A Novel Link between Fic (Filamentation Induced by cAMP)-mediated Adenylylation/AMPylation and the Unfolded Protein Response*

Background: Adenylylation/AMPylation by Fic proteins alters cellular signaling. HYPE, the sole human Fic protein, is an adenylyltransferase. Results: BiP is identified as a substrate for HYPE. HYPE adenylylates BiP and is required for UPR induction. Conclusion: HYPE regulates ER homeostasis. Significance: Adenylylation/AMPylation is a new mode of UPR regulation. This is the first demonstration of a physiological role for human HYPE. The maintenance of endoplasmic reticulum (ER) homeostasis is a critical aspect of determining cell fate and requires a properly functioning unfolded protein response (UPR). We have discovered a previously unknown role of a post-translational modification termed adenylylation/AMPylation in regulating signal transduction events during UPR induction. A family of enzymes, defined by the presence of a Fic (filamentation induced by cAMP) domain, catalyzes this adenylylation reaction. The human genome encodes a single Fic protein, called HYPE (Huntingtin yeast interacting protein E), with adenylyltransferase activity but unknown physiological target(s). Here, we demonstrate that HYPE localizes to the lumen of the endoplasmic reticulum via its hydrophobic N terminus and adenylylates the ER molecular chaperone, BiP, at Ser-365 and Thr-366. BiP functions as a sentinel for protein misfolding and maintains ER homeostasis. We found that adenylylation enhances BiP's ATPase activity, which is required for refolding misfolded proteins while coping with ER stress. Accordingly, HYPE expression levels increase upon stress. Furthermore, siRNA-mediated knockdown of HYPE prevents the induction of an unfolded protein response. Thus, we identify HYPE as a new UPR regulator and provide the first functional data for Fic-mediated adenylylation in mammalian signaling.


The maintenance of endoplasmic reticulum (ER) homeostasis is a critical aspect of determining cell fate and requires a properly functioning unfolded protein response (UPR). We have discovered a previously unknown role of a post-translational modification termed adenylylation/AMPylation in regulating signal transduction events during UPR induction. A family of enzymes, defined by the presence of a Fic (filamentation induced by cAMP)
domain, catalyzes this adenylylation reaction. The human genome encodes a single Fic protein, called HYPE (Huntingtin yeast interacting protein E), with adenylyltransferase activity but unknown physiological target(s). Here, we demonstrate that HYPE localizes to the lumen of the endoplasmic reticulum via its hydrophobic N terminus and adenylylates the ER molecular chaperone, BiP, at Ser-365 and Thr-366. BiP functions as a sentinel for protein misfolding and maintains ER homeostasis. We found that adenylylation enhances BiP's ATPase activity, which is required for refolding misfolded proteins while coping with ER stress. Accordingly, HYPE expression levels increase upon stress. Furthermore, siRNA-mediated knockdown of HYPE prevents the induction of an unfolded protein response. Thus, we identify HYPE as a new UPR regulator and provide the first functional data for Fic-mediated adenylylation in mammalian signaling.
Protein folding and maturation take place in the endoplasmic reticulum (ER), 2 where proteins enter as unfolded polypeptides and are either folded into mature proteins or misfolded into aggregates for degradation. When there is an imbalance between the rate of unfolded proteins entering the ER and the rate of protein folding by ER chaperones, ER stress is induced. Under such a condition, the ER activates an unfolded protein response (UPR), mediated by the following three ER transmembrane receptors: inositol-requiring protein-1 (IRE1), activating transcription factor-6 (ATF6). and protein kinase RNA-like ER kinase (PERK) (1,2). The UPR signaling cascade works at three different levels. First, it lowers protein synthesis and translocation into the ER, helping the cell to transiently adapt to the stress. Second, it up-regulates the transcription of ER chaperones and other protein folding machinery to increase the rate of protein folding. Third, when these responses cannot establish homeostasis, programmed cell death is triggered.
Here, we show that the human Fic protein, HYPE (Huntingtin yeast interacting protein E), is a player in the UPR cascade. HYPE belongs to the Fic (filamentation induced by cAMP) family of enzymes that uses cellular sources of nucleotides to carry out various post-translational modifications (3)(4)(5). Predominant among these post-translational modifications is the covalent addition of AMP to target substrates such as mammalian Rho GTPases (4 -8). Fic proteins are now known to utilize ATP, UTP, GTP, and CTP and their derivatives to adenylylate (AMPylate), uridinylate (UMPylate), phosphocholinate (PCylate), as well as phosphorylate protein targets in vitro and/or in vivo (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). Fic proteins are defined by an HXFX(D/E)(G/A)N(G/K)RXXR motif, with the invariant His being required for enzymatic activity (3,7). Their activity is often regulated by an inhibitory helix, defined by a conserved (S/T)XXXE(G/N) sequence, that occupies the Fic active site to prevent nucleotide docking (15). Removal of the helix from the active site by mutating the conserved Glu relieves inhibition.
The human genome encodes a single Fic protein, called HYPE/FicD (5). HYPE was identified in a yeast two-hybrid screen where it interacted with the polyQ region of huntingtin, mutations within which cause Huntington disease (16). HYPE is highly conserved among eukaryotes, with human HYPE displaying high amino acid sequence similarity to its eukaryotic orthologs as follows: mouse (Mus musculus), 89.5%; zebrafish (Danio rerio), 76%; fruit fly (Drosophila melanogaster), 55%; and worm (Caenorhabditis elegans), 45.4%.
Using Rho GTPases, RhoA, Rac1, and Cdc42, as artificial substrates in an in vitro adenylylation assay, we previously showed that HYPE preferentially uses ATP as a nucleotide source and functions as an adenylyltransferase (4). However, compared with bacterial Fic proteins like Histophilus somni IbpA that adenylylate Rho GTPases, enzymatic activity associated with WT HYPE was extremely weak (4,5). Accordingly, overexpression of HYPE in HeLa cells failed to induce a cytotoxic phenotype typical of that seen with IbpA (5). We reasoned that HYPE's weak enzymatic activity against Rho GTPases suggests that Rho GTPases may not be its physiological targets and/or that HYPE's active site was blocked by its inhibitory helix. Accordingly, a mutation of HYPE's Glu-234 to Gly in the context of HYPE's Fic domain alone (aa 187-437) was shown to relieve this inhibition by removing the inhibitory helix from the Fic active site (15).
Here, we have characterized HYPE's enzymatic activity in vitro and in tissue culture cells, and we show that uncontrolled HYPE activity induces apoptosis. Furthermore, we demonstrate that HYPE localizes to the lumen of the endoplasmic reticulum and adenylylates an ER resident chaperone, BiP, at Ser-365 and Thr-366. BiP is critical for protein quality control and maintaining ER homeostasis (1,2,17). Our findings indicate that HYPE expression is up-regulated upon induction of UPR, and knockdown of HYPE prevents activation of UPR. We postulate that HYPE regulates UPR progression by adenylylating BiP. Our study is the first functional demonstration of the physiological significance of Fic-mediated adenylylation in mammalian cell signaling.
Proteomic Analysis-Protein bands were excised from the gel, chopped into small (ϳ1 mm) pieces, destained with 25 mM NH 4 HCO 3 in 50% (v/v) acetonitrile, and digested as described previously (18). Resulting peptides were loaded into a capillary C18 column (15 cm ϫ 75 m, Agilent, Santa Clara, CA) coupled to a nano-HPLC system (1100 series, Agilent) equilibrated with 5% (v/v) mobile phase B (mobile phase A, 0.1 formic acid in water; mobile phase B, 0.1% (v/v) formic acid in acetonitrile), and eluted with the following gradient: 5% B for 5 min, 5-40% B in 50 min, 40 -90% B in 8 min, and hold at 90% B for 6 min. The flow rate was set at 300 nl/min. Eluting peptides were analyzed online by electrospray ionization on an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific, San Jose, CA). Full-MS scans were collected at 30,000 mass resolution at 400 m/z in a range of 300 -2000 m/z, and the eight most intense ions were selected for collision-induced dissociation in the linear trap (35% normalized collision energy), before being dynamically excluded for 1 min.
