Small-Molecule FICD Inhibitors Suppress Endogenous and Pathologic FICD-Mediated Protein AMPylation

Abstract

The AMP transferase, FICD, is an emerging drug target fine-tuning stress signaling in the endoplasmic reticulum (ER). FICD is a bifunctional enzyme, catalyzing both AMP addition (AMPylation) and removal (deAMPylation) from the ER-resident chaperone BiP/GRP78. Despite increasing evidence linking excessive BiP/GRP78 AMPylation to human diseases, small molecules that inhibit pathogenic FICD variants are lacking. Using an in vitro high-throughput screen, we identify two small-molecule FICD inhibitors, C22 and C73. Both molecules significantly inhibit FICD-mediated BiP/GRP78 AMPylation in intact cells while only weakly inhibiting BiP/GRP78 deAMPylation. C22 and C73 also inhibit pathogenic FICD variants and improve proinsulin processing in β cells. Our study identifies and validates FICD inhibitors, highlighting a novel therapeutic avenue against pathologic protein AMPylation.

Introduction

Protein AMPylation is a post-translational protein modification (PTM) that regulates protein function by the covalent attachment of an AMP moiety to accessible hydroxyl groups of Thr, Ser, and Tyr side chains.¹ This ATP-dependent process is catalyzed by a dedicated set of enzymes called AMPylases. AMPylases can be broadly classified into two groups: enzymes that possess a highly conserved fic domain (Fic),^2,3 catalyzing the transfer of AMP, and non-Fic enzymes, such as SelO and DrrA, that catalyze AMPylation through a Fic-independent mechanism.⁴⁻⁶ This study focuses on the development of small molecular inhibitors specific for the human Fic AMPylase FICD, which regulates the endoplasmic reticulum (ER) heat shock 70 protein chaperone binding immunoglobulin protein (BiP).⁷⁻⁹

Human FICD, also referred to as Huntingtin yeast-interacting partner E (HYPE), localizes to the ER lumen and is N-glycosylated on Asn275.⁸ Structurally, FICD consists of a single transmembrane domain (residues 24–44), two TPR domains TPR1 (residues 105–135) and TPR2 (residues 140–170), and the conserved, catalytic Fic domain (residues 215–432) joined to the TPR motifs by a short linker (residues 170–215) (Supplementary Figure S1). The TPR motifs dictate FICD’s target recruitment.^7,10−13 The Fic core comprises the conserved catalytic loop and the flap.¹⁴ FICD’s Fic core harbors the highly conserved Fic motif H₃₆₃PFIDGNGRTSR₃₇₄, while the flap (residues 311–324) is involved in positioning of target residues. FICD possesses an autoinhibitory helix (α-inh) containing the inhibitory motif (T/S)V(A/G)IE₂₃₄N.^11,15 FICD catalyzes AMP transfer to target hydroxyl groups via the conserved His363 in the Fic motif, which acts as a base to attack the phosphodiester bond of an ATP molecule, resulting in AMP transfer and the concomitant release of a pyrophosphate group (PP_i). Unlike most enzymes, FICD is bifunctional and catalyzes AMPylation as well as the removal of AMP from modified proteins (deAMPylation) using a single catalytic site.^7,13,16 The switch between AMPylation and deAMPylation states involves changes in enzyme oligomerization/monomerization and an exchange of metal ions coordinating FICD’s active site.¹⁶ Cellular signals that facilitate this switch remain poorly characterized but may involve changes in ER calcium levels.¹⁷

FICD regulates the ER stress response via reversible BiP AMPylation.^7,13,18 Published work is consistent with the model that under unstressed conditions, FICD AMPylates and generates a pool of primed (AMPylated) yet chaperoning-impaired BiP. The emergence of ER stress, however, results in rapid BiP deAMPylation, concomitant with the induction of the unfolded protein response (UPR^ER).^19,20

Two recent studies describe pathologic ficd mutations with clinical implications because of dysregulated ER proteostasis.^21,22 Homozygous FICD^R371S expression in human patients is linked to infancy-onset diabetes mellitus and neurodevelopmental impairments,²¹ whereas homozygous FICD^R374H expression leads to progressive motor neuron degeneration and peripheral neuropathy.²² Both mutations cause FICD to lose its deAMPylation activity while slightly increasing or retaining AMPylation activity. This results in excessive BiP AMPylation impairing UPR^ER signaling with the concomitant accumulation of misfolded and aggregated polypeptides.

In this study, we develop a fluorescence polarization-based high-throughput screen to discover small molecules that reduce FICD-mediated protein AMPylation. We employ this platform to screen 84,480 small molecules obtained from two separate small-molecule libraries, and identify a total of 81 putative FICD inhibitors. Using orthogonal in vitro and cell-based assays, we identify two compounds (C22 and C73) that significantly inhibit endogenous FICD-mediated BiP AMPylation while weakly inhibiting BiP deAMPylation. Molecular modeling implies that C22 and C73 may bind to the dimer interface of endogenous FICD and prevent the dimeric deAMPylase-competent FICD from adopting an AMPylase-competent conformation. We show that both compounds are noncytotoxic small molecules that do not trigger the UPR^ER and are effective against pathogenic FICD mutants in vitro. Finally, we demonstrate that C22 improves proinsulin folding and secretion in pancreatic β cells by reducing basal BiP AMPylation. Our study establishes FICD as a druggable target and suggests that targeting FICD may benefit multiple protein misfolding diseases.

Materials and Methods

Protein Expression and Purification

Human His₆-tagged _45–457FICD constructs (WT, E234G, R371S, and R374H) were cloned into pETDuet-1 plasmids. The plasmids were transformed and expressed inEscherichia coli (E. coli) BL21 or BL21-DE3 cells (Stratagene) and grown in a TB medium containing 50 μg/mL kanamycin to an optical density of 0.8–1. Protein expression was induced by adding 0.4 mM IPTG for 16–20 h at 18 °C. Thereafter, bacteria were collected by centrifugation, and bacterial pellets were sonicated in A lysis buffer (50 mM HEPES, 250 mM NaCl, 10 mM imidazole, and 1× protease inhibitor cocktail, pH 8.0). Lysates were cleared by centrifugation at 15,000g for 30 min. Supernatants were poured over nickel resin pre-equilibrated with a lysis buffer. Thereafter, the resin was washed with a wash buffer (50 mM HEPES, 250 mM NaCl, and 30 mM imidazole, pH 8). His-tagged proteins were eluted in an elution buffer (50 mM HEPES, 250 mM NaCl, and 350 mM imidazole, pH 8). Fractions containing FICD were analyzed for purity by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), pooled, and dialyzed in a dialysis buffer overnight (50 mM HEPES and 150 mM NaCl, pH 8.0).

Human His₆-SUMO-tagged BiP was expressed and purified as described previously.²³ Briefly, pSMT-WT BIP, kindly gifted by Dr. Liu (Virginia Commonwealth University), was expressed in E. coli BL21 cells grown in a TB medium containing 50 μg/mL kanamycin to an optical density of 0.6. Protein expression was induced by adding 1 mM IPTG for 5–6 h at 30 °C. Thereafter, bacterial pellets were sonicated in a lysis buffer and centrifuged at 15,000g for 30 min. Supernatants were poured over nickel resin pre-equilibrated with a lysis buffer. Thereafter, the resin was washed with a wash buffer. His-tagged proteins were eluted with an elution buffer (50 mM HEPES pH 8.0, 250 mM NaCl, and 250 mM imidazole). Fractions containing BiP were verified for purity by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), pooled, and dialyzed in a dialysis buffer (50 mM HEPES and 150 mM NaCl, pH 8.0).

Purified FICD and BiP protein concentrations were measured spectrophotometrically at 280 nm using the Lambert–Beer law.²⁴ Thereafter, FICD and BiP aliquots were flash-frozen in liquid nitrogen and stored at −80 °C in a storage buffer (50 mM HEPES, 150 mM NaCl, and 10% (v/v) glycerol, pH 8.0).

