AMPylation and Endoplasmic Reticulum Protein Folding Homeostasis

Abstract

The endoplasmic reticulum (ER)-localized Hsp70 chaperone, BiP, undergoes a rapid, reversible and inactivating post-translational modification. This covalent modification complements the slower, conventional unfolded protein response (UPR) in matching the supply of active Hsp70 chaperone to the protein folding demand within the ER lumen. Long believed to be ADP-ribosylation, we now know this modification to be AMPylation (adenylylation) of BiP's threonine 518. Here, we review the discovery of the responsible enzyme (the Fic domain–containing protein FICD), the structural and biochemical basis of the inactivating modification and the discovery of FICD's dual role as the enzyme that both AMPylates and deAMPylates BiP. The structural basis of BiP recognition by FICD and recent in vitro insights into oligomeric state-mediated regulation of FICD's antagonistic enzymatic activities are also reviewed, the latter in the context of how such a regulatory system may arise in cells. Last, we consider the physiological significance of BiP AMPylation and speculate on the fitness benefits of this metazoan-specific adaptation.

Hsp70 chaperones are conserved in all kingdoms of life and have nonredundant roles in protein folding homeostasis (proteostasis, Fig. 1; Rosenzweig et al. 2019). Most eukaryotes have a single gene encoding an endoplasmic reticulum (ER)-localized Hsp70. First identified as a protein that associates with unassembled immunoglobulin chains in cultured mammalian cells, the ER Hsp70 came to be known as BiP (Haas and Wabl 1983). This was followed by recognition of BiP's broader role in the translocation of nascent chains into and folding within the ER lumen (Munro and Pelham 1986; Flynn et al. 1989; Matlack et al. 1999).

Figure 1.

The allosteric Hsp70 chaperone cycle. (A) Cartoon representation of the chaperone cycle. Hsp70s can exist in two major conformations. In the presence of MgATP, Hsp70s principally assumes a domain-docked conformation (Hsp70:ATP) with low substrate affinity. In the presence of MgADP, or in the absence of bound nucleotide, Hsp70s favor a domain-undocked, linker-extended, and lid-closed state (Hsp70:ADP) that possesses high substrate affinity (with slow kinetics). This equilibrium can be modulated by changing solvent nucleotide content and levels of unfolded substrate (orange) and by co-chaperone proteins—nucleotide exchange factors (NEFs) and J-domain proteins (JDPs). The schematized (sub)domain architecture is annotated: substrate-binding domain-α/β (SBDα/β); nucleotide-binding domain (NBD). (B) Structural view of the Hsp70 chaperone cycle. Hsp70 domains are color-coded as in A with co-chaperones colored in lavender and pink. All structures are presented in the same view, aligned to the Hsp70 ATP-state (Hsp70:ATP) via its SBDβ (derived from PDB 5O4P). The Hsp70:ATP structure is an in silico constructed chimera utilizing the crystal structures of an SBDα-truncated AMPylated BiP (PDB 5O4P) and BiP(Δℓ_3,4):ATP (PDB 5E84). The Hsp70•JDP structure is a chimeric structure derived from the Escherichia coli Hsp70 complexes of DnaK:ATP•DnaJ (PDB 5NRO) and DnaK:ATP•NR (PDB 7KZU). The SBD-bound unfolded (NR) substrate peptides are displayed as orange space-filling models and NBD-bound nucleotides are shown using a gray ball-and-stick representation. The ADP-state Hsp70 (Hsp70:ADP) is a chimera of an SBDα-truncated BiP:ADP structure (PDB 7A4V) with an SBDα subdomain and NR substrate peptide (orange) derived from an isolated BiP SBD-NR fusion (PDB 5E85). The latter structure (PDB 5E85) is also spliced into the Hsp70:ADP•NEF chimera (which is largely based on the structure of lid-truncated bovine Hsc70 complexed with the yeast [Hsp110] NEF protein Sse1; PDB 3C7N). Note, for illustrative purposes, the domain-undocked Hsp70 states are depicted at a 70% scale relative to the domain-docked chaperone states. (A and B adapted from Perera 2021, with permission from the author.)

BiP/HSPA5 gene expression is regulated by a conserved transcriptional program that is sensitive to changes in the protein folding environment within the ER. This homeostatic process, known as the unfolded protein response (UPR), has been subject to intense scrutiny (for review, see Mori 2022). Less well-appreciated is the evidence for post-translational mechanisms that couple changes in ER proteostasis to BiP activity. Because of their short latency and rapid reversibility these post-translational processes have the potential to complement the well-studied UPR gene expression program: They can fill the temporal gap between the sensing of ER stress and the resultant, UPR-mediated change in effector protein levels.

The basic attributes of this post-translational strand of the UPR have been reviewed previously in this series (Preissler and Ron 2018). Here we shall focus on the reversible covalent modification of BiP in response to changing levels of ER stress: AMPylation of BiP(Thr518) by the ER-localized enzyme FICD. On the way, we have assimilated literature on BiP modification that spans four decades.

THE CIRCUITOUS PATH TO THE DISCOVERY OF BiP AMPylation

Covalent modification of the protein now known as BiP was recognized even before its role as a chaperone. Two-dimensional gel electrophoresis of detergent-soluble proteins extracted from the postnuclear supernatant of fibroblasts labeled with tritiated [³H]-adenosine demonstrated [³H] incorporation in a single conspicuous spot (Carlsson and Lazarides 1983). The label could be removed by incubation in vitro with a phosphodiesterase, yielding ³H-AMP. The Hendershot and Lee laboratories subsequently established that the labeled protein was BiP, and furthermore that it also incorporated ³²P from inorganic phosphate pools (Hendershot et al. 1988). This dual labeling led to the hypothesis that BiP was mono-ADP-ribosylated.