Data were processed and searched against the human SwissProt database (22,723 sequences, downloaded on July 9, 2013) using the MaxQuant package (version 1.4.0.8) in two rounds (19). In the first round, peptides were searched using cysteine carbamidomethylation (ϩ57.0215 Da) and 20 ppm and 0.5 Da for precursor and fragment mass tolerances, respectively. After recalibration, peptides were re-searched using the following parameters: 1) trypsin digestion in both peptide termini because of missed cleavage sites; 2) methionine oxidation (ϩ15.9949 Da) and adenylylation (ϩ329.0525 Da) of serine, threonine, and tyrosine as variable modifications; 3) cysteine carbamidomethylation as fixed modification; and 4) 4.5 ppm and 0.5 Da for precursor and fragment mass tolerances, respectively. The data were filtered with a false discovery rate of Յ0.01 in both peptide and protein levels.
Mammalian Cell Culture Assays-Human epithelium HEK293A, HEK293T, and HeLa cells and human breast cancer MCF7 cells were cultured in Dulbecco's modified Eagle's medium (Sigma) supplemented with 5% (v/v) fetal bovine serum (Life Technologies) and 100 g/ml penicillin/streptomycin (Sigma) at 37°C with 5% CO 2 . FuGENE HD (Promega) was used for transfections according to the manufacturer's directions. Cells were transfected for 10 h or 30 h, as indicated for each experiment. For detection of apoptosis, cells were incubated with DMSO for 3 h or 25 M Z-VAD-fmk (R&D Systems) or DMSO for 2 h, prior to treatment with 1 M staurosporine (Sigma). Samples were analyzed either by microscopy or by Western blot analysis following a sample preparation.
For induction of unfolded protein response, HEK293T cells were treated with 1 g/ml thapsigargin (Sigma) or 2.5 g/ml tunicamycin (Calbiochem) for specific time points. Cells were harvested, washed, and prepared for RNA isolation.
For siRNA-mediated knockdown of HYPE, HEK293T cells were transfected with 10 M FicD siRNA (Life Technologies, s22000 and s21999) for 48 h, using FuGENE HD (Promega) according to the manufacturer's directions. Cells were harvested, washed, and prepared for RNA isolation.
Fluorescence protease protection assay was done as described previously (20). Mitoneet-GFP and BiP-GFP were used as mitochondrial and ER lumen-specific controls (21).
ER-enriched fractions were prepared from fresh rat livers as described previously (22). The postnuclear supernatant was spun at 15,000 ϫ g to pellet the ER/microsome-enriched fraction, which was further separated over Histogenz density gradients. Equal amounts of protein from each fraction were separated by SDS-PAGE and analyzed by Western blotting. The ER fraction was then treated with trypsin as described previously (23). Primary antibodies against HYPE (Abcam), BiP (Cell Signaling), calnexin (Cell Signaling), and mitochondrial complex I (Mitosciences) were used to detect these proteins in the tissue fractions.
For microscopy, HEK293A cells transfected with HYPE or HYPE mutants were observed by fluorescence microscopy (Zeiss Axiovert 200 M and AMG EVOS). EGFP was excited at 488 nm; fluorescence was monitored between 510 and 520 nm. Cell morphology was observed using differential interference contrast (DIC) microscopy (Zeiss Axiovert 200 M). Fluorescence and DIC images were overlaid using the Zeiss software.
Immunoprecipitation and Western Blotting-For immunoprecipitation, HEK293T and MCF7 cells were lysed with modified RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% (v/v) Nonidet P40, containing 1.0 mM Pefabloc (Roche Applied Science), 1.0 mM benzamidine hydrochloride, 1 M leupeptin, 1 M E64). The homogenates were pre-cleared with 25 l of protein A/G beads (Pierce), centrifuged at 14,000 ϫ g at 4°C for 10 min, and the supernatant was incubated with anti-HYPE antibody (Abcam) or anti-BiP antibody (Cell Signaling) for 2-4 h at 4°C with gentle rotation followed by incubation with 25 l of protein A/G beads (Pierce) for 30 min at 4°C with gentle rotation. Beads were washed three times with ice-cold modified RIPA buffer, resuspended in 20 l of 1ϫ SDS loading buffer, and denatured at 95°C for 5-7 min.
Endoglycosidase H Treatment-HEK293A cells were collected, resuspended in cell lysis buffer (50 mM HEPES, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 100 mM NaF, 1 mM DTT, 100 M PMSF, 10% (v/v) glycerol, 1% (v/v) Triton X-100, pH 7.4), and kept on ice for 15 min. The supernatant was mixed with anti-FLAG M2-agarose (Sigma) and incubated at 4°C with gentle shaking for 1 h. The precipitate was washed three times and heated in Glycoprotein denaturing buffer (New England Biolabs) at 100°C for 10 min. The samples were then either left untreated or treated with endoglycosidase H (New England Biolabs) for 1 h at 37°C as per the manufacturer's protocol. The products were analyzed by Western blot.
RNA Isolation, RT, and qRT-PCR-Total cell RNA was extracted with the RNeasy mini kit (Qiagen). 5 g of RNA were treated with DNase (New England Biolabs) and reverse-transcribed with random primers using the Superscript III First Strand Synthesis System (Life Technologies). 2 l of a 1:10 dilution of cDNA were used for quantitative PCR with specific primers and SYBR Green PCR master mix 2 (Applied Biosystems) for 40 cycles in a CFX96 real time PCR system (Bio-Rad), according to the manufacturer's instructions, HYPE, GRP78, CHOP, and Xbp1 C t was normalized to GAPDH C t for each condition. Primers used for analysis of various genes are listed in Table 2. UPR induction was done as indicated.
RNase Binding Assay-Bovine pancreatic RNase A (Sigma) was chemically coupled to CNBr-activated Sepharose (GE Healthcare) according to the manufacturer's protocol at 0.2 mol per ml of resin volume. When mentioned, 30 nmol of RNase A matrix was denatured in 1.3 ml of 6 M guanidine hydrochloride, 0.1 M DTT, 10 mM Tris, pH 8.6, at room temperature overnight. Before use, the denaturant and DTT were washed away, and free thiols were blocked with 50 mM N-ethylmaleimide. 10 g of His-tagged wild-type BiP or His-SUMOtagged BiP S365A/T366A was incubated with 1 g of wild-type, E234G, or E234G/H363A HYPE in an adenylylation reaction containing 5 mM HEPES, pH 7.5, 1 mM manganese chloride tetrahydrate, 0.5 mM EDTA, and 1 M ATP for 15 min. The proteins were then incubated with 10 l of either the native or denatured RNase A resin in a total volume of 500 l of Buffer K at room temperature, with rotation for 1 h. The resins were washed with cold Buffer K, resuspended in SDS loading buffer, separated by SDS-PAGE, and probed for the respective proteins by Western blot.
ATPase Assay-ATPase assay was conducted using the phosphate assay kit (ab65622, Abcam). This assay measures the abundance of free inorganic phosphate (P i ) released from ATP hydrolysis reaction, which forms a chromogenic complex with malachite green and ammonium molybdate that gives an absorption peak at 650 nm. Sample was prepared using 1 g of wild-type or E234G or E234G/H363A HYPE ⌬102 and 10 g of unmodified or adenylylated BiP with 1 mM ATP in a 50-l volume and incubated for 30 min. The sample was diluted to a final volume of 200 l in water; 30 l of phosphate reagent was added and incubated for 30 min. The absorbance was read at 650 nm using Tecan. Phosphate concentration was calculated from a standard curve and represented as a bar chart.
