Fic-mediated deAMPylation is not dependent on homodimerization and rescues toxic AMPylation in flies

Protein chaperones play a critical role in proteostasis. The activity of the major endoplasmic reticulum chaperone BiP (GRP78) is regulated by Fic-mediated AMPylation during resting states. By contrast, during times of stress, BiP is deAMPylated. Here, we show that excessive AMPylation by a constitutively active FicE247G mutant is lethal in Drosophila. This lethality is cell-autonomous, as directed expression of the mutant FicE247G to the fly eye does not kill the fly but rather results in a rough and reduced eye. Lethality and eye phenotypes are rescued by the deAMPylation activity of wild-type Fic. Consistent with Fic acting as a deAMPylation enzyme, its activity was both time- and concentration-dependent. Furthermore, Fic deAMPylation activity was sufficient to suppress the AMPylation activity mediated by the constitutively active FicE247G mutant in Drosophila S2 lysates. Further, we show that the dual enzymatic activity of Fic is, in part, regulated by Fic dimerization, as loss of this dimerization increases AMPylation and reduces deAMPylation of BiP.

AMPylation is a post-translational modification involving the covalent attachment of AMP to the hydroxyl group of a threonine, serine, or tyrosine residue. This modification was first discovered by Stadtman and colleagues (1) studying the regulation of glutamine synthetase by a post-translational modification with AMP. Interestingly, this modification can be reversed by a repeated catalytic site in the second domain of glutamine synthetase adenyltransferase (2,3). Five decades later, AMPylation was rediscovered in studies with a bacterial virulence factor called VopS that encodes a Fic (filamentationinduced by cAMP) domain-containing protein (4). These iden-tified AMPylators contain a Fic domain with a conserved HPFX(D/E)GN(G/K)R motif required for catalytic activity (5). Bacterial genomes appear to encode multiple Fic domain proteins with varied transferase activities (6,7), whereas metazoan genomes encode only a single Fic domain-containing protein with a conserved topology and function.
In previous studies, we identified a Drosophila Fic domaincontaining protein, Fic, that includes a single transmembrane domain upstream of conserved tetratricopeptide repeats (TPRs) 2 and a Fic domain. These domains are highly conserved in metazoan Fic domain proteins and are oriented with the TPR and Fic domains located within the ER lumen (8). Loss of Drosophila Fic results in defects in synaptic transmission in the visual system (8). We observed that the ER-localized Drosophila Fic is responsible for AMPylation of the chaperone BiP (GRP78) (9). During the unfolded protein response (UPR), BiP is both deAMPylated and transcriptionally up-regulated to help cells cope with ER stress (10). We found that BiP is AMPylated by Fic in resting, unstressed cells and deAMPylated by Fic in cells responding to ER stress. Since these initial studies, other groups have observed AMPylation of BiP by conserved metazoan Fic proteins, known as FicD or HYPE in mammals and Fic-1 in Caenorhabditis elegans (11)(12)(13). Thus, BiP, like other chaperones, is regulated by ATP and substrate binding but additionally by transcription and the conserved post-translational modification AMPylation (9).
One key factor to understanding the regulation of AMPylation by Fic is the role of a conserved autoinhibitory helix found in most Fic domain proteins. This inhibitory helix contains a conserved glutamate residue that is critical for the regulation of AMPylation activity of Fic domain proteins (9,14,15). This glutamate typically forms a salt bridge with a conserved arginine in the Fic domain active site. The salt bridge between these glutamate and arginine residues blocks the coordination of the ␥ phosphate of ATP in the active site, thus inhibiting the AMPylation activity of the Fic domain (14). Mutation of this conserved glutamate in Fic (Fic E247G in Drosophila) and in other eukaryotic Fic proteins results in constitutively active AMPylating enzymes with apparent reduced substrate specificity (9,12,16). However, it is unclear how the inhibitory salt bridge is regulated in vivo to permit AMPylation of substrates, including BiP. Structural studies have suggested that a dimerization event might be involved in the regulation of Fic-mediated AMPylation activity (12,15). In humans and C. elegans, mutations in the predicted dimer interface of FicD and Fic-1 resulted in decreased auto-AMPylation, indicating dimerization may enhance AMPylation activity (12,15).
To further investigate the regulation of Fic activity in vivo, we generated transgenes for the expression of wild-type and mutant Fic proteins. We found that expression of constitutively active Fic E247G has no detectable phenotype in flies, as long as endogenous Fic is also present. By contrast, Fic E247G overexpression in a fic null fly is lethal. Similarly, eye-specific expression of Fic E247G caused a rough eye phenotype, but only in the absence of endogenous Fic. We hypothesize that the Fic E247G toxicity in fic null flies is due to constitutive AMPylation without corresponding deAMPylation activity mediated by wildtype Fic. Supporting this hypothesis, we observed deAMPylation activity by wild-type Fic and found that this deAMPylation activity by wild-type Fic is active in the presence of the constitutive AMPylating activity of Fic E247G . Furthermore, we found that dimerization of Fic was not essential for its deAMPylation activity, as a single mutation (I271D) in Fic that disrupts dimerization retains the ability to deAMPylate endogenous BiP. Interestingly, we found that disruption of Fic dimerization leads to increased AMPylation of recombinant BiP, indicating that dimerization could play a regulatory role in this enzymatic activity. In summary, we demonstrate that the balance of AMPylation and deAMPylation by Fic is essential for cell survival and is regulated, in part, by the dimerization state of Fic.