Fluorescence Polarization Assay

The binding kinetics of _45–457FICD^E234G to the fluorescent ATP analog N⁶-(6-aminohexyl)-ATP-5-FAM (Jena Biosciences, cat. no. NU-805-6FM) was determined by incubating increasing concentrations of the enzyme (0.75–2.5 μM), which was dissolved in an AMPylation buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 1 mM EDTA, and 1 mM DTT), with 250 nM FL-ATP (final concentration). A Multidrop nano 384-well reagent dispenser (Thermo Fisher, USA) was used to add FICD, dissolved in an AMPylation buffer, and FL-ATP (dissolved in ultrapure Milli-Q water) to a single 384-well black-bottom, black-walled microplate. The total reaction volume was 20 μL. The plate was snap-centrifuged at 1000g for 60 s, and the initial fluorescence polarization measurements were recorded using a BMG Pherastar plate reader fitted with 485/530 nm filters before it was incubated in the dark at 37 °C. Thereafter, the plate was loaded onto the plate reader and assessed for fluorescence polarization every 15 min from the beginning of the incubation until 120 mins had elapsed. Samples containing FL-ATP in an AMPylation buffer were used for setting the desired fluorescence gain adjustment.

High-Throughput Screening Setup

A Multidrop nano 384-well reagent dispenser was used to pipet 2 μM FICD^E234G dissolved in an AMPylation buffer into columns 1–22. 2 μM _45–457WT FICD was similarly added to column 23 as the negative control. The FICD^E234G enzyme was then incubated with compounds or 1% (v/v) DMSO (positive control) for 10–15 min at room temperature (RT). The Pintool Sciclone ALHD 3000 (PerkinElmer) equipment was used to transfer 200 nL of DMSO-dissolved compounds from 2 mM source plates into 384-well black, flat-bottom, black-walled microplates, to obtain a final concentration of 20 μM in a total volume of 20 μL. Compounds were sourced from the repurposing (FDA-approved drugs for other indications) and DART libraries maintained by the Centre for Chemical Genomics (CCG) at the University of Michigan. The compounds were added to columns 3–22 of each plate. Lastly, the Multidrop reagent dispenser was used to pipet 1 μL of FL-ATP (at a final concentration of 250 nM) into the entire plate. Plates were then incubated for 60 min at 37 °C in the dark. Post incubation, plates were loaded onto a BioTek stacker and scanned using the BMG Pherastar plate reader, in succession, to obtain fluorescence polarization values using 485/530 nm filters.

Z′ and S/B (signal/background) values were determined by fitting the data to eqs 1 and 2, respectively,

1

2

where μ_p and σ_p are the means and the standard deviations of the positive control samples and μ_n and σ_n are the means and the standard deviations of the negative control samples, respectively.

Concentration–Response Curves (CRC)

A Multidrop nano 384-well reagent dispenser was used to pipet 1 μM FICD^E234G (positive control) or WT FICD dissolved in an AMPylation buffer into designated microplate wells, which already contained either 200 nL of DMSO-dissolved compounds or an equivalent volume of 1% DMSO. Each compound was used at eight concentrations determined in accordance with semilog fold dilutions starting from 30 nM. The reaction mixture was incubated for 10–15 min at RT. One μL of FL-ATP (final concentration of 250 nM) was dispensed using the automated reagent dispenser into the whole plate, which was incubated for 90 min at 37 °C in the dark. It was subsequently transferred to the BMG Pherastar plate reader fitted with 485/530 nm filters to record fluorescence polarization. The same setup was used to obtain concentration–response curves for the commercially obtained Closantel analogs.

IC₅₀ values were determined by fitting polarization values to eq 3:

3

where Y is the polarization signal; X is the log concentration of the inhibitor (μM), and IC₅₀ is the concentration of the inhibitor that elicits a response halfway between bottom and top. This is not the same as the response at Y = 50. The hill slope describes the steepness of the family of curves.

Tissue Culture

A549 (ATCC, CCL-185), HeLa (ATCC-CRM-CCL2), and SK-N-SH (HTB-11) cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS) and a 1% penicillin–streptomycin mixture (GM) at 37 °C in 5% CO₂ until they reached approximately 80% confluency. Cells were washed 1x with PBS , trypsinized, resuspended in GM, and plated in either 6- or 96--well plates for subsequent assays.

Min6 (mouse insulinoma) cells were cultured in a DMEM medium (25 mM glucose) supplemented with 10% FBS, 100 IU/mL penicillin and 100 μg/mL streptomycin, and 0.05 mM β-mercaptoethanol (GM6).

Neonatal primary cardiomyocytes were isolated from C57B6/J mice (postnatal day 1–3) as previously described.²⁵ Nonadherent cardiomyocytes were washed from the dish and replated in collagen-precoated 96-well plates. Isolated cells were maintained in a plating medium for 24 h.

BiP AMPylation Kinetics in A549 Cells

Cells were grown in 6-well plates. After the cells reached approximately 60% confluency, GM was removed, and cells were incubated for 15, 30, 45, and 60 min in sterile PBS to assess BiP AMPylation levels.

When working with compounds, we preincubated cells with molecules C55, C83, C84, C522, C22, C73, and C34 in GM for approximately 12 h. GM was removed, and cells were exposed to PBS supplemented with the compounds for 60 min. Cells preincubated with 0.5% (v/v) DMSO in GM or PBS served as negative controls.

When determining whether preincubation was sufficient for affecting BiP AMPylation, we preincubated cells with either 10 μM FICD inhibitors (C22 or C73) or DMSO and subsequently exposed the cells to PBS for another 60 min supplemented with either the FICD inhibitors or DMSO only.

When determining whether supplementing FICD inhibitors in PBS is sufficient and necessary to affect BiP AMPylation, we did not preincubate the cells with compounds. Instead, cells were grown in GM until they reached approximately 70% confluency. Then, GM was removed, and PBS was added to the cells for 60 min. FICD inhibitors or DMSO was added to PBS at 0 (immediately), 15, and 30 min post PBS addition.

BiP DeAMPylation Kinetics in A549 Cells

Cells were grown to approximately 70% confluency. Next, cells were incubated with sterile PBS for 60 min. Post incubation, PBS was removed, and GM was added for 1, 3, 5, 10, and 15 min.

When working with FICD inhibitors, cells were incubated for 5 or 15 min in GM supplemented with either 10 μM compounds (C22 or C73) or 0.5% (v/v) DMSO. Post GM incubation, cells were washed once with PBS to remove residual GM.

In all AMPylation and deAMPylation experiments, post incubation, cells were harvested and sonicated in a cell lysis buffer (20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1% NP-40, 2 mM EDTA, and 1x protease inhibitor cocktail). The cell lysates were centrifuged at 10,000g for 10 mins at 4 °C. The supernatant was carefully removed and used in a bicinchoninic protein assay (BCA) to assess protein concentrations. A 4× Laemmli sample loading buffer (BioRad) was added to the supernatant, and the mixture was boiled for 5 min at 95 °C. Ten μg of the supernatant was loaded onto two separate 10% SDS-PAGE gels and resolved. Proteins were then transferred to PVDF membranes and blocked with 5% (w/v) milk or bovine serum albumin (BSA) in Tris-buffered saline supplemented with 0.1% (v/v) Tween-20 (TBST). One of the membranes was blotted with mouse anti-Thr AMP (17G6, Biointron) while the other with mouse anti-BiP (Proteintech). Mouse HRP-conjugated anti-GAPDH (Proteintech) and mouse anti-α-tubulin (Developmental Studies Hybridoma Bank) were used as loading controls. Membranes were incubated with primary antibodies (1:1000, diluted in TBST with 5% BSA or milk) at 4 °C overnight and then incubated with an HRP-conjugated-secondary antibody (1:5000) for 1 h at RT. The membranes were incubated with a ProSignal Dura ECL reagent (Prometheus) at RT for 2 min and imaged using an Invitrogen iBright FL1500 imaging system. Signals were quantified using ImageJ2 software.²⁶

