Discovery of α-Amidobenzylboronates as Highly Potent Covalent Inhibitors of Plasma Kallikrein

Hereditary angioedema (HAE), a rare genetic disorder, is associated with uncontrolled plasma kallikrein (PKa) enzyme activity leading to the generation of bradykinin swelling in subcutaneous and submucosal membranes in various locations of the body. Herein, we describe a series of potent α-amidobenzylboronates as potential covalent inhibitors of PKa. These compounds exhibited time-dependent inhibition of PKa (compound 20 IC50 66 nM at 1 min, 70 pM at 24 h). Further compound dissociation studies demonstrated that 20 showed no apparent reversibility comparable to d-Phe-Pro-Arg-chloromethylketone (PPACK) (23), a known nonselective covalent PKa inhibitor.

P lasma prekallikrein (PK) is a serine protease which circulates as a bound complex with its substrate, highmolecular-weight kininogen (HK). 1 PK has structural homology to factor XI (FXI), with both containing four apple domains at the N-terminus. 2 The PK−HK complex in plasma 3 can be activated by a single cleavage from coagulation factor XIIa (FXIIa) resulting in the activated form, plasma kallikrein (PKa).PKa performs a double cleavage of high-molecularweight kininogen (HK) to liberate the vasoactive peptide bradykinin, which binds the B2 bradykinin receptors (B2R) on endothelial cells, leading to activation of signaling events causing vascular permeability and an inflammatory response (Figure 1).
B2R is an essential G protein-coupled receptor (GPCR) involved in regulating homeostasis of the cardiovascular system as a vasodepressor. 4Binding of bradykinin to B2R promotes vasodilation, increased endothelial permeability, and capillary bed leakage, which in severe cases of bradykinin overproduction displays the symptoms of an acute angioedema attack. 5Accordingly, significant interest has developed around PKa as a drug target for the treatment of bradykinin-mediated angioedema such as hereditary angioedema (HAE).HAE is a rare disorder, occurring in three main subtypes; type I (quantitative) and type II (qualitative) originate from a deficiency and dysfunction in the endogenous C1 inhibitor protein (C1-INH), respectively, due to mutations of the C1-INH SERPING1 gene.Type III occurs as a result of mutation on the factor XII gene, which affects the PKa-mediated BK formation. 61-INH is the endogenous inhibitor of PKa, FXIIa and other proteases in the plasma, and HAE patients with reduced C1-INH functional levels are unable to effectively block PKa enzyme activity.Intravenously administered recombinant human C1-INH as well as plasma-derived C1-INH (pdC1-INH) have delivered safe and efficacious treatments of HAE attacks.HAE facilitates spontaneous and uncontrolled activation of the KKS.7 Patients have the potential to exhibit symptoms, such as subcutaneous and laryngeal swelling, and can also be accompanied by severe abdominal pain and obstruction of major organs.8 The PKa inhibitors berotralstat 1 is approved for prophylactic use and sebetralstat 2 9 is in late-stage clinical studies for the acute on-demand treatment of HAE attacks (Figure 2).Studies with sebetralstat have demonstrated that pharmacological inhibition of PKa suppresses the KKS, resulting in suppression of both the cleavage of the PKa substrate HK and the feedback activation of the KKS in plasma, thereby inhibiting the generation of bradykinin.10,11 Approved acute treatments of HAE attacks involve injected or infused therapies and include icatibant (Firazyr) and C1-INH (Ruconest, Berinert).Berotralstat (Orladeyo) is the only approved oral therapy acting as an inhibitor of PKa for the prophylactic treatment of HAE. 12 Clinical studies showed a 150 mg daily dose of berotralstat is efficacious as an oral prophylactic for HAE. 12 Other HAE therapies such as icatibant (Firazyr), a peptidomimetic injectable drug, is indicated in the EU for the symptomatic treatment of acute attacks in patients with HAE. 13 While other therapies treat the symptoms of acute attacks in patients with HAE through antagonistic suppression of the B2R, there is a need for further highly effective oral prophylactic treatments to address the outbreak of angioedema events.14 Increasingly, boron-containing molecules 15,16 are emerging as structures that target key nucleophilic residues within biological targets, promoting the formation of a covalent bond to maximize potency and elicit prolonged physiological responses, compared with conventional reversible inhibitors (Figure 3).17−19 Taniborbactam 3 and vaborbactam 4 possess boronic ester motifs designed to covalently inhibit serine β-lactamases as a combination therapy alongside the administration of antibiotics.This is a strategy not yet seen in the design of clinical candidates targeting PKa.Lewis acids, such as boron, possess a vacant p-orbital, capable of accepting a lone pair of electrons to generate a stable borate anion.Interestingly, this signifies that, while capable of undergoing covalent interactions with nucleophilic residues, the nature of this interaction is reversible, which may be beneficial to the toxicology profile of such inhibitors, with off-target toxicity being a frequent concern for covalent inhibitors.19 Bortezomib 5 is an FDA-approved inhibitor of the 26S proteasome used for the treatment of various cancers, operating via the covalent interaction of the boronic acid moiety with the active site of the proteosome and showing high potency (IC 50 = 2.4 nM) in in vitro cellular assays.20,21 Taniborbactam 3 is under clinical development as a combination therapy for the treatment of carbapenem-resistant bacterial infections via the inhibition of serine-and metallo-βlactamases. Crisaborole 9, used for topical treatment of fungal infections, is an NSAID inhibitor of the fungal phosophodiesterase 