A Ca2+-regulated deAMPylation switch in human and bacterial FIC proteins

FIC proteins regulate molecular processes from bacteria to humans by catalyzing post-translational modifications (PTM), the most frequent being the addition of AMP or AMPylation. In many AMPylating FIC proteins, a structurally conserved glutamate represses AMPylation and, in mammalian FICD, also supports deAMPylation of BiP/GRP78, a key chaperone of the unfolded protein response. Currently, a direct signal regulating these FIC proteins has not been identified. Here, we use X-ray crystallography and in vitro PTM assays to address this question. We discover that Enterococcus faecalis FIC (EfFIC) catalyzes both AMPylation and deAMPylation and that the glutamate implements a multi-position metal switch whereby Mg2+ and Ca2+ control AMPylation and deAMPylation differentially without a conformational change. Remarkably, Ca2+ concentration also tunes deAMPylation of BiP by human FICD. Our results suggest that the conserved glutamate is a signature of AMPylation/deAMPylation FIC bifunctionality and identify metal ions as diffusible signals that regulate such FIC proteins directly.


Results. 112 113
EfFIC is an AMPylator. 114 115 Enterococcus faecalis FIC belongs to class III FIC proteins, which are comprised of a single 116 FIC domain and carry an autoinhibitory glutamate in their C-terminal α-helix. We determined 117 crystal structures of unbound, phosphate-bound, AMP-bound and ATPγS-bound wild-type 118 EfFIC (EfFIC WT ) and of unbound and sulfate-bound EfFIC carrying a mutation of the 119 catalytic histidine into an alanine (EfFIC H111A ) ( Table 1 and Table S1). These structures 120 were obtained in different space groups, yielding 32 independent copies of the EfFIC 121 monomer with various environments in the crystal. All EfFIC monomers resemble closely to 122 each other and to structures of other class III FIC proteins ( Figure 1A). Notably, the C-123 terminal α-helix that bears the inhibitory glutamate shows no tendency for structural 124 flexibility, even in subunits that are free of intersubunit contacts in the crystal. The glutamate 125 has the same conformation as in other glutamate-bearing FIC protein structures ( Figure 1B) 126 and is stabilized by intramolecular interactions and, when present, by interactions with the 127 nucleotide cofactor (Figure 1C). Two crystal structures were obtained in co-crystallization 128 with a non-hydrolyzable ATP analog (ATPγS), for which well-defined electron density was 129 observed for the ADP moiety ( Figure S1A). The positions of the α and β phosphates of 130 ATPγS in these structures depart markedly from those seen in ATP bound unproductively to 131 wild type NmFIC 7 , or bound non-canonically to CdFIC 11 ( Figure 1D). In contrast, they 132 superpose well to cofactors bound in a position competent for PTM transfer 7,8 ( Figure 1E). 133 This observation prompted us to assess whether EfFIC is competent for AMPylation, using 134 autoAMPylation as a convenient proxy in the absence of a known physiological target 135 (reviewed in 20 ). Using [α-32 P]-ATP and autoradiography to measure the formation of 136 AMPylated EfFIC (denoted AMP* EfFIC WT ), we observed that EfFIC WT has conspicuous 137 autoAMPylation activity in the presence of Mg 2+ (Figure 1F). Mutation of the inhibitory 138 glutamate into glycine (E190G) increased AMPylation, indicating that AMPylation of 139 EfFIC WT is not optimal (Figure 1F). We conclude from these experiments that wild-type 140 EfFIC has canonical features of an AMPylating FIC enzyme, and that the inhibitory glutamate 141 mitigates this activity. 142 of EfFIC bound to AMP (EfFIC WT -AMP) ( Table 1 and Table S1). AMP superposes exactly 148 to the AMP moiety of AMPylated CDC42 in complex with the FIC2 domain of the IbpA 149 toxin 5 (Figure 2A). Electron-rich density was observed next to AMP in the active site, 150 corresponding to a calcium ion present in the crystallization solution to the exclusion of all 151 other metal ions ( Figure S1B). Ca 2+ has 6 coordinations with distances in the expected 2.1-152 2.9 Å range, arranged with heptahedral geometry in which one ligand, which would be 153 located opposite to one phosphate oxygen, is missing. It interacts with the phosphate of AMP, 154 with the acidic residue in the FIC motif (Glu115), and with the inhibitory glutamate (Glu190) 155 through a water molecule ( Figure 2B). The position of Ca 2+ in the EfFIC WT -AMP structure 156 differs from that of Mg 2+ observed in other FIC protein structures in complex with ATP 157 (Figure 2C), raising the intriguing issue that it may have an alternative catalytic function. 158 Inspired by the recent observation that animal FICD proteins have deAMPylation enzymatic 159 activity 18 ,24 , we analyzed whether EfFIC would have deAMPylation activity in the presence 160 of Ca 2+ . Remarkably, addition of Ca 2+ to EfFIC WT that had been previously autoAMPylated in 161 the presence of Mg 2+ and [α-32 P]-ATP induced conspicuous deAMPylation ( Figure 2D). 162