Proinsulin Secretion and Folding

Min6 cells were grown to 80% confluency and then fed fresh GM supplemented with C22 (20 μM) or DMSO for 16 h. Thereafter, both cells and media were collected, and cells were lysed in a RIPA buffer supplemented with a protease inhibitor cocktail. Cell lysates were clarified by centrifugation for 15 min at 12,000 rpm. Before electrophoresis, samples were boiled at 95 °C in an SDS-gel sample buffer under either nonreducing or reducing (200 mM DTT) conditions, and then resolved either on a nonreducing 15% SDS-PAGE in Tris-glycine buffer or a reducing 4–12% gradient NuPage gels. Proteins were then transferred to nitrocellulose membranes and blotted with mouse anti-proinsulin (Novus Biologicals), rabbit anti-BiP (Thermo), and mouse anti-AMPylated-BiP (17G6, Biointron) antibodies. Mouse anti-β-actin (Proteintech) was used as the protein loading control. Membranes were incubated with primary antibodies (1:1000, diluted in TBST with 5% BSA) at 4 °C overnight and then incubated with a HRP-conjugated secondary antibody (1:5000) for 1 h at RT.

In Vitro Cytotoxicity Assay

A 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT) assay was performed to determine the effect of compounds on the viability of HeLa (ATCC-CRM-CCL2), A549 (CCL-185), SK-N-SH (HTB-11) cells and neonatal murine cardiomyocytes. Immortalized cells were washed 2x with PBS, trypsinized, resuspended in GM, and plated in a 96-well plate at a seeding density of 1.0 × 10⁴ cells/well. When the cells attained approximately 70% confluence, they were incubated with the indicated concentrations of C22 or C73 and kept at 37 °C in 5% CO₂ for approximately 24 h. Cells incubated with 1% (v/v) DMSO alone served as negative controls. Final DMSO concentrations were kept below or at 1% (v/v). Post incubation, cells were washed once with sterile PBS. Ten μL stock MTT solution (5 mg mL^–1 in PBS) was diluted in 100 μL of PBS and added to the cells, which were further incubated at 37 °C in 5% CO₂ for approximately 1–2 h. The purple formazan crystals thus formed were dissolved in 100 μL of sterile DMSO, and the absorbance of the resulting mixture was measured using the Agilent BioTek Epoch 2 spectrophotometer at a sample wavelength of 540 nm and a reference wavelength of 630 nm. The lethal dose 50 (LD₅₀) values for C22 and C73 were determined by performing nonlinear regression analysis on the sigmoidal dose response curves obtained by fitting the data using GraphPad Prism (version 9.3.1, GraphPad Software). All experiments were carried out in triplicate.

RNA Isolation, Processing, and Quantitative PCR (qPCR)

Total RNA was isolated from 1 × 10⁶ A549 cells treated with 0.5% (v/v) DMSO or 5 μM FICD inhibitors (C22 or C73) for approximately 24 h using an RNA miniprep kit (Zymo Research) and quantified using a NanoDrop One Microvolume UV–vis spectrophotometer (Thermo Fisher). Cells exposed to 25 μg/mL tunicamycin, a well-known endoplasmic reticulum (ER) stress inducer, served as positive controls because tunicamycin induces the unfolded protein response (UPR) in the ER. One μg of RNA was reverse-transcribed using the high-capacity cDNA reverse transcription kit (Applied Biosystems) to obtain complementary DNA (cDNA). Twenty ng of cDNA was mixed with an ABclonal SYBR Green Master Mix (AbClonal) and appropriate primers for the human genes of interest (Supplementary Table S1), then plated in a single 384-well plate. The total volume of the reaction mixture was 20 μL. The plate was centrifuged at 600g for 2 min at RT and loaded onto a QuantStudio 5 real-time PCR system. PCR conditions were determined following the AbClonal protocol. Samples representative for each treatment, as described above, were pipetted in triplicate for each gene. Amplicons were quantified by comparison of 3-average ΔΔCT. Fold changes in transcript levels were computed relative to β-actin (ACTB).

Coimmunoprecipitation Assay

Cells were treated as described and lysed in RIPA buffer containing 1x protease inhibitor cocktail for 15 min on ice and sonicated at 20% for 30 s on ice. The cell lysate was centrifuged for 15 min at 4 °C, and protein quantification was performed using the DC protein assay kit (BioRad). Cell lysates were precleared using Pierce Protein A magnetic beads (Thermo Fisher) from for 1 h at 4 °C, and the beads were removed using a magnetic rack. The cell lysate (∼150 μg) was incubated with 3 μL of the Myc-tag antibody (Cell Signaling) overnight at 4 °C. Protein G magnetic beads (20 μL) were added to the mixture for 2 h at 4 °C. The immunoprecipitated proteins were isolated from the cell lysate mixture using the magnetic rack. The isolated beads were washed 3x with the cell lysis buffer. The bound proteins were released from the beads by boiling them in a SDS loading buffer for 10 min. These proteins, along with 5% input samples, were then resolved on 10% Tris-glycine SDS-PAGE gels and transferred onto PVDF membranes overnight at 4 °C. The membranes were probed for Myc-tag and associated proteins. This assay was performed thrice, and the bands were quantified using the Image Lab software from BioRad.

In Silico Docking and Molecular Dynamics (MD) Simulation

Rigid Docking of C22 and C73 to FICD Variants

The FASTDock program²⁷ was used to identify putative binding sites on each of the target proteins discussed in this study. The default set of 18 chemical probes were used, and 2000 best docked poses were retained for each probe. For each probe, the top 5 clusters were considered for the next steps. CDOCKER²⁸ was used to dock compounds C22 and C73 at the top 5 putative binding sites. For each compound, the following docking protocol was applied: at the 5 binding sites, 10 different rotamers of each compound were generated using OpenBabel, and docked while keeping the protein receptor region fixed (represented by a grid).²⁸ Thereafter, these 10 poses were rescored by applying the Fast Analytical Continuum Treatment of Solvation (FACTS) model,²⁹ which accounts for the desolvation penalty associated with each pose at the binding site. The FACTS-rescored docking scores were averaged across all poses. This protocol was independently repeated 10 times at each site for both compounds, thereby yielding an average dock score computed from a total of 100 poses for each compound. The site with the most favorable (negative) averaged FACTS-rescored dock score was considered the top binding site, and the best binding pose at that site exhibited the highest (most negative) FACTS-rescored dock score. 2D diagrams of protein–ligand interactions was generated using PlexView (https://playmolecule.org/PlexView/).

Molecular Dynamics (MD) Simulations

The stability of the FICD inhibitors at their (previously obtained) respective binding poses was analyzed using MD simulations. The CHARMM molecular simulation program³⁰ was used to perform these simulations. Receptors bound to C22 or C73 were placed in a bath of explicit water models (TIP3P model), and counterions (Na⁺ or Cl^–) were added to neutralize the charge of the solvated system. The following scheme was used to perform all MD simulations in this study. First, the solvated system energy was minimized to eliminate bad contacts introduced during system preparation. During minimization, which included 2000 steps of steepest descent minimization followed by 1000 steps of ABNR minimization, the receptor and compound (ligand) heavy atoms were harmonically restrained to their initial poses with a force constant of 10 kcal/mol/Ų. Next, an equilibration step was carried out under constant temperature (310 K) and volume conditions for 1 ns. The previous restraints were retained. Third, the restraints were reduced to 5 kcal/mol/Ų but were still imposed on the receptor and ligand heavy atoms, and the next phase of equilibration was carried out under constant temperature and constant pressure (1 atm) for another 1 ns. A penultimate phase of constant temperature and pressure (NPT) equilibration was initiated with restraint forces reduced to 1 kcal/mol/Ų for 1 ns. This was followed by another 1 ns of NPT equilibration with a very small restraint of 0.1 kcal/mol/Ų. Finally, the production simulation was run for 5 ns under NPT conditions without any restraints.