4 (PDE4) enzyme, achieving 490 nM potency with similar inhibition against the release of cytokines TNF-α, IL-2, and IFN-γ with high selectivity.23 Demonstrably, there is significant value emerging from the incorporation of boron warheads into pharmaceutical agents, translating to potential for high efficacy.24,25 The discovery and development of PKa inhibitors has recently been extensively reviewed.26 However, to the best of our knowledge, the only boronic-acid-containing PKa inhibitor to date is compound 10 (PKa IC 50 45 nM), a nonselective compound that is also reported to be an inhibitor of FXIIa (IC 50 190 nM), trypsin (IC 50 52 nM), plasmin (IC 50 394 nM), and FXIa (IC 50 13 nM) (Figure 4).27 Applying this covalent inhibitor design approach to PKa presented the challenge of positioning the warhead appropriately to ensure productive interaction with the catalytic serine of the Ser195, His57, Asp102 catalytic triad.28,29 This was explored through a dual pocket binding approach, gaining inspiration from a literature PKa inhibitor 11 whose binding mode spans the S1 and S4 binding pockets within the PKa active site.30 Compound 11 (Figure 5), 31 a competitive, reversible inhibitor of PKa (IC 50 = 2 nM) was of particular interest due to the positioning of Ser195 relative to the main scaffold.This sparked interest in the installation of a boron warhead on the benzylic position of analogues of 11 to induce a potential covalent interaction with Ser195 within the binding site of PKa.
In initial studies, we investigated a truncated fragment of 11, omitting the terminal methylpyrazole unit and replacing the  aminopyridine S1 group with the nonbasic 3-chlorophenyl group.These changes were carried out to enhance ease of synthesis and compound stability due to possible incompat-ibility between boronates and nucleophilic groups, such as the amine on the aminopyridine S1 group.
Molecular docking of truncated fragment 12 into the PKa active site showed comparable positioning of the benzylic carbon, extending from the S1 pocket, suggesting 12 to be a suitable candidate to bear a covalent warhead and for subsequent structure−activity relationship (SAR) investigations.Herein, the synthesis and isolated-enzyme potency data for a series of novel sub-nM PKa inhibitors are reported.
Initial synthetic explorations focused on the installation of a boron warhead at the highlighted benzylic position of 12.This was achieved by performing a Matteson homologation sequence of arylboronic pinacol esters 13a−b to yield the corresponding α-chloroboronate intermediates 14a−b, fol-    30 (C) Overlay of docked structure of 11 (teal) with 12 (green) in PKa.Docking experiments were performed using OEDOCKING Hybrid docking. 32owed by displacement with LiHMDS and acidic silyl deprotection to afford the corresponding α-aminoboronate hydrochloride salts 15a−b (Scheme 1). 33,34idation of 15a with 16a gave access to α-amidoboronate 17, and the corresponding boronic acid 18 was obtained under mild conditions via biphasic transesterification with pentylboronic acid (Scheme 2). 35olated enzyme potency was determined against PKa and FXIIa, using FXIIa as a test for the selective binding interaction with PKa.IC 50 measurements were calculated following enzyme and inhibitor preincubation over a 60 min time-course to elucidate possible covalent binding character, under the premise that observed compound potency should increase over time in the case of a covalent inhibitor due to increasing active site occupancy (Table 1).
Surprisingly, 18 showed no activity against either enzyme; after consideration, this was thought to be likely due to its instability in DMSO and aqueous media, causing protodeborylation, a process seen during purification and observed by LCMS.Likewise, intermediate 15a, also unstable over time in DMSO, showed no activity.Curiously, 17a demonstrated activity (64 μM at 60 min) after no initial observation of enzyme activity at 5 min preincubation, whereas 17b showed improved activity at both time points.These initial results provided promise for the exploration of further analogues.
Following these findings, corresponding analogues of 17a and 17b with the extended S4 pyrazole, bearing structural similarity to 10 were synthesized along with further S1 and S4 variants (Scheme 3).Amidation of the α-aminoboronate hydrochloride salts proceeded well to give the 19−22.The low yields obtained generally reflect the difficulty of purification and tendency of these compounds to decompose in reversephase purification media via cleavage of the pinacol ester and subsequent protodeborylation of the boronic acid as observed by LCMS analysis of purified fractions.Along with the 3chlorophenyl S1 substituent, the 2-fluoro-4-methylphenyl analogue was investigated, due to known high affinity for the PKa S1 binding site and their lack of nucleophilicity compared with more conventional S1 groups of this type, such as aminopyridine. 30ompounds 19−22 showed much improved biological activity through a combination of improved S1 binding (17a vs 22) and elongation of the S4 fragment to include a methylpyrazole or methylpyridone unit (19−21), thought to undergo a key S4 π-stacking interaction with Tyr174 (Table 2).
Occupation of the S4 and S1 pockets may have the effect of "anchoring" the scaffold into place, enabling productive interaction between the boron warhead and Ser195.In the case of each compound (19−22), a clear trend of increasing biological activity against time was visible, indicative of covalent binding, with the highest activity being observed for   The isolated enzyme activity of 20 was seen to continually increase over 24 h, achieving an IC 50 value of 0.29 nM at 2 h and 0.07 nM at 24 h (Figure 6).This suggests that once bound within PKa, 20 forms a stable complex with the protein without the propensity to protodeborylate or dissociate from PKa.