Some ambiguity remained, as bulk phospho-amino acid analysis of hydrolyzed BiP suggested incorporation of ³²P directly onto pSer, pThr (Hendershot et al. 1988), and even pTyr residues (Leustek et al. 1991). The extent of heterogeneity in BiP labeling was likewise ambiguous. For example, Carlsson and Lazarides’ two-dimensional gels revealed the presence of two predominant isoforms of BiP in chicken embryo fibroblasts, with pIs between 5.2 and 5.3, with tritiated label confined to the acidic form (Fig. 1 in Carlsson and Lazarides 1983). The existence of only two species was also supported by one-dimensional isoelectric focusing of BiP from cultured pituitary cells (Laitusis et al. 1999). However, Toledo and colleagues identified up to four isoforms of BiP extracted from bovine kidney MDBK cells (Fig. 4 in Leustek et al. 1991).

Figure 4.

A monomerization-linked increase in the flexibility of FICD's gatekeeper Glu234 enfeebles BiP deAMPylation. (A) The deAMPylation-competent active site from the complex of monomeric FICD^L258D-H363A•BiP-AMP is shown (PDB 7B7Z). Glu234 plays a crucial role in positioning a catalytic water (*) for in-line nucleophilic attack into the Pα-Oγ(Thr518) phosphodiester bond. Additionally, Glu234 interacts with the primary hydration shell of the catalytically essential Mg²⁺ ion. The increased flexibility of monomeric FICD(Glu234) also facilitated the crystallization of an alternative, deAMPylation-incompetent, FICD^L258D-H363A•BiP-AMP structure (PDB 7B80, salmon-colored structure shown in the insets). The inset panels depict orthogonal views of the two overlaid active sites, alongside His363 from FICD^L258D:MgATP (PDB 6I7K). Note that Glu234 of the deAMPylation-incompetent structure is displaced away from the position of the putative catalytic water (only present in the deAMPylation-competent Michaelis complex). The former's more outwardly facing Glu234 results in a shift in the bound Mg²⁺ coordination complex into a location that is sterically incompatible with the presence of a catalytic water. (B) The proposed reaction scheme for FICD-catalyzed deAMPylation of BiP(Thr518-AMP). This mechanism represents a general acid–catalyzed S_N2-type hydrolytic reaction. Charge development on the α-phosphate is stabilized by the Fic motif's anion-binding nest (not shown) and Mg²⁺ (the positioning of which is also influenced by Glu234 confirmation). The depressed pK_a of Glu234 renders it unlikely to participate in general base catalysis, but it may act as a (transient) late-stage proton acceptor (as depicted). (B adapted from Perera 2021, with permission from the author.)

Nevertheless, a consensus emerged regarding the physiological context of BiP's covalent modification: Conditions that increased the levels of ER stress were associated with reduced BiP labeling, whereas conditions that interrupted the supply of client proteins into the ER (lowering levels of ER stress) increased the levels of BiP modification (Carlsson and Lazarides 1983; Ledford and Jacobs 1986; Hendershot et al. 1988; Leno and Ledford 1989, 1990; Laitusis et al. 1999; Chambers et al. 2012). Especially informative were findings that the modified form of BiP was largely excluded from the complexes BiP ordinarily forms with its clients (Hendershot et al. 1988; Freiden et al. 1992); this suggested that the modification(s) were inhibitory with respect to BiP's chaperone activity.

Despite the evidence for their potential importance to ER proteostasis, the biochemical details of these putative inactivating modification(s) remained obscure for many years. Peptide mapping experiments conducted in mouse plasmacytoid cell lines revealed that essentially all the ³²P-labeling of endogenous BiP was localized to a carboxy-terminal cyanogen bromide fragment: Thr434 to Met541 (Gaut 1997). The same cyanogen bromide fragment was subsequently shown to be the only site of labeling from pools of [³H]-adenosine in CHO cells. Furthermore, the label incorporated into the same fragment from ³²P inorganic phosphate pools was resistant to alkaline phosphatase treatment (Chambers et al. 2012). These findings were consistent with ADP-ribosylation somewhere in BiP's carboxy-terminal substrate-binding domain. Laboring under that assumption and armed with the knowledge that hydroxylamine-resistant mono-ADP-ribosylation occurs on arginines and cysteines, we systematically mutated all five arginine residues located within the Thr434 to Met541 CnBr BiP fragment (to lysines) and discovered that an Arg470Lys mutation abolished nearly all [³H]-adenosine and ³²P labeling of BiP (Chambers et al. 2012).

These finding were deemed most consistent with Arg470 being a major site of regulated ADP-ribosylation of BiP. However, two problems remained: There was no mass-spectrometric evidence for the existence of the predicted ADP-ribosylated peptide, and there was no candidate enzyme for the predicted reaction. Fortunately, the landscape around BiP modification was about to change. In 2009, the Orth and the Dixon laboratories reported that VopS and IbpA, bacterial enzymes and pathogenicity effector proteins when secreted into mammalian host cells, catalyzed the transfer of AMP from ATP onto the hydroxyl group–bearing side chains of specific residues on host Rho GTPases (Worby et al. 2009; Yarbrough et al. 2009). AMPylation (or adenylylation) had long been known in eukaryotes, but it was hitherto believed to only represent activated intermediates of various enzymes (e.g., nucleic acid ligases) or a step in creating reactive substrate intermediates (e.g., in ubiquitin conjugation) and not a stable regulatory modification.

VopS and IbpA belong to a large family of bacterial proteins that share a common “filamentation induced by cyclic AMP” or Fic domain. The Dixon laboratory also noted that animal genomes have a single copy of a highly conserved gene encoding a Fic domain protein (Worby et al. 2009). Previously identified (in a yeast two-hybrid screen) as Huntingtin-interacting protein E (HYPE) the protein, now known as FICD, is a homodimeric type II transmembrane protein with an ER-luminal catalytic Fic domain (Bunney et al. 2014; Sanyal et al. 2015). An early study of the phenotype arising from FICD inactivation in the fly eye was interpreted as reflecting an extracellular role for the Fic domain (Rahman et al. 2012). However, subsequent findings pointed to BiP being an AMPylation target of ER-localized FICD. Intriguingly, BiP AMPylation followed the pattern previously observed in its labeling from [³H]-adenosine and ³²P pools, in that the level of AMPylated BiP decreased upon ER stress (Ham et al. 2014).

