Optimization of Ketobenzothiazole‐Based Type II Transmembrane Serine Protease Inhibitors to Block H1N1 Influenza Virus Replication

Human influenza viruses cause acute respiratory symptoms that can lead to death. Due to the emergence of antiviral drug‐resistant strains, there is an urgent requirement for novel antiviral agents and innovative therapeutic strategies. Using the peptidomimetic ketobenzothiazole protease inhibitor RQAR‐Kbt (IN‐1, aka N‐0100) as a starting point, we report how substituting P2 and P4 positions with natural and unnatural amino acids can modulate the inhibition potency toward matriptase, a prototypical type II transmembrane serine protease (TTSP) that acts as a priming protease for influenza viruses. We also introduced modifications of the peptidomimetics N‐terminal groups, leading to significant improvements (from μM to nM, 60 times more potent than IN‐1) in their ability to inhibit the replication of influenza H1N1 virus in the Calu‐3 cell line derived from human lungs. The selectivity towards other proteases has been evaluated and explained using molecular modeling with a crystal structure recently obtained by our group. By targeting host cell TTSPs as a therapeutic approach, it may be possible to overcome the high mutational rate of influenza viruses and consequently prevent potential drug resistance.


Introduction
The influenza virus is the causative agent of an infectious disease affecting the respiratory tract.[4] Influenza viruses belong to the Orthomyxoviridae family and possess a segmented genome consisting of negative-sense single-stranded RNA.These viruses are divided into four genera according to their antigenic properties: A, B, C, and D. [5,6] Considered one of the most common and virulent, the type A virus is further classified into several subtypes based on 18 hemagglutinin (HA) and 11 neuraminidase (NA) variants. [4]otably, specific subtypes of influenza A viruses have caused global pandemics associated with more severe symptoms and increased mortality rates, the most deadly being the 1918 H1N1 pandemic (Spanish flu), with up to 50 million deaths, and the most recent being the 2009 H1N1 pandemic known as the Swine flu, causing 280,000 deaths. [7]accination is currently the first-choice treatment to limit influenza infection, but antigenic drift forces constant vaccine updates and can lead to poor immunity if medical authorities fail to predict dominant variants or when new strains emerge.Thus, developing and implementing alternative therapeutic strategies is crucial.Different classes of direct-acting antivirals (DAA) targeting influenza A viral proteins have been approved by the U.S. Food and Drug Administration for the treatment of infection: NA inhibitors (zanamivir, oseltamivir, peramivir), M2 channel inhibitors (amantadine, and rimantadine), and more recently a molecule (baloxavir) targeting the viral cap-dependent endonuclease. [4,8]Inhibition of NA prevents the release of viral particles bound to sialic acids terminally linked to glycoproteins and glycolipids expressed on the cell surface.In the case of M2 inhibition, it hinders the uncoating of the virus following endosomal acidification and induces a premature conformational conversion of HA during its transportation to the cell surface. [9]Finally, endonuclease inhibitors limit viral mRNA synthesis and genome replication within the host cell.Although many therapeutic strategies are still used successfully, influenza viruses have acquired resistance against all current drugs with varying degrees due to their high mutation rates.Hence, several mutations have been found for NA, M2 channels, and other viral genes that help the virus escape treatment and induce drug resistance. [4,10,11]This acquired resistance against current treatments and the constant threat of new variants' emergence highlight the importance of discovering novel antiviral targets and drug candidates as a future response to seasonal flu and potential pandemics.
The cleavage and activation of the viral surface glycoprotein precursor HA0 into HA1 and HA2 by host cell proteases is a required step in the maturation of influenza viruses. [12]This cleavage exposes the otherwise hidden fusion peptide in HA2, which subsequently inserts into cell membranes following endosome acidification, thus allowing the eventual release of the viral genetic material in the host cell.This process represents an opportunity to develop a host-directed antiviral, which could circumvent the development of resistance observed with DAAs.The host proteases cleaving HA0 recognize monobasic or multibasic sequences in a loop between HA1 and HA2.[14][15] On the other hand, host trypsin-like proteases cleave monobasic sequences (X-X-X-R/K-#X) found in HA of low pathogenicity human influenza A viruses (IAVs) such as H1N1, H2N2, and H3N2. [12]Amongst those proteases are some members of the type II transmembrane serine protease (TTSP) family, comprised of 18 distinct human enzymes. [16]These proteases share a catalytic domain with similar specificity expressed at the plasma membrane of cells.[18] Reports have shown that TTSP members such as TMPRSS2, [19,20] TMPRSS4, [21] TMPRSS13, [22,23] HAT, [24] DESC1, [23] and matriptase [25,26] can cleave and activate HA0, thereby facilitating influenza cell entry, at least in vitro.][32][33][34][35][36] Here we performed a structure-activity relationship study using the tetrapeptide RQAR-Kbt (IN-1) as a protease inhibitor scaffold, matriptase as a prototypical TTSP, and the H1N1 influenza A strain as the target to optimize compounds with antiviral activities.