Endogenous Fic expression is protective against the constitutive AMPylating activity of Fic E247G
To study the effect of Fic in Drosophila, we used the CRISPR-Cas9 system to generate a fic null mutant, referred to as fic 30C . The CRISPR-Cas9 -induced fic 30C indel introduced a 101-base pair deletion resulting in a frameshift and early stop codon after leucine 155 of the coding sequence, creating a null fic allele ( Fig.  1A and supplemental Fig. S1). As observed previously with other fic mutants, homozygous fic 30C flies are viable with no externally visible phenotypes (8). Next, to assess the effect of Fic expression, we generated transgenic flies with C-terminal V5 and His 6 -tagged Fic proteins under control of the Gal4/UAS system.   Figure 1. Recessive toxicity of Fic E247G expression. A, structure of the Drosophila fic gene and protein with the regulatory Glu-247, dimerization Ile-271, and catalytic His-375 residues. The CRISPR/Cas9-induced fic 30C mutation induces a 101-base pair deletion, generating an early stop codon N-terminal to the conserved TPR domain. B, a representative Western blot shows expression of UAS-Fic transgenes (wild type, H375A, and E247G) under control of the ubiquitous Da-Gal4 driver in a wild-type background. Quantification of Western blots from three separate experiments showed no significant difference in expression of the UAS-Fic transgenes. C, Da-Gal4 homozygous females were crossed to the indicated UAS-Fic/TM6b, Hu males either in a wild-type or fic 30C background. The graph represents percentage of progeny inheriting the indicated UAS-Fic transgene together with the Gal4 driver. The expected Mendelian percentage is 50%. Bars, mean of three independent crosses; dots, values for independent crosses; error bars, S.D. The total number of flies scored for each genotype was at least 150. D, representative SEM images of eyes expressing the indicated UAS-Fic transgene with LongGMR-Gal4 in a wild-type or fic 30C background. Scale bar, 100 m. E, representative SEM images of eyes expressing the indicated UAS-Fic transgene along with LongGMR-Gal4, UAS-Fic E247G in a fic 30C background. Scale bar, 100 m.

Fic-mediated deAMPylation
All UAS-Fic transgenes were inserted at the same locus on chromosome 3. These flies, in contrast to flies previously generated with randomly inserted UAS-Fic transgenes (8), exhibit undistinguishable expression levels for all three Fic proteins (Fig. 1B).
We initially tested the effect of Fic transgene overexpression on the viability of wild-type flies. Flies homozygous for the ubiquitous Da-Gal4 driver on chromosome 3 (Da-Gal4/Da-Gal4) were crossed to flies containing the chromosome 3 balancer TM6B, Hu and one copy of a UAS-Fic transgene encoding either wild-type Fic, constitutively active Fic E247G , or catalytically inactive Fic H375A . From this cross, 50% of progeny were expected to be flies ubiquitously expressing the transgene. Viability of flies was determined by scoring adult flies for the presence or absence of a visible dominant chromosome 3 marker (TM6B, Hu). In these experiments, no difference was observed between the expected and observed numbers of flies expressing the wild-type UAS-Fic transgene compared with those with the balancer chromosome (Fig. 1C). Likewise, expression of either UAS-Fic E247G or UAS-Fic H375A transgenes under control of the Da-Gal4 driver did not affect viability of flies in a fic wild-type background.
Next, we assessed the effect of Fic transgene overexpression in fic 30C null flies. For this purpose, fic 30C/30C ; Da-Gal4/Da-Gal4 flies were crossed to fic 30C/30C ; UAS-Fic/TM6B flies, fic 30C/30C ; UAS-Fic H375A /TM6B flies, or fic 30C/30C ; UAS-Fic E247G /TM6B flies. From these crosses, 50% of progeny were expected to be flies ubiquitously expressing the transgene. As observed in a fic wild-type background, we found no difference between the expected and observed viability of fic 30C flies expressing the UAS-Fic or UAS-Fic H375A transgenes under Da-Gal4 control (Fig. 1C). However, expression of the UAS-Fic E247G transgene under the ubiquitous Da-Gal4 promoter was lethal in the fic 30C null background. These data indicate that expression of the wild-type Fic protein from the endogenous fic locus is sufficient to suppress lethality caused by Fic E247G overexpression.