To further assess the covalent binding character of 19−22, a series of matched pairs 24−27, omitting the boronate warheads, were synthesized (Scheme 4).
No significant change in activity was observed over the Due to the limitations encountered in the synthesis and purification of the α-amidobenzylboronates (19−22), a second series of boronate-containing compounds was explored, where the amide group was replaced by a ketone to give compounds 31 and 32.The compounds were prepared through the reaction of pinacol diboronate with intermediates 29 and 30, 38 synthesized from common intermediate 28 through an Aldol reaction with commercial benzaldehydes (Scheme 5).
Intermediates 29−30 showed very weak or no inhibition of PKa, and this may be a result of either the geometry of the exit vector from the S1 binding pocket imposed by the rigid alkene linking group or the removal of a key hydrogen bonding interaction observed between the amide N−H and the protein (Ser597).In contrast, compounds 31 and 32 demonstrated reasonable levels of PKa activity.However, they did not display the time course activity exhibited by 19−23, suggesting noncovalent reversible inhibition, with 32 having an IC 50 of ∼40 nM, throughout the time course experiment (Table 2).
−41 Boronates 19−20, covalent reference inhibitor 23, and compounds 24 and 26 were subjected to a 62.5-fold   dilution after 10 min preincubation, and the inhibitor dissociation was monitored using the fluorogenic substrate. 42t was shown that 23 and 20 showed no apparent reversibility over the 600 s time frame of the kinetic dissociation experiment, suggesting that they are covalent inhibitors of PKa.Interestingly, 19 showed slow dissociation over the time course, suggesting possible reversible covalent inhibition, whereas 24 and 26 showed rapid dissociation from PKa.In the same experiment, boronate 32 displayed slow dissociation over the experimental time course, whereas compound 31 demonstrated fast dissociation from PKa.The known highly potent reversible PKa inhibitor FE99026 33 (IC 50 ∼ 4−6 nM), 43 demonstrated slow dissociation from PKa (Figure 7).Time-dependent potency determination was carried out against FXIa, thrombin, trypsin, and plasmin to further evaluate the selectivity of compounds (Table 3).FXIa shares the closest structural homology with PKa of all trypsin like proteases with 58% similarity in sequence identity. 44Selected compounds were assayed against FXIa alongside PKa, and all compounds showed a distinct preferential activity toward PKa over FXIa.Compound 20 showed modest inhibitory activity of FXIa, achieving an IC 50 of 12,230 nM after 60 min preincubation time.In addition, inhibitory potency was seen to increase at each time point, suggesting a mild covalent interaction within the FXIa active site.Compound 20 also showed some time-dependent inhibition against plasmin, achieving an IC 50 of 12,090 nM after 60 min preincubation time.
In all cases, enzyme potency was seen to have a >1000-fold preference toward PKa over FXIa, thrombin, trypsin, and plasmin.For FXIa, this selective behavior could reflect the difference in structure of the P4 sites between PKa and FXIa with Tyr174 in PKa being substituted for a Glu residue in the case of FXIa, 45 thus disenabling a key π-stacking interaction in the P4 binding site.PPACK 23, was shown to have timedependent inhibition against thrombin, trypsin, and plasmin.
It was demonstrated that compounds 19−22 covalently inhibit PKa through analysis of time-dependent binding kinetics and through comparison of activity with homologues 24−27.Compound 20 displayed the most potent binding activity against PKa, showing picomolar activity at 24 h.In the compound dissociation assay, 20 demonstrated no apparent reversibility, suggesting that 20 may be a covalent inhibitor.Additionally, high PKa selectivity for 20 was observed against FXIa, thrombin, trypsin, and plasmin.
Molecular docking experiments were conducted to visualize the potential binding mode of 20.However, it soon became apparent that the pinacol boronate group was too large to fit into the Ser195 subpocket.Having previously experienced hydrolysis of the pinacol boronate during purification, model substrate 17b was used to study the stability of the pinacol boronate in PBS buffer by NMR (see Supporting Information Figure S1).17b was rapidly hydrolyzed to pinacol plus the boronic acid 35, in which 35 existed in equilibrium with the oxaborolane 34 (Scheme 6), an observation recently reported for the synthesis of aminoboronic acids as LONP1 inhibitor compounds. 46 docking study was performed on oxaborolane 36 and boronic acid 37, through docking into the Ala-195 mutant, followed by visualization of the docking results in PKa.Compound 37 showed interaction with Ser195, whereas 36 did not bind (Figure 8).
A docking study was performed with ketone 32 as assisted hydrolysis via ketone enolization, which would be disfavored  due to the presence of the electropositive β-boronate.In this case, docking results showed 32 bound to the PKa active site with the pinacol boronate exposed to solvent, giving further rationalization for the noncovalent interaction of 32 (see Supporting Information Figure S2).
The novel class of α-amidobenzylboronates is the first reported class of covalent inhibitors targeting plasma kallikrein.Their discovery adds to the available toolbox of compounds to study this important serine protease and represents a stepchange toward the design of covalent inhibitors targeting plasma kallikrein.We acknowledge their complex synthesis and purification.However, we feel that these preliminary studies add to the toolbox of available covalent warheads for future serine protease inhibitor design.