Statistical Analysis-Data were analyzed using t test and twoway ANOVA (for Fig. 6) with Tukey's post hoc correction. Specifically, we conducted a two-way ANOVA with "siRNA treatment" and "stress (UPR) treatment" as the fixed main effects, the "siRNA treatment ϫ stress treatment" as the interaction parameter, and the "number of dead cells" as the response variable. Our data met the assumptions of ANOVA. Separate models were used for the 48-and 72-h trials, because these were run as separate experiments. Because each two-way ANOVA indicated that the main effects and the interaction were all highly significant, we then conducted post hoc comparisons between control siRNA under UPR and siRNA1 under UPR; control siRNA under UPR and siRNA3 under UPR; control siRNA under no UPR and siRNA1 under no UPR; control siRNA under no UPR; and siRNA3 under no UPR. The p values that were used to evaluate statistical significance were adjusted for multiple comparisons using Tukey's correction.

RESULTS
HYPE ⌬N45 -E234G Is a Highly Active Adenylyltransferase That Autoadenylylates at Thr-183, Ser-79, and Thr-80 in Vitro-HYPE is a 52-kDa protein that is expressed in all tissues, albeit at very low levels (5). SMART (Simple Molecular Architecture Research Tool) analysis of the domain architecture of HYPE predicts an N-terminal signal peptide and/or a hydrophobic stretch of amino acids that could specify a transmem- brane sequence (Fig. 1A). This is followed by a tetratricopeptide repeat (TPR) domain that is typically implicated in proteinprotein interactions and a Fic domain at its C terminus (Fig.  1A). In addition, HYPE contains an intrinsic inhibitory helix between its TPR and Fic domains (Fig. 1A). All previous reports on HYPE activity have been with its Fic domain only. To test the enzymatic activity of HYPE in the context of both its Fic and TPR domains, we mutated Glu-234 in its inhibitory helix to Gly

Role of HYPE in UPR
using site-directed mutagenesis. Next, His 6 -SUMO-tagged wild-type (WT; aa 46 -458) protein, its constitutively active HYPE ⌬N45 -E234G mutant, and the corresponding catalytically inactive HYPE ⌬N45 -E234G/H363A protein were bacterially expressed and purified. Amino acids 1-45 of HYPE's hydrophobic N terminus were truncated to facilitate solubility. The enzymatic activity of the purified proteins was tested by assessing the ability of each protein to autoadenylylate in an in vitro adenylylation assay using [␣-32 P]ATP as a nucleotide source. As shown in Fig. 1B, HYPE ⌬N45 -E234G displayed significantly higher autoadenylylation activity compared with HYPE ⌬N45 -WT. In contrast, HYPE ⌬N45 -E234G's catalytically inactive mutant (HYPE ⌬N45 -E234G/H363A) failed to display any autoadenylylation activity. The presence of the His 6 -SUMO tag did not affect the adenylyltransferase activity of HYPE, as the autoadenylylation levels between tagged (uncleaved) versus untagged (cleaved) versions of HYPE ⌬N45 -WT or HYPE ⌬N45 -E234G were similar (Fig. 1C).
We also assessed HYPE-E234G's adenylyltransferase activity in the presence of various divalent cations. HYPE was maximally active in a buffer containing Mn 2ϩ ( Fig. 1D and "Experimental Procedures"), displaying nearly a 2-fold increase in activity over the Mg 2ϩ -containing buffer and a 4-fold increase over the Ca 2ϩ -containing buffer (Fig. 1E). All Fic proteins characterized so far have been shown to use Mg 2ϩ in vitro. The need for divalent cations for HYPE's adenylyltransferase activity was confirmed by chelating the autoadenylylation reactions with EDTA, which significantly reduced autoadenylylation (Fig. 1, D  and E).
Mass spectrometric analysis of autoadenylylated HYPE ⌬N45 -WT and HYPE ⌬N45 -E234G HYPE revealed Thr-183, Ser-79, and Thr-80 as common sites of adenylylation between the two proteins (Fig. 1F). To validate this observation, we made Ala substitutions in the context of HYPE ⌬N45 -WT at each of the sites of modification and tested the HYPE ⌬N45 -T183A (Site 1) mutant and HYPE ⌬N45 -S79A/T80A (Site 2) double mutants for their ability to autoadenylylate. As a control, we tested additional mutations in the Site 2 peptide at Ser and Thr residues that were not found to be adenylylated (HYPE ⌬N45 -T76A/S77A double mutant; control site). The samples were analyzed by autoradiography, and the signal for each sample was quantified. The autoadenylylation signal for HYPE ⌬N45 -WT was set as 1 arbitrary unit (AU). As expected, the HYPE ⌬N45 -T76A/S77A control was autoadenylylated at levels similar to HYPE ⌬N45 -WT ( Fig. 1G; AU 1.19 for HYPE ⌬N45 -T76A/S77A). Furthermore, the combined mutation in Ser-79 and Thr-80 abrogated autoadenylylation at levels similar to the mutation at Thr-183 ( Fig. 1G; AU 0.36 for HYPE ⌬N45 -S79A/T80A and AU 0.39 for HYPE ⌬N45 -T183A). These data indicate that, in vitro, HYPE is autoadenylylated at Sites 1 and 2 at similar levels. We therefore constructed and tested a HYPE ⌬N45 -S79A/T80A/T183A triple mutant, which contained Ala substitutions at both Sites 1 and 2. Interestingly, the HYPE ⌬N45 -S79A/T80A/T183A triple mutant also failed to fully eliminate autoadenylylation (AU 0.32), suggesting that, similar to what is seen with other posttranslational modifications such as phosphorylation and ubiquitination, additional sites can be autoadenylylated when HYPE's preferred sites of modification are missing (24).
Constitutive Activity of HYPE Induces Caspase-dependent Apoptosis-To test the cellular effect of HYPE activation, we transfected human cell lines, HEK293A and HeLa, with C-terminally GFP-tagged full-length WT HYPE (HYPE FL -WT) or its E234G (HYPE FL -E234G) or E234G/H363A (HYPE FL -E234G/ H363A) mutants expressed from a mammalian expression vector. Results with HEK293A are shown. Cells were transfected for 30 h and observed by microscopy. Expression of WT HYPE resulted in a normal cellular morphology similar to the control GFP-transfected cells, in agreement with our previously published results ( Fig. 2A) (5). Furthermore, HYPE displayed perinuclear localization. In contrast, cells transfected with HYPE FL -E234G displayed a cell rounding phenotype, reminiscent of the cytotoxic phenotype associated with bacterial Fic proteins such as IbpA ( Fig. 2A) (5). This cell rounding phenotype was abrogated when cells were transfected with the catalytically inactive HYPE FL -E234G/H363A mutant, indicating that this phenotype was due to HYPE's enzymatic activity. Because WT HYPE does not induce a cell rounding phenotype, we reasoned that this is because WT-HYPE is intrinsically inhibited by its inhibitory helix and that simply overexpression is not sufficient to relieve this inhibition. Thus, the physiological signal involved in activating HYPE remains to be determined. Furthermore, these data clearly show that HYPE's activity needs to be tightly regulated within the cell, as constitutive expression of active HYPE is detrimental to the cell.