Tissue-specific expression of Fic E247G exhibits cell autonomous and recessive toxicity
To compare the effect of tissue-specific expression of the overactive Fic E247G with wild-type Fic, we used the photoreceptor-specific LongGMR-Gal4 promoter for specific expression in the eye. This allowed us to detect eye-specific phenotypes in the surviving adult flies. Consistent with the results from ubiquitous expression, LongGMR-Gal4 -driven expression of either wild-type Fic or Fic E247G transgenes resulted in no externally visible defects in a fic wild-type background (Fig. 1D). Expression of wild-type UAS-Fic in the fic 30C null flies did not interfere with normal eye development. However, in fic 30C null flies, expression of the overactive AMPylating Fic E247G mutant resulted in flies with a severe eye phenotype characterized by smaller eyes with disturbed ommatidial structures. To confirm that this rough eye defect was caused by the loss of wild-type Fic in the fic 30C null flies, we tested the effect of LongGMR-Gal4, UAS-Fic E247G expression in hemizygous fic 30C /Df(2L)BSC296 flies (supplemental Fig. S1). As observed in fic 30C null flies, overexpression of UAS-Fic E247G in the eye resulted in a rough eye phenotype.
Because the presence of endogenous Fic in flies appears to counterbalance the lethality and eye morphology defects of UAS-Fic E247G expression, we next asked whether overexpression of UAS-Fic or UAS-Fic H375A transgenes would rescue the rough eye phenotype of fic 30C ; LongGMR-Gal4, UAS-Fic E247G flies (Fig. 1E). As in flies expressing endogenous Fic, overexpression of the wild-type UAS-Fic transgene results in normal eye morphology. However, overexpression of the catalytically dead UAS-Fic H375A transgene failed to rescue the rough eye phenotype.
We hypothesized that the lethality and rough eye phenotypes of UAS-Fic E247G are caused by toxicity of this enzyme. To test this, TUNEL assays were performed on late third instar larval eye discs of wild-type and fic 30C null flies expressing UAS-Fic or UAS-Fic E247G transgenes under the control of LongGMR-Gal4 (supplemental Fig. S2). As predicted, eye discs from fic 30C animals expressing UAS-Fic E247G were heavily positive for TUNEL staining, indicating cell death. However, eye discs from wildtype and fic 30C animals in which wild-type UAS-Fic was expressed show no significant TUNEL staining.
These experiments indicate that expression of the overactive AMPylating Fic E247G in the absence of wild-type Fic is toxic to cells. We have observed in previous in vitro studies that Fic E247G is an enzyme with hyperactivity and reduced specificity (9). Indeed, other in vitro and structural studies have also found that mutations to the autoinhibitory helix of eukaryotic Fic enzymes result in overactive and promiscuous AMPylators (11,12). Furthermore, the nature of the developmental eye defects under control of LongGMR-Gal4 suggests that the toxicity of dFic E247G is cell-autonomous. Taken together, these data indicate that the catalytic activity of wild-type Fic protein counterbalances the unregulated AMPylating activity of the Fic E247G mutant.

Recombinant Fic mediates deAMPylation of in vitro AMPylated BiP
Recently, it was found that FicD, the mammalian homolog of Fic, could both AMPylate and deAMPylate BiP and that disruption of the autoinhibitory ␣ helix of FicD could inhibit the deAMPylation activity of FicD (17). Based on these recent findings and our fly studies, we hypothesized that both the AMPylation and deAMPylation activity of Fic are conserved functions for this enzyme and are essential for proper regulation and survival of a cell.
To address this hypothesis, we investigated whether an AMPylated BiP substrate could be deAMPylated by Fic. For these assays, recombinant [ 32 P]AMP-BiP T229A was generated by in vitro AMPylation of purified His 6 -BiP T229A , a mutant with diminished ATP hydrolysis activity (18). We compared the in vitro deAMPylation activity of recombinant wild-type Fic and the catalytically dead Fic H375A using [ 32 P]AMP-BiP T229A as a substrate ( Fig. 2A). The addition of recombinant wild-type Fic resulted in a concentration-dependent reduction of the [ 32 P]AMP signal from [ 32 P]AMP-BiP T229A protein ( Fig. 2A, lanes [2][3][4][5], whereas the addition of the catalytically dead Fic H375A did not reduce this signal ( Fig. 2A, lanes 6 -9). This indicates that wild-type Fic, but not Fic H375A , is able to deAM-Pylate recombinant AMPylated BiP.