Results and Discussion
In our previous work, we identified the tetra-peptidomimetic molecule RQAR-Kbt (IN-1, aka N-0100), based on the matriptase autocatalytic sequence, as a highly potent inhibitor of matriptase (K i 0.011 nM). [31]However, this compound only moderately reduced H1N1 replication in the human pulmonary Calu-3 cell line (EC 50 5.6 μM) [25] and we hypothesized that this was partly due to the low in cellulo metabolic stability.Using the structure of IN-1 as a starting point for the development of more efficacious influenza inhibitors, we implemented the following modifications as depicted in Figure 1, to generate new compounds and potentially improve stability: (1) introduction of unnatural amino acids to modulate affinity; (2) N-alkylation of the amide bonds to explore the importance of backbone amides on inhibition; and (3) modifications of the N-terminus.

Chemistry
The synthesis of tetrapeptide derivatives (compounds 15 a-m) was performed using arginine-hydroxybenzothiazole and the corresponding tripeptides, following the procedure described previously. [31,35]Briefly, the tripeptides are constructed on a solid phase using 2-chlorotrityl chloride resin, which allows for cleavage while retaining the amino acid side chain protecting groups.The amine group of the warhead is then coupled in solution to the tripeptide carboxylate using the standard HATU coupling procedure.Finally, the alcohol of the warhead is oxidized to the corresponding ketone to form the serine trap, and the protecting groups are removed under acidic conditions.The warhead can be synthesized from commercially available benzothiazole and Boc/Pbf bisprotected arginine.The preparation of desaminoarginine 3 (Scheme 1) started with 5-aminovaleric acid.The carboxylic acid was protected as a methyl ester under acidic conditions to give intermediate 1.
Then, guanidinylation was carried out using bis-Boc-thiourea in the presence of triethylamine and mercuric chloride to generate intermediate 2. [37] Hydrolysis of the methyl ester with lithium hydroxide eventually yielded the desaminoarginine derivative 3 in three steps with a 64 % overall yield.Selective Nmethylation of amino acids was performed on 2-chlorotrityl chloride resin, following the method described in Scheme 2A, [38] inspired by Fukuyama et al.. [39] The method involved the activation of the primary amine by introducing an orthonitrobenzylsulfonyl group (o-NBS) in the presence of collidine to yield analog 5. Subsequent Mitsunobu reaction [39] using PPh 3 , DIAD, and MeOH gave intermediate 6. Denosylation with mercaptoethanol and DBU generated analog 7, which was coupled with the last Boc-protected amino acid in the presence of HATU, HOAt, and DIPEA.The tripeptide N-methylamide 8 was obtained after the cleavage reaction of the resin using HFIP in DCM.The functionalization of the intermediate N-tripeptides with a capping group was carried out on the 2-chlorotrityl chloride resin, as shown in Scheme 2B.Acetylation of the Nterminus was performed using an acetic acid solution in the presence of HATU and DIPEA, while sulfonylation of the Nterminus required MsCl or PhSO 2 Cl with HATU and DIPEA.Tripeptides with a capping group on the N-terminus were cleaved from the resin by acidolysis.

