Discovery and validation of a novel inhibitor of HYPE-mediated AMPylation

Abstract

Adenosyl monophosphate (AMP)ylation (the covalent transfer of an AMP from Adenosine Triphosphate (ATP) onto a target protein) is catalyzed by the human enzyme Huntingtin Yeast Interacting Partner E (HYPE)/FicD to regulate its substrate, the heat shock chaperone binding immunoglobulin protein (BiP). HYPE-mediated AMPylation of BiP is critical for maintaining proteostasis in the endoplasmic reticulum and mounting a unfolded protein response in times of proteostatic imbalance. Thus, manipulating HYPE’s enzymatic activity is a key therapeutic strategy toward the treatment of various protein misfolding diseases, including neuropathy and early-onset diabetes associated with two recently identified clinical mutations of HYPE. Herein, we present an optimized, fluorescence polarization-based, high-throughput screening (HTS) assay to discover activators and inhibitors of HYPE-mediated AMPylation. After challenging our HTS assay with over 30,000 compounds, we discovered a novel AMPylase inhibitor, I2.10. We also determined a low micromolar IC50 for I2.10 and employed biorthogonal counter-screens to validate its efficacy against HYPE’s AMPylation of BiP. Further, we report low cytotoxicity of I2.10 on human cell lines. We thus established an optimized, high-quality HTS assay amenable to tracking HYPE’s enzymatic activity at scale, and provided the first novel small-molecule inhibitor capable of perturbing HYPE-directed AMPylation of BiP in vitro. Our HTS assay and I2.10 compound serve as a platform for further development of HYPE-specific small-molecule therapeutics.

Introduction

Filamentation induced by cyclic-adenosyl monophosphate (AMP) (Fic) proteins are an evolutionarily conserved family of enzymes. To modulate cell signaling, most Fic proteins studied to date catalyze a post-translational modification called AMPylation (adenylylation), which entails the covalent attachment of an AMP to target protein substrates at serine, threonine, or tyrosine.1, 2 AMPylation is enabled by the Fic motif (HXFX(D/E)(G/A)N(G/K)RXXR) within the structurally conserved active site of the Fic domain. The leading Fic motif histidine initiates AMPylation by deprotonating the target hydroxyl group, followed by a nucleophilic attack of the activated oxygen on the alpha phosphate of Adenosine Triphosphate (ATP). The reaction is completed by electron shuttling to the pyrophosphate-leaving group.

The human genome encodes for a single Fic protein called Huntingtin Yeast Interacting Partner E (HYPE) or FICD. HYPE resides in the endoplasmic reticulum (ER), where it AMPylates the heat shock chaperone binding immunoglobulin protein (BiP).3, 4 AMPylation of BiP renders it an inactive chaperone during homeostasis. However, in response to the accumulation of misfolded proteins in the ER, BiP is deAMPylated to become an active chaperone.3, 5 In addition to BiP, we recently identified alpha-Synuclein (aSyn) as a target of HYPE AMPylation.⁶ aSyn is a presynaptic protein, misfolded aggregates of which are associated with neuropathies like Parkinson’s disease. We showed that AMPylated aSyn saw a reduction of several cytotoxic phenotypes associated with Parkinson’s disease, such as fibrillation and membrane permeability. Interestingly, manipulation of BiP AMPylation was also shown to indirectly influence the aggregative state of aSyn and other disease-linked proteins (i.e. β-amyloid in Alzheimer’s disease and PolyQ in Huntingtin’s disease).⁷

Following the discoveries of BiP and aSyn, a wave of putative substrates for HYPE AMPylation is being revealed.⁸ This HYPE AMPylome encompasses proteins involved in disease pathways ranging from cancer to diabetes to neurodegeneration. Importantly, two clinical mutations of HYPE were recently reported, both of which result in a rare disease phenotype involving motor neuron defects and early-onset diabetes.9, 10 Given this critical central role for HYPE in cellular homeostasis, we wanted to target it for broad therapeutic intervention. To this end, we previously devised a pilot fluorescence polarization (FP)-based, dual high throughput screen (HTS) to identify activators and inhibitors of HYPE-mediated AMPylation.¹¹ After screening nearly 10,000 compounds, we found that calmidazolium and aurintricarboxylic acid can reliably enhance and inhibit AMPylation in vitro, respectively. Despite these bioactivities, the ultimate therapeutic utility of these compounds was limited by their multiple nonspecific cellular targets. Calmidazolium, for example, interacts with no less than five other proteins, including its namesake, the ER calcium-binding resident calmodulin.11, 12, 13, 14, 15 Likewise, aurintricarboxylic acid interferes with ribonuclease and topoisomerase II, among others.11, 16, 17, 18 The promiscuous nature of these hits is not surprising, given their origination from a pilot screen library (LOPAC¹²⁸⁰) containing pharmacologically active compounds.¹⁹ Nevertheless, the ability of our high throughput assay to deliver hit compounds encouraged us to proceed with screening a larger compound library for novel, HYPE-specific molecules. We therefore expanded our pilot screen to over 30,000 compounds for activators and inhibitors and included a central nervous system (CNS)-compatible library to account for HYPE’s role in neurodegeneration.

Results and discussion

Assay design and optimization

Though endogenously expressed HYPE exhibits dual catalytic activity, in vitro, wild type (WT) HYPE is largely restricted to deAMPylation.⁵ These basal levels of AMPylase activity make WT HYPE an ideal candidate for discovering activators.4, 11 Conversely, HYPE is rendered a hyperactive AMPylase upon mutation of the regulatory glutamate to glycine (E234G HYPE).²⁰ Mechanistically, E234 regulates AMPylation by forming a salt bridge with R374, which in turn, precludes an AMPylation-competent orientation of ATP in the Fic domain active site. These catalytic differences were exploited in our HTS assay design by using hypoactive AMPylase WT HYPE to screen for potential activators and hyperactive E234G as the basis for discovering inhibitors. WT and E234G HYPE’s relative AMPylation efficiencies remain similar for substrate and autoAMPylation. We, therefore, were able to conduct a streamlined, efficient screen with HYPE as its own substrate.

As a nucleotide source, we turned to a fluorescently polarizable (FP) ATP analog, Fl-ATP (Figure 1(a)).11, 12, 21 After light is polarized by a filter, the rapid rotation of free Fl-ATP in solution will depolarize this light, leading to a low FP signal. However, if Fl-AMP is attached to a large protein (i.e., via AMPylation), this rotation is considerably slower and remains polarized with a high FP signal. Of note, the covalent attachment of Fl-AMP to a protein substrate is not the only way to slow down its rotation. Some compounds are prone to assembling in large aggregates which can trap Fl-ATP to produce a false positive AMPylation signal. To control for this potentially interfering occurrence, we added Triton X-100 (TX-100). This nonionic detergent is often used in small molecule screens to prevent aggregates from forming. However, an excess amount of detergent could have a deleterious effect on protein stability and catalysis. We, therefore, assayed a range of TX-100 concentrations to assess their impact on AMPylation, starting with 0.1%, equal to that of our pilot screen validation buffer data not shown. While 0.1% TX-100 proved suitable for our validation reactions, we wanted to ensure it would not interfere with AMPylation on our FP microplate screening format. We see that all measured concentrations yield acceptably high levels of E234G HYPE AMPylation and low levels of WT HYPE AMPylation. In fact, we find decreasing levels of detergent to enact a slight inhibitory effect on E234G HYPE AMPylation while the WT signal is slightly boosted at the lowest concentration. These marginal catalytic differences prompted us to use a 0.1% TX-100 concentration, which would minimize the chance of compound aggregates.

Fig. 1.