Fic-mediated deAMPylation Recombinant Fic mediates deAMPylation of AMPylated BiP in cell lysates
Next, we asked whether the addition of recombinant Fic could deAMPylate endogenous BiP from Drosophila Schneider 2 (S2) lysates. AMPylation of endogenous BiP was observed in S2 lysates by Western blot analysis using the anti-AMP-threonine antibody (19) (Fig. 2B). As observed previously, treating cells with cyclohexamide (CHX) increased levels of BiP AMPylation in S2 cells due to the fact that this compound reduces ER stress by halting protein translation (Fig. 2B, lanes 5-8) (9). By contrast, treatment of S2 cells with DTT, which reduces disulfide bonds, causes up-regulation of the UPR, resulting in the loss of BiP AMPylation (Fig. 2B, lanes 9 -12) (9).
To validate our Western blot detection of AMPylation and to determine the site(s) on which endogenous BiP was AMPylated in S2 cells, we purified endogenous BiP using concanavalin A beads and performed LC-MS/MS analysis to identify AMPylated residues (supplemental Fig. S3). From this analysis, we found that endogenous BiP was AMPylated at a single site, Thr-518, a conserved threonine in the substrate-binding domain. This site has previously been identified in mammalian cells (13). In earlier studies, we identified Thr-366, a conserved residue in the ATPase domain of BiP, as a site of AMPylation (9); however, under our current conditions, only Thr-518 AMPylation was observed on endogenous BiP in these lysates. Of note, the Thr-518 AMPylation was detected in untreated and CHX-treated S2 cells but not found in DTT-treated S2 cells, consistent with our Western blot analysis of BiP AMPylation (9) (Fig. 2B).
To determine whether Fic has deAMPylation activity, we tested the ability of recombinant protein to remove AMPylation from endogenous BiP in Drosophila (S2) lysates. Recombinant proteins were added to solubilized organelle fractions of lysates and incubated for 1 h at 30°C. In both untreated and CHX-treated cells, the addition of wild-type recombinant Fic resulted in the loss of BiP AMPylation, as indicated by Western

Fic-mediated deAMPylation
blot analysis using the anti-AMP-threonine antibody (Fig. 2B,  lanes 2 and 6). No difference was observed in DTT-treated samples, as no AMPylated BiP was present to be modified (Fig. 2B,  lane 10). The addition of recombinant overactive AMPylating Fic E247G or the catalytically dead Fic H375A mutant proteins did not alter the steady-state levels of BiP AMPylation in these lysates (Fig. 2B, lanes 3, 7, and 11 and lanes 4, 8, and 12, respectively). We also tested the ability of recombinant Fic to deAMPylate BiP from lysates treated with tunicamycin, a UPR-inducing drug, and 4-phenylbutyrate, an ER stress inhibitor (supplemental Fig. S4). We found that the addition of recombinant wild-type Fic to lysates treated with 4-phenylbutyrate did result in loss of BiP AMPylation, whereas tunicamycin-treated lysates contained no observed AMPylated BiP to modify (supplemental Fig. S4).
Of note, the recombinant Fic E247G remained auto-AMPylated when added to the lysate, indicating that endogenous Fic did not recognize Fic E247G as a substrate for deAMPylation (Fig.  2B, lanes 3, 7, and 11). Another interesting observation is that recombinant Fic is able to efficiently deAMPylate BiP from CHX-treated lysates. Under CHX conditions, endogenous Fic becomes an AMPylator of BiP. This suggests that in vivo some other factor(s) control the enzymatic activity of Fic and the balance of BiP AMPylation. In vitro, the additional factor(s) may not be present or may be unable to efficiently regulate the recombinant Fic added to these lysates.

Fic-mediated deAMPylation is enzymatic
The fact that the catalytic histidine His-375 is required for deAMPylation of BiP strongly suggests that the deAMPylation of BiP is an enzymatic process rather than a block of Fic AMPylation activity. To determine whether the deAMPylation activity of Fic is enzymatic, deAMPylation of endogenous BiP was assayed over a range of concentrations and times. The levels of observed endogenous BiP AMPylation decreased with the addition of higher concentrations of recombinant Fic (Fig.  2C). Furthermore, when a moderate concentration of Fic (50 nM) was added to the lysate and incubated over an extended period of time, the levels of BiP AMPylation decreased in a time-dependent manner (Fig. 2D). These data show a time and concentration dependence of the deAMPylation activity mediated by Fic, thereby supporting our proposal that Fic is an enzyme that deAMPylates the modified BiP substrate.