163
In the above setup, the AMPylation and deAMPylation activities are potentially acting 164 concurrently. To characterize the deAMPylation reaction selectively, the hyperactive 165 EfFIC E190G mutant was autoAMPylated in the presence of Mg 2+ , purified to remove ATP, PPi 166 and Mg 2+ such that no AMPylation remains possible, then its deAMPylation was triggered by 167 addition of EfFIC WT or EfFIC mutants in the presence of Ca 2+ . The level of AMPylated EfFIC 168 (denoted AMP-FAM EfFIC) was quantified by fluorescence using ATP-FAM, an ATP analog 169 fluorescently labeled on the adenine base. Robust deAMPylation was observed upon addition 170 of EfFIC WT and Ca 2+ (Figure 2E, EfFIC WT panel). No spontaneous deAMPylation of AMP-171 FAM EfFIC E190G was observed in the absence of EfFIC WT (Figure 2E, control panel), indicating 172 that the deAMPylation reaction occurs in trans. We used this deAMPylation setup to identify 173 residues critical for deAMPylation ( Figure 2E, mutant panels). Mutation of the catalytic 174 histidine (H111A) and of the metal-binding acidic residue in the FIC motif (E115A) impaired 175 deAMPylation of AMP-FAM EfFIC E190G . Likewise, EfFIC E190G , which carries the mutation of the 176 inhibitory glutamate, was unable to catalyze deAMPylation, consistent with the absence of bifunctional enzyme, that AMPylation and deAMPylation are borne by the same active site, 179 and that the inhibitory glutamate is involved in the deAMPylation reaction. 180