Fluorescence polarization (FP) of Fl-ATP monitors HYPE-mediated AMPylation. (a) FP AMPylation assay design. When unattached to WT HYPE or compound-inhibited E234G HYPE (purple), Fl-ATP undergoes rapid rotation to depolarize plane-polarized light, giving basal FP signal. Attachment of Fl-AMP via autoAMPylation (compound-activated WT HYPE or E234G HYPE) permits light to remain polarized for a high FP signal. (b) Structure of Fl-ATP (N6-(6-amino)hexyl-ATP-5-carboxyl-fluorescein). (c) Assay reproducibility assessment. (a) WT and E234G HYPE autoAMPylation reactions were measured for FP on separate 384-well microplates. Z′ = 0.67 and S/B = 3.6. Each dot represents a single AMPylation reaction in a separate well. Black dash lines represent three standard deviations (SD) plus or minus the control means. Abbreviations used: AMP, adenosyl monophosphate; HYPE, Huntingtin Yeast Interacting Partner E.

HYPE-directed AMPylation is dependent on a divalent cation as a cofactor. Via an in-gel, in vitro AMPylation using α-³²P-ATP we previously established a hierarchy of catalytic efficiencies among divalent cations with manganese (Mn²⁺) at the top, followed by magnesium (Mg²⁺).⁴ Despite the apparent prevalence of Mg²⁺, differences can arise between the assay formats (in-gel versus FP microplate) and nucleotide source (radiolabeled versus fluorescently labeled). Moreover, recent structures of HYPE and in complex with its substrate BiP elucidate an essential role for Mg²⁺ in coordinating the alpha and beta phosphates of ATP.22, 23 In the direct assessment of cofactor preference, we conducted side-by-side WT and E234G HYPE FP microplate AMPylation reactions differing only by the presence of Mn²⁺ versus Mg²⁺. Interestingly, we observe a slight signal increase in the Mg²⁺-containing buffer over the one with Mn²⁺ for E234G HYPE AMPylation, and the opposite for WT HYPE (i.e., higher Mn²⁺ signal than Mg²⁺; data not shown). During buffer optimization for our pilot screen, we noticed a similar trend for WT HYPE AMPylation, although there was no significant difference in E234G HYPE AMPylation between the cations.¹¹ These differences could result from the presence of TX-100, which was absent in our pilot HTS screening buffer.

Quality assessment and library selection

After establishing optimal buffer conditions, we wanted to assess the quality and scalability of this assay for a full, automated HTS campaign. To this end, we turned to Z′ and signal/background (S/B) analysis (Figure 1(c)).²⁴ We plated 384 WT and 384 E234G HYPE AMPylation reactions on separate microplates with our HTS buffer containing 0.1% TX-100 and 1 mM Mg²⁺ as a cofactor. Further, we maintained all other assay parameters as in our original pilot HTS, including enzyme concentration (400 nM), Fl-ATP concentration (25 nM), and reaction time (10 min). The low Fl-ATP concentration was specifically selected to not bias our screen away from compounds having an ATP-competitive mode of action; whereas the 10-min time point ensured our reactions were within HYPE’s linear range of AMPylation. Under these assay conditions, we observed excellent Z′ (i.e., 1 > Z′ ≥ 0.5) and S/B values: 0.67 and 3.6, respectively. These were comparable to, yet lower than, our pilot screen statistics (Z′ = 0.88 and S/B = 4.6). These discrepancies could stem from our novel buffer conditions (i.e., TX-100 and Mg²⁺). Alternatively, the addition of dimethyl sulfoxide (DMSO) may have had a dampening effect on the signal (particularly the E234G HYPE signal) as WT is already basal. Despite being the standard solvent for dissolving polar and nonpolar compounds found in chemical libraries, DMSO is not without its downsides. We and others have reported DMSO to impede the enzymatic activities of multiple proteins.11, 25

With a robust screening assay in hand, we next sought out an appropriate collection of compounds. In our pilot screen, we used a diverse set of compounds from four common HTS libraries, including Microsource Spectrum and LOPAC¹²⁸⁰. While these libraries may be suitable for a pilot screen owing to the high probability of potential hits, their utility for discovering a HYPE-specific modulator is limited. This is because these bioactive compounds act against a broad range of cellular targets.²⁶ We, thus, selected from two libraries containing novel compounds that are less likely to have off-target effects (Table 1). The first of these libraries, DIVERset from Chembridge, was optimized for compound effectiveness by three dimensional conformational analysis, and stringent application of drug-like and functional group filters. The second library, CNS-Set also by Chembridge, uses a series of physicochemical filters (e.g., Lipinski Rules, Polar Surface Area, etc.) to enhance the probability of blood–brain barrier penetration and bioavailability. Taken together, the more than 30,000 compounds from these 2 libraries ensure that potential hits will be amenable to downstream drug discovery efforts. The CNS-Set library was selected in anticipation of HYPE-centric treatments for neurodegeneration and neurogenesis.6, 7, 8

Table 1.

Compound library information.

Library name	Company	Number of compounds	Chemical properties
DIVERset	ChemBridge	14,080	Diverse
CNS-Set	ChemBridge	16,000	BBB penetrant
Total	---	30,080	---

Abbreviations used: BBB, blood–brain barrier; CNS, central nervous system.

WT HYPE activator high-throughput screen

In the first half of our dual HTS, we challenged over 30,000 of these compounds against WT HYPE toward the discovery of novel AMPylation enhancers (Figure 2(a)). Here, we define our hit threshold at 20% activation normalized to internal controls (WT HYPE with DMSO as a low control and E234G HYPE as a high control), which yields an overall hit rate of 0.12% activators (i.e., 36 hits out of 30,080 compounds). Among these 36 hits, 15 stemmed from the DIVERset library, while the other 21 were from the CNS-Set. To control for autofluorescing false positive compounds, or those intrinsically capable of quenching fluorescence emanating from Fl-ATP(/AMP), we manually filtered compounds containing outlier parallel intensities.

Fig. 2.

WT HYPE activator HTS. (a) WT HYPE FP autoAMPylation reaction incubated with compounds from either DIVERset or CNS-Set libraries. All samples were normalized to WT and E234G HYPE controls on the same plate in the DMSO-containing buffer. Each dot represents a different compound incubated with a single AMPylation reaction in a separate well. Black dashes represent ±3 SD from the mean. Red dashes represent the 20% hit definition, with all dots above this threshold being considered hit compounds. Minor tick marks on the x-axis represent a single 384 well plate. (b) Plate-to-plate variability of the high (E234G HYPE) and low (WT HYPE) DMSO internal controls. Left and right y-axes are for Z′ and S/B values, respectively. Each dot represents the calculated Z′ or S/B values from 32 E234G HYPE high controls and 32 WT HYPE low controls within a single plate. (c) Coefficient of variance assessment. Left and right y-axes are for WT and E234G HYPE coefficient of variation (CV), respectively. Each dot represents the calculated CV value from 32 E234G HYPE high controls and 32 WT HYPE low controls within a single plate. Abbreviations used: AMP, adenosyl monophosphate; CNS, central nervous system; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E; SD, standard deviation.

Excellent Z′ values and high S/B values were observed throughout our activator HTS, denoting a quality assay, with minimal plate-to-plate variability (Figure 2(b)). Our coefficient of variance analysis also returned values consistently below the 10% limit, which defined an excellent assay. Notably, while all E234G HYPE high controls gave sub-10% CVs (most being sub-5%), the WT HYPE low controls displayed considerably more variability, with some falling outside of the sub-10% range. This difference likely results from the fact that at low, basal levels of WT HYPE AMPylation signal, slight variance in FP values get magnified relative to those of the much higher E234G signal. Nevertheless, we took care not to advance hits from plates having outlier quality statistics.

WT HYPE activator microplate validation

The hits from our activator HTS were cherry-picked for validation in an independent replication of the same HTS assay, likewise on microplate format. Upon second-pass selection, these hits yielded a modest R² value of 0.44 with respect to the initial HTS (Figure 3(a)). Interestingly, grouping hits based on their libraries of origin revealed an even higher correlation among CNS-Set compounds (Figure 3(b)) and radically lower correlation among DIVERset molecules. This could be explained by the higher initial bioactivities of CNS-Set compounds, thus reducing the likelihood of false positives.

