Lysine trimethylation regulates 78-kDa glucose-regulated protein proteostasis during endoplasmic reticulum stress

The up-regulation of chaperones such as the 78-kDa glucose-regulated protein (GRP78, also referred to as BiP or HSPA5) is part of the adaptive cellular response to endoplasmic reticulum (ER) stress. GRP78 is widely used as a marker of the unfolded protein response, associated with sustained ER stress. Here we report the discovery of a proteostatic mechanism involving GRP78 trimethylation in the context of ER stress. Using mass spectrometry–based proteomics, we identified two GRP78 fractions, one homeostatic and one induced by ER stress. ER stress leads to de novo biosynthesis of non-trimethylated GRP78, whereas homeostatic, METTL21A-dependent lysine 585–trimethylated GRP78 is reduced. This proteostatic mechanism, dependent on the posttranslational modification of GRP78, allows cells to differentially regulate specific protein abundance during cellular stress.

The endoplasmic reticulum (ER) 6 is critical for protein biosynthesis and folding (1,2). Disturbed ER homeostasis is characterized by accumulation of mis-and/or unfolded proteins and referred to as ER stress (ERS). To counteract the detrimental effects of ERS, cells have evolved a highly conserved unfolded protein response (UPR). The UPR is primarily an adaptive response that attenuates protein translation, induces ER-associated protein degradation, and increases ER chaperone expression (1,2). When prolonged ERS persists, cells ultimately initiate cell death programs (3)(4)(5).
The ER-resident chaperone protein GRP78 plays a critical role in sensing ERS and triggering the UPR (1,2). In particular, accumulating mis-and/or unfolded proteins place an increas-ing demand on GRP78 that results in its dissociation from the three transmembrane receptors PKR-like ER kinase (PERK), activating transcription factor 6 (ATF6), and inositol-requiring enzyme 1 (IRE1), leading to their activation and triggering the UPR (1,2,6,7). The UPR induces an up-regulation of the chaperone GRP78 to enhance the folding capacity of the ER.
Although numerous studies have used the increase in GRP78 protein abundance, assayed by Western blotting, as a marker of ERS (8), little is known about GRP78 proteostasis and the role of posttranslational modifications in this process. We were intrigued by initial proteomics experiments in mouse podocytes undergoing ERS that did not reveal the expected up-regulation of GRP78 as detected by mass spectrometry. Subsequent work in podocytes, a mouse pancreatic beta islet cell line (MIN6), and HEK cells led to the discovery that, in podocytes and MIN6 cells, ERS triggers the de novo synthesis of nontrimethylated GRP78 and simultaneous degradation of existing, lysine-trimethylated GRP78. This posttranslational modification requires the activity of N-lysine methyltransferase 21A (METTL21), and the critical residue for trimethylation is lysine 585 of GRP78.