To determine the mechanism of HYPE-mediated cell rounding, we assessed HYPE transfected cells for necrosis, apoptosis, FIGURE 1. HYPE-E234G is a highly active adenylyltransferase that autoadenylylates Thr-183, Ser-79, and Thr-80, residues that lie outside the Fic domain. A, schematic representation of full-length (WT) HYPE indicating its hydrophobic transmembrane (TM) domain with a potential signal sequence (SS, gray box), followed by a tetratricopeptide domain (TPR, blue box) and its Fic domain (red box). HYPE's conserved Fic motif of sequence HPFIDGNGRTSR and its inhibitory domain (yellow) of sequence TVAIEG are indicated. Bottom panel, scheme of N-terminally truncated HYPE (⌬N45). The hydrophobic first 45 amino acids of HYPE were truncated to facilitate solubility of the purified protein. B, autoadenylylation activity of HYPE ⌬N45 -E234G is more robust than wild-type HYPE ⌬N45 , whereas HYPE ⌬N45 -H363A/E234G is catalytically dead. Bacterially expressed His 6 -Sumo-HYPE ⌬N45 -WT, His 6 -Sumo-HYPE ⌬N45 -E234G, or His-Sumo-HYPE ⌬N45 -H363A/E234G was incubated in an in vitro adenylylation assay using [␣-32 P]ATP. Samples were separated on SDS-PAGE and visualized by autoradiography (top panel) and Coomassie staining (bottom panel). Autoradiographs obtained at 1-and 5-h exposure times are shown, because the weak autoadenylylation activity of WT HYPE is evident only at the later time point. C, autoadenylylation activity of untagged and tagged HYPE ⌬N45 wild-type and HYPE ⌬N45 -E234G were similar. Bacterially expressed His 6 -Sumo-HYPE ⌬N45 -WT or His 6 -Sumo-HYPE ⌬N45 -E234G was incubated in an in vitro adenylylation assay using [␣-32 P]ATP. Samples were separated on SDS-PAGE and visualized by autoradiography (top panel) and Coomassie staining (bottom panel). Where indicated the His 6 -Sumo tag was cleaved off of HYPE. D, autoadenylylation activity of HYPE ⌬N45 -E234G with various divalent cations as cofactors. HYPE ⌬N45 -E234G was incubated with [␣-32 P]ATP in buffers containing 1 mM Ca 2ϩ , Co 2ϩ , Zn 2ϩ , Mg 2ϩ , or Mn 2ϩ , in the presence or absence of EDTA. HYPE displays maximum activity in the presence of Mn 2ϩ . This activity is completely eliminated when Mn 2ϩ is chelated by the addition of 20 mM EDTA. E, quantitative analysis of HYPE's autoadenylylation levels plotted as fold activity. F, tandem mass spectra (MS/MS) fragmentation patterns of peptides from autoadenylylated HYPE ⌬N45 -WT indicate the presence of AMP on residues Ser-79, Thr-80, and Thr-183. Thr-183 is labeled as Site 1, whereas Ser-79 and Thr-80 are together labeled as Site 2. G, in vitro adenylylation assay with purified WT and autoadenylylation site mutants of HYPE, validating Ser-79, Thr-80, and Thr-183 as autoadenylylation sites. Thr-76 and Ser-77 (labeled as control site) are residues within the peptide containing AMP modifications on Site 2 and serve as a negative control. and autophagy. Although we found no evidence for autophagy and necrosis (data not shown), we found that expression of HYPE FL -E234G induced apoptosis, as measured by PARP cleavage (Fig. 2B). PARP is cleaved and inactivated by caspases upon induction of programmed cell death, and it is used as an indicator of apoptosis. Compared with the 10-h time point, HYPE-mediated apoptosis was apparent only at 30 h posttransfection (Fig. 2B). Cells expressing HYPE FL -E234G displayed cPARP similar to the positive control, i.e. cells treated with the kinase inhibitor staurosporine. HYPE FL -WT induced extremely low levels of PARP cleavage, although the HYPE FL -E234G/H363A catalytically inactive mutant did not induce PARP cleavage. Furthermore, treatment of cells with the Z-VAD-fmk pan-caspase inhibitor prevented PARP cleavage in HYPE FL -E234G-transfected cells, indicating that constitutive activation of HYPE induces apoptosis in a caspase-dependent manner (Fig. 2C). Similar results were obtained with HeLa cells (data not shown).
Autoadenylylation appears to play a limited role in apoptosis and HYPE's function in cells. Specifically, we tested HYPE's autoadenylylation Site 1 and Site 2 mutants for their ability to induce cytotoxicity in mammalian cells. Each of the Site 1 (T183A), Site 2 (S79A/T80A), and Site 1/Site 2 (S79A/T80A/ T183A) mutants was generated in a HYPE FL -E234G background and tested for their ability to induce apoptosis, as assayed by microscopy and PARP cleavage (Fig. 2, D and E). The level of cPARP detected by Western analysis was quantified as a ratio of cPARP/HYPE expression for each corresponding sample. PARP cleavage by HYPE FL -E234G was set at AU ϭ 1. As expected, HYPE FL -E234G induced apoptosis when expressed in HEK293A cells for 30 h, although HYPE FL -E234G/H363A did not (AU ϭ 0.05). Surprisingly, based on cellular morphology, all the autoadenylylation site mutants induced a cell rounding phenotype similar to HYPE FL -E234G (Fig. 2D). Western blot analysis confirmed that all the autoadenylylation mutants displayed PARP cleavage at 30 h (Fig. 2E), although their levels of cPARP varied (T183A ϭ 0.41; S79A/T80A ϭ 1.65; S79A/T80A/T183A triple mutant ϭ 1.44). Although T183A-HYPE displayed somewhat lower cPARP levels, this difference was not evident for the triple mutant.
Rho GTPases Are Not the Physiological Targets for HYPE-We have previously shown that bacterial Fic proteins, like IbpA from H. somni, induce cytoskeletal collapse in host cells by adenylylating and inhibiting mammalian Rho GTPases (5,14). The apoptosis phenotype that we observed in Fig. 2A was reminiscent of the cell rounding phenotype associated with adenylylation of Rho GTPases, like RhoA, Rac1, and Cdc42, by IbpA. Also, Rho GTPases like Rac1 have previously been used as artificial substrates for assessing HYPE's adenylyltransferase activity, with most success when tested with just HYPE's Fic domain (4,5,15). We therefore assessed the ability of the activated HYPE ⌬N45 -E234G mutant to adenylylate Rho GTPases in the context of its TPR (putative substrate binding) domain. For this, we incubated GST-tagged Rac1 with His 6 -SUMO-tagged HYPE ⌬N45 -WT and HYPE ⌬N45 -E234G in an in vitro adenylylation reaction using [␣-32 P]ATP and a Mn 2ϩ containing buffer (see under "Experimental Procedures"). His 6 -SUMO-tagged IbpA-Fic2 (25) was used as a positive control and incubated with Rac1 in a similar reaction except Mg 2ϩ was substituted for Mn 2ϩ , which is optimal for IbpA activity. As shown in Fig. 2F, Rac1 adenylylation is not significantly increased even upon incubation with HYPE ⌬N45 -E234G, although IbpA-Fic2 robustly modifies Rac1. Increased autoadenylylation of HYPE ⌬N45 -E234G confirms that it is active. Thus, Rho GTPases are unlikely to be physiological targets for HYPE. Accordingly, we failed to detect adenylylated Rac1 or Cdc42 when cells were cotransfected with GFP-tagged full-length HYPE-E234G and FLAG-tagged Rac1 or Cdc42, and the Rho GTPase pulled out with anti-FLAG M2 affinity agarose was analyzed by Western blot analysis using anti-AMPylated Thr and anti-AMPylated Tyr antibodies (data not shown). Thus, HYPE-mediated apoptosis is unlikely to be a result of HYPE adenylylating cytoplasmic Rho GTPases.
