Design, synthesis, and evaluation of peptide‐imidazo[1,2‐a]pyrazine bioconjugates as potential bivalent inhibitors of the VirB11 ATPase HP0525

Helicobacter pylori (H. pylori) infections have been implicated in the development of gastric ulcers and various cancers: however, the success of current therapies is compromised by rising antibiotic resistance. The virulence and pathogenicity of H. pylori is mediated by the type IV secretion system (T4SS), a multiprotein macromolecular nanomachine that transfers toxic bacterial factors and plasmid DNA between bacterial cells, thus contributing to the spread of antibiotic resistance. A key component of the T4SS is the VirB11 ATPase HP0525, which is a hexameric protein assembly. We have previously reported the design and synthesis of a series of novel 8‐amino imidazo[1,2‐a]pyrazine derivatives as inhibitors of HP0525. In order to improve their selectivity, and potentially develop these compounds as tools for probing the assembly of the HP0525 hexamer, we have explored the design and synthesis of potential bivalent inhibitors. We used the structural details of the subunit–subunit interactions within the HP0525 hexamer to design peptide recognition moieties of the subunit interface. Different methods (cross metathesis, click chemistry, and cysteine‐malemide) for bioconjugation to selected 8‐amino imidazo[1,2‐a]pyrazines were explored, as well as peptides spanning larger or smaller regions of the interface. The IC50 values of the resulting linker‐8‐amino imidazo[1,2‐a]pyrazine derivatives, and the bivalent inhibitors, were related to docking studies with the HP0525 crystal structure and to molecular dynamics simulations of the peptide recognition moieties.