The nonbonded interactions within the system were truncated at 12 Å after the application of potential switching function starting at 10 Å. Particle mesh Ewald (PME) was used to treat long-range electrostatics with k = 0.32 Å^–1, order = 4, a grid size of 0.8–1.2 Å, and a force error tolerance of 10^–5. SHAKE³¹ was employed to constrain the distances of hydrogen–heavy atom bonds after performing hydrogen mass repartitioning.³² The temperature was regulated using a Langevin thermostat with a friction coefficient of 5 ps^–1. Pressure, in the constant pressure simulations, was isotopically regulated using an MC barostat with volume changes attempted every 25 steps. The integration time step was set to 2 fs.

Force field parameters of the receptors, TIP3P water model, and counterions were taken from the CHARMM36 force field.³³⁻³⁵ Parameters for the FICD inhibitors were obtained using the CGenFF program (v 2.5.1).³⁶

MM/PBSA Calculations

End point free energy calculations using the MM/PBSA technique were carried out using the snapshots generated from the production phase of the simulations. Only the last 4.5 ns of the 5 ns of the production was used to obtain snapshots every 40 ps. We used the single-trajectory protocol of MM/PBSA where the energy terms associated with the receptor–ligand complex, the receptor, and the ligand were all derived from the same snapshot as opposed to individual simulations conducted for each of these systems separately. The molecular mechanical (MM) energy terms and the surface area (SA) terms were determined using the CHARMM program. The SA term was used to compute the nonpolar component of the solvation energy by coupling it with a surface tension value of 0.005 kcal/mol/Å.² The Poisson–Boltzmann (PB) framework was used to compute the polar component. Delphi³⁷ was used for the PB calculations. The internal and external dielectric values were set at 1 and 80, respectively. The salt concentration was set to 0 outside the solute’s (receptor, ligand, or the complex) molecular surface, thereby reducing the Poisson–Boltzmann equation to Poisson equation only. The molecular surface itself was drawn using a solvent probe radius of 1.4 Å to emulate water. The solute atoms’ charges and radii were taken directly from the set used in the MD simulations.

Statistical Analysis

Statistical analysis was performed using the GraphPad Prism (version 10.2.2) software. Unpaired t tests with Welch’s correction and two-way ANOVA tests were performed. Figure legends specify the utilized tests for each data panel. p values were computed to determine statistical significance. If a p value is less than 0.05, then it is flagged with one star (*). If a p value is less than 0.01, then it is flagged with two stars (**). If a p value is less than 0.001, then it is flagged with three stars (***), and if a p value is less than 0.0001, then it is flagged with four stars (****).

Rigor and Reproducibility

All hit validation assays were performed at least in triplicate. For Western blot-based quantifications, figures in the main manuscript show one representative Western blot. The other replicates are shared in the supplementary figures.

Results

In Vitro Fluorescence Polarization Screening Identifies Putative FICD Inhibitors

To identify molecules that inhibit FICD-dependent protein AMPylation, we optimized an in vitro fluorescence polarization (FP) assay initially developed to identify FICD activators and VopS inhibitors^38,39 (Supplementary Figure S2A). In this assay, we used a fluorescent ATP analog N⁶-(6-aminohexyl)-ATP-5-FAM (FL-ATP) (Supplementary Figure S2B) to measure auto-AMPylation of mutant FICD^E234G (Supplementary Figure S2C). Unlike wild-type (WT) FICD, which preferentially adopts a deAMPylase-competent but AMPylation-deficient dimeric conformation, FICD^E234G acts as a deAMPylation-deficient, constitutive AMPylase resulting from conformational changes in the FICD inhibitory helix that relieve autoinhibition.^8,9,11 To determine optimal reaction conditions, we tested different concentrations of FICD^E234G (Supplementary Figure S2D) and buffer components (Supplementary Figure S2E). We found that using 1 μM FICD and 0.25 μM FL-ATP in a buffer containing 50 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 1 mM EDTA, and 1 mM DTT provided a reliable dynamic window enabling us to screen for both FICD^E234G activators and inhibitors, which is consistent with previous studies on FICD-mediated protein AMPylation in vitro.(16,18,20,39−44) We then validated these reaction conditions, comparing FP measurements of autoinhibited FICD WT and FICD^E234G in a 384-well plate setup, which confirmed assay reproducibility and reliability (Z-factor = 0.83) (Figure 1A).

Figure 1.

High-throughput screening (HTS) assay. (A) Assay variability was assessed by computing Z′. 2 μM WT or FICD^E234G diluted in an AMPylation buffer was incubated with 250 nM FL-ATP. Wells containing only FL-ATP served as negative controls. (B) Randomized-well activity scatter plot of compounds from the repurposing library. Compounds were screened, in singlets (black), to identify putative FICD inhibitors. 1 μM WT FICD (red) or FICD^E234G (teal) was pipetted onto a 384-well plate. Thereafter, compounds (at a final concentration of 20 μM) or DMSO (1% v/v) were added into test and control wells, respectively. An auto-AMPylation reaction was initiated by adding FL-ATP to each well, and the reaction plate was incubated in the dark at 37 °C for 90 min. (C) Z′ scores and S/B ratio for 15 plates representing the repurposing library. (D) AMPylated and total BiP levels of A549 cells maintained in GM or treated with sterile PBS for the indicated time points. (E) Quantification of (D). (F) AMPylated and total BiP levels of A549 cells treated with PBS for 1 h in the presence of either 0.5% (v/v) DMSO (control) or C22 (at indicated concentrations). (G) Quantification of (F). (H) AMPylated and total BiP levels of A549 cells treated with PBS for 1 h in the presence of either 0.5% (v/v) DMSO (control) or C73 (at indicated concentrations). (I) Quantification of (H). GAPDH was used as the protein loading control. A two-way ANOVA was performed to assess statistical significance between control and treated samples. Western blots of two remaining biological replicas for panels (D), (F), and (H) are shown in Supplementary Figure S4.

Using this optimized assay, we screened two distinct small-molecule libraries: a commercially available (Selleck) repurposing library composed of 5120 FDA-approved compounds (Figure 1B) and a custom library consisting of 79,360 compounds from the Dart Neuroscience Collection. We maintained an average Z′ score of >0.6 and an average signal-to-baseline ratio (S/B) of >3 (Figure 1C) during our screening, offering a large dynamic window to identify potential FICD inhibitors. In our primary screening, we defined hits as compounds that inhibit FICD^E234G activity by ≥20% or show a standard deviation (S.D.) ≥3 from WT FICD activity as primary hits. Using these criteria, we identified 4019 compounds in our primary screening (4.7% hit rate).