HYPE Localizes to the Lumen of the ER-To identify HYPE's substrates, we sought to determine HYPE's subcellular localization, which would help in narrowing down potential substrates. C-terminally GFP-tagged HYPE was expressed in HeLa cells via transfection, and the immunofluorescence was visualized by microscopy. These samples were also stained for various organellar markers like Mitotracker Red for mitochondria, LAMP1 for the lysosome, GM130 for the Golgi, EEA1 for endosomes, and calreticulin for the ER (Fig. 3A). HYPE-GFP colocalized with calreticulin, indicating that HYPE localizes to the ER. No colocalization was observed with any of the other organellar markers. We further confirmed the ER localization of endogenous HYPE by fixing several mammalian cell lines, including COS7 and HeLa cells, staining for HYPE with ␣-HYPE antibody, and visualizing the immunofluorescence by microscopy. Immunofluorescence assays confirmed that endogenous HYPE also colocalizes with the ER marker calreticulin, thus confirming that HYPE localizes to the ER (Fig. 3B).
Because HYPE's N terminus is hydrophobic, we hypothesized that it may harbor an ER-targeting signal. We therefore fused HYPE's aa 1-45 to GFP, transfected HeLa cells with this fusion construct, and assessed GFP localization by immunofluorescence microscopy. Unlike the GFP vector control, where GFP localizes to both the cytoplasm and nucleus, the HYPE(1-45)-GFP fusion protein targeted to the ER as determined by colocalization with calnexin (Fig. 3C, compare upper panels with the two lower panels). Thus, the ER localization signal sequence for HYPE is located in its first 45 amino acids (Fig.  3C). We next extended these analyses by testing cellular fractions of rat liver tissue to confirm that endogenous HYPE localizes to the ER. Specifically, a freshly harvested rat liver was homogenized and subjected to a series of gradient centrifugations to isolate subcellular fractions, as described previously (21,23). Each fraction was analyzed by Western blot using antibody to HYPE and other organellar control antibodies. Endogenous HYPE consistently colocalized with subcellular fractions that were enriched for the ER (ER and Mito-ER, Fig. 3D).
HYPE is predicted to be a single pass transmembrane protein. Therefore, we sought to determine the topology of HYPE in the ER. For this, we transfected HeLa cells with HYPE C-terminally tagged with GFP and assessed its orientation in the ER membrane using an FPP assay (Fig. 4). The FPP assay uses the restricted proteolytic digestibility of GFP-tagged membrane proteins to reveal their intramembrane orientation. In the FPP assay, the cholesterol-binding drug digitonin is used to selectively permeabilize the plasma membrane, which allows cytosolic contents to diffuse across the plasma membrane, whereas intracellular organelles are retained. A subsequent proteinase K treatment then destroys the fluorescence associated with any GFP molecule attached to a protein of interest facing the cytosol but not with an organellar protein that is membrane-asso- ciated or lumenal. HeLa cells were transfected with a vector expressing GFP alone (control for cytoplasmic protein) or GFPtagged proteins like Mitoneet (a mitochondrial protein that faces the cytoplasm), BiP (an ER lumenal protein), or HYPE (Fig. 4A). FPP analysis of these samples shows that HYPE's expression/fluorescence patterns mimic the ER lumen protein, BiP, indicating that HYPE likely faces the ER lumen (Fig. 4A). Furthermore, trypsin treatment of the ER tissue fractions revealed that HYPE could be digested only after pretreatment with detergent in a manner similar to the ER lumen protein BiP, confirming that HYPE is a type II ER membrane protein (Fig.  4B).
HYPE Is N-Glycosylated Predominantly at Residue Asn-275-Based on HYPE's orientation in the ER lumen, we predicted that HYPE would be N-glycosylated. To test this, we transfected HEK293A cells with FLAG-tagged WT or E234G-HYPE. The human secreted kinase Fam20C was used as a positive control for glycosylation, although the vector alone served as a negative control (26). At 30 h following transfection, cell lysates were bound to M2-agarose beads to isolate expressed FIGURE 4. HYPE localizes to the ER lumen and is glycosylated predominantly at Asn-275. A, localization of HYPE in the lumen was determined by FPP analysis. GFP-tagged HYPE or controls (GFP only; GFP-tagged Mitoneet, a mitochondrial protein that faces the cytoplasm; and GFP-tagged BiP, an ER protein that resides in the lumen) were transfected into HeLa cells, and the fluorescence was observed following treatment with digitonin or digitonin ϩ trypsin. HYPE's fluorescence pattern matches that of the known ER lumen protein, BiP. B, ER fraction of rat liver tissue treated with trypsin in the presence and absence of detergent shows that HYPE is not susceptible to proteolytic cleavage in the absence of detergent, indicating luminal topology of HYPE. BiP (ER lumen protein) and calnexin (ER membrane protein facing cytosol) are used as controls. C, FLAG-tagged WT, E234G, and glycosylation site mutants of HYPE (N275Q, N446Q, and N275Q/N446Q-HYPE) were transfected into HEK293A cells. The expressed protein was pulled down using anti-FLAG M2 agarose beads and either left untreated or treated with endoglycosidase H. Samples were then analyzed by Western blotting with antibody to the FLAG epitope. WT, E234G, and N446Q HYPE are heavily glycosylated, whereas N275Q-HYPE displays much less glycosylation. The N275Q/N446Q-HYPE double mutant displays no glycosylation, indicating that although both sites are glycosylated, the bulk of the glycosylation occurs on Asn-275. Fam20C, a known glycosylated protein, was used as a positive control.
FLAG-tagged proteins. Recovered proteins were then separated by SDS-PAGE and analyzed by Western blot by probing with anti-M2 FLAG antibody. As indicated in Fig. 4C, both WT-and E234G-HYPE, as well as the Fam20C positive control, displayed a higher mobility shift that could be eliminated by treating the sample with endoglycosidase H, an endoglycosidase that cleaves sugar moieties from N-glycosylated proteins.
Computational programs predict Asn-275 and Asn-446 as potential sites of glycosylation on HYPE. We therefore mutated both these sites to a Gln, which should eliminate glycosylation at these sites without altering the protein structure. FLAGtagged N275Q-and N446Q-HYPE, as well as an N275Q/ N446Q-HYPE double mutant, were transfected into HeLa cells and assessed for glycosylation as described above. Although the mutation of Asn-275 significantly reduced the level of glycosylation, total glycosylation of HYPE could be eliminated only upon mutating both Asn-275 and Asn-446 (Fig. 4C). These data further confirm HYPE's location in the ER lumen and show that HYPE is glycosylated predominantly at Asn-275. Also, the data in Figs. 2F, 3, and 4 taken together support our findings that, given HYPE's topology within the lumen of the ER, cytoplasmic Rho GTPases, RhoA, Rac1, and Cdc42 are not HYPE's physiological targets.
HYPE Expression Is Up-regulated in Response to Activation of a UPR-The human UPR pathway consists of three branches regulated by ER transmembrane receptor proteins IRE1, ATF6, and PERK. The UPR integrates these three pathways to activate an ER stress response involving both translational and transcriptional mechanisms that function to relieve ER stress. However, upon sustained stress, the UPR pathway activates a proapoptotic signaling cascade. The transcriptional induction of UPR target genes is mediated by cis-acting promoter elements like the UPR element and the ER stress-response elements (I and II). Analysis of HYPE's promoter revealed the presence of potential UPR element at ϩ150, Ϫ49, Ϫ112, Ϫ308, and ER stress-response element II at Ϫ142 (Fig. 5A). HYPE's ER localization and its stress-responsive promoter elements led us to ask whether HYPE was involved in UPR. To assess the effect of UPR induction on HYPE expression, we induced UPR in mammalian cells using the ER calcium channel blocker thapsigargin (Tg) and the N-linked glycosylation blocker tunicamycin (Tn), which are drugs known to induce a UPR stress response. RNA was isolated from lysates of untreated and Tg-or Tn-treated samples and assessed by quantitative RT-PCR for expression of HYPE and other UPR target genes like BiP, CHOP, XBP1, and negative controls, like actin or proencephalin. As expected from transcriptomics data, HYPE levels were very low in untreated cells. However, HYPE expression was significantly up-regulated at 6 h post-induction and remained high until 24 h post-induction by Tg (Fig. 5B) and Tn (Fig. 5C).