| INTRODUCTION
Since its isolation in 1983, 1 Helicobacter pylori (H. pylori) has been identified as the most common human bacterial infection, present in approximately half of the world's population. 2 This type of Gramnegative bacteria is found in the human stomach and causes illnesses such as gastric ulcers, gastritis, and various cancers including mucosaassociated lymphoid tissue (MALT)-lymphoma and gastric adenocarcinoma. 3,4 Although the majority of those infected are asymptomatic, H. pylori-positive patients have a 10%-20% lifetime risk of developing ulcer disease and a 1%-2% risk of developing gastric cancer, 5 and as such, the bacteria have been classified as a category 1 carcinogen. 6 It has been estimated that between 2008 and 2015, the proportion of noncardiac gastric cancer attributable to H. pylori increased from 74.7% to 89.0%. 7 The current standard treatment of H. pylori infections is based upon triple therapy, 8,9 consisting of a proton pump inhibitor and a choice of two antibiotics or quadruple therapy [8][9][10] in which bismuth compounds are also used. The success of these therapies is unfortunately under pressure due to rising antibiotic resistance and off-target effects caused by prolonged antibiotic treatment. 11 Multi-drug resistant H. pylori (resistance to ≥3 antibiotics of different classes) ranges from ≤10% in Europe to >20% in India and >40% in Peru. 12 So far, no eradication therapy can provide high eradication rates (>90%), and a variety of approaches to targeting H. pylori including antivirulence therapeutics, mucolytic agents, and antibacterial agents are currently being investigated. 13 Gram-negative bacteria have evolved a range of secretion systems to transport substrates across their cell membranes. They can release small molecules, proteins, and DNA into extracellular space or can inject these substrates into a target cell. 14 Seven classes of double membrane-spanning secretion systems, Type I-Type VII, have so far been identified. The Type IV secretion system (T4SS) is of particular interest, as it mediates the transfer of plasmid DNA between bacterial cells, thus contributing to the spread of antibiotic resistance genes.
Inhibitors of the T4SS are therefore of interest as antimicrobial agents with the potential to slow the development of antimicrobial resistance. 15 H. pylori can be grouped into two classes, 16,17 and the more virulent type I strains contain the cytotoxin-associated genes pathogenicity island (cagPAI) 18 and are referred to as CagA + strains. The cagPAI consists of 31 genes, the majority of which code for T4SS, 5,16 which in H. pylori is responsible for penetrating the gastric epithelial cells and facilitating the translocation of toxic bacterial factors into host cells. [19][20][21] T4SS are multifunctional macromolecular nanomachines, incorporating 12 different types of protein subunit with specific roles in the complex. 22 The VirB11 ATPase HP0525 is a key component of this complex, which provides energy to power the system and is required to drive CagA secretion and delivery. It is also hypothesized to act as a molecular switch that controls the export of DNA and the assembly of the T4SS pilus. Hence, there has been much interest in the structure and function of this protein. There have been extensive studies into the crystal structures of the H. pylori VirB11 homolog HP0525 bound to both ADP 23 and the non-hydrolysable ATPγS as well as the nucleotide-free, apo-form. 24  Selective inhibitors of VirB11 have the potential to combat the proteins and pathways that lead to pathogenic, symptomatic colonization of the stomach, and may also be useful chemical biology tools to elucidate the pathways of assembly of the T4SS macromolecular complex. However, only a small number of small molecule inhibitors of H. pylori HP0525 have so far been reported. 15 Thiadiazolidine-3,-5-diones 26 and the non-competitive inhibitor 4-(5-methylpyridin-2-yl) oxybenzoic acid 27 have been shown to inhibit VirB11 ATPase, whereas heterocyclic 2-pyridone inhibitors 28 have been shown to disrupt T4SS apparatus biogenesis by attenuating the delivery of peptidoglycan and CagA to host cells. Unsaturated 2-alkynoic fatty acids have also been shown to inhibit conjugation in TrwD, a VirB11 homolog. 29 We have previously reported the synthesis and high throughput screen of a series of novel 8-amino imidazo[1,2-a]pyrazine derivatives against VirB11 ATPase HP0525. 30 Biochemical evaluation showed moderate to good potency, highlighting these compounds as competitive inhibitors of HP0525 and potential antibacterial agents. However, as these molecules probably bind at the ATPase active site, to avoid inhibiting other ATPases in mammalian cells, we wished to try to improve their selectivity for VirB11 ATPase. We were also motivated to exploit the potential of these compounds as chemical biology tools for probing the assembly of the HP0525 hexamer. In order to achieve both of these aims, in this paper, we have explored the design and synthesis of potential bivalent inhibitors of the VirB11 ATPase HP0525.
Affinity and specificity of an inhibitor can be greatly enhanced by taking advantage of hydrophobic/hydrophilic features near the enzyme active site. By linking an active site binding compound to a moiety that interacts with these features on the enzyme surface, it is possible to differentiate among enzymes. The resulting compounds are known as bivalent inhibitors. Bivalent molecules comprising a small molecule and a peptide have been frequently studied, as this approach has been particularly valuable in designing selective kinase inhibitors, 31 in which an ATP-binding site directed small molecule is tethered to a protein substrate site of the protein kinase. Examples of potent and selective bivalent inhibitors include ATPγS-IRS727 peptide bioconjugates as insulin receptor protein tyrosine kinase (IRK) inhibitors 32 and high-affinity nonselective kinase inhibitor K252a-protein kinase inhibitor protein (PKI) bioconjugates as inhibitors of protein kinase A (PKA). 33 Bivalent inhibitors can also be designed to selectively disrupt protein-protein interactions (PPIs). 34 In this work, we have used the imidazo[1,2-a]pyrazine compounds previously synthesized to target the ATP binding site of HP0525, whereas a peptide sequence has been rationally designed to disrupt the peptide-peptide interactions of the subunit-subunit interface and, in doing so, disrupting hexamer formation. All resins were pre-swelled in DMF for at least 30 min prior to synthesis start. Standard Fmoc solid phase peptide synthesis (SPPS) was employed. The total volume of all reagents in each step was 1.5 ml.
All reagents were dissolved in HPLC grade DMF.  CTD-CTD interactions. Image generated using chimera. 25 (B)4hematic representation for the mode of action of VirB11 ATPases. NTD (light pink), CTD (light blue). ATP binding (blue) and ADP (yellow). In 1 (nucleotide-free form), the CTD maintains the pseudo-scaffold and the NTD is mobile. Binding of three ATP molecules then locks three of the subunits into a rigid conformation (state 2). Hydrolysis of the first three ATPs to ADP together with binding of a further three ATP molecules to the remaining subunits leads to state 3. Hydrolysis of the final ATP molecules leads to a rigid hexamer (state 4). Finally, release of the six ADPs allows the hexamer to revert to its nucleotide-free asymmetric form (state 1). Reproduced with permission (The EMBO Journal) from Savvides et al. 24 reagents were removed by filtration under vacuum, and the resin washed with DMF (6 Â 1.5 ml).