Screening Funnel Confirms Seven Compounds as FICD Inhibitors

To validate our primary hits, we followed a classic small-molecule testing paradigm (Supplementary Figure S3A). First, we retested all 4019 primary hits in a secondary screening, which was performed in triplicate. These experiments confirmed 646 of the initial 4019 compounds to reduce FICD^E234G auto-AMPylation. Of the 646 compounds, we selected 126 compounds that reduced FICD^E234G activity in both screens by at least 20%, showed good accessibility for side-chain modifications, and were not reported to exhibit pan-assay interference (PAINs). Of these 126 compounds, 55 belonged to the repurposing library and were tested in concentration–response (CRC) experiments, in which 45 showed dose-dependent FICD inhibition. The subsequent exclusion of known modulators of human enzymes further shrank our hit list to 6 molecules. Next, we used MScreen,⁴⁵ a compound management and HTS data analysis tool, to search for analogs exhibiting at least 40% structural similarity to the selected 6 hits. We identified 133 additional analogs, which were subsequently tested in triplicate. After these experiments, we were left with 19 compounds that reduced FICD^E234G auto-AMPylation by at least 20%. CRC experiments confirmed 13 of these 19 compounds as dose-dependent FICD^E234G inhibitors. We then repeated these tests using fresh, commercially sourced, chemical matter which confirmed 7 compounds (Supplementary Figure S3B) exhibiting IC₅₀ < 20 μM (Supplementary Figure S3C and Supplementary Table S2). Among those 7 compounds were liothyronine (C522), a marketed therapeutic for treating hypothyroidism and myxedema coma, as well as structurally related compounds, C83, C84, and C53. Closantel (C22 and C73), an antiparasitic used by veterinarians, was also identified. Interestingly, closantel shares some structural similarities to liothyronine in that they both contain a polyiodinated phenolic moiety and have multiple halogens appended to other regions of the molecule. In contrast, compound C47, another putative FICD inhibitor that we identified, belongs to the viridacatumtoxin family of fungus-derived antibiotics and is structurally dissimilar from the liothyronine and closantel series.

Compounds C22 and C73 Efficiently Inhibit FICD-Mediated BiP AMPylation in Intact Cells

Our initial in vitro assays tested for inhibition of auto-AMPlyation of FICD^E234G, a mutant enzyme version with constitutive AMPylation activity. In the next step, we determined if the seven putative FICD inhibitors reduced the activity of endogenous FICD in intact cells. The ER-resident chaperone, BiP, represents the most reliably AMPylated FICD target.^{8,9,18,19,22,40,43,44,46} Because of the wealth of information on FICD-mediated BiP AMPylation and the availability of an antibody that reliably detects AMPylated BiP,⁴⁷ we next assessed our FICD inhibitors in cell-based BiP AMPylation assays. We incubated A549 cells with the compounds for approximately 12 h in DMEM supplemented with 10% fetal bovine serum (GM). Afterward, we replaced GM with sterile PBS containing the compounds, and incubated the cells for another hour. In the absence of FICD inhibitors, the PBS treatment acutely depleted cells of nutrients, which led to a progressive increase in BiP AMPylation that reached saturation after approximately 45 min (Figure 1D,G). In contrast, we observed that compounds C22 (Figure 1E) and C73 (Figure 1F) promoted a concentration-dependent inhibition of FICD-mediated BiP AMPylation in response to nutrition shortage. Compounds C47, C55, C83, C84, and C522 did not affect BiP AMPylation levels (Supplementary Figure S4). At the highest tested compound concentration (10 μM), we observed ∼ 80% decrease in BiP AMPylation levels for both C22 (Figure 1H) and C73 (Figure 1I). These results confirm C22 and C73 as potent cell-permeable FICD inhibitors.

FICD Inhibitors C22 and C73 Have a Favorable Cytotoxicity Profile and Do Not Induce the UPR^ER

Our cell-based assays showed that using compounds C22 and C73 at concentrations of 10 μM led to efficient FICD inhibition. To determine whether they are cytotoxic at micromolar doses, we incubated HeLa, A549, and SK-N-SH cells with increasing concentrations of each compound. We found that both compounds were well-tolerated by all three immortalized cell lines, exhibiting LD₅₀ (lethal dose₅₀) values between 88 and 190.3 μM (Supplementary Table S3 and Supplementary Figure S5A–C). Repeating the experiments using primary murine cardiomyocytes, a cell type particularly sensitive to small-molecule toxicity⁴⁸⁻⁵⁰ (Supplementary Figure S5D) confirmed the limited cytotoxicity of compounds C22 and C73. Since FICD is a key regulator of the unfolded protein response in the endoplasmic reticulum (UPR^ER),^7−9,51 we next assessed whether C22 and C73 induce UPR^ER signaling. We performed quantitative PCR (qPCR) to evaluate expression levels of genes regulated by the UPR^ER. We found that while the treatment with tunicamycin, a known ER stressor,^52,53 induced a strong UPR^ER, the exposure to C22 and C73 at inhibitory concentrations did not change the expression of selected genes involved in regulating ER homeostasis, inflammation, and apoptosis including FICD, BiP, activating transcription factor 4 (ATF4), spliced (s) and total (t) X-box binding protein 1 (XBP1), activating transcription factor 6 alpha (ATF6A), CASPASE4,and CASPASE6(54) (Supplementary Figure S5E). We next tested whether preincubation of cells with C22 for 6 h would affect tunicamycin and thapsigargin-induced UPR^ER induction. Thapsigargin is a SERCA inhibitor promoting ER stress by depleting ER calcium stores.⁵⁵ Our experiments showed that pretreating cells with C22 did not lead to a significant change in tunicamycin- and thapsigargin-induced UPR^ER, only resulting in a nonsignificant trend to slightly attenuate ATF6A and ATF4 expression in thapsigargin-treated cells (Supplementary Figure S5F–J). Finally, we assessed CHOP protein levels in cells exposed to FICD inhibitors, which confirmed that neither FICD inhibitor promoted enhanced CHOP expression (Supplementary Figure S5K). These results provide strong evidence that FICD inhibitors C22 and C73 are tolerated across cell types at concentrations well outside the therapeutic window and do not disturb UPR^ER-mediated cellular processes.

C22 and C73 May Bind Both Monomeric and Dimeric FICD In Silico

In unstressed cells, FICD preferentially resides in a deAMPylation-competent dimeric conformation. During our cell-based experiments, we exposed cells before and during nutrient depletion stress to FICD inhibitors, which decreased overall BiP AMPylation. We considered two possible explanations for this finding: First, the compounds could bind and stabilize the dimeric deAMPylation complex, thus increasing BiP deAMPylation. Second, the compounds could bind to monomeric FICD once available, preventing de novo BiP AMPylation. To understand whether FICD inhibitors show a preference for binding monomeric or dimeric FICD, we assessed their binding affinities to apo, dimeric WT FICD (PDB ID 4U04) and monomeric FICD^L258D (PDB ID 6I7J) by in silico docking. First, we identify all potential binding sites in the WT dimeric FICD using FASTDock.²⁷ Then, we used CDOCKER²⁸ to dock the FICD inhibitors to these binding cavities and rank the top 5 binding sites for both C22 and C73 (Supplementary Table S4andSupplementary Figure 6A). Two of these putative binding sites were particularly interesting: the smaller, potentially more flexible dimeric interface¹¹ (site #1) and residues near the tetratricopeptide repeat (TPR-II) domain (site #2) (Supplementary Figure 6B). C22 displays a stronger binding affinity (−28.86 ± 4.89 kcal/mol) for site #1 than site #2 (−20.83 ± 2.75 kcal/mol), while C73 displays similar binding affinities of −17.95 ± 5.25 and −15.232 ± 9.99 kcal/mol to sites #1 and #2, respectively. Moreover, equilibrium MD simulations performed in the presence of explicit water molecules show that both compounds retain their respective binding modes with marginal deviations from the initial docked pose. This is evidenced by the fact that the structural root-mean-square deviation (RMSD) (of only the heavy atoms) for both compounds did not exceed 6.5 Å from the initial docked poses for majority of the MD simulation time course (Supplementary Figure 6C).