HYPE Is Required for UPR Induction via ATF6 and PERK-
The UPR-induced expression of HYPE led us to ask whether HYPE might play a role in the UPR pathway. We therefore knocked down HYPE in HEK293T cells using siRNA specific to HYPE. Scrambled siRNA was used as a control. As shown in Fig.  5D, although siRNA treatment did reduce HYPE expression levels, this decrease was difficult to quantify because HYPE is known to be expressed at very low levels in untreated cells. Therefore, to assess the effectiveness of our siRNA treatment, we induced UPR in control and siRNA-treated samples with Tg for 6 h, and then measured HYPE expression by qRT-PCR. Fig.  5D shows that unlike control samples, which showed the expected up-regulation of HYPE expression upon UPR induction, cells treated with siRNA targeting HYPE failed to induce HYPE expression even upon Tg treatment, indicating efficient silencing of HYPE expression. We further confirmed the effectiveness of the siRNA-mediated knockdown of HYPE by immunoprecipitating HYPE from cells treated with control or HYPEspecific siRNA and by checking protein levels of HYPE by Western analysis. Again, siRNA treatment eliminated HYPE at the protein level, whereas the control treatment did not (Fig.  5D, lower panel). In these siRNA-mediated HYPE knockdown cells, we next induced UPR by treating with Tg for 6 h and assayed UPR progression by qRT-PCR analysis of UPR effector genes, BiP, CHOP, spliced XBP1 (Xbp1(s)), and total XBP1 (Fig.  5E). Xbp1(s) up-regulation serves as a readout for IRE1 activation, and BiP up-regulation is a readout primarily of ATF6 activation, whereas CHOP is up-regulated via the activation of PERK.
We observed that in the absence of HYPE, UPR progression failed to up-regulate canonical UPR target genes like BiP and CHOP, although, interestingly, levels of XBP1(s) compared with total XBP1 remained unaltered between the control and HYPE siRNA samples (Fig. 5E). These data provide strong evidence that HYPE positively regulates UPR, with a stronger influence on the ATF6 and PERK branches of the UPR. We are currently mapping out the details of the exact UPR branches affected by HYPE activity.
HYPE Is Needed for Cell Survival under UPR Induction-Our findings presented in Fig. 5 indicate that HYPE positively regulates UPR progression. As mentioned earlier, UPR activation entails an early response that involves up-regulation of chaperones such as BiP to increase folding of misfolded proteins. However, upon sustained stress, UPR triggers activation of proapoptotic pathways. Accordingly, the inability of cells to mount a normal UPR makes them more susceptible to stress resulting in increased cell death. Because the absence of HYPE prevents UPR activation, we hypothesized that knockdown of HYPE would reduce cell survival in response to ER stress. To test this, HEK293T cells were transfected in duplicate with siRNA to HYPE or with the scrambled siRNA control. Tg was added at 1 h post-transfection, and both siRNA and UPR treatments were allowed to progress for up to 48 or 72 h. One set of duplicate samples was assessed via qRT-PCR to confirm knockdown of HYPE and induction of UPR using BiP and CHOP as readouts, as described for Fig. 5 (data not shown). Cell survival was determined as a measure of the total number of dead cells that stained positive for trypan blue in control or HYPE siRNA-treated samples at 48 and 72 h. As predicted, a significantly greater number of dead cells were observed in samples treated with HYPE siRNA at both 48 and 72 h of UPR induction, with a more pronounced increase at 72 h (Fig. 6). These data indicate that the inability to manifest a normal UPR in the absence of HYPE affects cell survival, and the data provide overwhelming evidence that HYPE plays a critical role in regulating UPR.
HYPE Adenylylates the ER Chaperone BiP-To gain insight into the mechanism by which HYPE regulates UPR, we sought to identify protein interactors and potential substrates for HYPE. Using MCF7 breast cancer cells stably expressing C-ter-minally FLAG-tagged H363A-HYPE as a substrate trap, we conducted pulldown assays using anti-FLAG M2-agarose followed by mass spectrometry and identified the ER molecular chaperone BiP (binding immunoglobulin protein) as an interacting partner for HYPE (Fig. 7A). Specifically, analysis of indicated bands on the silver-stained gel shown in Fig. 7A revealed bands corresponding to BiP (ϳ78 kDa), IgG (ϳ25 kDa; known to bind to BiP), HYPE monomer (ϳ50 kDa), and HYPE dimer (ϳ100 kDa). HYPE's interaction with BiP was confirmed by Western blot analysis of the pulled down samples using antibody to HYPE and BiP (Fig. 7B). BiP is a 78-kDa heat shock protein that, like HYPE, also resides in the ER lumen. It has a canonical Hsp70 architecture consisting of an N-terminal ATPase domain and a C-terminal substrate-binding domain (27). By binding to misfolded proteins generated in response to cellular stresses, BiP maintains ER homeostasis and determines cell fate in response to ER stress. To ensure that the HYPE-BiP interaction was not an artifact of overexpressing HYPE, we repeated immunoprecipitation of BiP from untreated HEK293T cells and assessed the eluted interacting partners by Western blot analysis using antibody against HYPE (Fig. 7C). Because HYPE expression levels are very low in cells, only BiP could be detected in our input samples. However, a significant amount of HYPE could be clearly detected in the immunoprecipitated eluate, confirming that endogenous HYPE interacts with BiP.