Fig. 3.

WT HYPE activator validation. (a) Comparison between initial HTS compound activity (x-axis) and second pass, cherry-pick (CP) validation (y-axis) with the same FP AMPylation assay. All validation compounds were assayed on the same plate and normalized to internal high (E234G HYPE) and low (WT HYPE) DMSO controls. Each dot represents the same compound incubated in two independent AMPylation reactions. (b) Data reprocessed from (a) to show compound validation correlation differences in the DIVERset library. (c) Data reprocessed from (a) to show compound validation correlation differences in the CNS-Set library. Abbreviations used: AMP, adenosyl monophosphate; CNS, central nervous system; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E.

We next advanced several hits for follow-up validation using, chemical, diversity, biological activity, and commercial availability as selection criteria (Figure 4(a)–(d), Table 2). Of note, two of our hits (A2.6 and A2.7) had strikingly similar structures, which differed only in the presence of an oxy-methyl group in A2.6. This suggests that these compounds may target HYPE at the same residues and with a similar mode of action. To control specifically for false positive compounds with intrinsically high fluorescence properties, we performed microplate FP AMPylation reactions with or without enzyme. We hypothesized that true positive hits would have a low FP signal in the absence of HYPE and a higher signal with HYPE; conversely false positives would maintain a high “AMPylation” signal irrespective of an AMPylase. Indeed, we observed several false positive activators among our initially selected hits (data not shown). After reselecting more hits from our reservoir of cherry-picked compounds, we were able to find four putative activators lacking intrinsic FP (data not shown).

Fig. 4.

Chemical structures of selected WT HYPE activators. (a) Compound A2.5. (b) Compound A2.6. (c) Compound A2.7. (d) Compound A2.8. Abbreviation used: HYPE, Huntingtin Yeast Interacting Partner E.

Table 2.

WT HYPE activator hit information.

Compound ID	Chemical formula	MW (Da.)	Library ID	Library	% Act. HTS	% Act. Cherry-pick	Mean % Act.
A2.5	${\mathrm{C}}_{22}{\mathrm{H}}_{20}{\mathrm{N}}_4{\mathrm{O}}_6\mathrm{S}$	468	S342-0409	CNS set	30	10	20
A2.6	${\mathrm{C}}_{23}{\mathrm{H}}_{22}{\mathrm{N}}_4{\mathrm{O}}_7\mathrm{S}$	499	S342-0417	CNS set	21	13	17
A2.7	${\mathrm{C}}_{22}{\mathrm{H}}_{18}{\mathrm{N}}_4{\mathrm{O}}_7\mathrm{S}$	482	S342-0376	CNS set	25	14	19.5
A2.8	${\mathrm{C}}_{17}{\mathrm{H}}_{22}{\mathrm{N}}_2{\mathrm{O}}_3$	302	2895-0212	CNS set	37	52	44.5

Abbreviations used: CNS, central nervous system; HYPE, Huntingtin Yeast Interacting Partner E.

Concentration–response curves were conducted for these hits using increasing concentrations of our top putative activators under standard FP microplate reaction conditions (Figure 5 (a)–(e)). Three of our top four hits (A2.5–A2.7) display dose dependencies with EC₅₀ values in the mid-micromolar range. A2.8, however, has a poor response even at our highest compound concentration of 500 μM, which precludes EC₅₀ quantification. This lack of strong concentration-dependent response suggests that A2.8 may be a false positive activator.

Fig. 5.

WT HYPE activator concentration-response curve. (a) FP plots from WT HYPE AMPylation reactions incubated with increasing concentrations of DMSO-dissolved activators from 0 to 1000 μM. All data were fitted to Eq. 5 (see Materials and methods) to determine EC₅₀ values. (b)–(e) Reprocessed graphs of each individual activator (A2.5–A2.8) fitted to Eq. 5. Abbreviations used: AMP, adenosyl monophosphate; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. EC, Effective concentration.

WT HYPE activator in-gel counterscreen

To further validate these activator hits, we employed orthogonal biochemical assays capable of directly tracking AMPylation in vitro. Besides monitoring FP, the Fl-ATP fluorophore is suitable for in-gel fluorescence AMPylation assays.11, 12 Apart from the FP-based microplate assay (which indirectly correlates changes in AMPylation status to changes in fluorophore rotation), in-gel assays monitor the direct posttranslational modification of protein substrates separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). After challenging our in-gel fluorescence AMPylation assay with hits from the FP microplate assays, we fail to see the activation of AMPylation (Figure 6(a) and (b)). Our top pilot screen hit, A1 (calmidazolium), did, however, prove bioactive above our DMSO control.

Fig. 6.

WT HYPE activator validation and IbpA-2 cross reactivity using in-gel AMPylation assays. (a) Representative fluorescence in-gel autoAMPylation assay with WT HYPE or E234G HYPE (last lane). Fluorescence image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (b) Quantification of (a) normalized to internal DMSO controls. Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed to determine statistical significance. (c) Representative radioactive in-gel AMPylation assay with WT IbpA-Fic2, Cdc42 and α-³²P-ATP. WT HYPE-mediated AMPylation of T229A BiP-AMP with or without A1 was used as an activation control. Phosphor screen image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (d) Quantification of (c) normalized to internal DMSO controls (first lane). Quantified data represented as the mean ± SEM of three independent experiments. Unpaired t-tests were performed between DMSO (first lane) and all other IbpA-2 + Cdc42 samples incubated with compound to determine statistical significance. Abbreviations used: AMP, adenosyl monophosphate; BiP, binding immunoglobulin protein; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. SEM, Standard Error of the Mean. IbpA, Immunoglobulin binding protein.

Given the relatively large size of the fluorescein ring conjugated to Fl-ATP, we wanted to control for possible interference with the AMPylation reaction. We, therefore, used a radiolabeled α-³²P-ATP, which can monitor the covalent attachment of AMP without the bulky fluorescent handle.²⁷ However, even with this more physiologically relevant nucleotide, no activation of AMPylation is apparent (data not shown). We previously reported differences in AMPylation signal sensitivity between an FP microplate format and in-gel assays, with the latter being less capable of detecting weaker signals.¹¹ This is especially problematic for the hypoactive AMPylase WT HYPE. Unlike our hyperactive AMPylase E234G HYPE mutant, WT HYPE is a robust deAMPylase, which further compounds the issue of detecting stable AMPylation (or manipulation thereof).3, 5 Moreover, our stringent selection filtering of putative hits (e.g., intrinsic parallel intensities, etc.), left us with less potent activators to begin with.

As a potential solution to detecting a weak WT HYPE AMPylation signal, we turned to evolution. HYPE is a member of the highly structurally and functionally conserved Fic domain protein family. One of the best-studied Fic proteins, the bacterial effector Immunoglobulin binding protein (IbpA), is known to AMPylate host cell Rho GTPases during its pathogenesis with high efficiency.2, 28, 29 To validate our activators independent of WT HYPE, we challenged to top putative activators against IbpA AMPylation of Cdc42 using radiolabeled α-³²P-ATP (Figure 6(c) and (d)). Since the IbpA toxin houses two separate Fic domains, we simplified our experimental design by targeting Cdc42 with a construct solely containing the second of the two domains, IbpA-2. Consistent with our WT HYPE data, none of the putative activators increased IbpA-2′s AMPylation of Cdc42 beyond our DMSO controls. The lack of AMPylation enhancement by our top pilot screen WT HYPE activator, A1, is also consistent with previous results.¹¹

E234G HYPE high-throughput screen

Under the same reaction parameters as our WT HYPE activator HTS, we rescreened the 30,080 compounds from DIVERset and CNS-Set libraries (Table 1) for inhibition of E234G HYPE. Using the same 20% hit definition, we saw an overall hit rate of 0.38% (i.e., 114 hits out of 30,080 compounds), more than 3 times that of our activator screen. Given the ease with which catalysis can be disrupted versus enhanced, this relatively higher rate of inhibitor hits is as expected.³⁰ Among these 114 hits, 32 stemmed from the DIVERset library, while 82 were from the CNS-Set. To control for auto-fluorescing false positive compounds, or those intrinsically capable of quenching fluorescence emanating from Fl-ATP(/AMP), we manually filtered compounds containing outlier parallel intensities.