Molecular mechanics Poisson–Boltzmann surface area (MM/PBSA) calculations are considered more accurate in estimating binding free energies of protein–ligand complexes than scoring algorithms of most docking programs.⁵⁶ Hence, we used MM-PBSA scores as indicators of absolute binding energies for poses obtained from MD simulations of our docked FICD–inhibitor complexes. We found that residues near the TPR-II domain, which previously constituted the second ranked site for C22, displayed an approximately 2-fold higher MM-PBSA score of −28.70 ± 4.91 kcal/mol as compared to the smaller dimeric interface (−14.74 ± 5.02 kcal/mol). This change can be attributed to a stronger binding mode because of the cation−π interactions between the guanidinium group on the side chain of ARG180 (packed against the TPR-II domain) and the two aryl groups of C22 (Figure 2A). In contrast, the only prominent interaction is a hydrogen bond (H-bond) between the nitrile group in C22 and SER288 of one of the FICD monomers at site #1 (Figure 2A), which could explain the lower affinity for that site. C73, the deprotonated form of C22, in accordance with our previous docking results, shows slightly favorable binding at site #2 (−269.61 ± 16.70 kcal/mol) compared to site #1 (−256.38 ± 7.31 kcal/mol). This slightly increased affinity is a result of the favorable cation−π interactions as well as a H-bond with the ARG180 side chain at site #1 compared to H-bonds with the side chains of ARG308 and ARG293 via its nitrile group at site #2 (Figure 2B).

Figure 2.

In silico docking and MD simulations of FICD inhibitors bound to dimeric and monomeric FICD. (A,B) Top two putative binding sites on the apo dimeric WT FICD (PDB ID 4U04) for C22 (A) and C73 (B). #1 and #2 denote the top 2 sites. The inhibitory helix, catalytic core, and TPR-II domains are highlighted in light magenta, salmon, and green, respectively. Prominent contacts between the compounds and their neighboring residues at these sites are shown in the cartoon representation. Contacts were drawn using PlexView and derived from the snapshot with maximum contacts across the last 100 snapshots (amounting to the first 4.5 ns of production simulation time). (C) Top binding site for C22 and C73 to the monomeric FICD^L258D (PDB ID 6I7J). Prominent contacts are illustrated on the side. Yellow-orange, gray, purple, red, and green circles represent aromatic, apolar, negatively charged, positively charged, and polar interacting residues. Dotted black lines and dotted orange lines represent hydrogen bonds and cation−π interaction, respectively.

In the next step, we used the same protocol to identify and rank putative binding sites of C22 and C73 on the strictly monomeric FICD variant,¹⁶ FICD^L258D (Figure 2C). We found C22-bound monomeric FICD with a heightened affinity to the top binding site as indicated by a higher MM/PBSA score (−43.33 ± 6.52 kcal/mol) as compared to the top 2 binding sites on the dimeric FICD. At this site, the oxygen atoms of C22 formed H-bonds with ASN274 and LYS271 (Figure 2C). These H-bonds, in conjunction with the lack of cation−π interactions as observed in the dimeric form, explained the stronger binding affinity for the monomeric FICD. In contrast, C73 displayed a reduced binding affinity to monomeric FICD at the top binding site compared to the top 2 sites on the dimer (−124.28 ± 9.23 kcal/mol) and engages in a single H-bond with ASN274 (Figure 2C).

C22 and C73 Inhibit Both DeAMPylation- and AMPylation-Competent FICD States in Intact Cells

To further characterize C22 and C73, we preincubated A549 cells with 10 μM C22 and C73 for approximately 12 h, allowing them to bind and stabilize deAMPylation-competent WT FICD (Supplementary Figure S7). We utilized both regular DMEM and commercially available serum-free cell growth media (Opti-MEM) to exclude that the FBS we used in GM reduced bioavailability of the inhibitors.⁵⁷ Afterward, we exposed these cells to PBS containing DMSO or compounds at a 10 μM concentration. We found that preincubating cells with the FICD inhibitors in Opti-MEM, prior to nutrient depletion, was sufficient to prevent BiP AMPylation. In contrast, cells preincubated with C22 or C73 in GM show no reduction in AMPylated BiP. Cells continuously exposed to C22 or C73 in both media showed significant reductions in AMPylated BiP levels (Supplementary Figure S8 and Figure 3A,B). These results are consistent with our in silico results and confirm that both C22 and C73 are capable of stabilizing the deAMPylation-competent FICD conformation, thereby preventing its switch to an AMPylation-competent state. To determine if C22 and C73 further inhibited the AMPylase-competent FICD conformation, we devised an experiment in which we delayed the addition of FICD inhibitors until after the induction of nutrient depletion stress. This delay provided the cells with enough time to at least initiate the conformational switch required to transition from FICD-mediated deAMPylation to AMPylation. The addition of either C22 or C73 at the onset of PBS-induced nutrient shortage significantly reduces BiP AMPylation. Delaying inhibitor addition by 15 min attenuated the decline in BiP AMPylation, while a 30 min delay in C22 or C73 addition had no effect on BiP AMPylation levels (Figure 3C,D). Taken as a whole, these results support a model in which C22 and C73 are acting during the transition phase when the enzyme is adopting an AMPylation-competent conformation. This could occur through at least two mechanisms: First, C22 and C73 could stabilize FICD in a deAMPylation-competent conformation and prevent the transition. Second, the compounds could prevent AMPylation-competent FICD from engaging in BiP AMPylation.

Figure 3.

Effect of FICD inhibitors on deAMPylation- and AMPylation-competent FICD states. (A) AMPylated and total BiP levels in A549 cells preincubated with the 10 μM FICD inhibitors in GM or OM and subsequently exposed to PBS supplemented with 10 μM FICD inhibitors for 15 min. Cells preincubated with 0.5% (v/v) DMSO and exposed to PBS supplemented with 0.5% (v/v) DMSO served as positive controls, while cells grown in GM or OM supplemented with 0.5% (v/v) DMSO but not exposed to PBS served as negative controls. From left: Lanes 1–2 represent negative controls, lanes 3–4 represent positive controls, lanes 5–6 represent cells preincubated with C22 and exposed to PBS supplemented with C22, while lanes 7–8 represent cells preincubated with but not subsequently exposed to C22 during PBS treatment. Lanes 8–12 follow the exact order as lanes 5–8 but represent cells treated with C73. (B) Quantification of (A). (C) AMPylated and total BiP levels in A549 cells exposed to PBS supplemented with 10 μM C22 or C73 at the indicated time points and lysed 60 min post PBS addition. Cells treated with 0.5% (v/v) DMSO in PBS for approximately 60 min served as the positive control. (D) Quantification of (C). (E) AMPylated and total BiP levels of PBS-treated cells incubated in GM for the indicated time points. (F) Quantification of (E). (G) AMPylated and total BiP levels of PBS-treated cells incubated in GM with or without 10 μM C22 or C73, for the indicated time points. (H) Quantification of (G). α-Tubulin was used as the protein loading control. Cells treated with 0.5%(v/v) DMSO in PBS served as the positive control. GM indicates cells never exposed to PBS, which served as negative controls. BiP-AMP/BiP signal intensity ratios were computed relative to positive controls. A two-way ANOVA was performed to assess statistical significance between control and treated samples. The asterisk (*) indicates the protein band used for quantification purposes; the lower protein band is nonspecific for BiP AMPylation. Western blots of two remaining biological replicas for panels (A), (C), (E), and (G) are shown in Supplementary Figure S7.

C22 and C73 Only Weakly Inhibit BiP DeAMPylation in Intact Cells

FICD is a bifunctional enzyme preferentially adopting a deAMPylase-compentent conformation in the absence of stress.^19,20 To further characterize the mode of action of compounds C22 and C73, we tested whether these molecules could inhibit FICD-mediated BiP deAMPylation in intact cells. Our assay was based on the finding that starvation-mediated BiP AMPylation in A549 cells was quickly reversed upon incubating the cells in a complete medium (Figure 3E). The incubation of nutrient-depleted cells in GM for 5 min was sufficient to significantly reduce AMPylated BiP levels, with a maximal ∼60% decrease occurring after 15 min (Figure 3F). The addition of both C22 and C73 delayed but did not prevent BiP deAMPylation from occurring (Figure 3G,H). These results indicate that C22 and C73 are weak modulators of FICD-mediated BiP deAMPylation in intact cells.