Given that our data thus far indicate that HYPE interacts with BiP and regulates UPR, we asked whether HYPE adenylylated BiP. We bacterially expressed and purified N-terminally truncated (lacking BiP's hydrophobic signal peptide) His 6 -SUMO-BiP (aa 19 -637), and we incubated it with WT-HYPE, constitutively active E234G-HYPE, or catalytically inactive E234G/H363A-HYPE in an in vitro adenylylation assay. E234G-HYPE robustly adenylylated BiP, whereas WT HYPE had weak adenylyltransferase activity toward BiP (Fig. 7, D and  E). E234G/H363A-HYPE did not adenylylate BiP (Fig. 7D). The autoradiograph in Fig. 7E was quantitated, and the level of adenylylation signal on BiP when incubated with WT-HYPE was set as an AU ϭ 1. As shown in Fig. 7E, HYPE autoadenylylation also does not appear to be a prerequisite for BiP adenylylation, as both the HYPE autoadenylylation mutants, S79A/ T80A-HYPE and T183A-HYPE, modified BiP, albeit at levels lower than WT-or T76A/S77A-HYPE (S79A/T80A-HYPE, AU ϭ 0.696; T183A-HYPE, AU ϭ 0.282; T76A/S77A-HYPE, AU ϭ 0.997). LC-MS/MS analysis confirmed that HYPE adenylylated BiP and identified AMP moieties on Ser-365 and Thr-366 of BiP (Fig. 7F). Both these sites lie within BiP's ATPase domain. Site-directed mutagenesis of BiP at Ser-365 and Thr-366 confirmed these residues as the sites of adenylyla- FIGURE 5. HYPE plays an important role in the unfolded protein response pathway. A, schematic of HYPE promoter containing UPR element sites at ϩ150, Ϫ49, Ϫ112, Ϫ308, and ER stress-response element II site at Ϫ142. B, HYPE gene expression is up-regulated in response to UPR induction by 1 g/ml thapsigargin; C, 2.5 g/ml tunicamycin. qRT-PCR analysis was done for HYPE, UPR genes (BiP, CHOP, Xbp1(s); positive control), and actin (negative control). All values are normalized to GAPDH internal control. Each histogram from left to right corresponds to 1.5 and 5 min and 2, 6, 12, 24, and 48 h post-induction. D, upper panel, RNAi knockdown of HYPE using two siRNA constructs, followed by UPR induction using 1 g/ml Tg analyzed by qRT-PCR (same as B) shows down-regulation of HYPE in the HYPE siRNA cells but not in the scrambled control. Lower panel, control (C; scrambled siRNA) or HYPE-specific siRNA (S1 and S3)-treated samples were immunoprecipitated for HYPE using a mouse polyclonal antibody to HYPE and analyzed by Western blot using a rabbit polyclonal antibody to HYPE. HYPE protein could not be detected in siRNA-treated samples but was detected in the control sample. E, qRT-PCR analysis of siRNA-mediated HYPE knockdown samples indicates decreased BiP and CHOP expression in the absence of HYPE, whereas Xbp1 remains unaffected. Actin and proenkephalin are used as negative controls. FIGURE 6. HYPE is needed for cell survival during ER stress. HEK293T cells knocked down for HYPE via two different siRNA treatments (S1 and S3) were compared with untreated cells for their ability to survive upon UPR induction using 1 g/ml thapsigargin for 48 h (A) or 72 h (B). Scrambled siRNA was used as a control. As expected, sustained UPR induction (ϩUPR) induced cell death in both control and HYPE siRNA treated samples compared with the unstressed cells (ϪUPR). Cell death, however, was significantly more pronounced in cells treated with siRNA targeting HYPE. Tables respective for the 48-and 72-h panels indicate details of a two-way ANOVA statistical analysis, with "siRNA treatment" (siRNA_treat) and "stress (UPR) treatment" (stress_treat) as the fixed main effects, the "siRNA treatment X stress treatment" (siRNA_treat*stress-tre) as the interaction parameter, and the "number of dead cells" as the response variable. * on the histograms denotes p Ͻ 0.0001. tion, as S365A/T366A-BiP when incubated with E234G-HYPE in an in vitro adenylylation reaction failed to be radiolabeled (Fig. 7G). Thus, BiP is a bona fide target of HYPEmediated adenylylation.
Drosophila Fic, dFicD, was recently shown to adenylylate BiP preferentially using Ca 2ϩ as a divalent cation (28). We thus compared HYPE's ability to adenylylate BiP in the presence of Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ cations. Unlike dFicD, HYPE activity against BiP was highest in the presence of Mn 2ϩ and Mg 2ϩ but not Ca 2ϩ (Fig. 7H). Specifically, BiP was robustly modified in the presence of both Mg 2ϩ and Mn 2ϩ , with slightly higher activity with Mn 2ϩ , whereas HYPE failed to adenylylate BiP in the presence of Ca 2ϩ (Fig. 7H).
BiP is known to be ADP-ribosylated by an unknown enzyme at residues Arg-470 and Arg-492 of its substrate binding domain. We then tested whether HYPE could utilize nicotinamide adenine dinucleotide (NAD) as a nucleotide source for ADP-ribosylating BiP. WT BiP was incubated with E234G-HYPE in a reaction buffer containing 32 P-labeled NAD as a nucleotide source. However, HYPE failed to ADP-ribosylate BiP (data not shown). Thus, BiP undergoes different post-translational modifications as follows: ADP-ribosylation at its substrate binding domain and adenylylation at its ATPase domain.

HYPE-mediated Adenylylation of BiP Does Not Affect BiP's Binding to Misfolded Proteins-Our data thus far establish that
BiP is a bona fide endogenous substrate of HYPE and led us to hypothesize that HYPE may regulate UPR by post-translationally modifying BiP. We therefore assessed the consequence of adenylylation on BiP activity and function. As mentioned earlier, BiP is an Hsp70 chaperone that uses its ATPase activity to bind and refold misfolded proteins during UPR. Using purified wild-type or S365A/T366A-BiP (aa 19 -637) with purified denatured or native RNase A, we found that adenylylation of BiP does not affect its binding to denatured RNase A (Fig. 8). Specifically, we incubated native or denatured RNase A with WT or S365A/T366A-BiP that was either adenylylated by HYPE or left untreated and compared the ability of adenylylated versus unmodified BiP to bind denatured RNase A. Fig. 8B shows that adenylylated BiP bound to denatured RNase A at levels similar to nonadenylylated BiP. Furthermore, S365A/ T366A-BiP retained the ability to bind misfolded proteins, as seen from its binding to denatured RNase A (Fig. 8, A and B). Hence, adenylylation of BiP does not affect its function of binding misfolded proteins. Interestingly, HYPE itself was observed to bind misfolded proteins in an activity independent manner, because WT-, E234G-, and E234G/H363A-HYPE bound to RNase A (Fig. 8A). Furthermore, this interaction was enhanced in the presence of WT BiP (Fig. 8B).
HYPE-mediated Adenylylation Enhances BiP's ATPase Activity-The adenylylation sites for BiP (Ser-365 and Thr-366) lie in its ATPase domain. Therefore, we assessed the consequence of BiP adenylylation on its ATPase activity. WT BiP was incubated with WT-, E234G-, or E234G/H363A-HYPE in an in vitro adenylylation reaction. Following adenylylation, the ATPase activity of modified BiP was compared with unmodified BiP as measured colorimetrically by the release of P i following ATP hydrolysis (see "Experimental Procedures"). Proteins used in this assay were first cleaved to remove their purification tags and then purified further by ion exchange and size exclusion to ensure no background ATPase activity was observed from residual bacterial chaperone contaminants from the protein purification process (see "Experimental Procedures"). As shown in Fig. 8C, ATPase activity of BiP was significantly higher when incubated with WT-HYPE and even higher when incubated with E234G-HYPE. Incubation with E234G/H363A-HYPE did not significantly alter BiP's ATPase activity, nor did any of the HYPE proteins display significant ATPase activity (Fig. 8C). These data indicate that adenylylation enhances BiP's ATPase activity and fits with all our data indicating a positive role for HYPE in UPR activation and progression.

DISCUSSION
We have identified a previously unknown role for Fic-mediated adenylylation in regulating the UPR pathway. Our data show that HYPE is up-regulated upon UPR induction (Fig. 5, B and C). Furthermore, our siRNA-mediated knockdown of HYPE reveals that HYPE is required for normal induction and progression of UPR likely via the ATF6 and PERK pathways (Fig. 5E). Consequently, cells knocked down for HYPE fail to cope with ER stress and show decreased survival (Fig. 6). Additionally, sustained expression of HYPE activity induces caspase-mediated apoptosis (Fig. 2), in keeping with the programmed cell death seen in response to sustained UPR activation.
We ruled out cytoplasmic Rho GTPases, which have been used as in vitro substrates for HYPE, as physiological targets for HYPE by showing that activation of HYPE in the context of both its TPR and Fic domains does not enhance modification of Rac1 (Fig. 2F). Additionally, we show that HYPE is targeted via its N terminus to the lumen of the endoplasmic reticulum (Figs. 3 and  4), providing additional proof that the physiological target of HYPE must also localize to the ER lumen. By this same logic, histone H3, which has previously been used as an in vitro substrate for HYPE, is also not likely to be a physiologically relevant target (29). In agreement with HYPE's ER localization, we show for the first time that HYPE is glycosylated predominantly at Asn-275 (Fig.  4C). Furthermore, we identify BiP, a sentinel chaperone in the activation of UPR, as a substrate for HYPE and identify Ser-365 and Thr-366 as sites of adenylylation on BiP (Fig. 7).