Fic deAMPylation can compete with unregulated Fic E247G AMPylation in cell lysates
Interestingly, no increased BiP AMPylation was observed in any of the lysates incubated with Fic E247G (Fig. 2B, lanes 3, 7,  and 11). We predict that the lack of additional AMPylation could be explained by the lack of additional sources of ATP and the presence of low levels of deAMPylation activity by endogenous Fic. This would be consistent with our observations in wild-type flies, where the activity of overexpressed exogenous Fic E247G is overcome by low levels of endogenous wild-type Fic (Fig. 1). Because these lysates contain very low levels of Fic (9), a small amount of Fic could indeed counterbalance the unregulated AMPylating activity of Fic E247G without altering the appropriate levels of BiP AMPylation in the cell, as observed in flies overexpressing the UAS-Fic E247G transgene.
We predicted that the addition of an exogenous ATP source may alter the balance of AMPylation and deAMPylation activity when Fic E247G is added to S2 lysate. Consistent with this hypothesis, when ATP was added to S2 lysates, the addition of recombinant Fic E247G resulted in an observed increase in BiP AMPylation (Fig. 2E, lane 3), whereas the addition of recombinant Fic was still able to efficiently deAMPylate BiP (Fig. 2E,  lane 2). To determine whether recombinant Fic could also compete with the AMPylation activity of recombinant Fic E247G , equal amounts of recombinant wild-type Fic and Fic E247G were incubated in S2 lysates supplemented with ATP (Fig. 2E, lane  4). The addition of Fic together with Fic E247G to the S2 lysate resulted in deAMPylation of endogenous BiP, indicating that recombinant wild-type Fic can outcompete Fic E247G . Because Fic E247G retains high AMPylating activity, the ability of Fic to counterbalance Fic E247G enzymatic activity indicates that the deAMPylation activity of Fic is also quite high in S2 lysates. Furthermore, the apparent increased efficacy of endogenous Fic also suggests that additional factors or modifications to endogenous Fic must occur in the ER to control the activity of this bifunctional enzyme.
We also investigated whether competition between wildtype Fic and constitutively active Fic E247G occurs during in vitro AMPylation of a substrate. For these assays, purified recombinant BiP ATPase domain, His 6 -BiP , was used as a substrate for Fic-mediated AMPylation, as this was previously shown to be modified (9). AMPylation assays were performed by incubating Fic E247G with BiP  in AMPylation buffer with ATP at 30°C for 1 h. After incubation with 100 nM Fic E247G , AMPylation of 4 M BiP(27-407) was detected by Western blot analysis using the anti-AMP-Thr antibody (supplemental Fig. S5 (A and B) Table S1). This is consistent with our previously published observations on Fic in vitro AMPylation of BiP (9).
To determine whether recombinant Fic could compete with Fic E247G for BiP  in vitro, increasing concentrations (from 25 to 800 nM) of Fic were added to AMPylation reactions containing 100 nM Fic E247G and 4 M BiP(27-407) (supplemental Fig. S5A, lanes 4 -9). The addition of increasing amounts of Fic to the reaction resulted in decreasing BiP(27-407) AMPylation (supplemental Fig. S5A, lanes 4 -9). As a control, we observed no effect on the in vitro AMPylation of BiP  treated with increasing concentrations of Fic H375A (supplemental Fig. S5B, lanes 4 -9).
Appreciably more Fic was required to completely counteract the AMPylation of BiP  by Fic E247G in vitro. One possible explanation for this is that Fic E247G is a more efficient enzyme than wild-type Fic in vitro. This somewhat contradicts our observation that endogenous Fic can counterbalance the toxic effects of transgenic expression of Fic E247G in vivo. Of note, in previous studies, we observed that the truncated form of BiP is a better in vitro substrate for AMPylation than full-

Fic-mediated deAMPylation
length BiP (9). Therefore, deAMPylation of this truncated form of BiP could be somewhat compromised. Alternatively, Fic deAMPylation activity may be enhanced by other factors in vivo, such as changes in post-translational modifications or effects on the dimerization state of the Fic enzyme that are not present in this in vitro reaction.

Fic AMPylation of BiP is attenuated by Fic dimerization
Because the AMPylation activity of human FicD and C. elegans Fic-1 was shown to be reduced in dimerization mutants (12,15), we wanted to know whether the AMPylation and deAMPylation activity of Drosophila Fic was also affected by dimerization. To test this, we mutated an isoleucine (Ile-271) in the predicted dimer interface of Fic shown to be critical for dimerization in C. elegans Fic-1 and human FicD (12,15). To prevent dimerization artifacts from the GST tag, we utilized an N-terminal His 6 -tagged Fic expression system. As predicted, purified Fic and Fic I271D eluted by size exclusion chromatography as a dimer and monomer, respectively (Fig. 3A). This was also true for purified Fic E247G and Fic E247G,I271D . Furthermore, native gel migration of purified Fic I271D and Fic E247G,I271D was significantly faster than Fic, Fic E247G , and Fic H375A , consistent

Fic-mediated deAMPylation
with a change in the oligomeric state of these mutants (Fig. 3B).
Initially, we tested the AMPylation activity of our dimer mutant. We hypothesized that, as observed in human and C. elegans, disruption of Fic dimerization would lead to decreased AMPylation activity of the enzyme. Because wildtype Fic exhibits very low AMPylation activity, we compared the AMPylation activity of the constitutively active AMPylating mutant Fic E247G with its corresponding dimerization mutant Fic E247G,I271D . We observed that auto-AMPylation levels of recombinant Fic E247G,I271D were markedly lower than recombinant Fic E247G by Western blot analysis (Fig. 3C). To determine whether dimerization of Fic alters its AMPylation activity for the substrate BiP, AMPylation assays were performed in the presence of [␣-32 P]ATP. Several concentrations (100 -400 nM) of recombinant Fic, Fic I271D , Fic E247G , and Fic E247G,I271D were added to 4 M His 6 -BiP T229A and incubated for 30 min at 30°C (Fig. 3D). Incubation of BiP without Fic enzyme or with wildtype Fic showed no [ 32 P]BiP AMPylation signal by autoradiograph (Fig. 3D, lanes 1-4). However, incubation of BiP with Fic E247G showed a clear [ 32 P]BiP AMPylation signal (Fig. 3D,  lanes 8 -10), indicating that Fic E247G is an active AMPylating enzyme. Strikingly, the addition of Fic I271D to BiP also resulted in [ 32 P]BiP AMPylation signal (Fig. 3D, lanes 5-7), indicating that Fic I271D is also an active AMPylating enzyme. Furthermore, the addition of the Fic E247G,I271D mutant resulted in a substantially stronger [ 32 P]BiP signal than either Fic E247G or Fic I271D (Fig. 3D, lanes 11-13), suggesting that the AMPylating activity of the double mutant is significantly higher than that of either mutant alone. This also suggests that the regulatory effects of the autoinhibitory helix salt bridge and the dimerization interface have independent but synergistic effects on Fic AMPylation activity.