These findings were reproduced shortly thereafter in the Mattoo laboratory (Sanyal et al. 2015) and in ours (Preissler et al. 2015b). In addition to confirming that BiP is an AMPylation target of FICD, we established that inactivation of the FICD gene abolished all trace of BiP modification (in CHO and pancreatic acinar cells). Furthermore, intact protein mass spectrometry of endogenous BiP revealed only two species in unstressed cells with masses of unmodified and singly AMPylated BiP; with the latter being entirely absent in FICD knockout cells (Preissler et al. 2015b).

These discoveries reconciled previous observations: AMPylation explains protein labeling from pools of both [³H]-adenosine and inorganic [³²P]-phosphate. The recovery of labeled AMP by treatment of modified BiP with phosphodiesterase (Carlsson and Lazarides 1983) is consistent with hydrolysis of the phosphodiester bond between BiP(Thr518) and the AMP moiety. Even the defect in BiP labeling wrought by mutation in Arg470 (Chambers et al. 2012) can be explained by sensitivity of FICD-mediated BiP AMPylation to the latter's allosteric state (see below).

Nonetheless, important discrepancies exist between the findings of the different groups. The Orth and Mattoo laboratories initially reported on the presence of an AMPylated peptide centered on BiP(Thr366), a residue that forms part of the α-helical base of the nucleotide-binding pocket in BiP and homologous Hsp70 nucleotide-binding domains (Fig. 2A, lower inset; Yang et al. 2015). Our lab, by contrast, found no evidence for AMPylation on Thr366 and instead identified Thr518 as an AMPylated residue (Preissler et al. 2015b), a finding confirmed by others (Broncel et al. 2015; Casey et al. 2017; Fauser et al. 2021). Thr518, unlike Thr366, is located within an unstructured loop on the far reaches of BiP's substrate-binding domain surface (Fig. 2A, upper inset). Moreover, it is encompassed in the Thr434 to Met541 CnBr fragment previously found to be labeled with ³²P and [³H]-adenosine in vivo (Gaut 1997; Chambers et al. 2012). Furthermore, whereas FICD prominently AMPylated BiP Thr518, we observed no AMPylation of Thr366 either in vivo or in vitro. Importantly, paired SILAC measurements showed that endogenous FICD-mediated AMPylation of BiP Thr518 (in vivo) was associated with ∼40% depletion of the peptide containing the nonmodified Thr518—a stoichiometry of modification that is in agreement with iso-electric focusing measurements (Carlsson and Lazarides 1983; Laitusis et al. 1999; Chambers et al. 2012; Preissler et al. 2015b). There is no corresponding information on the stoichiometry of BiP AMPylation on Thr366. Based on these observations it seems reasonable to conclude that at least in plasmacytoid cells (studied by Gaut 1997) or CHO and AR42J cells (studied by Preissler et al. 2015b) BiP AMPylation occurs almost exclusively on Thr518.

Figure 2.

AMPylation locks, and FICD binds, BiP in its ATP-like state. (A) Despite being nucleotide-free, BiP-AMP adopts a domain-docked, lid-open, ATP-state conformation similar to other ATP-bound Hsp70 structures. Shown here is an alignment (via the NBD) of BiP:ATP (PDB 5E84; colored according to its domain architecture as in Fig. 1), Escherichia coli DnaK:ATP (PDB 4B9Q; gray), and AMPylated BiP lacking bound-nucleotide (BiP-AMP:Apo; PDB 5O4P; yellow). The modified Thr518 of BiP-AMP had incomplete electron density corresponding to the AMP moiety, as a result only the α-phosphate is modeled. Structural superpositions of BiP:ATP (PDB 5E84) and BiP:ADP (PDB 7A4V) are shown aligned by the NBD (lower inset) or SBDβ (upper inset). These insets highlight the region surrounding Thr366 and the loop containing Thr518, respectively. In both cases, semiopaque surface representations of BiP:ATP are also displayed. Note, Thr366, in both the BiP ADP- and ATP-state forms part of a structured α-helical turn at the base of the adenosine-binding pocket of the NBD (hydrogen bonds formed by Thr366 are shown with pink dashed lines). Conversely, Thr518 in BiP's ATP-state (in contrast to Thr518 in the ADP-state) is part of an extended and unstructured loop (ℓ_7,8), in which Thr518 does not engage in a hydrogen bond network. (A adapted from Perera 2021, with permission from the author.) (B) The overall architecture of dimeric FICD fully engaged with two BiP molecules, cartooned within the context of the ER membrane. The displayed heterotetrameric deAMPylation complex is based on the monomeric FICD-containing (heterodimeric) deAMPylation complex (PDB 7B7Z) with a full-length SBDα (green) aligned from the BiP:ATP structure (PDB 5E84). An isolated FICD dimer structure is also superposed (gray). (C) FICD recognizes the domain-docked ATP-state of BiP. The monomeric FICD-containing deAMPylation complex (PDB 7B7Z) is shown with a semiopaque surface representation overlaying AMPylated BiP and colored by its (sub)domain architecture. Selected residues forming intermolecular contacts are shown as sticks. FICD's TPR domain engages a tripartite (NDB-linker-SBDβ) surface that is specific to BiP's domain-docked, ATP-like state (see Fig. 1B). Simultaneously, the close apposition of ATP-state BiP's NBD and SBD also facilitates engagement of the Thr518-bearing SBDβ(ℓ_7,8) with the catalytic Fic domain. (B and C reprinted from Perera et al. 2021 under the terms of the Creative Commons CC BY license.)