| Amino acid coupling
The reaction syringe was added Fmoc-protected amino acid

| Peptide cleavage and side chain deprotection
The resin was washed with CH 2 Cl 2 (3 Â 3 ml), MeOH (3 Â 3 ml) and The resultant precipitate in solution was stored at À20 C for 30 min prior to being spun at 4000 rpm for 10 min at 4 C to produce a crude peptide pellet. The supernatant Et 2 O was decanted off, and the peptide washed a further three times with Et 2 O. The crude peptide pellet was then re-dissolved in minimum water and freeze-dried for storage prior to purification.

| General peptide purification
The peptides were analyzed and purified via reverse phase HPLC Analytical HPLC traces, ES+, and MALDI spectra of peptides 1-9 can be found in the supporting information.
An aliquot of a stock solution of 17 in DMF (2.87 mM) was added such that the volume added corresponded to 3 eq, and the reaction mixture stirred at RT for 3 h. The solvent was removed in vacuo and the crude material purified via reverse phase preparative HPLC.

| "Click" chemistry (on resin)
The peptides were synthesized according to standard SPPS, with all side chain and the terminal Fmoc protecting groups left on. The peptide was also left bound to the resin, and all subsequent reactions were carried out in the syringe; 164 (43.3 mg, 0.080 mmol, 4 eq) in DMF (2 ml) was added to the reaction syringe followed by the addition of a suspension of CuI (76.0 mg, 20 eq) and sodium ascorbate (158 mg, 40 eq) in H 2 O/ t BuOH (2:1; 750 μl) to give a total solvent composition of DMF/H 2 O/ t BuOH (8:2:1). The mixture was agitated at RT for 16 h, followed by evacuating the solvent and washing with DMF (3 Â 3 ml), MeOH (2 Â 3 ml), CH 2 Cl 2 (3 ml), and DMF (2 Â 3 ml). The terminal Fmoc was removed, followed by cleavage of the peptide from the resin and deprotection of the side-chain protecting groups using conditions stated above.

| Cysteine-maleimide conjugation (on resin)
The peptides were synthesized according to standard SPPS, with all side chain and the terminal Fmoc protecting groups left on. The peptide was also left bound to the resin, and all subsequent reactions were carried out in the syringe. The S t Bu protecting group on R240C was selectively deprotected by: washing the resin-bound peptide with CH 2 Cl 2 (3 Â 3 ml); soaking in EtOH/CH 2 Cl 2 /H 2 O (4:6:1; 3 ml); purging with Ar; adding nBu 3 P (50.0 μl, 0.200 mmol, 10 eq); and agitating for 3 h at RT. The syringe was evacuated and washed with CH 2 Cl 2 (2 Â 3 ml), MeOH (2 Â 3 ml), CH 2 Cl 2 (2 Â 3 ml), and DMF (2 Â 3 ml). Maleimide conjugation to the free thiol in R240C was carried out by adding 5 (42.0 mg, 0.060 mmol, 3 eq) in DMF (3 ml) and agitating at RT for 16 h. The syringe was evacuated and washed with DMF (3 Â 3 ml). The terminal Fmoc was removed, followed by cleavage of the peptide from the resin and deprotection of the side-chain protecting groups using conditions stated above.
Analytical HPLC traces, ES+, and MALDI spectra of conjugated peptides 28, 29, and 30 can be found in the supporting information.

| Biological assays
The HP0525 protein was produced in the E. coli strain BL21 Star (DE3) (Invitrogen) as described previously. 39 The protein concentration was estimated spectroscopically using a NanoDrop (Thermo Scientific) and a calculated extinction coefficient at 280 nm, based on the amino acid composition. The ATPase activity of HP0525 was measured, with and without a specific amount of compound present, using an in vitro ATPase colorimetric assay kit (Innova Biosciences).
The assay was performed in 96-well ELISA microplates (Greiner Bio-One), using a multipipette/robot. 250 μM ATP; and 10% DMSO) was added to each assigned well, followed by the addition of 1 μl of compound (at 0.5; 5 or 50 mM to achieve the final concentrations of 5; 50 and 500 μM in the reaction, respectively) in DMSO (or 1 μl DMSO to controls). The solutions were mixed carefully by pipetting. The reaction was started by the addition of 50 μl of 0.106 μM HP0525 to each well (except the negative control, see text below), and the reaction plate was directly transferred to 37 C for 30 min of incubation. The reaction was stopped by the addition of the gold mix according to the standard protocol of the kit. The absorbance at 620 nm was measured after 30 min at RT. For each compound, the percentage of absorbance relative to non-inhibited HP0525 was calculated, after subtracting the absorbance value of the negative control. In the negative control, the protein was added after the gold mix, as described in the standard protocol of the kit, which when used as a blank corrects for all free P i not produced by the enzyme during the 30 min incubation at 37 C. A known inhibitor of HP0525 (CHIR02) 26 was used as a control inhibitor. All measurements were made in duplicate.
A selection of the compounds was assayed as above at additional concentrations ranging between 1 and 50 μM (the measuring points were optimized to cover the range to fit a sigmoidal dose response curve, and at the same time ensure compounds were not precipitating) from which IC 50 values were calculated ( Figure S1). Each compound was screened on three separate plates, and each plate was read twice, and a mean was calculated for each concentration. The data were normalized by subtracting the negative control (0% active) and relating it to the positive control (100% active). The software GraphPad Prism 5 was used to generate two dose-response curves