Figure 1 .
Figure 1.Schematic of (left) contact activation system (CAS) and (right) kallikrein−kinin system (KKS).Both initiated by factor XIIa, the cascade leading to bradykinin production and the positive feedback mechanism within the KKS are shown.

Figure 3 .
Figure 3. Examples of literature boron-containing compounds under FDA approval or in clinical development.

Figure 4 .
Figure 4.Chemical structure of the only known nonselective boronic acid containing PKa inhibitor compound 10.

Figure 5 .
Figure 5. (A) Truncation of 11 to the structure of interest 12. (B) Docking of 12 into PKa active site (PDB 6O1S) with key residues shown in bold.30(C) Overlay of docked structure of 11 (teal) with 12 (green) in PKa.Docking experiments were performed using OEDOCKING Hybrid docking.32

Figure 6 .
Figure 6.Extended time course for 20 against plasma kallikrein.ΔF = change in fluorescence, and RFU = relative fluorescence units.

Figure 8 .
Figure 8. Docking of boronic acid 37 into the PKa active site (Ala-195 mutant).Docking results were subsequently visualized in PKa.Docking experiments were performed using OEDOCKING Hybrid with results visualized in PKa active site (PDB 6O1S) using the PyMOL Molecular Graphics System, Version 2.0 Schrodinger, LLC.

Table 1 .
Biological Activity of Compounds 15a, 17a, and 18 against Plasma Kallikrein and FXIIa IC 50 (μM) a a Reagents and conditions: a) carboxylic acids 16a−c HATU, DIPEA, MeCN, 0 °C to rt, 3 h; for 16b, see ref 36; for 16c, see ref 9. 20 (0.33 nM at 60 min preincubation time).Interestingly, 20 had no effect on the inhibition of FXIIa activity over similar time course experiments, showing a propensity for high PKa selectivity.Compound 23 is a known nonselective covalent inhibitor of PKa 37 and was selected as a covalent control for method validation.
1hour time course experiments for 24−27, indicating reversible, noncovalent inhibitors of PKa.Compound 24 showed no PKa inhibition, whereas 25 and 26 showed comparable potency of 278 and 196 nM after 1 h preincubation time, respectively.Compound 27 showed the highest activity of this series, showing an IC 50 of 35 nM after a 1-hour preincubation (Table 2).Overall, this study showed that 24−27 exhibit significantly lower biological activity against PKa in comparison to 19−22, as well as demonstrating noncovalent binding over the 1-hour time course experiments, providing further evidence that compounds 19−22 bind covalently to PKa.

Table 2 .
Biological Activity of Compounds 19−27 and 29− 32 against Plasma Kallikrein PKa IC 50 (nM) a Data are expressed as mean of 2 experiments, where each experimental curve was performed in triplicate. a