In FICD knockout CHO cells BiP has a single mass, corresponding to the unmodified protein, with no detectable acidic forms as measured by isoelectric focusing (Preissler et al. 2015b). This argues that FICD-mediated AMPylation is the only significant modification of the chaperone in this cellular context. Cytosolic chaperones have long been known to undergo phosphorylation (for reviews, see Griffith and Holmes 2019; Nitika et al. 2020) and the ER chaperone PDI has recently been shown to undergo regulatory phosphorylation (Yu et al. 2020). Therefore, it remains possible that in some circumstances BiP too may undergo functionally important phosphorylation. Noteworthy in this regard is the finding that a Legionella pneumophila effector protein, LegK4, targets cytosolic Hsp70s for inactivating phosphorylation on Thr495 (a residue homologous to BiP Thr518) (Moss et al. 2019). Thus, the functional consequences of BiP AMPylation by FICD may be accessible by alternative biochemical events.

THE FUNCTIONAL CONSEQUENCES OF BiP AMPylation

As noted above, more AMPylated BiP is found as ER stress wanes (Laitusis et al. 1999; Chambers et al. 2012; Ham et al. 2014; Preissler et al. 2015b) and AMPylated BiP is preferentially excluded from complexes with client proteins in vivo (Hendershot et al. 1988). These observations hint at AMPylation's role in ER proteostasis: inactivating idle BiP to match the supply of active chaperone to fluctuations in demand. These notions are further supported by native gel immunoblotting that tracks the disposition of BiP from cells subjected to various physiological conditions. In rapidly growing cultured cells BiP is distributed between complexes with its heterogenous clients and discrete BiP oligomers, with the latter likely representing inactive BiP–BiP complexes (Preissler et al. 2015a). At any one time very little BiP is monomeric. However, AMPylation—whether entrained physiologically by cutting off the supply of unfolded proteins to the ER, or imposed genetically by enforced expression of a deregulated allele of FICD—results in an appreciable pool of monomeric BiP (Preissler et al. 2015b). Consistent with BiP's role as a repressor of UPR signaling, deregulated FICD overexpression also conspicuously activates the UPR (Preissler et al. 2015b)—although a contribution toward UPR induction from promiscuous AMPylation of other targets cannot be excluded (Truttmann et al. 2016b).

AMPylation (on Thr518) weakens BiP's ability to form a stable complex with a substrate peptide in vitro: Higher dissociation rates of bound peptide from BiP-AMP (compared with unmodified BiP) were observed both in the absence of nucleotide or in the presence of ADP (Preissler et al. 2015b). ADP binding (or the apo state) otherwise allosterically imposes a domain-undocked conformation on Hsp70s, one in which substrates are firmly engaged in the substrate-binding domain (Fig. 1; Bertelsen et al. 2009). Interestingly, AMPylation of Thr518 biases BiP toward the opposite conformation, one normally assumed by the ATP-bound chaperone (Fig. 2A). Here the two domains are docked and substrates may exchange rapidly. This bias is observed crystallographically, as nucleotide-free and ADP-bound AMPylated BiP both crystallized in a domain-docked conformation typical of ATP-bound Hsp70s (Preissler et al. 2017b), and in solution, by protease protection assays (Preissler et al. 2017b) and nuclear magnetic resonance (NMR) studies (Wieteska et al. 2017).

These in vitro observations suggest that exclusion of modified (AMPylated) BiP from complexes with its substrates, observed in vivo (Hendershot et al. 1988; Freiden et al. 1992), is likely a reflection of the direct effects of AMPylation on substrate binding. But the effect of AMPylation extends beyond this feature: Hsp70s engage their substrates via a cycle in which the initial association occurs when the chaperone is in the “high–on rate,” ATP-bound, domain-docked state. Following instruction from a J-domain protein (JDP) co-chaperone, Hsp70s hydrolyze their bound ATP and transition to the domain-undocked state in which the substrate is firmly engaged (Fig. 1). This cycle, which results in a kinetically imposed ultra-affinity of Hsp70s for their substrates, is completed by exchange of the bound ADP for ATP and substrate release (Misselwitz et al. 1998; De Los Rios and Barducci 2014). By interfering with the ability of ER-localized J-domain co-chaperones to stimulate ATP hydrolysis, AMPylation disrupts this cycle before its energetically costly step (Preissler et al. 2017b). Thus, AMPylation disfavors not only tight binding of substrates by BiP molecules that have hydrolyzed their bound ATP, but also encourages BiP to remain in an ATP-bound state, in which substrates are bound loosely and exchange readily. Triggered by conditions associated with declining burden of substrates, BiP AMPylation promotes an energetically efficient, laissez faire regime in the ER.

Thr518, on the loop linking β7 and β8 (ℓ_7,8) of BiP's SBD, is far removed from the site engaged by the effector J-domain of co-chaperones. In crystal structures, the AMPylated side chain is exposed to the solvent and makes no contacts with other BiP residues (Fig. 2A). Furthermore, the affinity of the J-domain for its binding site in BiP is unaffected by AMPylation (Preissler et al. 2017b). However, the transition of Hsp70s from the domain-docked ATP-bound state to the domain-undocked ADP-bound state, following ATP hydrolysis, entails a major rearrangement of SBD(β8), the preceding ℓ_7,8, and with it the disposition of the side chain of Thr518 (Figs. 1B and 2A, upper inset). Little is known about the intermediates in this transition nor of the role of the J-domain in bringing them about, but there is evidence that disruption of contacts between the SBD and NBD derepress Hsp70's ATPase activity and that structural changes in the SBD are required (for review, see Kityk et al. 2015). It is thus tempting to speculate that the bulky, amphiphilic, solvent-exposed side chain of AMPylated Thr518 disfavors a conformation that ℓ_7,8 normally assumes as part of this allosteric path, thereby accounting for both the bias in BiP's disposition toward the domain-docked conformation and AMPylation's ability to antagonize (JDP-stimulated) ATP hydrolysis.