Fic deAMPylation activity does not require dimerization
Next, we tested the deAMPylation activity of our dimer mutant on AMPylated BiP. Because both enzymatic activities of Fic use the same active site, we predicted that, like AMPylation, the deAMPylation activity of Fic would also be altered by the disruption of dimerization. To determine whether dimerization is required for deAMPylation activity, we performed in vitro deAMPylation assays using recombinant [ 32 P]AMP-BiP T229A as a substrate as described above. Recombinant Fic, Fic I271D , Fic E247G , Fic E247G,I271D , and Fic H375A were each added to [ 32 P]AMP-BiP T229A and incubated for 60 min in the absence of ATP (Fig. 4A). In the samples treated with wild-type Fic or the monomeric Fic I271D , the [ 32 P]AMP signal from [ 32 P]AMP-BiP T229A was greatly reduced compared with the no-enzyme control (Fig. 4A, lanes 1-3), whereas the Fic E247G -, Fic E247G,I271D -, and Fic H375A -treated samples showed no reduction in [ 32 P]AMP signal of BiP T229A (Fig. 4A, lanes 4 -6). Both wild-type Fic and monomeric Fic I271D were further tested in a concentration-dependent manner, and we observed that the addition of these enzymes resulted in a concentration-dependent reduction of the [ 32 P]AMP signal from [ 32 P]AMP-BiP T229A protein (Fig. 4B). However, the enzymatic activity of monomeric Fic I271D appeared to be slightly compromised, as more enzyme was required to fully deAMPylate the substrate (Fig. 4B, lanes 6 -9) compared with wild-type Fic (Fig. 4B, lanes  2-5).
To determine whether dimerization is required for deAMPylation of endogenous BiP, recombinant His 6 -Fic proteins were incubated with organelle fractions of Drosophila S2 cell lysates for 1 h at 30°C (Fig. 4C). Consistent with the observed deAM-Pylation activity toward recombinant BiP, both wild-type Fic and the monomeric Fic I271D proteins were able to deAMPylate endogenous BiP, whereas Fic E247G , Fic E247G,I271D , and Fic H375A were not able to deAMPylate BiP. Again, compared with wildtype Fic, the deAMPylation activity of Fic I271D is slightly

Fic-mediated deAMPylation
reduced, as higher concentrations of Fic I271D enzyme are required to completely deAMPylate endogenous BiP (Fig. 4D). When a limiting concentration of Fic I271D (50 nM) is added to the lysate and incubated for 90 min, the levels of BiP AMPylation decreased in a time-dependent manner (supplemental Fig.  S6, lanes 8 -14); however, deAMPylation of BiP is not taken to completion as was observed with wild-type Fic (supplemental Fig. S6, lanes 1-7). Taken together, these data indicate that Fic I271D retains its deAMPylation activity despite also having increased AMPylation activity.

Fic dimerization is not required to rescue Fic E247G toxicity
Because Fic E247G and Fic E247G,I271D both display increased AMPylation activity, we wanted to know the effects of these mutant proteins in vivo. We generated UAS-Fic I271D and UAS-Fic E247G,I271D transgenic flies with C-terminal V5 and His 6 tags as described above. As with UAS-Fic, UAS-Fic E247G , and UAS-Fic H375A , Western analysis of the monomeric UAS-Fic I271D and UAS-Fic E247G,I271D proteins indicates stable expression of these transgenes in flies (Fig. 5A).
Next, we assessed the effect of Fic transgene overexpression in fic 30C null fly eyes. Consistent with the results from previous experiments, LongGMR-Gal4 -driven expression of either Fic E247G or Fic E247G,I271D transgenes resulted in no externally visible defects in a fic wild-type background (Fig. 5B). Furthermore, expression of the monomeric Fic I271D in fic 30C null flies did not interfere with normal eye development. However, expression of the UAS-Fic E247G,I271D transgene under the LongGMR-Gal4 promoter was lethal in the fic 30C null background. Overexpression of toxic proteins under the LongGMR-Gal4 promoter has previously been shown to cause lethality, most likely due to LongGMR-Gal4 expression in larval and pupal neurons outside of the visual system (20). The lethality of UAS-Fic E247G,I271D under the LongGMR-Gal4 promoter suggests that the observed increase in in vitro AMPylation is retained in vivo because corresponding UAS-Fic E247G flies are viable in the same genetic background. Furthermore, these data indicate that expression of the wild-type Fic protein from the endogenous fic locus is sufficient to suppress lethality caused by Fic E247G,I271D overexpression, supporting the hypothesis that both deAMPylation and AMPylation are required for cell survival.
Next, we asked whether overexpression of the UAS-Fic I271D transgene would rescue the rough eye phenotype of fic 30C ; LongGMR-Gal4, UAS-Fic E247G flies (Fig. 5C). Again, as with the UAS-Fic transgene (Fig. 1E), overexpression of the monomeric UAS-Fic I271D transgene suppressed the UAS-Fic E247Ginduced eye phenotype. Taken together, these data suggest that the remaining deAMPylation activity of Fic I271D is sufficient to counterbalance the enhanced AMPylation activity of this enzyme in vivo.