C22 Derivative C34 Promotes BiP AMPylation Inhibition Potency in Intact Cells

C22, and its sodium salt, C73 are both halogenated salicylanilide molecules. To identify C22 analogs with improved FICD inhibition activity, we tested several commercially available analogues using our well-established FP assay (Table 1). Interestingly, the benzamide moiety was intolerant to most structural changes that we evaluated. The removal of the 3- and 5-iodo functional groups (compound 2 (C51)) resulted in an approximately 40-fold reduction in potency (Table 1). The incorporation of a 5-chloro-2-hydroxy benzamide (compound 3 (C32)) modestly improved inhibition potency compared to C51 (Table 1, Supplementary Figure S9, and Figure 4A). Adding an unsubstituted benzamide (compound 4 (C50)) and 3-nitro-2-methyl benzamide (compound 5 (C52)) resulted in a complete loss of potency (Table 1 and Figure 4A). While only a limited number of benzamide analogues were tested, compounds 2–5 indicate that the 2-hydroxyl moiety is essential for compound potency, and halogenated substitutions at the 3- and 5-positions are required for compounds to be effective FICD inhibitors. Of all six tested analogs, compound 6 (C34) was of particular interest. The replacement of racemic nitrile with a carbonyl moiety while retaining the 3- and 5-iodo groups resulted in C34 exhibiting significantly improved potency compared to compounds 2–5 and a similar potency compared to our benchmark compounds, C22 and C73 (Table 1 and Figure 4A). In cell-based assays, we further observed that C34 promoted a concentration-dependent decrease in FICD-mediated BiP AMPylation (Figure 4B). We observe ∼75% decrease in BiP AMPylation levels at 1 μM and a near complete inhibition of BiP AMPylation at 5 and 10 μM (Figure 4C). This was superior to the FICD inhibition profile of both C22 (Figure 1D) and C73 (Figure 1F), which reduced BiP AMPylation levels by approximately 80% at the highest tested concentration (10 μM). These results indicate that functional group flexibility in the biaryl region of the scaffold could be further exploited to rationally improve these FICD inhibitors.

Table 1. Commercially Available Closantel Analogs^a.

^aCompound 1 is closantel, while compounds 2–6 are closantel derivatives. All compounds share the same core structure depicted above the table. IC₅₀ values (μM), as shown in the table, were obtained by fitting the concentration response data using a nonlinear regression method in GraphPad Prism. The asterisk (*) indicates extrapolated values. The best fit curve should be interpreted with caution as it showed a low R-squared value.

Figure 4.

Concentration response assessment of compound C34. (A) One μM FICD^E234G was incubated with varying concentrations of C34 for 10–15 min at RT, following which FL-ATP was added. The reaction was incubated for 90 min at 37 °C in the dark. Each dot represents the mean of duplicate mP measurements with the arrows representing the standard error of means (SEM). (B) AMPylated and total BiP levels of A549 cells treated with PBS for approximately 60 min in the presence of either 0.5% (v/v) DMSO (control) or C34 (at indicated concentrations). Tubulin was used as the protein loading control. (C) Quantification of (B). A two-way ANOVA was performed to assess statistical significance between control and treated samples. Western blots of the two remaining biological replicas are shown in Supplementary Figure S8.

FICD Inhibitors C22 and C73 Inhibit Human-Pathogenic FICD Variants In Vitro

Recent reports link two mutations in the FICD active site to infancy-onset diabetes and motor neuron degeneration.^21,22 In both cases, the pathogenic FICD variants, FICD^R371S and FICD^R374H, excessively AMPylate BiP. We thus tested whether FICD inhibitors C22 and C73 suppress FICD^R371S and FICD^R374Hin vitro. Using our FP assay, we observed that both FICD^R371S (Figure 5A and Supplementary Figure S10A,B) and FICD^R374H (Figure 5B and Supplementary Figure S10A,B) showed auto-AMPylation activity, yet substantially less than FICD^E234G. Incubating both pathologic FICD mutants with increasing concentrations of FICD inhibitors did not significantly decrease auto-AMPylation levels except at the highest tested concentration (100 μM) where C22 and C73 elicited auto-AMPylation reduction by approximately 30 and 20%, respectively (Figure 5C). Despite limited inhibition of auto-AMPylation, both C22 and C73 significantly reduced BiP AMPylation in vitro at a molar ratio of 20:1 (inhibitor:protein) (Figure 5D,E). These results indicate that Arg371Ser and Arg374His mutations have a distinct impact on FICD auto- and BiP AMPylation and define C22 and C73 as potent inhibitors of pathologic FICD variants.

Figure 5.

Effect of C22 and C73 on pathologic protein AMPylation in vitro. Auto-AMPylation activity of (A) FICD^R371S and (B) FICD^R374H was assessed by incubating FICD variants at indicated concentrations with 250 nM FL-ATP. FP was measured at indicated time points. Wells containing WT FICD and FICD^E234G served as negative and positive controls, respectively. Triplicate measurements were averaged for each FICD concentration and plotted as a function of time. (C) One μM FICD^E234G and 5 μM FICD^R371S and FICD^R374H were incubated with varying concentrations of C22 and C73 for 1 h and FP measured thereafter. FICD mutants incubated with DMSO served as negative controls. Triplicate measurements were averaged for each compound concentration and plotted for each FICD mutant. (D) AMPylated and total levels of recombinant human BiP in the presence of FICD inhibitors or DMSO. WT FICD (±) BiP served as negative controls. The red asterisk (*) indicates the protein band used for quantification purposes, while the purple asterisk (*) indicates nonspecific AMPylation of a protein contaminant, which copurified with recombinant FICD. (E) Quantification of (D). A two-way ANOVA was performed to assess statistical significance between control and treated samples. Western blots of two remaining biological replicas are shown in Supplementary Figure S9.

C22 Promotes Proinsulin Folding and Secretion in Pancreatic β-Cells

Recent work suggests a novel link between infancy-onset diabetes and excessive FICD-mediated BiP AMPylation.²¹ Since BiP’s chaperone function is required for proinsulin folding⁵⁸ and FICD-mediated BiP AMPylation inhibits its chaperone activity,^8,9,13,51 we investigated whether reducing BiP AMPylation using FICD inhibitors improves proinsulin folding and secretion in Min6 pancreatic β-cells. To this end, we treated Min6 cells with 20 μM FICD inhibitors (the inhibitor concentration was optimized as shown in Supplementary Figure S10C,D) or DMSO for 16 h, and both conditioned media (M) and cell lysates (C) were probed for proinsulin levels (Figure 6A). The proinsulin protein content in the complete system (M+C) was unaffected by the 16 h treatment with both C22 (Figure 6B) and C73 (Supplementary Figure S10E). Interestingly, we observed ∼8-fold increase in proinsulin levels in M as compared to C when cells were treated with C22 (Figure 6C) and a roughly 2-fold increase when cells were exposed to C73 (Supplementary Figure S10E–G). To elucidate the reason for such an observation, we investigated whether C22 affected proinsulin protein folding by measuring the abundance of aberrant disulfide-linked proinsulin complexes. Two of the most readily quantifiable misfolded forms of proinsulin are its disulfide-linked dimer and trimer forms, and the ratio of these aberrant forms to monomeric proinsulin was markedly improved by C22 treatment (Figure 6D,E). We hypothesized that C22-mediated reduction in BiP AMPylation may contribute to improved proinsulin folding by increasing the pool of active BiP. Indeed, we observed an approximately 2.5-fold decrease in AMPylated BiP levels in cells treated with C22 as compared to cells treated with DMSO (Figure 6F,G). Cells treated with C73 showed only a moderate decrease in BiP AMPylation (Supplementary Figure S10H,I). We further found that treatment of Min6 cells with C22 significantly increased BiP binding to proinsulin (Figure 6H–J), implying that a reduced FICD activity might enable increased BiP client binding. These results indicate that the FICD inhibitor C22 promotes proinsulin folding and anterograde trafficking out of the ER by increasing the pool of chaperone-competent BiP, which reduces misfolded or aggregated proinsulin levels.