Identification of a distinct, ERS-inducible GRP78
Podocytes are the only postmitotic cell population of the kidney, and therefore their adaptations to ERS may be particularly important for their survival as well as kidney health (9,10). To better understand the molecular pathways involved in podocyte ERS, we analyzed podocytes in the presence or absence of the ERS inducer thapsigargin by MS-based proteomics, using isobaric tandem mass tag (TMT10, see "Experimental Procedures") labeling for quantification, to characterize the ERS-associated podocyte proteome (Fig. 1A). Although GADD34, ATF4, and CHOP were up-regulated in the setting of ERS, as expected (9, 10), GRP78 was not significantly up-regulated (Fig.  1A). This result was contradictory to prior studies based on Western blot analyses for GRP78 protein abundance under the same conditions (9,10).
We first confirmed the induction of GRP78 after thapsigargin-induced ERS at the mRNA (Fig. 1B) and protein levels (Fig. 1C). We next asked how MS proteomics experiments led to differences in GRP78 protein abundance detection compared with Western blot analyses. To identify the underlying reason for the observed differences, we conducted a series of Western blot experiments using three different anti-GRP78 antibodies directed against different epitopes ( Fig. 2A). In contrast to the anti-GRP78 ERS antibody (3177, Cell Signaling Technology) that we (9, 10) and others (6) have routinely applied as a marker of ERS, anti-GRP78 antibodies directed against two other regions, on the N terminus (ab32618, Abcam) or the C terminus (MABC675, EMD Millipore), aligned with the proteomics data, showing no change in total GRP78 (GFP78 TOT ) after treatment with thapsigargin ( Fig. 2B) as well as the ER stressor tunicamycin (an N-glycosylation inhibitor) (supplemental Fig. 1, A and B). The specificity of these antibodies was confirmed in GRP78-silenced podocytes (Fig. 2B). Anti-GRP78 TOT (C-terminal) and anti-GRP78 ERS antibodies were subsequently used to assess these two distinct GRP78 proteotypes in other cell types. Similar to podocytes, thapsigargin treatment did not alter GRP78 TOT levels in MIN6 cells (Fig.  2C). In contrast, we noted increased levels of GRP78 TOT as well as strong up-regulation of GRP78 ERS in HEK cells (Fig. 2C). Taken together, these results suggested that GRP78 protein abundance is regulated and fine-tuned and hinted at putative mechanisms for tight modulation of GRP78 proteostasis in specialized, differentiated cells such as MIN6 pancreatic beta cells and kidney podocytes.
Given that the anti-GRP78 ERS antibody was produced by a synthetic peptide corresponding to the region surrounding G584, we hypothesized that a modification in its vicinity could differentiate this distinct GRP78 ERS fraction. This prompted us to take a closer look at the individual -fold changes of the 58 unique peptides corresponding to GRP78 identified by proteomics. Although 57 of the peptides showed no change in abundance upon treatment with thapsigargin, a single peptide corresponding to residues Leu 587 -Lys 597 (LSSEDKETMEK) was found to be consistently up-regulated in two independent proteomics experiments with two biological replicates each ( Fig. 3A and supplemental Table 1).
Three lines of evidence led us to the hypothesis that a posttranslational modification may affect anti-GRP78 ERS antibody binding: a region of GRP78 consisting of 10 amino acids was exclusively up-regulated, as detected by MS proteomics (Fig.  3A); an antibody against the same region specifically detected GRP78 ERS in three different cell types (Fig. 2C); and the two distinct GRP78 fractions are differentially regulated at the protein level in the setting of ERS (Fig. 2, B and C). We became interested in a GRP78 lysine residue conserved throughout evolution from yeast to human (Fig. 3B), corresponding to Lys 586 in mouse and Lys 585 in human GRP78 (Fig. 3A), and shown previously to undergo methylation (11)(12)(13). We thus looked more deeply into the proteomics data, allowing for lysine methylation in the database search. This analysis revealed two unique peptides (Leu 583 -Lys 597 and Leu 583 -Lys 592 ) that contained the trimethylated Lys 586 (represented in blue, Fig. 3, A-C) and were down-regulated under conditions of ERS (Fig. 3, A and B). The identified single up-regulated peptide, Leu 587 -Lys 597 (Fig. 3, A and C, and supplemental Table 1) was formed by tryptic cleavage after unmodified Lys 586 . Trypsin cannot cleave C-terminal

Lys 585 trimethylation and GRP78 proteostasis
to trimethylated Lys 586 , which is why this modified residue was internal in both peptides observed to contain this site (Fig. 3, A and C). Cleavage was not observed after lysine 592 because it is bounded on the N-and C-terminal side with acidic amino acids that are known to hinder cleavage by trypsin. Taken together, these experiments suggest that two proteotypes of GRP78 are present, one that is trimethylated at Lys 586 and one that is not, and that this difference could be responsible for the protein fraction detected by the C-terminal GRP78 TOT antibody but not the anti-GRP78 ERS antibody.