The biochemical and structural features reviewed above all center around the inactivating nature of BiP AMPylation. Nonetheless, it is important to note the possibility that the AMPylation of BiP may also result in gain-of-function features. For example, AMPylated BiP may serve to sequester J-domain proteins or other co-chaperones away from the post-translationally unmodified and chaperoning-competent pool of BiP. Such features, in concert with the direct role for AMPylation in inhibiting BiP chaperoning activity, may well figure in the biochemical consequences of FICD action.

THE BIOCHEMISTRY AND REGULATION OF FICD-MEDIATED BiP AMPylation AND deAMPylation

The core of the Fic domain and the catalytic Fic motif is highly conserved across enzymatically functional Fic protein homologous. As such the fundamentals of AMPylation catalysis are shared across this protein family (Luong et al. 2010; Xiao et al. 2010; Palanivelu et al. 2011). Here we focus on the specific features of FICD that underlie its role in physiological regulation of BiP activity.

FICD Is Both a BiP AMPylase and deAMPylase

The level of AMPylated BiP is rapidly reduced upon induction of ER stress (Laitusis et al. 1999). A clue to the nature of this process was provided by a peculiarity of FICD: Although cells lacking FICD lost all trace of AMPylation, trans-rescue of the deficiency was conspicuous only with an enzymatically deregulated allele of FICD (Glu234Gly, see below) and overexpression of the wild-type enzyme systematically failed to rescue (Ham et al. 2014; Preissler et al. 2015b). This raised the prospect that FICD may have the capacity to both AMPylate and deAMPylate BiP, with the latter activity dominating at high levels of expression of the wild-type enzyme.

Indeed, this proved to be the case: When incubated with recombinant wild-type FICD, AMPylated BiP was rapidly deAMPylated. The deregulated FICD^E234G (in stark contrast to its hyper-AMPylation ability) is bereft of deAMPylation activity, whereas other residues of the active site Fic motif proved essential for the catalysis of both activities (Casey et al. 2017; Preissler et al. 2017a). The deAMPylation reaction proved to be hydrolytic, generating AMP and unmodified BiP (Preissler et al. 2017a, and see below). FICD is also responsible for BiP deAMPylation in vivo, as both mammalian and fly cells lacking endogenous FICD are hypersensitive to overexpression of the deregulated FICD^E234G allele (Casey et al. 2017; Preissler et al. 2017a).

Escherichia coli glutamine synthetase adenylyl transferase provides a precedent for a bifunctional gene encoding both an AMPylase and deAMPylase, but these activities are segregated to different domains of the polypeptide (Xu et al. 2010). FICD is unusual in exploiting the same active site for two (mutually antagonistic) activities. A question then emerges as to how these functionally opposed activities are regulated by the physiological state of the cell.

Substrate-Level Regulation of BiP AMPylation

It was previously documented that AMPylation biases BiP toward its domain-docked ATP-like state and that FICD preferentially binds and AMPylates BiP in its domain-docked conformation (Preissler et al. 2015b; Perera et al. 2019). Recent co-crystal structures of the deAMPylation complex of AMPylated BiP bound to FICD have served to rationalize these observations (Perera et al. 2021). FICD's ability to recognize BiP's ATP-state stems from its TPR domain, which, in an analogous fashion to the co-chaperone J-domain (Fig. 1B), proves necessary and sufficient for binding the tripartite BiP interface of NBD-linker-SBDβ that is specific to domain-docked BiP (Fig. 2B and 2C; Fauser et al. 2021; Perera et al. 2021). Furthermore, intermolecular β-sheet engagement of FICD's Fic domain with BiP's SBDβ(ℓ_7,8) is only possible when BiP's Thr518-bearing loop is extended (Perera et al. 2021)—a conformation unique to the ATP-state Hsp70 configuration (Fig. 2A, upper inset). Moreover, the simultaneous interaction of FICD with both BiP's SBDβ(ℓ_7,8) and BiP's NBD can only occur when BiP is in an ATP-like, domain-docked state (Figs. 1B, 2B, and 2C).

FICD's substrate specificity has important ramifications for the regulation of BiP AMPylation in response to changes in abundance of unfolded proteins within the ER. An excess of chaperones (over unfolded proteins) elevates the level of free BiP:ATP (the substrate of FICD-mediated AMPylation) and, therefore, favors its ability to compete with BiP-AMP for the occupancy of FICD's active site. Conversely, an increase in the burden of unfolded proteins will titrate unmodified BiP into its domain-undocked substrate-bound state (which cannot be bound or AMPylated by FICD). Under these stressful conditions the inactive pool of AMPylated BiP (which is intrinsically biased toward the domain-docked Hsp70 conformation) remains a substrate for FICD-mediated deAMPylation. Substrate-level regulation of BiP AMPylation/deAMPylation thus constitutes a mass action–driven feedback loop.

Allosteric Regulation of FICD through Fic Domain Dimerization

Diverse Fic proteins possess a regulatory glutamate (Glu234 in FICD) whose engagement in the active site blocks AMPylation. This arises from an interaction with a conserved Fic motif arginine—Arg374 in FICD (Fig. 3A; Engel et al. 2012; Goepfert et al. 2013; Bunney et al. 2014). Glu234 competes electronically and sterically with Arg374's ability to engage the γ-phosphate of ATP and thus inhibits binding of ATP in an AMPylation-competent conformation (Fig. 3; compare competent [salmon] and incompetent [blue] structures; Perera et al. 2019). Indeed, Glu234 side chain removal (by mutation to glycine) unmasks FICD's AMPylation activity (Ham et al. 2014; Preissler et al. 2015b; Sanyal et al. 2015) but abolishes all FICD-mediated deAMPylation (Casey et al. 2017; Preissler et al. 2017a). The Michaelis complex of AMPylated BiP and FICD rationalizes the dependence of deAMPylation on Glu234—engagement of its side chain in the active site facilitates hydrolytic BiP deAMPylation by activating and aligning a catalytic water molecule for in-line nucleophilic attack into the Pα-Oγ(Thr518) phosphodiester bond (Fig. 4; Perera et al. 2021).

Figure 3.