As in our activator screen, here we maintained excellent Z′ values and high S/B values throughout our inhibitor HTS, denoting a quality assay, with minimal plate-to-plate variability (Figure 7(b)). We did, however, have a single plate (Figure 7(b), plate #9) with an unacceptably low Z′ value. Since no 20% hits happened to be on this plate, this anomaly was inconsequential to our overall screening process. We also observed coefficient of variance values consistently below the 10% limit, defining an excellent assay, with only one exception for WT HYPE plate #9 (Figure 7(c)). As in our activator HTS, all E234G HYPE high controls gave sub-5% CVs, while the WT HYPE low controls yielded higher values with variability. The reason for this increased variance is likely the same (i.e., low, basal levels of WT HYPE AMPylation signal amplify minor variance in FP values relative to those of the much higher E234G signal).

Fig. 7.

E234G HYPE inhibitor HTS. (a) E234G HYPE FP autoAMPylation reaction incubated with compounds from either DIVERset or CNS-Set libraries. All samples were normalized to WT and E234G HYPE controls on the same plate in DMSO-containing buffer. Each dot represents a different compound incubated with a single AMPylation reaction in a separate well. Black dashes represent ±3 SD from the mean. Red dashes represent the 20% hit definition, with all dots above this threshold being considered hit compounds. Minor tick marks on the x-axis represent a single 384 well plate. (b) Plate-to-plate variability of the high (E234G HYPE) and low (WT HYPE) DMSO internal controls. Left and right y-axes are for Z′ and S/B values, respectively. Each dot represents the calculated Z′ or S/B values from 32 E234G HYPE high controls and 32 WT HYPE low controls within a single plate. (c) Coefficient of variance assessment. Left and right y-axes are for WT and E234G HYPE CV, respectively. Each dot represents the calculated CV value from 32 E234G HYPE high controls and 32 WT HYPE low controls within a single plate. Abbreviations used: AMP, adenosyl monophosphate; CNS, central nervous system; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E; SD, standard deviation.

E234G HYPE inhibitor microplate validation

Hits from our inhibitor screen were then cherry-picked in an independent replication of the same HTS microplate assay. Strikingly, little correlation was observed from the initial HTS to the cherry-pick validation (Figure 8(a)), although this correlation was slightly improved for the DIVERset library (Figure 8(b) and (c)). These all-around poor correlations could, in part, result from the low hit definition of 20%, as a more stringent threshold leads to increased reproducibility.¹¹

Fig. 8.

E234G HYPE inhibitor validation. (a) Comparison between initial HTS compound activity (x-axis) and second pass, cherry-pick (CP) validation (y-axis) with the same FP AMPylation assay. All validation compounds were assayed on the same plate and normalized to internal high (E234G HYPE) and low (WT HYPE) DMSO controls. Each dot represents the same compound incubated in two independent AMPylation reactions. (b) Data reprocessed from (a) to show compound validation correlation differences between in DIVERset library. (c) Data reprocessed from (a) to show compound validation correlation differences between in CNS-Set library. Abbreviations used: AMP, adenosyl monophosphate; CNS, central nervous system; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E.

Using chemical diversity, biological activity, and commercial availability as selection criteria, we forward processed our top hits for validation (Figure 9(a)–(l), Table 3). Inspectional analysis of these compounds’ chemical structures revealed several motifs, including a two-ringed, purine-like structure seen in I2.4–I2.7, and I2.12 (Figure 9(d)–(g), (l)). This feature is structurally reminiscent of ATP’s nitrogenous base. Moreover, these results are consistent with our pilot screen, in which we identified several ATP analogs as top inhibitor hits.¹¹ Global structural motifs can be found amongst other putative hits: I2.5 and I2.6 (Figure 9(e) and (f)), by example, differ only in the position and variety of alkyl halides on their terminal benzene ring. Taken together, these data hint at similar mechanisms of action among the hits.

Fig. 9.

Chemical structures of selected E234G HYPE inhibitors. (a) Compound I2.1. (b) Compound I2.2. (c) Compound I2.3. (d) Compound I2.4. (e) Compound I2.5. (f) Compound I2.6. (g) Compound I2.7. (h) Compound I2.8. (i) Compound I2.9. (j) Compound I2.10. (k) Compound I2.11. (l) Compound I2.12. Abbreviation used: HYPE, Huntingtin Yeast Interacting Partner E.

Table 3.

E234G HYPE inhibitor hit information.

Compound ID	Chemical formula	MW (Da.)	Library ID	Library	% Inh. HTS	% Inh. Cherry-pick	Mean % Inh.
I2.1	${\mathrm{C}}_{18}{\mathrm{H}}_{26}{\mathrm{N}}_4$	298.4	74731416	DIVERset	32.5	29.7	31.1
I2.2	${\mathrm{C}}_{16}{\mathrm{H}}_{16}{\mathrm{N}}_2{\mathrm{O}}_4$	300.3	88812993	DIVERset	20.3	22.7	21.5
I2.3	${\mathrm{C}}_{21}{\mathrm{H}}_{17}{\mathrm{N}}_5{\mathrm{O}}_2$	371.4	57920321	DIVERset	35.0	30.7	32.9
I2.4	${\mathrm{C}}_{13}{\mathrm{H}}_{14}{\mathrm{N}}_4\mathrm{O}$	242.3	8561-07647	CNS-Set	27.8	31.0	29.4
I2.5	${\mathrm{C}}_{20}{\mathrm{H}}_{15}{{\mathrm{F}}_2\mathrm{N}}_5\mathrm{O}$	379.4	J094-0169	CNS-Set	23.5	32.9	28.2
I2.6	${\mathrm{C}}_{20}{\mathrm{H}}_{15}{\mathrm{BrFN}}_5\mathrm{O}$	440.3	J094-0175	CNS-Set	32.7	29.7	31.2
I2.7	${\mathrm{C}}_{15}{\mathrm{H}}_{10}{\mathrm{BrN}}_3\mathrm{OS}$	360.2	5089-0088	CNS-Set	20.9	34.2	27.6
I2.8	${\mathrm{C}}_{18}{\mathrm{H}}_{20}{\mathrm{N}}_6{\mathrm{O}}_2$	352.4	80384857	CNS-Set	25.2	29.1	27.2
I2.9	${\mathrm{C}}_{21}{\mathrm{H}}_{18}{\mathrm{N}}_4{\mathrm{O}}_2$	358.4	S343-0959	CNS-Set	27.2	27.8	27.5
I2.10	${\mathrm{C}}_{14}{\mathrm{H}}_9{\mathrm{N}}_5{\mathrm{O}}_2$	279.3	P901-0269	CNS-Set	24.5	29.7	27.1
I2.11	${\mathrm{C}}_{15}{\mathrm{H}}_{18}{\mathrm{N}}_2\mathrm{O}$	242.3	D014-0004	CNS-Set	19.9	24.6	22.3
I2.12	${\mathrm{C}}_{16}{\mathrm{H}}_{11}{\mathrm{FN}}_4\mathrm{O}$	294.3	8561-07646	CNS-Set	27.1	33.6	30.4

Abbreviations used: CNS, central nervous system; HYPE, Huntingtin Yeast Interacting Partner E.

