is indicative of ER stress caused by infection. LLO destruction occurred in both cases, resulting in an increased pool of qualitatively indistinguishable free glycans (). Importantly, normal MEFs under HSV-1–induced ER stress had LLO levels decreased by ~50% (), similar to that observed with PERK−/− MEFs under conventional ER stress (). There was little effect at 5 h.p.i., when synthesis of a representative glycoprotein (gC) was barely detectable (Figure S6A), but substantial effects were identified at 8–11 h.p.i. At this level of infection, when envelope glycoprotein synthesis was abundant, there was good temporal agreement between increases of M6P, increases of free glycans, and LLO loss (). To evaluate the plausibility of this approach for characterizing LLO destruction in the future with other enveloped viruses studied elsewhere, we demonstrated that results with HSV-1–infected MEFs prepared and analyzed in MAL's laboratory () were also observed with methanolic suspensions of HSV-1-infected MEFs prepared by IM's laboratory and shipped to the MAL's laboratory for analysis (Figure S6B). Taken together, we propose that viral infection is a biological context for M6P signaling and LLO destruction, possibly to suppress levels of LLOs required by the pathogen.