FICD monomerization increases gatekeeper Glu234 flexibility permitting AMPylation-competent binding of MgATP. (A) Increased gatekeeper Glu234 flexibility, afforded by FICD monomerization, facilitates binding of ATP in an AMPylation-competent conformation. Structures of monomeric FICD^L258D (PDB 6I7K; salmon) and dimeric wild-type FICD (PDB 6I7G; blue), both crystallized in the presence of Mg²⁺ and ATP, are superposed via their catalytic Fic domains. Note the stark difference in the mode of ATP binding exhibited by the two FICDs. (B) Overlaid van der Waals (VDW) radii from the above structures, now modeled along with the AMPylation target region from BiP(SBDβ). The SBDβ(ℓ_7,8) (green) is positioned such that the BiP target residue hydroxyl group Thr518(Oγ) is 3 Å from ATP(Pα) and directly in-line with the Pα-O3α phosphoanhydride bond. A comparison of the left and middle panels shows that only the monomeric FICD^L258D:MgATP complex is compatible with BiP target residue engagement in which there is no significant VDW overlap with the Fic domain flap (Thr518(Oγ)-Val316(Cγ1)) and in which there also exists the potential for hydrogen bonding between BiP(Thr518) and FICD(His363). The right panel demonstrates that only the monomeric FICD^L258D co-crystal structure contains a Glu243 compatible with an AMPylation-competent mode of ATP binding, in which ATP's γ-phosphate is coordinated by Arg374 (see A). (C) A reaction scheme for the FICD-catalyzed AMPylation of BiP(Thr518). Fic proteins facilitate a general base catalyzed S_N2-type adenylyltransferase reaction. Partial covalent bonds in the proposed transition state (‡) are depicted with dashed lines, and polar interactions with hashed lines. Note, extra (partial; δ) negative charge, delocalized through the α- and β-phosphates, is stabilized by Mg²⁺ coordination and the Fic domain LR-type anion-binding nest (G³⁶⁸NG³⁷⁰, not shown). (C adapted from Perera 2021, with permission from the author.)

Glu234 is poised to serve as a gatekeeper that reciprocally regulates the two antagonistic activities of the bifunctional FICD enzyme. Biophysical and cellular clues emerged as to how the disposition of such a gatekeeper residue might be regulated endogenously. FICD forms a dimer in solution (Bunney et al. 2014), but high-level expression of dimerization-competent FICD was unable to promote a pool of AMPylated BiP in FICD^–/– CHO cells. In contrast, low-level expression of the same FICD, or high-level expression of a constitutively monomeric FICD^L258D, promoted a conspicuous pool of AMPylated BiP in FICD^–/– cells (Perera et al. 2019).

These in vivo observations are consistent with a model in which the functionally opposed activities of FICD are reciprocally regulated by its oligomeric state, which is necessarily concentration dependent. In vitro, too, dimeric FICD is an avid deAMPylator with poor AMPylation activity, whereas monomerization, either by mutagenesis or dilution, simultaneously impairs FICD-mediated deAMPylation and significantly stimulates FICD-mediated AMPylation (Casey et al. 2017; Perera et al. 2019, 2021). Correspondingly, dimeric FICD has greater affinity for AMPylated BiP (relative to monomeric FICD), with the converse being true with respect to the engagement of unmodified BiP (Perera et al. 2019, 2021).

These biochemical observations have structural counterparts: In many Fic proteins the regulatory glutamate is presented to the active site from an α-helix (referred to as α_inh; Engel et al. 2012). FICD's Glu234-containing α_inh abuts the dimerization interface. Both isolated FICD proteins co-crystallized in the presence of MgATP, and deAMPylation complexes of FICD and BiP-AMP implicate oligomerization-linked modulation of α_inh and gatekeeper Glu234 flexibility in the workings of FICD's regulatory switch (Figs. 3 and 4, respectively). Increased flexibility of Glu234, afforded by loss of the stabilizing effect of the dimer interface or by mutagenesis along the hydrogen bond network linking the dimerization interface to α_inh, permits AMPylation-competent binding of MgATP (Fig. 3A and 3B), while simultaneously reducing the propensity of Glu234 to assume the conformation required to align a catalytic water molecule for deAMPylation (Fig. 4A, compare the inset salmon and yellow deAMPylation complexes, and Fig. 5A).

Figure 5.

A hypothetical basis for FICD's oligomeric-state responsiveness to changing endoplasmic reticulum (ER) conditions. (A) Cartooned slices through the monomeric FICD:MgATP•BiP AMPylation complex and the dimeric FICD•BiP-AMP deAMPylation complex. Blue arrows within each active site represent the flow of electron pairs. Monomeric FICD possesses a more flexible α_inh and gatekeeper Glu234. Therefore, monomeric FICD is a good AMPylase and a poor deAMPylase, whereas dimeric FICD is a good deAMPylase and poor AMPylase (see Figs. 3 and 4). (A adapted from Perera et al. 2021 under the terms of the Creative Commons CC BY license.) (B) The ER lumen under conditions of low (left) and high (right) unfolded protein load is schematized. Under conditions of low ER luminal unfolded protein load (left), a limited amount of BiP is engaged by unfolded substrate and the ER's ATP/ADP ratio remains high. This results in a high FICD occupancy of BiP:ATP (the AMPylation substrate) and ATP. ATP biases FICD toward a more unfolded state of its α_inh and in so doing allosterically destabilizes the FICD dimer interface (schematized as an inhibitory arrow pointing toward the dimer interface). The combination of substrate-level regulation of BiP:ATP availability and an oligomeric state switch toward a more monomeric FICD population facilitates AMPylation (and functional inactivation) of excess BiP chaperone. In the stressed ER (right), the ATPase activity of BiP will be stimulated as unmodified BiP molecules act to chaperone the increased burden of unfolded client proteins (see Fig. 1). Furthermore, under these conditions the concentration of AMPylation substrate (BiP:ATP) will be reduced. If the net ER luminal ATPase rate exceeds the ATP/ADP exchange rate of the ER lumen with the cytosol/mitochondria, the ER's ATP/ADP ratio will fall. This will reduce FICD's ATP occupancy whilst increasing its ADP occupancy. ADP may itself stabilize a more folded α_inh/inwardly orientated Glu234 conformation, which results in an allosteric reduction in FICD's dimerization K_D,eff—increasing the concentration of dimeric FICD. Dimeric FICD efficiently deAMPylates the inactive pool of AMPylated BiP (BiP-AMP:ATP), liberating more unmodified BiP back into the functional chaperone cycle.