We next performed concentration–response curves for these hits using increasing concentrations of our top putative inhibitors under standard FP microplate reaction conditions (Figure 10(a)). Though most putative inhibitors failed to display robust dose-dependent inhibition of AMPylation, I2.10 showed high potency, with low micromolar IC₅₀ values (Figure 10(a)). Indeed, I2.10's bioactivity was on par with that of our top validated pilot screen hit I2 (aurintricarboxylic acid; Figure 10(c)). Unlike I2, however, I2.10 is a novel chemical with no known biological targets, which holds promise for potential drug candidacy.

Fig. 10.

E234G HYPE inhibitor concentration-response curve. (a) FP plots from E234G HYPE AMPylation reactions incubated with increasing concentrations of DMSO-dissolved inhibitors from 0 to 200 μM. All data were fitted to Eq. 6 (see Materials and methods) to determine IC₅₀ values. (b) Reprocessed graphs of the top inhibitor, I2.10. (c) Reprocessed graph of the top inhibitor from the pilot HTS, I2, for comparison. All data were fitted to Eq. 5 (see Materials and methods) to determine IC₅₀ values. Abbreviations used: AMP, adenosyl monophosphate; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E.

E234G HYPE inhibitor in-gel counterscreen

Putative inhibitors were then subjected to similar in-gel fluorescence and radiolabeled validations as our activators to directly probe for protein AMPylation. Starting with our fluorescent in-gel assay, we see that I2.10 consistently perturbs HYPE autoAMPylation in vitro (Figure 11(a) and (b)), using Fl-ATP as a nucleotide source. The bioactivity of I2.10 is mirrored under the more physiological conditions of AMPylation with an α-³²P-ATP substrate (which lacks the bulky fluorophore; Figure 11(c) and (d)).

Fig. 11.

E234G HYPE inhibitor validation using in-gel AMPylation assays. (a) Representative fluorescence in-gel autoAMPylation assay with E234G HYPE or WT HYPE (last lane). Fluorescence image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (b) Quantification of (a) normalized to internal DMSO controls. Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed: **P < 0.01. (c) Representative radioactive in-gel autoAMPylation assay with E234G HYPE or WT HYPE (last lane). Phosphor screen image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (d) Quantification of (c) normalized to internal DMSO controls. Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed: *P < 0.05; ***P < 0.005. Abbreviations used: AMP, adenosyl monophosphate; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. SEM, Standard Error of the Mean.

Despite an ever-expanding AMPylome, the ER heat shock chaperone BiP remains HYPE’s only fully validated protein substrate to date. We, therefore, assessed the ability of our novel top inhibitor, I2.10, to manipulate HYPE-promoted AMPylation of BiP. Intracellularly, BiP exists in two conformation states and fluctuates between these depending on its position in the chaperone cycle. During homeostasis, BiP adopts an open conformation that favors AMPylation and disfavors BiP binding to its unfolded protein substrates. Alternatively, triggering the unfolded protein response under stress conditions leads to deAMPylation of BiP, thus permitting it to bind to and fold its misfolded protein clients in a closed conformation. Given this chaperone cycle-dependent conformational preference, and the fact that WT BiP is thought to dynamically fluctuate between conformations, we opted to perform our in vitro AMPylation assays with a mutant BiP construct that’s constitutively locked in an open, AMPylation-competent state. Using this T229A BiP, and Fl-ATP as a cosubstrate, we were able to confirm the inhibitory properties of I2.10 (Figure 12 (a) and (b)). Further, I2.10's inhibitory efficacy against BiP AMPylation was reproducible using the more physiological α-³²P-ATP (Figure 12(c) and (d)). As with our autoAMPylation results, I2.10 inhibited BiP AMPylation comparable to the I2 positive control.

Fig. 12.

E234G HYPE inhibitor validation of BiP AMPylation with in-gel assays. (a) Representative fluorescence in-gel T229A BiP AMPylation assay with E234G HYPE or WT HYPE (last lane). Fluorescence image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (b) Quantification of (a) normalized to internal DMSO controls. Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed: *P < 0.05. (c) Representative radioactive in-gel T229A BiP AMPylation assay with E234G HYPE or WT HYPE (last lane). Phosphor screen image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (d) Quantification of (c) normalized to internal DMSO controls. Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed: *P < 0.05. Abbreviations used: AMP, adenosyl monophosphate; BiP, binding immunoglobulin protein; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. SEM, Standard Error of the Mean.

To assess the cross-reactivity of our putative inhibitors, we conducted in vitro AMPylation assays using IbpA-2 and Cdc42 with radiolabeled α-³²P-ATP. We observed no manipulation of AMPylation under these conditions (Figure 13). Consistent with our pilot screen, the I2 control also had no impact on AMPylation of Cdc42. This may be due to sequence and/or structural differences between the Fic domains of HYPE and IbpA-2. In fact, comparative structural analyses of the HYPE:BiP and IbpA:Cdc42 cocrystal structures reveal only partial superimposition among their respective Fic domain active sites.²³

Fig. 13.

Cross reactivity of inhibitors on IbpA-2 AMPylation of Cdc42. (a) Representative radioactive in-gel AMPylation assay with WT IbpA-Fic2, Cdc42 and α-³²P-ATP. E234G HYPE-mediated AMPylation of T229A BiP-AMP with or without I2 was used as an inhibition control. Phosphor screen image (upper) shows direct protein AMPylation. Coomassie (lower) displays protein loading. (b) Quantification of (a) normalized to internal DMSO controls (first lane). Quantified data represented as the mean ± SEM of three independent experiments. Unpaired t-tests were performed between DMSO (first lane) and all other IbpA-2 + Cdc42 samples incubated with compound to determine statistical significance. Abbreviations used: AMP, adenosyl monophosphate; BiP, binding immunoglobulin protein; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. SEM, Standard Error of the Mean.

E234G HYPE, though a hyperactive AMPylase, is deAMPylation deficient owing to the lack of formation of a supportive salt bridge between E234 and R374. Moreover, this regulatory glutamate is essential in deAMPylation, as it coordinates a catalytic water molecule within the Fic active site to nucleophilically attack the phosphodiester bond joining AMP to the target residue. WT HYPE, with its intact E234, is thus a robust deAMPylase. Given the HYPE’s role in deAMPylation-mediated activation of BiP’s chaperoning activity during the response to cellular stress (i.e., unfolded protein response induction), we wanted to determine our hit compounds’ ability to manipulate deAMPylation. We hypothesized that since AMPylation and deAMPylation occur at the same Fic catalytic pocket (albeit by slightly different mechanisms), compounds affecting HYPE’s AMPylase activity could likewise modulate the reversing post-translational modification. To test this, we conducted a standard in vitro two-step deAMPylation assay: (1) E234G HYPE and Fl-ATP are used to AMPylate T229A BiP to completion, followed by incubation with deAMPylation competent WT HYPE in the presence or absence of compound (Figure 14(a)). Under these conditions, we witnessed no enhancement or inhibition by our hit compounds of the current or pilot HTS. Intriguingly, we did see a lower deAMPylation signal in most of the inhibitors than in the activators or DMSO controls; however, none of these differences were statistically significant (Figure 14(b)).

Fig. 14.

Compound impact on HYPE-mediated deAMPylation. (a) Representative fluorescence in-gel deAMPylation assay with WT HYPE used a deAMPylase and T229A BiP-AMP used as a deAMPylation substrate. All T229A BiP was initially AMPylated by E234G HYPE and Fl-ATP to completion. Buffer/DMSO (first lane) and catalytically dead E234G/H363A (EG/HA) HYPE (last lane) served as negative controls for deAMPylation (last lane). Fluorescence image (upper) shows direct protein de/AMPylation. Coomassie (lower) displays protein loading. (b) Quantification of (a) normalized to internal Buffer/DMSO controls (first lane). Quantified data represented as the mean ± SEM of four independent experiments. Unpaired t-tests were performed between WT HYPE/DMSO (second lane) and all other samples incubated with compound to determine statistical significance. Abbreviations used: AMP, adenosyl monophosphate; BiP, binding immunoglobulin protein ; DMSO, dimethyl sulfoxide; HYPE, Huntingtin Yeast Interacting Partner E. SEM, Standard Error of the Mean.