181
The AMPylation and deAMPylation reactions are differentially regulated by metals. 182

183
The above results raise the issue of the nature of signals acting on the bifunctional active site 184 of EfFIC to regulate AMPylation/deAMPylation alternation. Previous work showed that 185 AMPylation of Escherichia coli DNA gyrase by NmFIC, which shares 56% sequence identity 186 with EfFIC, was highly sensitive to the toxin concentration, with a sharp drop of activity 187 above 250 µM 21 . We used purified AMP-FAM EfFIC E190G to analyze whether the deAMPylation 188 activity of EfFIC WT would be similarly inhibited by increasing concentrations of EfFIC WT (1-189 2000 nM). As shown in Figure 3A, deAMPylation increased with EfFIC WT concentration,190 indicating that this reaction is not adversely affected by EfFIC concentration. Alternatively,191 we reasoned that the distinct electrochemical properties of Ca 2+ and Mg 2+ (reviewed in 25 ) 192 may allow them to support AMPylation and deAMPylation differentially. Remarkably, Ca 2+ 193 was unable to support AMPylation, contrary to Mg 2+ ( Figure 3B). In contrast, both Mg 2+ and 194 Ca 2+ supported potent deAMPylation ( Figure 3C, left panel). Importantly, mutation of the 195 inhibitory glutamate eliminated the ability of EfFIC to use Ca 2+ for deAMPylation, while the 196 mutant retained partial deAMPylation in the presence of Mg 2+ (Figure 3C, right panel). To 197 understand how Ca 2+ affects AMPylation and deAMPylation differentially, we determined the 198 crystal structure of EfFIC WT -ATPγS-Ca 2+ . In this structure, Ca 2+ is heptacoordinated to the α-199 and β-phosphates of ATPγS (of which again only the ADP moiety is visible), to the inhibitory 200 glutamate and to 4 water molecules ( Figure 3D). Strikingly, Ca 2+ is shifted by 3.2 Å from 201 Mg 2+ bound to NmFIC in an equivalent configuration 7 ( Figure 3E) and it lacks an interaction 202 with the acidic residue from the FIC motif, which explains why it inhibits AMPylation. 203 Together, these observations point towards a 3-position metal switch differentially operated 204 by Mg 2+ and Ca 2+ , which includes a position competent for AMPylation, a position that 205 inhibits AMPylation and a position competent for deAMPylation. To test this hypothesis, we 206 measured the apparent AMPylation efficiency of EfFIC WT at different Mg 2+ /Ca 2+ ratio. As 207 shown in Figure 3F, AMPylation is prominent when Mg 2+ exceeds Ca 2+ , while Ca 2+ in excess 208 over Mg 2+ favors deAMPylation. We conclude from these experiments that competition 209 between Mg 2+ and Ca 2+ regulates the balance between AMPylation and deAMPylation 210 efficiencies and that this regulatory metal switch is implemented by differential usage of the 211 inhibitory glutamate and the acidic residue in the FIC motif for binding metals. activity of human FICD 18 , which features a glutamate structurally equivalent to the inhibitory 217 glutamate in EfFIC (see Figure 1B, 10 ) that is critical for deAMPylation 18 . We analyzed 218 whether, as observed in EfFIC, Mg 2+ and Ca 2+ metals could also affect FICD activity, using 219 fluorescent ATP-FAM to monitor BiP AMPylation. No measurable AMPylation of BiP by 220 FICD WT was observed, neither with Mg 2+ nor Ca 2+ , although FICD WT itself showed weak 221 autoAMPylation in the presence of both metals ( Figure 4A). Alternatively, we used 222 FICD E234G , in which the conserved glutamate is mutated to glycine, to produce AMPylated 223 BiP. Remarkably, while purified AMP-FAM BIP was efficiently deAMPylated by FICD WT in the 224 presence of Mg 2+ , no deAMPylation was measured in the presence of Ca 2+ ( Figure 4B). To 225 determine whether FICD does not bind Ca 2+ or is unable to use it for deAMPylation, we 226 carried out an Mg 2+ /Ca 2+ competition experiment in which FICD WT and purified AMP-FAM BiP 227 were incubated at increasing Ca 2+ concentration and a fixed Mg 2+ concentration. As shown in 228 Figure 4C, deAMPylation efficiency decreased as the Ca 2+ /Mg 2+ ratio increased, suggesting 229 that Ca 2+ inhibits deAMPylation by competing with Mg 2+ . We conclude from these 230 experiments that Ca 2+ binds to FICD in an inhibitory manner, which allows it to tune the 231 deAMPylation efficiency of FICD towards the BiP chaperone. In this study, we sought after a diffusible signal able to regulate directly the large group of 236 AMPylating FIC proteins in which the AMPylation activity is repressed by a conserved 237 glutamate within the active site. Combining crystallography and PTM assays, we first show 238 that bacterial EfFIC is a bifunctional enzyme that encodes AMPylating and deAMPylating 239 activities and that both reactions use the same active site. Next, we discover that the balance 240 between these opposing activities is controlled by a metal switch, in which each reaction is 241 differentially supported and inhibited by Mg 2+ and Ca 2+ in a manner that the Mg 2+ /Ca 2+ ratio 242 determines the net AMPylation level. Furthermore, we identify the inhibitory glutamate and 243 the acidic residue in the FIC motif as residues essential for the metal switch. Finally, we show that deAMPylation of the endoplasmic reticulum BiP chaperone by human FICD is also 245 dependent on the Ca 2+ /Mg 2+ ratio, with high Ca 2+ concentration inhibiting deAMPylation. 246