Discussion
Overall, our studies have shown that uncontrolled Fic-mediated AMPylation is lethal, and endogenous wild-type Fic can rescue flies from this lethality. Additionally, we have demonstrated that deAMPylation is likely to be the default activity of wild-type Fic and that this enzyme must change in a way so that net AMPylation of a substrate can occur. In these studies, we have also found that Fic dimerization attenuates the AMPylation activity of Fic toward its substrate BiP but is not critical for the deAMPylation activity of Fic.
Wild-type Fic appears to principally exhibit deAMPylation activity. During low ER stress, the activity of Fic appears as an AMPylator, thereby promoting the AMPylation of a portion of the available BiP and generating a subpopulation of inactive BiP that is held in reserve in the ER. When ER stress is elevated, Fic-mediated deAMPylation of BiP during ER stress would then increase the pool of active BiP, allowing for a more rapid and more acute response to stress. It is important to note that BiP resides in multiple conformational states (21) that are regulated by ER stress, which may also impact its accessibility as a Fic substrate.
Exactly how this activity is switched in vivo has yet to be identified; however, there are two independent mutations in Fic that alter its AMPylation and deAMPylation activity on BiP (Fig. 6). The first mutation, Fic E247G , is a well-characterized overactive AMPylating mutant. Mutation of this conserved glutamate results in the disruption of an inhibitory salt bridge in the active site that promotes the proper coordination of ATP in

Fic-mediated deAMPylation
the binding pocket and AMPylation activity (9,14,15). Importantly, disruption of this salt bridge also completely disrupts deAMPylation activity of the enzyme. The second mutation, Fic I271D , has been observed in this study and by others to disrupt homodimerization of Fic (12,15). Disruption of Fic dimerization increases AMPylation activity for its substrate BiP while reducing Fic auto-AMPylation and without significantly attenuating Fic's deAMPylation activity. Interestingly, despite the increased AMPylation activity of Fic I271D , flies overexpressing this mutant do not display the toxicity associated with Fic E247G . This observation further supports the hypothesis that both deAMPylation and AMPylation are required for cell survival. Indeed, these two mutations appear to work by distinct molecular mechanisms, as combining these two mutations resulted in a synergistic effect on the AMPylation activity of Fic E247G,I71D .
It is intriguing that dimerization would inhibit the AMPylation activity of Fic but not significantly alter the deAMPylation activity of this enzyme. It is tempting to speculate that the dimerization state of Fic could be a key regulatory factor working with another unknown component or modification in vivo. Future studies on how the bimodal activities of Fic are regulated within the ER will be critical to understanding how Fic regulation of BiP is controlled in the ER.

Cell culture and fly work
Drosophila melanogaster Schneider 2 (S2) cells were cultured with standard protocols and grown at 27°C in Schneider's medium supplemented with 10% fetal bovine serum and antibiotics (22). Flies were maintained using standard conditions. The Bloomington Stock Center provided Da-Gal4 (BS 55850) and LongGMR-Gal4 (BS 8121) driver lines as well as the fic deficiency stock Df(fic) ϭ Df(2L)BSC296 and OreR.

Generation of fic 30C null mutation
CRISPR-mediated mutagenesis (23) was performed with tools and the protocol available from the O'Connor-Giles, Wildonger, and Harrison laboratories (23). Mutagenesis was performed in ywv;;Nanos-Cas9 using two gRNAs targeting fic-Exon1 (gRNA1, CCGGAAGTCCTACTGCGCTA; gRNA2, ACCGTCAGCGCACGGCAGA). Injections were performed by Rainbow Transgenic Services (Camarillo, CA). The surviving F1 progenies were crossed to balancer stocks, and subsequently, potential founder F2 males were individually crossed to the balancer stock and screened for CRISPR-induced indels. Mutations were detected by PCR amplification of the target region, reannealing, T7 endonuclease treatment, and polyacrylamide gel electrophoresis analysis of potential cleavage products. Sanger sequencing of PCR products detected the precise sequence of candidate deletions.