| Circular dichroism spectroscopy
The experiments were performed at the ISMB Biophysics Centre.
The peptides were dissolved in water; their CD spectra were recorded thrice at 1 mg/ml (β9-αF-β10 3 at 0.5 mg/ml) in a Quartz Suprasil cell (pathlength: 0.1 mm). Data were processed with CDTool45 and Dichroweb46. Outliers were excluded, spectra zeroed, averaged, and converted from millidegree units into Δε units, which is a molar unit; the weighed amount of the sample is important for this conversion. Given the weighting errors stated above, data processing was performed for the corresponding range of different sample concentrations. Different algorithms for analysis of the results were tested in Dichroweb. 44 CONTIN was chosen as the data analysis algorithm. SP175_S3 was chosen as a reference set (others were tested before making this choice). Given α-helical content was calculated for different protein concentrations, the results are provided as a range.

| Bivalent inhibitor design
To design bivalent inhibitors that would be selective for the VirB11 ATPase, we sought to combine the structural insights gained from the PPIs between the subunits of the hexameric HP0525 complex 24  In order to design a mimic of the PPIs at the subunit interface, we aimed to rationally design a peptide that resembles the αF helix and F I G U R E 2 (A) Crystal structure of ATPγS bound to HP0525 (1NLY), highlighting the targeted peptide region at the subunit-subunit interface. ATPγS is in yellow. 24 (B) Structure of the β9-αF-β10 peptide region showing the different secondary structures and the potential first point of attachment for conjugation. Image generated using PyMOL 45 neighbouring residues (Figure 2a). We envisaged that by anchoring the small molecule in the nucleotide binding site, the peptide part of the chimera would interact with the subunit-subunit interface, possibly displacing the native segment and in the process disrupting hexamer formation. The peptide sequence highlighted consists of the β9 sheet followed by a loop into the αF helix followed by a further loop into the β10 sheet (Figure 2b). It was unclear what length of peptide would be required to interact with the native peptide region, and therefore, three peptides with varying lengths were chosen for investigation: αF-loop 1, αF-β10 2, and β9-αF-β10 3 (Table 1). Furthermore, as Arg240 is in close proximity to the nucleotide binding site, this seemed to be the ideal site for conjugation to the imidazo[1,2-a] pyrazine partner. By substituting this arginine for other amino acids capable of conjugating to other moieties, a range of bivalent inhibitors could be synthesized. We therefore prepared variants of these three peptides with Arg240 substituted by Cys (4, 5, 6), S-allyl cysteine (SAC) (7,8,9), or azidolysine (AzLys) (10,11,12). Preliminary docking studies suggested that a single PEG unit should be of sufficient length to extend from the carbamate linked to the toluoyl moiety, to the peptide sequence that will make interactions with the adjacent subunit.
Thus, our bivalent inhibitor reagents were designed to consist of a small molecule inhibitor, based on imidazo[1,2-a]pyrazine, that will bind to the nucleotide binding site, linked (via a PEG chain) to a peptide, based on the αF helix of HP0525, that could then substitute the native αF helix and disrupt the opening and closing mechanism of the hexameric portal (Figure 1b).

| Synthesis
There were a number of possible approaches to the synthesis of these bivalent inhibitors. We adopted the strategy of first synthesizing the
In the case of R240C peptides a non-acid labile protecting group (acetamidomethyl -Acm) 50,51 was used on the other cysteine residues so that the peptide produced has only one thiol free for maleimide conjugation.