Simultaneous de novo synthesis of GRP78 ERS and degradation of baseline GRP78
To understand the proteostasis of GRP78, we asked whether ERS converts baseline GRP78 abundance into GRP78 ERS or whether the effect of ERS is restricted to a de novo synthesized GRP78 ERS fraction, as also suggested by the strong induction of GRP78 mRNA (Fig. 1B). Podocytes were treated with thapsigargin in the presence or absence of the translation inhibitor cycloheximide. Cycloheximide blocked the thapsigargin-mediated GRP78 ERS induction, whereas GRP78 TOT levels decreased, indicating that baseline, pre-ERS GRP78 may be degraded during ERS (Fig. 4A). In support of this, adding hydroxychloroquine, an inhibitor of lysosomal protein degradation, resulted in the preservation of GRP78 TOT abundance. The restored GRP78 TOT did not correlate with an increase in GRP78 ERS , confirming that these are two distinct GRP78 fractions (Fig.  4A). The proteasome inhibitor lactacystin did not block the degradation of baseline GRP78 (supplemental Fig. 1C). Together, these data show that GRP78 ERS is de novo synthesized under ERS, whereas baseline GRP78 is likely degraded in the lysosome.
The proteomics experiment suggested that homeostatic, pre-ERS GRP78 is trimethylated on Lys 586 . We confirmed this using a lysine trimethylation-specific antibody (K-3Me) that detected a 78-kDa protein corresponding in molecular size and abundance to GRP78 TOT (Fig. 4A). Although GRP78 is a chaperone that binds unfolded proteins during ERS, it has also been reported that it can form pools of inactive oligomers ready to be activated as needed (14). We therefore wondered whether extended heat denaturing could further dissociate GRP78-containing protein complexes to increase the fraction of linearized, monomeric GRP78 detectable by Western blotting. To this end, we heat-denatured protein lysates for 4 -24 h (Fig. 4B) and noted that, indeed, this correlated with a progressively increased abundance of linearized, monomeric GRP78 (Fig. 4B) (14). In this experiment, GRP78 abundance in control, non-ERS cells correlated with the protein abundance recognized by the K-3Me antibody at 78 kDa, which was absent in ERS-induced podocytes (Fig. 4B), in keeping with the notion that only homeostatic, pre-ERS GRP78 is trimethylated. To further investigate whether the observed lysine trimethylation protein band at 78 kDa corresponds to baseline, pre-ERS GRP78, we looked for the 78-kDa lysine trimethylation protein band in GRP78-depleted cells. As expected, the 78-kDa lysine trimethylation protein band was diminished after GRP78 silencing (Fig.  4C). Together, these results confirmed the MS data that lysine trimethylation is responsible for the differences between these two distinct GRP78 fractions.
Based on previous studies, a candidate enzyme for trimethylation of Lys 586 in mouse GRP78 (or Lys 585 in human GRP78) is METTL21A (11,12), which has been annotated as a cytosolic enzyme but also localizes to the ER (Fig. 4E), likely because of its C-terminal ER-targeting peptide REDL (mouse) or KEDL (human). We asked whether depleting podocytes of METTL21A would lead to a reduction in GRP78 trimethylation and whether this would affect GRP78 proteostasis. In METTL21A-silenced podocytes, GRP78 TOT levels remained constant, but GRP78 ERS became detectable in the absence of thapsigargin/ERS, as assessed by several UPR markers (Chop (Fig. 4C) and Chop, Xbp1s, and Gadd34 (Fig. 4D)). In addition, Lys 585 trimethylation and GRP78 proteostasis the amount of detectable GRP78 ERS correlated with the Mettl21a knockdown efficiency (Fig. 4D). These findings supported Lys 586 (Lys 585 ) trimethylation of GRP78 by METTL21A.
Last, to further evaluate Lys 585 as the residue responsible for trimethylation and as a marker to distinguish baseline GRP78 from GRP ERS , we generated the trimethylation-resistant point mutant myc-tagged human GRP78(K585R) and expressed it in HEK cells. Overexpression itself induced an ERS response, as indicated by the presence of endogenous GRP78 ERS (Fig. 4F).
In immunoprecipitation experiments using anti-myc, anti-GRP ERS detected wild-type, myc-tagged GRP78 but not the point mutant myc-GRP78(K585R) (Fig. 4F). This experiment thus confirmed that Lys 585 is essential for anti-GRP78 ERS binding, and it is therefore a critical residue to distinguish nonmethylated GRP78 ERS from baseline, trimethylated GRP78 (Fig. 4F).