Figure 6.

Effect of C22 and C73 on proinsulin secretion and folding in Min6 pancreatic β-cells. (A) Min6 cells were treated with 20 μM C22 or DMSO for 16 h and the media (M) and cell lysate (C) probed for proinsulin levels. β-actin was used as the protein loading control. (B) Total proinsulin levels in M+C as quantified from (A). (C) Proinsulin levels in M compared to C (M/C ratio) as quantified from (A). (D) Nonreducing SDS-PAGE showing the proinsulin monomer and higher-order oligomers. Numbers on the left represent the protein ladder molecular weights. HSP90 was used as the protein loading control. (E) Quantification of (D). (F) Min6 cells treated with C22 or DMSO were lysed and probed for AMPylated BiP levels. β-actin was used as the protein loading control. The asterisk (*) indicates the protein band used for quantification purposes; the upper protein band is nonspecific for BiP AMPylation. (G) Quantification of (F). (H) HEK 293 cells were transfected with plasmids encoding for myc-Ins2 (C96Y) or clover (control), incubated for 24 h, and treated with 20 μM C22 or DMSO for 16 h. Cells were collected and myc-Ins2 (C96Y) were retrieved from total cell lysates by immunoprecipitation (IP). The figure discloses both total input (pre-IP) and IP fractions. (I,J) Quantifications of (H). Statistical significance between control and treated groups was assessed by performing an unpaired t test with Welch’s correction. Data are presented as means ± SD.

Discussion

In this study, we define a pair of small molecules, C22 and C73, as novel FICD inhibitors with limited cytotoxicity. We show that these molecules inhibit BiP AMPylation catalyzed by both wild-type and pathogenic FICD variants, and highlight their potential to improve proinsulin folding and secretion in pancreatic β cells. We further demonstrate that the compounds are amenable to rational medicinal chemistry-based improvements and present a proposed mode of target engagement supported by in silico and cell-based work.

Recent advances in our understanding of FICD biology provide a compelling premise for the development of FICD inhibitors for therapeutic considerations. FICD-mediated cycles of BiP AMPylation and deAMPylation regulate ER homeostasis. Pathogenic mutations in the FICD active site disrupt this equilibrium, leading to a loss of deAMPylation function, which results in the accumulation of AMPylated BiP.^21,22 Additional work indicates that the loss of endogenous FICD activity mitigates pressure overload-induced cardiac hypertrophy by inducing a robust UPR^ER response and enhancing ER-phagy in cardiomyocytes.²⁵ In this study, we show direct evidence for enhanced proinsulin processing in response to FICD inhibition. In all these scenarios, FICD inhibitors are expected to provide immediate benefits by reducing BiP AMPylation to restore and/or boost UPR^ER signaling and ER homeostasis. A small number of putative in vitro FICD inhibitors have previously been described.^38,39,59 Unlike compound C22 and its sodium salt C73, however, these molecules were not tested against endogenous or pathogenic FICD variants. C22 and C73 are thus the first tool compounds to efficiently inhibit FICD in tissue culture models. C22, a halogenated salicylanilide, further shows functional group flexibility in the biaryl region of the scaffold, which amends itself for rational improvement.

Closantel sodium is a well-known antiparasitic drug primarily used to treat “liver flukes” (caused by a helminthFasciola hepatica) in animals and “river blindness disease” (caused by a filariumOnchocerca volvulus) in humans.^60,61 It kills Fasciola hepatica by selectively uncoupling its ATP-generating mitochondrial oxidative phosphorylation pathway.⁶² Moreover, closantel sodium significantly inhibits biofilm formation by the pathogenic methicillin-resistant Staphylococcus aureus.⁶³ Additionally, it inhibits dimerization of human taspase 1, an enzyme overexpressed in several tumors and a potential target for cancer drug development.⁶⁴ However, accidental consumption of closantel at high doses causes acute reversible blindness in pets and humans. This is attributed to retinal damage and photoreceptor depletion.⁶⁵⁻⁶⁷ Moving forward, medicinal chemistry efforts will be required to reduce retinal toxicity and mitigate off-target engagement of closantel derivatives.

Using a combination of in silico MD simulations and cell-based assays, we provide the first evidence that FICD inhibitors C22 and 73 may bind dimeric FICD, preferentially at the smaller dimeric interface or the TPR-II domain. If confirmed experimentally, then this interaction could prevent BiP AMPylation by either blocking the switch to an AMPylation-competent, monomeric state or abrogating BiP binding to FICD. We speculate that the compounds may also bind monomeric FICD, which would lead to inhibition of BiP AMPylation. The finding that C22 and C73 moderately inhibit FICD-mediated BiP deAMPylation further suggests that the compounds may abrogate the transition from an AMPylation-competent monomer to a deAMPylation-competent dimeric state. Determining the structure of C22-bound FICD will be a critical next step to confirm this proposed mode of action.

Using pancreatic cells, we show that FICD inhibitors C22 and C73 enhance anterograde trafficking and proinsulin folding while reducing aggregated or misfolded proinsulin. These results are significant, as to the best of our knowledge, no other small molecule with similar capacity to improve proinsulin processing is described. However, we have no direct evidence that the observed improvement in proinsulin processing is strictly mediated through FICD inhibition. Moving forward, it will thus be critical to test more specific and potent FICD inhibitors, including C34, in similar assays and to exclude the off-target activity of C22, C73, and related molecules. Future proof-of-concept studies using in vivo models for proinsulin misfolding will provide the first evidence for the therapeutic potential of using FICD inhibitors to mitigate autosomal dominant diabetes.

Limitations of the Study

This study takes the first steps toward understanding the structure–activity relationship (SAR) between FICD and compounds C22 and C73. The described SAR-based analog testing represents a glance at the potential of SAR-based medical chemistry to further improve compound efficacy. Additional efforts will likely lead to the development of a more potent lead compound. We also failed to source compound C34 in reasonable quantities preventing the testing of this most promising molecule in more assays. Testing the compounds in more cell lines, stress conditions, and ultimately in vivo are critical next steps toward defining their potential for future clinical use.

Significance

This study identifies two cell-permeable FICD inhibitors, C22 and C73, which inhibit FICD-mediated BiP AMPylation while exhibiting low cytotoxicity. Both compounds inhibit wild-type and pathologic FICD variants. This is significant, considering that the number of identified pathologic FICD variants is increasing but small molecules to target these mutant enzymes are lacking. Our in silico docking work implies two possible modes of action for C22 and C73, in line with the proposed model that FICD oligomerization is critical to controlling its AMPylation/deAMPylation activity. Our study also demonstrates that targeting FICD improves proinsulin folding and secretion. These results establish a first link between FICD activity and proinsulin processing in the ER and highlight a promising new application for FICD inhibitors to improve proinsulin processing in β cells. Taken as a whole, our study confirms FICD as a druggable enzyme and provides critical support for considering FICD as a target for multiple clinical indications.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschembio.4c00847.

• Supplementary tables disclosing qPCR primers (Table S1), IC₅₀ values of tested compounds (Table S2), LD₅₀ values of C22 and C73 in distinct cell types (Table S3), and in silico docking scores (Table S4); supplementary Figures S1–S10 disclosing additional experimental details, assay schematics, and materials, including images of additional Western blot replicates, the quantification of which are shown in the main figures (PDF)

Author Contributions

MCT supervised the project. BKC, AC, MA, JR, CLB and PDA designed and planned the experiments. BKC, AC, MMA, and SML conducted the experiments. BKC and MCT wrote the manuscript, and all authors edited and approved the final manuscript.

The authors declare no competing financial interest.