We propose that HYPE regulates UPR via its interactions with BiP. Accordingly, HYPE-mediated adenylylation of BiP enhances its ATPase activity (Fig. 8C), which is required for refolding misfolded proteins to cope with ER stress (30). These data suggest a positive role for HYPE in the UPR pathway, in keeping with our qRT-PCR data showing up-regulation of HYPE upon ER stress and the role of HYPE in cell survival under sustained ER stress.
During the course of submitting this manuscript, a report describing the Drosophila homolog of HYPE, dFicD, was published (28). In agreement with our findings, dFicD is also upregulated in response to UPR induction and adenylylates BiP at FIGURE 8. Adenylylation of BiP does not affect BiP's ability to bind misfolded proteins, but it does enhance BiP's ATPase activity. A, purified recombinant WT HYPE or its E234G and E234G/H363A mutants and WT BiP or its S365A/T366A mutant were incubated with native (N) or denatured (D) RNase A bound to Sepharose beads. The proteins were then eluted from beads, separated on SDS-PAGE, and probed for HYPE and BiP. HYPE binds similarly to native and denatured RNase A, although both WT and S365A/T366A BiP bind robustly to denatured RNase. B, unmodified or adenylylated WT and S365A/T366A BiP incubated with native or denatured RNase A show that irrespective of its modification status, BiP robustly binds to denatured RNase A, indicating that BiP's binding to misfolded proteins is not affected by HYPE-mediated adenylylation. C, unmodified or adenylylated WT BiP was assessed for ATPase activity using colorimetric detection of inorganic phosphate released by ATP hydrolysis. BiP modified by WT or E234G HYPE shows significantly higher ATPase activity than unmodified BiP. BiP incubated with E234G/H363A HYPE shows ATPase activity similar to BiP alone, suggesting that HYPE-mediated adenylylation of BiP is responsible for enhancing its ATPase activity. D, model for UPR regulation involving HYPE-mediated adenylylation of BiP. Upon ER stress, HYPE is up-regulated and adenylylates BiP. Adenylylation enhances BiP's ATPase activity, thus enabling it to more efficiently refold misfolded proteins. Adenylylation also enhances BiP's disassociation from PERK and ATF6, thus efficiently activating these pathways. Although our data thus far indicate that IRE1␣ is not affected by HYPE activity, more work is needed to confirm this *, p Ͻ 0.005.
Thr-366. However, there are some differences in the experimental strategies and results obtained between our study and the report on dFicD. For instance, although both HYPE and dFicD use divalent cations for their activity, unlike what is reported for dFicD (28), we failed to see an increase in BiP adenylylation by HYPE in the presence of Ca 2ϩ . Instead, Mn 2ϩ and Mg 2ϩ served as optimal cations, at physiologically relevant concentrations, for BiP adenylylation (Fig. 7H). The ability to use both Mg 2ϩ and Mn 2ϩ may give HYPE an advantage over other Mg 2ϩ -dependent enzymes; a similar property is seen for other proteins of the secretory pathways, such as the Fam20C kinase (26). Next, using the ATPase domain of BiP with or without its linker region as an adenylylation target and to measure ATPase activity, Ham et al. suggest that adenylylation disrupts the ATPase activity of BiP (28). This observation is surprising, because both HYPE and BiP are up-regulated during UPR and inhibiting BiP function would be detrimental to coping with stress. A more likely explanation is that the ATPase assays with Drosophila Fic were not done with full-length BiP, and the inhibitory effect of adenylylation was inferred indirectly. In contrast, using proteins purified and separated by ion exchange and size exclusion, to ensure purity and lack of ATPase activity from contaminating bacterial chaperones from the purification process, we show that adenylylation enhances the ATPase activity of BiP (Fig. 8C), which fits with our observation that HYPE positively regulates UPR.
Our studies further reveal that adenylylation of BiP does not alter its ability to bind to misfolded proteins (Fig. 8, A and B). Interestingly, however, we find that HYPE itself is able to bind to misfolded proteins, a function that is independent of its adenylyltransferase activity (Fig. 8A). Furthermore, its binding to misfolded proteins is enhanced in the presence of BiP (Fig.  8B). These findings provide useful insight to speculate that HYPE may form a complex with BiP and misfolded proteins under UPR stress.
During UPR activation, misfolded proteins are first bound by TPR domain containing proteins like Hsp40 before handing them off to BiP (31). This mechanism stalls unnecessary activation of UPR cascades, offering the cell a chance to recover from ER stress. These co-chaperones also regulate interaction between Hsp70 and Hsp90 via their TPR domains and coordinate the cycle of binding of misfolded proteins and release of folded polypeptides from the chaperone machinery (30,32). Based on our structural analysis of IbpA Fic in complex with its Cdc42 substrate, domains outside the Fic enzymatic core are likely to be important for substrate recognition (25). To this end, HYPE contains a TPR domain, which may be important for its protein-protein interactions. It is possible that HYPE interacts with misfolded proteins in complex with BiP and other co-chaperones, and adenylylation may be important for stabilizing this interaction.
We are currently mapping exactly which branches of UPR are directly affected by HYPE's activity. Our data so far show a more dramatic effect on PERK and ATF6. Why HYPE appears to affect predominantly the PERK and ATF6 pathways remains to be determined. One possibility is that HYPE may have targets other than BiP. This idea is supported by the fact that even though HYPE adenylylates BiP, the presence of BiP enhances HYPE's autoadenylylation levels (Fig. 7D). It is therefore possible that in addition to being adenylylated by HYPE, BiP may also function as a cofactor for HYPE. We are currently assessing other ER-localized members of the UPR pathway as adenylylation targets of HYPE.
Recently, the ER protein PARP16 was shown to regulate PERK and Ire1␣-mediated UPR activation by ADP-ribosylating these proteins and improving their disassociation from BiP (33). One possible explanation for HYPE's effect on the PERK and ATF6 pathways, but not Ire1␣ pathway, is that adenylylation of BiP specifically alters its ability to interact with PERK and ATF6. We hypothesize that adenylylation promotes BiP's disassociation from PERK and ATF6, thus activating these pathways (Fig. 8D). This also explains why sustained activation of HYPE, such as with expression of E234G-HYPE, induces apoptosis, which is controlled by PERK activation. The observed increase in BiP's ATPase activity upon adenylylation would also promote induction of this pathway.
The physiological signal that activates HYPE remains a critical question. BiP could be adenylylated in vitro only when incubated with E234G-HYPE but not WT-HYPE (Fig. 7, D and E). This observation suggests that simply an interaction between WT-HYPE and its substrate is not sufficient to relieve HYPE's intrinsic inhibition. Thus, substrate binding is not the signal for the enzymatic activation of HYPE. Autoadenylylation also does not appear to be a signal for regulating HYPE activity, as autoadenylylation mutants were minimally affected in their ability to induce apoptosis and adenylylate BiP (Figs. 2D and 7E).
Our study is the first characterization of the human Fic protein, and the first functional demonstration of the physiological significance of this enigmatic protein in eukaryotic signaling. Our knockdown studies and determination of the consequence of BiP modification further provide mechanistic insight into the role of Fic-mediated adenylylation in mammalian signal transduction.
Several lines of evidence support the activation of UPR during disease, for which BiP's ability to monitor and recover from these stress perturbations is critical. As a protein that modifies and activates BiP under stress conditions, HYPE poses as an attractive candidate for disease therapeutics.