247
The identification of a potent deAMPylation activity in a bacterial FIC protein (this study) 248 and in human FICD 18 , which depends on a structurally equivalent glutamate in these 249 otherwise remotely related FIC proteins, leads us to propose that the conserved glutamate is a 250 signature of the ability of FIC proteins to catalyze both AMPylation and deAMPylation. Our Finally, our discovery that deAMPylation of the BiP chaperone by human FICD is modulated 308 by changes in Ca 2+ concentration raises important questions with respect to the regulation and 309 role of FICD in endoplasmic reticulum (ER) functions. The ER is the major organelle 310 involved in Ca 2+ homeostasis (reviewed in 27-29 ), and the total Ca 2+ concentration in the ER is depletion of the ER Ca 2+ store induced by the drug thapsigargin, swiftly alters protein folding 313 processes and activates the unfolded protein response 31 . Large Ca 2+ fluctuations are therefore 314 considered as major determinants of ER stress responses (reviewed in 28,29  to efficient deAMPylation of BiP and up-regulation of its activity, which is a key feature of 328 the UPR. We propose that FICD is has features of an enzymatic sensor of Ca 2+ in the ER, 329 which could allow it to function as an integrator between Ca 2+ homeostasis in the ER and the

AMPylation activities 396
A : DeAMPylation is not inhibited at high EfFIC concentration. Purified AMP-FAM EfFIC E190G 397 was incubated with increasing concentrations of EfFIC WT for one hour in the presence of 398 100µM Ca 2+ . AMPylation levels were measured by fluorescence as in Figure 2D. 399 B : EfFIC WT uses Mg 2+ but not Ca 2+ for autoAMPylation. AutoAMPylation was carried out at 400 1mM Ca 2+ or Mg 2+ for one hour and was measured by autoradiography as in Figure 1F. Softwares used in this project were curated by SBGrid 40 . Crystallization conditions, data 488 collection statistics and refinement statistics are given in Table S1. All structures have been 489 deposited with the Protein Data Bank (PDB codes in Table S1). sample buffer and boiling for 5 min. For deAMPylation, AMPylation was performed as 498 above, then 1 mM EDTA was added with or without 10 mM CaCl 2 . Proteins were resolved by 499

SDS-PAGE and AMPylation was revealed by autoradiography. 500
EfFIC AMPylation and deAMPylation fluorescence assays were carried out using the 501 following protocols. AMPylation was carried out using a fluorescent ATP analog modified by 502 were carried out using 3 µM of AMPylated protein and 6 µM of freshly purified EfFIC proteins (except for the experiment depicted in Figure 3A) in a buffer containing 50mM Tris-510 HCl pH 8.0 and 150 mM NaCl, for 1h at 30°C. Reactions were stopped by addition of 511 reducing SDS sample buffer and boiling for 5 min. Proteins were resolved by SDS-PAGE and 512 modification by AMP-FAM was revealed by fluorescence using green channel (excitation: 513 488 nm, emission: 526 nm) on a Chemidoc XR+ Imaging System (BioRad). 514 FICD AMPylation and deAMPylation fluorescence assays were carried out using the 515 following protocols. AMPylation was carried out using fluorescent ATP-FAM. AMPylated International PhD program. We are grateful to the scientific teams at the PX1 and PX2 535 beamlines at the SOLEIL synchrotron (Gif-sur-Yvette, France) and from the ID29, ID30-A3 536 and ID30B beamlines at the European Synchrotron Research Facility (ESRF, Grenoble, 537