Cellular toxicity validation

Before a hit compound can be advanced to drug candidacy, it must prove nontoxic to its target cells. We, therefore, determined the cellular toxicity of our top hit, I2.10, using a standard MTT cell viability assay with standard in vitro cell line models: Henrietta Lacks (HeLa) and Human embryonic kidney (HEK) 293. MTT is a yellow dye that gets reduced to purple formazan crystals in the presence of Nicotinamide adenine dinucleotide phosphate proton donors.³¹ Since abundant Nicotinamide adenine dinucleotide phosphate production is a hallmark of cellular respiration, we were able to correlate cell viability to the generation of purple crystals by measuring the absorbance at 570 nm.

Given I2.10's robust bioactivity at 10 µM, we wanted to assess whether this concentration would be suitable for cellular treatment. Our previous top activator, A1, was used as a control due to its reported cellular toxicity.

After incubating HeLa cells in varying amounts of the compound, we observed no toxicity (100% cell viability) from I2.10 (Figure 15(a)), even at concentrations five times higher than its in vitro IC₅₀ (Figure 10(a) and (b)). Notably, our maximum experimental concentration of 50 µM corresponds to the in vitro saturation point of I2.10-mediated AMPylation inhibition (i.e., zero percent AMPylation). This is a promising finding and suggests the potential for total chemical ablation of cellular AMPylation with no deleterious effects on the health of the target cell.

Fig. 15.

Assessment of cellular toxicity (a) MTT cell viability assay of HeLa cells incubated with increasing concentrations of compounds for 48 h on a microplate then treated with MTT reagents to produce colorimetric signal. Cell viability was calculated from colorimetric absorbance signal at 570 nm in accordance with Eq. 8 (see Materials and methods). A1 was added as a positive control for toxicity. All experimental performed in three biological replicates with three technical replicates each. Quantified data represented as the mean ± SEM of three independent replicates. (b) As in (a) but with HEK 293 cells. HeLa, Henrietta Lacks. HEK, Human embryonic kidney. MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazoliumbromide. SEM, Standard Error of the Mean.

As expected, our A1 control impacted cell viability with as little as 1 µM (Figure 15(a)). Due to the promiscuous nature of this compound, we cannot unequivocally deduce the mechanisms of cytotoxicity.13, 14, 15 However, it is plausible that A1's reported activation of AMPylation is at play. This result has been genetically phenocopied by cellular transfection with the hyperactive E234G HYPE, which is known to induce caspase-dependent apoptosis.

The relative toxicity trends seen in HeLa cells remained consistent in HEK 293. I2.10 treated HEK 293 cells retain ∼80% cell viability at our bioactive concentration of 10 µM (Figure 15(b)). However, we did observe somewhat of a concentration-dependent effect in I2.10's toxicity in HeLa, dropping to ∼60% cell viability at 50 µM. Though further investigation is needed to elucidate this cell specificity in I2.10-induced toxicity, it is possible that these arise from inherent biochemical differences between HeLa and HEK 293 cells, potentially providing novel targets for I2.10.

Docking studies

To better understand how I2.10 may be interacting with HYPE, a computational approach was taken via docking studies. The molecule was docked using the Glide Extra Precision (XP) method that allows for flexible ligand and receptor docking. The apo-HYPE structure (PDB: 4U04) was prepared using the Protein Preparation Wizard available through the Maestro interface of the Schrodinger platform, and I2.10 was prepared using LigPrep. Next, the receptor grid was generated at the ATP-binding site defined by Fic moiety residues His363, Arg371, Arg374, and Tyr400. Then I2.10 was docked into the grid.

The top pose for I2.10 was selected for analysis by considering docking score and protein–ligand interactions with residues in the ATP-binding site. Most noticeably, I2.10 fits within the ATP-binding site pocket and interacts with residues within this site (Figure 16). Notably, I2.10 is mostly planar and hydrophobic in nature, leading to the predicted pose forming extensive hydrophobic interactions with aromatic residues within the ATP-binding pocket (Figure 16). These hydrophobic interactions are flanked by two accepted hydrogen bonds, one between the imidazole accepting an H-bond from His318 and the other between the furan oxygen accepting an H-bond from the side chain of Asn369. Interestingly, in the ATP-bound structure (PDB: 4U07), the aromatic adenine moiety participates in π–π interactions with Tyr400. The docked pose for I2.10 overlaid with the observed ATP-bound conformation (Figure 16(d)) shows the pyridine heterocycle from I2.10 aligning well with the adenine moiety from ATP and is in proximity with Tyr400. Thus, it is feasible, based on this docking study, that I2.10 binds to the ATP site and gains two hydrogen bonds but also maintains a similar interaction with Tyr400. While these docking studies are promising for elucidating how I2.10 may inhibit HYPE, further work is required to fully characterize the binding mode and validate these docking results. Nonetheless, the docked pose is informational for the design of these studies.

Fig. 16.

The docked pose of I2.10 in the apo-HYPE structure was generated using Glide. Key residues in the binding site are shown as sticks, with polar hydrogens shown for clarity. Predicted protein–ligand hydrogen bonds are depicted as yellow dashes. (a) Docked pose for I2.10 (orange sticks) in apo-HYPE (gray cartoon). (b) Two-dimensional protein–ligand diagram for the docked pose of I2.10 in apo-HYPE with arrows to represent hydrogen bond directionality and green residues representing a hydrophobic microenvironment. (c) Docked pose for I2.10 (orange sticks) in apo-HYPE (gray surface). (d) Docked pose for I2.10 (orange sticks) in HYPE (gray surface) aligned with holo-ATP-bound (cyan sticks) HYPE (PDB: 4U07, pale green cartoon). Panels (a), (b), and (d) were generated using PyMol. The ligand-interaction diagram in panel B was generated in Maestro. Abbreviations used: HYPE, Huntingtin Yeast Interacting Partner E. PDB, Protein Data Bank.

Conclusion

HYPE’s regulatory position at the nexus of various disease pathways makes it an attractive therapeutic target. Additionally, the discovery of two rare diseases associated with mutations in the HYPE/FicD active site that render a constitutively AMPylated and inactive form of BiP makes a case for identifying an inhibitor for HYPE-mediated AMPylation. However, to date, no novel compounds directed against HYPE-mediated AMPylation have been discovered. Here, we establish an optimized and scalable dual-screening assay capable of identifying novel modulators of in vitro HYPE AMPylation. We also present a robust and reproducible pipeline for validating hits using orthogonal biochemical and cellular assays. Indeed, we have discovered a novel inhibitor of HYPE-directed AMPylation in I2.10. We report consistent I2.10-induced reduction of HYPE autoAMPylation with low micromolar bioactivity. Moreover, I2.10 enacts a reliable inhibitory effect of AMPylation on HYPE’s primary substrate, the ER chaperone BiP. The high bioactivity and low cytotoxicity of I2.10 suggest it could act specifically against cellular targets like BiP and other emerging substrates. As it already possesses CNS amenable characteristics (e.g. blood–brain barrier penetrant), I2.10 is especially suited for development as a HYPE-specific drug for the treatment of clinical conditions of neurodegeneration and diabetes. Ongoing follow-up experiments in neuronal and diabetes models will determine the cellular efficacy of I2.10. Docking studies predict that I2.10 binds in the ATP-binding site by accepting two hydrogen bonds and interacting with residues known to interact with the adenine functionality on ATP. While these docking studies utilized the Glide Xp method to account for the largest degree of protein and ligand flexibility there is still a possibility that protein movement upon binding could influence the interactions with the ligand. Future studies to validate these docking results through site-directed mutagenesis or elucidation of a ligand-bound structure are important. Conversely, molecular dynamics (such as those employed previously to study chaperone interactions32, 33) may be used to gain a better understanding of the trajectory for which the ligand interacts with the protein and finds its final binding state. This would allow for large degrees of flexibility and may provide additional clues for further optimization. Further validation of binding pose will inform medicinal chemistry efforts to generate I2.10 derivatives with enhanced drug-likeness and bioactivity. I2.10's relatively low molecular weight (Table 3) and logP values make it well-positioned for hit-to-lead optimizations. Novel, optimized derivatives will then be fed back into our established validation pipeline.