| Bivalent inhibitor synthesis
Before attempting to conjugate malemide 17 to the target peptides, reaction conditions were optimized using a model system.  (Figure 4), but unfortunately no product was isolated when the longer peptides (αF-β10 and β9-αF-β10) were used.
Despite trialling a range of conditions, the on-resin cross metathesis of R240SAC peptides and 15 was unsuccessful.

| Biological results
Initially, the activity of the PEGylated-imidazo[1,2-a]pyrazines were tested for inhibition of HP0525 using an in vitro ATPase colorimetric assay. 30 The IC 50 data are shown in Table 2 Figure 5). Indeed, when studying the in silico lowest energy binding modes within the nucleotide binding site, they appear to adopt a conformation similar to that of

| Bivalent inhibitor reagents
The three peptide-imidazo[1,2-a]pyrazine conjugates (28,29,30) were tested at concentrations of 5, 50, and 500 μM, and the % absorbance at 620 nm analyzed to give an indication of their inhibitory effect ( Table 3). The results clearly indicate that the peptide conjugates only inhibit very weakly at a high concentration.

| Wild-type peptides
The stability of the peptide fold might be important for both activity and reactivity. In silico stability testing on the three WT peptides using molecular dynamics showed only that the β9-αF-β10 peptide 3 was immediately, and the αF-β10 peptide 2 unfolded, but at a slower rate.

| CONCLUSION
Improving the selectivity profiles of ATPase or kinase inhibitors that bind to the ATP binding site is crucial to avoiding toxicity due to off-target effects in vivo. A bivalent inhibitor approach, whereby the tight binding of ATP mimics is combined with targeting elements that bind outside the ATP binding site, has considerable potential.
Designing a peptide or a small molecule that can mimic or disrupt PPIs is challenging, and the parameters that will lead to a successful inhibitor are not well-understood. 54 In this work, we have succeeded in improving the activity of a lead 8-amino imidazo[1,2-a]pyrazine inhibitor of the VirB11 ATPase HP0525 by attachment of a PEG moiety, primed for bioconjugation.
We have established effective methodology for the bioconjugation of the 8-amino imidazo[1,2-a]pyrazine-PEG to short peptides, via either cysteine-maleimide reaction or alkyne/azide click chemistry. A possible reason for the conjugation failing for the longer length peptides could be the accessibility of the relevant amino acid. Where the conjugation was carried out on the solid phase, the R240 amino acid is close to the C-terminus and therefore close to the resin. If the longer length peptides are extensively folded on-resin, the thiol, allyl, or azide moieties might be considerably buried from their conjugation partner. The use of a lower loaded resin, 55 or of protecting groups such as Hmb or pseudoproline moieties designed to disrupt folding on resin of "difficult sequences, 56 could be beneficial, as could using alkyne/azide click chemistry for bioconjugation in solution.
However, conjugation to peptide fragments designed to mimic or disrupt the interactions between the subunits in the VirB11 ATPase HP0525 hexameric complex did not result in effective enzyme inhibitors. This was particularly disappointing, as functionalization of the 8-amino imidazo[1,2-a]pyrazine fragment with the PEGcarbamate linkers actually improved their inhibition relative to the parent compound 14. One possible explanation for this is that the linker between the 8-amino imidazo[1,2-a]pyrazine and the peptide fragment has the wrong length and/or structure to correctly orient the peptide fragment for binding to the next subunit. Whereas further detailed docking studies might help to identify improved linkers, a more powerful approach would undoubtably be to use a peptide aptamer "Trojan Horse" 57 or an enzyme-templated fragment elaboration strategy, 58,59 in which the enzyme itself selects the correct combination of linker and peptide fragment for effective bivalent binding from a small library. A second explanation is that the β9-αF-β10 segment of HP0525, and smaller fragments, do not fold in such a way as to form effective mimics of the HP0525 subunit interface, or that this segment cannot compete with the complete protein during assembly of the hexamer. This is partly supported by our CD and MD studies.
Approaches such as helix stabilization by stapling 60 might improve the binding, although introduction of a constraint to preorganize peptides to a particular secondary structure does not always increase the binding potency to the target. 61 Finally, peptide sequences with N-terminal capping and C-terminal amide modifications should be synthesized, in order to avoid introducing additional charges to the sequence. All of these additional peptide modifications would then need to be tested separately for their ability to fold in solution, access the subunit-subunit interface, and to inhibit hexamer assembly. However, the initial studies presented in this paper will enable both an enzyme-templated fragment elaboration strategy, and the pathway of the HP0525 hexameric complex assembly, to be further studied.