Oligomeric state-linked regulation of FICD's bifunctionality provides a mechanism for efficiently enabling stable accumulation or depreciation of BiP-AMP, whilst mitigating the potential for futile ATP hydrolysis (that would result from successive cycles of AMPylation and deAMPylation). Moreover, FICD oligomeric state regulation likely synergizes with substrate-level regulation of BiP AMPylation (outlined above) to facilitate dynamic buffering of ER chaperone capacity in response to changing unfolded protein abundance. However, unlike substrate-level control, oligomeric state adaptation requires an additional layer of regulation—one that modulates FICD's monomer–dimer equilibrium in response to changing conditions in the ER.

MODULATION OF FICD's MONOMER–DIMER EQUILIBRIUM BY THE ATP/ADP RATIO

ATP and ADP directly modulate FICD's oligomeric state in vitro: ATP increases the dimer dissociation rate and the effective dimerization dissociation constant, K_D,eff. Conversely, ADP decreases the dimerization K_D,eff and efficiently antagonizes the dimer-destabilizing effects of ATP (Fig. 5B; Perera et al. 2019). A complete mechanistic understanding of the coupling of nucleotide binding to dimerization is currently lacking. However, it seems likely that this is the corollary of the process described above, in which monomerization induces increased α_inh and Glu234 flexibility (to permit AMPylation-competent MgATP binding). Bidirectionality in this allosteric pathway would, in turn, facilitate an ATP binding-induced increase in α_inh and Glu234 flexibility and ultimately dimer interface destabilization (discussed further in Perera 2021).

Chaperones are the major ER-resident ATP-binding proteins on a mass basis (Dierks et al. 1996). Moreover, the chaperoning capacity and, by extension, ATPase activity of the ER is dominated by BiP (Bakunts et al. 2017). As the ER membrane spatially and temporally delineates the ER lumen from the ATP buffering systems of the eukaryotic cytosol, it is plausible that the ER's ATP/ADP ratio might dynamically respond to (and serve as a proxy for) changing levels of unfolded protein load within the compartment. Although fluctuations in ER ATP/ADP ratio have yet to be studied experimentally, the in vitro biochemical properties of FICD suggest a mechanism by which its oligomeric state and enzymatic activity may be coupled to the physiological state of the ER (Fig. 5B).

THE FITNESS BENEFITS OF FICD: EVOLUTIONARY AND PHYSIOLOGICAL CONSIDERATIONS

The FICD gene is conserved in all branches of the animal kingdom. Yet, the fitness-promoting features suggested by such conservation are not obvious in the experimental data available to date. FICD is nonessential: CHO cells and AR42J cells genetically engineered to eliminate the encoded protein are superficially indistinguishable from wild-type cells (Preissler et al. 2015b). Worms with a disruption to the catalytic domain of FICD are fertile and enjoy a normal life span under standard laboratory conditions (Truttmann et al. 2016a), and even mice lacking FICD are superficially indistinguishable from wild-type (McCaul et al. 2021, and our unpublished observations).

Predictably perhaps, given BiP's role in negative feedback in the UPR, neutering a mechanism used by cells to create a pool of inactive BiP resulted in a minor delay in the onset of UPR signaling in FICD knockout cells—a delay that was more prominent when application of stress followed exposure to conditions associated with enhanced FICD-mediated BiP AMPylation (Preissler et al. 2015b). A kinetic delay of this sort may have led the Mattoo laboratory to conclude that FICD plays a positive role in UPR signaling (Sanyal et al. 2015). Thus, although the notion that AMPylation can create a pool of readily available inactive BiP—to draw upon and defend proteostasis in times of need through BiP deAMPylation—seems an attractive fitness-enhancing principle, it appears that most animal cells have sufficient redundancy in their coping mechanisms to render this theoretical benefit experimentally inconspicuous.

An alternative benefit of a process that can rapidly inactivate surplus BiP may relate to the cost of over-chaperoning. A chaperone's ability to promote net protein folding depends on complex kinetic parameters that govern the fate of the chaperone-stabilized folding intermediates and the size of the pools of folding intermediates, folded products, and chaperones (González et al. 2002). Thus, although BiP's engagement of unfolded secreted protein precursors holds them in a folding-competent conformation, defends against toxic misfolding, and favors folding of the BiP substrate toward its native state, it also favors a competing process that channels proteins to degradation (Kabani et al. 2003). Levels of active BiP contribute to the outcomes of this quality control process, as BiP overexpression leads to degradation of precursor proteins that are otherwise secreted (Dorner et al. 1992). These considerations suggest that FICD limits the waste that would otherwise occur as cells cope with BiP surpluses arising from fluctuations in the burden of unfolded proteins entering the ER (Fig. 6A and 6B). This idea fits well with the temporal profile of AMPylation, which is most conspicuous as cells transition from high to low activity of the unfolded protein response (Laitusis et al. 1999; Preissler et al. 2015b).

Figure 6.