Materials and methods

Protein expression and purification

Recombinant His₆-small ubiquitin-like modifier tagged proteins were purified as previously described.⁶ Specifically, Δ102 (aa 103-458 HYPE constructs (WT, E234G, and E234G/H363A) and Δ19 T229A BiP (aa 20-654) proteins were cloned into pSMT3 plasmids and expressed in Escherichia coli BL21-DE3-RILP (Stratagene) in Luria Broth medium containing 50 μg/mL of kanamycin to an optical density of A₆₀₀ = 0.6. Protein expression was induced with 0.4 mM Isopropyl β-d-1-thiogalactopyranoside for 12–16 h at 18 °C. Lysis was performed on frozen pellets dissolved in lysis buffer (50 mM Tris, 250 mM NaCl, 5 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, pH 7.5). Lysed cells were centrifuged at 15,000 g for 50 min. Supernatants were poured over cobalt resin. The resin was washed with wash buffer (50 mM Tris, 250 mM NaCl, 15 mM imidazole, pH 7.5). Tagged proteins were eluted with elution buffer (50 mM Tris, 250 mM NaCl, 350 mM imidazole, pH 7.5). The His₆-small ubiquitin-like modifier tags were cleaved by incubating proteins with Ubl-specific protease 1 overnight at 4 °C. The protein mixture was diluted in wash buffer without imidazole and reapplied unto a cobalt column. The flow through containing cleaved protein was further purified by size exclusion chromatography in a buffer containing 100 mM Tris and 100 mM NaCl, pH 7.5. Fractions containing HYPE were verified for purity by SDS-PAGE and pooled together. Protein concentrations were measured spectrophotometrically at A₂₈₀. Proteins were flash frozen and stored at −80 °C in Storage Buffer (50 mM Tris, 300 mM NaCl, 10% [v/v] glycerol, pH 7.5).

Glutathione S-Transferase (GST)-tagged WT IbpA-Fic2 (IbpA-2) and Q61L Cdc42 were bacterially expressed and purified as previously described.²⁸

Assay development

Microplate optimization

A fluorescence polarization (FP)-based, dual high throughput screen (HTS) was used, as described in Figure 1 and reported previously.¹¹ Briefly, FP of Fl-ATP was used to monitor HYPE-mediated AMPylation. When unattached to WT HYPE or compound-inhibited for E234G HYPE, Fl-ATP undergoes rapid rotation to depolarize plane-polarized light, giving a basal FP signal. However, attachment of Fl-AMP via autoAMPylation (compound-activated WT HYPE or E234G HYPE) permits light to remain polarized for a high FP signal.

Buffer optimization experiments were performed by incubating 0.4 μM of Δ102 HYPE enzyme with 25 nM Fl-ATP (final concentrations) in HTS buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (pH 7.5), 1 mM MgCl₂, 0.1% TX-100, 1% DMSO) in the dark at room temperature for 10 min in 384-well black bottom, black wall microplates (20 μL reaction volume). Where specified, 1 mM MgCl₂ was replaced with an equivalent concentration of MnCl₂, or 0.1% TX-100 was replaced with 0.05% or 0.01% Trion X-100. Microplates were snap centrifuged at 1000 g for 10 s. Microplates were loaded onto a BioTek BioStack NEO2 plate-reader and assessed for fluorescence polarization with 485/530 nm filters and a 55/50 gain adjustment.

A Multidrop 384 reagent dispenser (ThermoFisher) was used to add 0.4 μM HYPE, then 25 nM Fl-ATP (final concentrations) to 384-well black/black microplates (20 μL reaction volume). Both enzyme and nucleotide were dissolved in HTS Buffer. AMPylation Reactions were incubated for 10 min at room temperature in the dark. Microplates were snap centrifuged at 1000 g for 10 s, then transferred to BioTek BioStack NEO2 plate-reader. Reactions were assessed for fluorescence polarization with 485/530 nm filters and a 55/50 gain adjustment.

Assay quality was determined using statistical parameters signal to baseline (S/B); control variability (CV) which measures the variability within a control group, and Z′, which measures assay variability based on the separation between positive and negative controls.11, 12, 24 Z′, S/B, and CV values were determined as previously described²⁴ by fitting the data to (1), (2), (3), respectively,

$$\mathrm{Z}\mathrm{’}=1-\frac{3\left({\sigma }_p+{\sigma }_n\right)}{\left|{\mathrm{\mu }}_p-{\mathrm{\mu }}_n\right|}$$ (1)

$$\mathrm{S}/\mathrm{B}=\frac{{\mathrm{\mu }}_p}{{\mathrm{\mu }}_n}$$ (2)

$$\mathrm{CV}=\frac{\sigma }{\mathrm{\mu }}\times 100$$ (3)

where µ_p and µ_n are the means of the positive (E234G HYPE) and negative (WT HYPE) controls, respectively; and σ_p and σ_n are the standard deviations of the positive and negative controls, respectively.²⁴

High-throughput screen

An Echo Liquid Handler (Labcyte) was used to transfer 200 nL of DMSO-dissolved compounds via acoustic-coupled ejection from 10 mM source plates into 384-well black, flat-bottom, black-walled microplates, for a final compound concentration of 10 μM. Compounds were sourced from the following libraries: DIVERset (Chembridge), CNS-Set (Chembridge). The compounds went into columns 3–22 of each plate, while columns 1, 2, 23, and 24 were reserved for equivalent volumes of 100% DMSO controls.

A Multidrop 384 reagent dispenser was used to pipette 0.4 μM of WT HYPE (final concentration) dissolved in HTS Buffer into columns 1–22 of the activator plates, and columns 1 and 2 of the inhibitor plates as positive controls. Similarly, 0.4 μM of E234G HYPE was added to columns 3–24 of the inhibitor plates, and columns 23 and 24 of the inhibitor plates as positive controls. Enzymes were then incubated with either compounds or DMSO for 10 min at room temperature.

A Multidrop 384 reagent dispenser was used to pipette 25 nM of Fl-ATP (final concentration) into all wells of each plate, giving a final reaction volume of 20 μL. Plates were then incubated for 10 min at room temperature in the dark, followed by snap centrifugation at 1000 g for 10 s. Microplates were loaded onto BioTek BioStack stacker and read in succession for fluorescence polarization with 485/530 nm filters and a 55/50 gain adjustment. Fluorescence polarization values were converted to fractional activation or fractional inhibition values according to (3), (4), respectively,

$$\mathrm{FA}=\frac{{(x-\mathrm{\mu }}_n)}{{\mathrm{\mu }}_n}$$ (4)

$$\mathrm{FI}=\frac{(x-{\mathrm{\mu }}_p)}{{\mathrm{\mu }}_p}$$ (5)

where µ_p and µ_n are the means of the positive (E234G HYPE) and negative (WT HYPE) controls, respectively; and x is the measured value of fluorescence polarization.

Microplate “cherry-picking” validations were performed as described above.