Discussion
This study revealed lysine trimethylation of GRP78 as a distinguishing characteristic of its baseline, pre-ERS fraction compared with a trimethylation-independent, inducible, ERS-associated fraction. We were led to this discovery by MS proteomics

Lys 585 trimethylation and GRP78 proteostasis
experiments on samples from cells under ERS, which revealed a posttranslationally modified fraction of GRP78.
An intriguing discovery in this study is that, in highly differentiated cells such as podocytes, ERS induces de novo biosynthesis of GRP78, whereas baseline METTL21A-dependent trimethylated GRP78 undergoes degradation (Fig. 4G). Of interest, ERS-mediated lysosomal degradation was also observed for other ER-resident chaperones (supplemental Fig.  1D), which is in agreement with a recent study (15). These findings suggest a substantial turnover of ER-resident proteins during chronic ERS. We speculate that this previously unrecognized mechanism may be a particularly important adaptive response in postmitotic cells whose defenses against cellular stressors must be heightened to secure survival (16). Although future work is needed to better understand cell-specific versus conserved mechanisms of GRP78 and other ER chaperone proteostasis, our results show that pancreatic beta cells (MIN6) and kidney podocytes exhibit elaborate and efficient regulation of GRP78 turnover as part of their adaptive response to ERS.
The absence of METTL21A-dependent trimethylation at Lys 585 on GRP78 ERS is intriguing, because this lysine is conserved among ER chaperone/heat shock proteins, and it has been shown to affect heat shock protein A8 (HSPA8) function (11,12). Moreover, mouse Lys 586 or human Lys 585 is located in the "lid" domain of GRP78, which is open in the ATP-bound state (low affinity for substrates) and closes upon ATP hydrolysis (stable substrate binding) (11,17). Therefore, it is intriguing to speculate that the lack of posttranslational modification may alter the conformation of GRP78 in a way that may be beneficial during ER stress. Future studies will address whether the trimethylation of Lys 585 /Lys 586 affects GRP78 substrate binding, localization, or susceptibility to lysosomal degradation. In conclusion, our study reveals a previously unrecognized complexity in GRP78 proteostasis under conditions of cellular stress, with implications for cellular response to stress across a wide range of cell types.

Proteomics
Total protein was isolated from podocytes treated with 2.5 M thapsigargin or vehicle (DMSO) for 16 h. Protein lysates, two biological replicates per condition, were prepared in 8 M urea, 75 mM NaCl, 1 mM EDTA in 50 mM Tris HCl (pH 8), 2 g/ml aprotinin (A6103, Sigma), 10 g/ml leupeptin (11017101001, Roche), and 1 mM PMSF (78830, Sigma), and deep coverage quantitative mass spectrometric analysis was performed as reported previously (24 -26) following TMT10 labeling and offline HPLC fractionation. All MS data were interpreted using the Spectrum Mill software package v5.0 prerelease (Agilent Technologies, Santa Clara, CA). Similar MS/MS spectra acquired on the same precursor m/z within Ϯ 60 s were merged. MS/MS spectra were excluded from searching when they failed the quality filter by not having a sequence tag length Ͼ 0 or did not have a precursor MHϩ in the range of 750 -6000. MS/MS spectra were searched against the UniProt Lys 585 trimethylation and GRP78 proteostasis mouse protein database. All spectra were allowed Ϯ 20 ppm mass tolerance for precursor and product ions, 30% minimum matched peak intensity, and "trypsin allow P" enzyme specificity with up to four missed cleavages. For the proteome search, carbamidomethylation at cysteine was searched as a fixed modification, and TMT10 labeling was required at lysine, but peptide N termini were allowed to be either labeled or unlabeled. Allowed variable modifications for whole proteome searches were acetylation of protein N termini and oxidized methionine. Allowed variable modifications for the trimethylation search were trimethylation of lysine and oxidation of methionine with a precursor MHϩ shift range of Ϫ200 to 100 Da. Identities interpreted for individual spectra were automatically designated as confidently assigned using the Spectrum Mill autovalidation module to use target-decoy-based false discovery rate estimates to apply score threshold criteria. Peptide level TMT ratios were calculated as the median of all PSM level ratios contributing to a protein subgroup.

Statistics
A representative experiment of at least three repeats is shown. Individual experiments were performed in triplicates, and data are expressed as means and standard deviations. Significance of differences was calculated with analysis of variance and Bonferroni post hoc tests using the Prism 6 program. The significance level was set to p Ͻ 0.05.