The fitness benefit of an FICD-mediated, post-translational unfolded protein response. (A) A schematized comparison of the progression of BiP availability and unfolded protein response (UPR) signaling, in a hypothetical secretory cell either possessing (WT) or lacking FICD (FICD^–/–), in response to a transient increase in secreted protein production (unfolded protein load). The bar marked t₀ (here and in C) indicates the latency period of a purely transcriptional UPR, on which FICD^–/– cells are solely reliant. Although the benefits of recruiting dormant BiP (by deAMPylation) may be offset by more chaperone at baseline, BiP AMPylation/ deAMPylation provides mitigation against surplus chaperone without increasing the risk of under-chaperoning and proteotoxicity. Note, the circled symbols 1 and 2 are used throughout the figure to refer to the recovery phase of WT and FICD^–/– cells, respectively. (A adapted from Preissler and Ron 2018, with permission from Cold Spring Harbor Laboratory Press © 2018.) (B) A cartoon contrasting conditions in the ER of WT and FICD^–/– cells during recovery from an acute burst of unfolded protein load. In FICD^–/– cells excess BiP may interfere with the ability of nascent proteins to achieve their fully folded state—thus increasing the rate of ER protein degradation. The size of the green arrows illustrates the resultant fluxes of secretory proteins out of the ER, either through ER–Golgi transport (ER export) or through protein degradation (ERAD). (C) Output of a mathematical model that compares nascent protein aggregation and degradation in a hypothetical pancreatic cell of an animal that has fasted overnight and fed in the early morning, which either contains (WT) or lacks (FICD^–/–) machinery for reversibly inactivating BiP. The model predicts a modest effect of FICD in lessening (dangerous) aggregation and a more conspicuous effect in minimizing (futile) degradation. (C adapted from Chambers et al. 2012, with permission from The Rockefeller University Press © 2012.)

The notion that reversibly inactivating idle BiP defends against over-chaperoning, and a resulting increase in secreted protein degradation, is supported by mathematical modeling (Fig. 6C; Chambers et al. 2012). Nonetheless, this hypothesis presently lacks direct experimental support. This may reflect the ease by which we can experimentally detect catastrophes of misfolding and the difficult accounting needed to uncover the consequences of interfering with the cellular economy. This problem is likely compounded by the fact that, unlike the natural world they aim to emulate, most experimental systems used in life science are not resource limited.

We must also consider exceptions to the bland phenotypic landscape of FICD elimination. In the fly eye, loss-of-function mutations in FICD have been reported to result in photosensitivity and blindness. Initially interpreted as reflecting the loss of an AMPylating enzyme on plasma membrane extension of glia cells (Rahman et al. 2012), in a subsequent study the same group attributed the phenotype to defective BiP AMPylation (Moehlman et al. 2018). This last interpretation is supported by evidence for enhanced UPR activity in the FICD knockout fly eye and by the observation that introduction of a BiP^T366A allele phenocopied the FICD knockout phenotype (Thr366 had been identified by the same group as the site of FICD-mediated BiP AMPylation in cultured Drosophila cells; Ham et al. 2014). Therefore, it remains possible that the fly eye has a special requirement for BiP AMPylation. Conversely, it may be the case that FICD has substrates other than BiP and that loss of their modification compromises the visual system selectively in flies (mice lacking FICD are reported to have normal vision; McCaul et al. 2021). However, it is also worth noting that unlike BiP AMPylation on Thr518, for which there is now ample biochemical and structural validation, our understanding of Drosophila BiP AMPylation on Thr366 (the biochemical correlate of the dramatic phenotypes observed in the fly eye) is less complete (see above). The significance of these intriguing genetic observations will hopefully be revealed in time.

Notwithstanding the mechanism(s) by which retention of the FICD gene is favored, it is interesting to speculate on the process by which it was acquired in the first place. Fic genes are found in all domains of life, but in Eukarya they are restricted to the animal kingdom. Their absence from plants, yeasts, and protozoa suggests that a horizontal transfer event from a bacterium to the last common metazoan ancestor is responsible for FICD's presence in our genomes (Khater and Mohanty 2015). The ancestral gene product may have already targeted BiP or that feature may have arisen later with the fusion of the TPR domains to the catalytic Fic domain. Given the cost of unopposed BiP AMPylation, it is tempting to speculate that deAMPylation—the default activity of modern FICD—might have been present from the outset. This has a precedent, as bacterial effectors include not only AMPylases but also a deAMPylase (Neunuebel et al. 2011; Tan and Luo 2011). Moreover, a bacterial Fic protein, and distant relation of metazoan FICD from Enterococcus faecalis has also been demonstrated to possess both AMPylation and deAMPylation activities that are both dependent on a homologous glutamate residue to FICD's Glu234 (Veyron et al. 2019). This observation suggests that Fic domain bifunctionality may indeed be a widespread, structurally conserved, and largely overlooked feature across a large branch of the Fic protein family.

Given that FICD's opposing activities can be regulated by a dimerization interface that is largely contributed by the core catalytic Fic domain, a region that is evolutionarily conserved and also used as a functional dimer interface in other bacterial Fic proteins, it is possible that the ancestral gene may have arrived with its regulatory switch already in place. Therefore, following modest refinements, such as a transmembrane domain to target and retain the protein in the ER, FICD may have been rendered fit-for-purpose and undergone early fixation in the genome of an ancestral animal.

CONCLUDING REMARKS

AMPylation emerges as the only well-documented and well-understood regulatory modification of BiP. Although it accounts for previous conflicting reports, interpreted at the time as BiP phosphorylation or ADP-ribosylation, other BiP modifications occurring in specialized cells or circumstances cannot be excluded.

The fundamentals of the enzymology of BiP AMPylation and deAMPylation by FICD are now understood; however, the details of the coupling between the regulation of FICD's antagonistic activities and the functional state of the ER remain to be resolved.

Stoichiometrically significant BiP AMPylation on Thr518 and its biochemical consequences are clearly documented; however, the fitness benefits of this regulatory event have yet to be demonstrated experimentally—as animals and cells lacking FICD do not display strong phenotypes.

Regulated BiP AMPylation/deAMPylation on Thr518 may limit a potential inefficiency in the workings of the ER that would otherwise arise from gratuitous degradation of precursors of secreted proteins; however, this remains an unproven speculation. The existence of other meaningful substrates of FICD, or the significance of reports of BiP AMPylation on a different residue (Thr366), also remains largely unsubstantiated.