Concentration-response curves

A Multidrop 384 reagent dispenser was used to pipette 0.4 μM of WT or E234G HYPE (final concentrations) dissolved in HTS Buffer into designated microplate wells. Exactly 10 μM of DMSO-dissolved compounds (final concentrations) or an equivalent volume of 100% DMSO was manually pipetted into their specified wells. Enzymes were incubated with compounds or DMSO for 10 min at room temperature. The reagent dispenser was then used to pipette 25 nM of Fl-ATP (final concentration) into each well, followed by a 10 min incubation at room temperature in the dark. Plates were snap centrifuged at 1000 g for 10 s, then transferred to BioTek BioStack NEO2 plate-reader. Reactions were assessed for fluorescence polarization with 485/530 nm filters and a 55/50 gain adjustment. EC₅₀ and IC₅₀ values were determined by fitting polarization values to (5), (6), respectively,

$$\mathrm{Y}=\frac{(Y_{\max }\times {\left[I\right]}^n)}{{{(\mathit{EC}}_{50}}^n+{[I]}^n)}+Y_{\min }$$ (6)

$$\mathrm{Y}=\frac{(Y_{\max }\times {\left[I\right]}^n)}{{{(\mathit{IC}}_{50}}^n+{[I]}^n)}+Y_{\min }$$ (7)

where Y_max is the maximum polarization signal; I is the μM concentration of activator or inhibitor; n is the Hill slope; EC₅₀ or IC₅₀ are the inflection point concentrations; and Y_min is the baseline polarization response.

Fluorescence in-gel AMPylation validation

A final concentration of 1 μM of WT HYPE was preincubated with 10 μM DMSO-dissolved compound (final concentration) or an equivalent volume of 100% DMSO at room temperature for 10 min. Reactions were run for 20 h at 30 °C in the dark and began with the addition of 1 μM Fl-ATP (final concentration). Alternatively, E234G HYPE AMPylation reactions were run at 10 min due to its hyperactivity. All other parameters were held constant. All reactions were performed HTS Buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 1 mM MgCl₂, 0.1% TX-100, pH 7.50). All reactions were quenched with 4× SDS loading dye, boiled for 5 min at 95 °C, and samples were run on 12% SDS-PAGE gels. Gels were imaged on a Typhoon 9500 FLA imager (General Electric) for fluorescence at 473 nm. Imaged gels were then stained with Coomassie Brilliant Blue, destained, and imaged for protein concentration. Where specified, 5 μM of T229A BiP (final concentration) was added to reactions before the addition of the compound.

Radiolabeled in-gel AMPylation validation

Final concentration of 1 μM of HYPE and (where specified) 5 μM of T229A BiP (final concentration) were preincubated with 10 μM DMSO-dissolved compound (final concentration) or an equivalent volume of 100% DMSO at room temperature for 10 min in HTS Buffer. Reactions were run for 10 min at 30 °C and began with the addition of 0.01 μCi/μL α-³²P-ATP (final concentration). All reactions were performed according to our previously described protocol for in vitro radioactive AMPylation assays.²⁷ Where specified, 5 μM of T229A BiP (final concentration) was added to reactions before the addition of the compound.

All bacterial Fic specificity assays were conducted with 1 μM GST-tagged IbpA-Fic2, 5 μM GST-tagged Q61L Cdc42, and 0.01 μCi/μL α-³²P-ATP (all final concentrations). All other reaction parameters were kept as per HYPE radiolabeled in-gel AMPylation assays.

Fluorescence in-gel deAMPylation validation

Final concentrations of 1 μM of E234G HYPE, 5 μM of T229A BiP, and 1 μM Fl-ATP in HTS buffer were reacted to completion for 90 min at 30 °C in the dark. After AMPylated T229A BiP substrate was generated, 5 μM of WT HYPE deAMPylase or catalytically dead E234G/H363A HYPE or an equivalent volume of Storage buffer were added to reactions, immediately followed by the addition of 10 μM of a compound of the equivalent volume of DMSO. deAMPylation reactions were returned to 30 °C in the dark for 60 min. All reactions were quenched with 4× SDS loading dye, boiled for 5 min at 95 °C, and samples were run on 12% SDS-PAGE gels. Gels were imaged on a Typhoon 9500 FLA imager (General Electric) for fluorescence at 473 nm. Imaged gels were then stained with Coomassie Brilliant Blue, destained, and imaged for protein concentration.

MTT cell viability validation

The cell lines were procured from American Type Culture Collection and cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum without antibiotics. Cells were grown and maintained in a humidified incubator at 37 °C with 5% CO₂. For the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay, cells were trypsinized and harvested at ∼80–90% confluency and seeded in 96 well plates at a density of 1 × 10⁴ cells per well. The seeded cells were incubated overnight and treated with 1 µM, 3 µM, 5 µM, 7 µM, 10 µM, 20 µM, 30 µM, or 50 µM (final concentration) of DMSO-dissolved compound for 48 h. To rule out vehicle-induced toxicity, only 0.5% of the final DMSO concentration was added to the cells. Post-incubation, the compound treatment was removed and replaced with 0.5 mg/mL (final concentration) of freshly prepared MTT solution. After incubating for 3.5 h at 37 °C, 5% CO₂ humidified incubator, MTT was carefully removed without aspirating the formazan crystals and replaced with an equivalent volume of DMSO to dissolve the formazan crystals. The plates were then read using a multi-plate reader at 570 nm. The absorbance was recorded, and % viability was calculated according to the following equation:

$$\%\mathrm{viability}=\frac{{\mathit{Ab}}_t}{A{b}_c}\times 100$$ (8)

where, Ab_t is the absorbance of compound-treated cells, and Ab_c is the absorbance of control cells.

Protein preparation and docking

All protein and ligand preparation and docking studies were run in the Schrodinger Small Molecule Drug Discovery Suite (Schrodinger, LLC, New York, NY, software release 2022-2) using computational tools available through its Maestro interface. The apo-HYPE (Protein Data Bank (PDB)): 4U04) crystal structure was prepared as follows. A protein reliability report was generated prior to preparation as a point of reference to ensure the improvement of the prepared structure. During the pre-processing step, termini were capped, missing side chains were filled in, bond orders were assigned, the Chemical Component Dictionary database was used, hydrogens were replaced, zero-order bonds to metals were created, disulfide bonds were created, missing loops were filled in using Prime,34, 35 and heteroatom states were generated with pH 7.4 ± 1.0 using Epik. The preprocessed structure was then optimized by sampling water orientations using program that predicts the pKa values of ionizable groups in proteins (version 3.0) and protein-ligand complexes (version 3.1 and later) based on the three dimensional structure pH 7.4. Minimization of the optimized structure was then performed by converging heavy atoms to an Root Mean Square Deviation of 0.3 Å using the OPLS4 forcefield, followed by the removal of waters with less than two hydrogen bonds to non-waters. A protein reliability report was generated for the final prepared structure to ensure no structural problems existed in proximity to the binding pocket. The hit compound I2.10 was prepared using LigPrep where possible heteroatom states were generated for pH 7.4 ± 1.0 using Epik. The rest of the LigPrep parameters were left at default settings.

A receptor grid was generated by defining the enclosing box at the centroid of the selected residues H363, N369, Y400, and N407. The prepared I2.10 structure was then docked into the prepared apo-HYPE structure using the receptor grid with Glide, using Extra Precision (XP) docking and no constraints, with the remaining docking settings at default.

Funding and support

This publication was made possible with support from the National Institute of General Medical Sciences of the National Institutes of Health (R01GM10092), an Indiana Clinical and Translational Research Grant (CTSI-106564), and a Purdue Institute for Inflammation, Immunology, and Infectious Disease Core Start Grant (PI4D-209263) to S.M.; an Overseas Visiting Doctoral Fellowship from the Science and Engineering Research Board of India (Purdue-India SERB OVDF Program) to H.C.; and support from Grant #UL1TR002529 (A. Shekhar, PI), May 18, 2018 to April 30, 2023, and Grant #TL1TR002531 (T. Hurley, PI), May 18, 2018 to April 30, 2023, from the National Institutes of Health, National Center for Advancing Translational Sciences, Clinical and Translational Sciences Award to A.C.

Declaration of generative AI in scientific writing

No generative AI was used in the drafting or completion of this manuscript.

Declarations of interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Data availability statement

No data was used for the research described in the article.