Generation of UAS-V5-Fic constructs
Fic cDNA sequences (Fic WT , Fic E247G , and Fic H375A ) were subcloned into a pUASt vector modified by insertion of an AttB sequence, sequence-verified, and injected into embryos (Best-Gene, Chino Hills, CA) at the 62E1 landing site (24). To generate the Fic I271D and Fic E247G,I271D mutants, gBlocks (IDT, Coralville, IA) containing the respective mutant sequences were synthesized, subcloned into the pUASt-V5-Fic-AttB vector via the HiFi Assembly Kit (New England Biolabs, Ipswich, MA), and injected into embryos at the same site. Da-Gal4 -driven expression levels of all constructs were determined by Western blotting. In brief, five flies were homogenized in 200 l of lysis buffer (10% SDS, 6 M urea, and 50 mM Tris-HCl, pH 6.8, plus 10% DTT), boiled for 2 min, and spun for 10 min at 20,000 ϫ g to remove debris. 10 l were separated by SDS-PAGE and transferred to nitrocellulose membranes. Blots were probed with anti-V5 (1:3000; R960-25 Invitrogen) and anti-Hook (1:8000 (25)) and detected using fluorescence-labeled antibodies and an Odyssey scanner (LI-COR Biosciences, Lincoln, NE).

Survival analysis of flies expressing Fic E247G
To assess the lethality of unregulated Fic, we expressed C-terminally V5-His 6 -tagged UAS-Fic, UAS-Fic H375A , or UAS-Fic E247G via the ubiquitous Da-Gal4 driver in either wild-type or fic 30C

Fic-mediated deAMPylation
cent of expected was calculated from the ratio of recovered flies of the relevant genotypes compared with half the total number (number observed/(total number/2)). Crosses were repeated three times. The total number of flies scored was at least 150 for all crosses.

Electron microscopy
SEM of fly eyes were obtained as described previously (26,27). In short, eyes were fixed in 2% paraformaldehyde, 2% glutaraldehyde, 0.2% Tween 20, and 0.1 M cacodylate buffer, pH 7.4, for 2 h. Samples were washed four times with increasing ethanol (25-100%) for 12 h each followed by a series of hexamethyldisilazane washes (25-100% in ethanol) for 1 h each. Flies were air-dried overnight and mounted on SEM stubs, and the bodies were coated in fast-drying silver paint. Flies were sputter-coated with a gold/pallidum mixture for 60 s and imaged at ϫ900 magnification, with extra high tension set at 3.0 kV on a scanning electron microscope (SIGMA, Zeiss). Stereoscopic light images were acquired on a V12 stereoscope (Zeiss) at ϫ100 zoom. Stacks of ϳ30 ϫ 10-m optical sections were merged and filtered with CZFocus software (Zeiss).

TUNEL staining
Cell death was detected with the TUNEL method as described previously (28). Briefly, eye discs from wandering third instar larvae were dissected into PBS, fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 in PBS (PBST), and incubated in TUNEL reaction mixture (In Situ Cell Death Detection Kit-TMR Red, Roche Applied Science) for 2 h. Tissues were mounted in Vectashield with DAPI (Vectorlabs, Burlingame, CA). Stacks of 20 ϫ 1-M confocal sections were acquired with an LSM 710 confocal microscope (Carl Zeiss) and equally processed for brightness and contrast using ImageJ (National Institutes of Health).

Mass spectrometry analysis
Samples were run on SDS-polyacrylamide gels and stained with Coomassie Blue dye to isolate proteins before analysis by mass spectrometry. Sample preparation included reduction and alkylation of proteins using DTT and iodoacetamide, overnight digestion with trypsin, and desalting via solid-phase extraction. LC-MS/MS experiments were carried out on a Thermo Scientific EASY-nLC 1200 liquid chromatography system coupled to a Thermo Scientific Orbitrap Fusion Lumos mass spectrometer. MS1 spectra were acquired in the Orbitrap (resolution ϭ 120,000). Peptide fragmentation subsequently took place via high-energy collision-induced dissociation, and MS2 spectra were acquired in the ion trap. Resulting MS/MS spectral data were searched using the Mascot search engine (Matrix Science) for peptide identification. Additionally, AMPylation PTM assignments were manually verified from fragmentation spectra. The spectral count totals displayed in supplemental Table S1 refer to the number of MS/MS spectra identified with a Mascot score above a threshold of 15.

Concanavalin A purification of BiP for mass spectrometry
Concanavalin A pulldowns were performed on ER/mitochondrial fractions of S2 cells prepared as described above. ER/mitochondrial fractions were suspended in RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 5 mM MgCl 2 , 5 mM MgCl 2 , 5 mM CaCl 2 , 1 mM PMSF, protease inhibitor mixture) for 30 min on ice and centrifuged at 10,000 ϫ g for 10 min at 4°C. The soluble lysate was incubated with equilibrated concanavalin A beads overnight at 4°C. The beads were washed three times with RIPA buffer. To elute endogenous BiP from the concanavalin A beads, the pulldowns were incubated with RIPA buffer ϩ 5 mM ATP at room temperature for 15 min. Eluate was transferred to a new tube, TCAprecipitated, and suspended in SDS loading buffer for SDS-PAGE and Western blot analysis.
Author contributions-A. K. C. conceived and executed experiments, analyzed data, and drafted and revised the manuscript. A. T. M. conceived and executed experiments, analyzed data, and revised the manuscript. J. Z. and K. A. S. performed mass spectrometric analysis of protein samples and revised the manuscript. H. K. and K. O. conceived experiments, analyzed data, and revised the manuscript.