Structure-activity relationship (SAR) of matriptase inhibitors
To gain insights into the molecular determinants behind peptidomimetic protease binding and identify potential aspects for developing better compounds while maintaining or improving enzymatic inhibitory activity, we conducted SAR studies using matriptase as the target enzyme.Despite different TTSPs being implicated in influenza replication, we chose matriptase as the primary target based on our results demonstrating that both siRNA knockdown of matriptase and pharmacological inhibition with IN-1 can reduce influenza infection. [25][42] Previous studies indicated that matriptase and other TTSPs cleave substrates with aromatic residues (phenylalanine, tyrosine) in the P2 position. [43]Moreover, we demonstrated that tetrapeptides RQFR-Kbt (IN-2) and RQYR-Kbt (IN-3) are potent matriptase inhibitors (Table 1). [35]These data highlighted the tolerance in the active site of matriptase for aromatic amino acids, suggesting the existence of hydrophobic interactions between Phe in P2 and Phe708 [99] (the corresponding standard chymotrypsin numbering is given in brackets throughout the text) in the S2 pocket of matriptase. [44]This represents a molecular space that could be explored for compound optimization regarding affinity and selectivity.
To better understand the SAR at the P2 position in matriptase inhibition, we introduced functionalized phenylalanine (Table 1).First, nitro-and halogen-substituents were introduced at the para-position of the phenylalanine ring.Compounds 15 a, 15 b, and 15 c displayed similar inhibition properties towards matriptase compared to IN-2, thus indicating that the presence of an electron-withdrawing group on the phenyl ring has a neutral effect on matriptase inhibition.Second, we explored the effect of incorporating an additional carbon on the side chain with homophenylalanine (hF) residue (compound 15 d).This yielded a more potent compound than IN-2 (K i 0.06 vs. 0.23 nM), supporting the importance of the side chain length of the P2 phenylalanine for compound efficiency, specifically in cases where the P2 substituent bears an aromatic pharmacophore.Using D-amino acids in peptidomimetics can increase resistance to enzyme degradation and offer opportunities to modulate compound selectivity. [45]Our previous works revealed that IN-1 (K i 0.011 nM), with l-amino acid at the P1 position, is 400-fold more potent than IN-4 (K i 4.6 nM), having D-amino acid at the same position (Table 2).To ascertain the importance of the stereochemistry at the P3 and P4 positions, l-amino acids were substituted by the corresponding D-amino acid, and their ability to inhibit matriptase was assessed in vitro (Table 2).
Position 3 (15 e) and 4 (15 f) showed higher tolerance to Damino acids compared to P1, with potency toward matriptase decreased by about 10-fold but still maintained at the subnanomolar scale.When D-isomers were introduced simultaneously at P3 and P4 (15 g), potency further decreased to a K i of 0.43 nM, 40-fold less compared to the parent IN-1, entirely composed of l-amino acids.These results suggest that although l-amino acid configuration is preferable for optimal inhibition at P3 and P4 positions, this is not as critical as in P1 position.Next, we investigated the importance of backbone amides by N-methylation at P3 and P4 positions.Compounds 15 h and 15 i exhibited decreased inhibition potency compared to IN-1 but were still potent, with K i values of 0.9 and 1.3 nM, respectively.This effect can be explained by steric hindrance generated by N-methyl, a conformational change, or a loss of hydrogen bonding in the active site of matriptase, emphasizing the  [31] . .
importance of the amide backbone on matriptase inhibition.Additionally, after various attempts at synthesis and purification, we were not able to obtain the N-methylated analogs and the corresponding stereoisomer at position 2 (D-amino acid).However, preliminary results with the obtained mixtures show that the compounds do not surpass those presented in Tables 2  and 3 (Data not shown).Finally, to determine the importance of the N-terminus on potency, we introduced different capping groups, such as acetyl and sulfonyl, or simply removed the alpha-amine of the N-terminal residue (Table 3).Compounds with acetyl (15 j), methylsulfonamide (15 k), and phenylsulfonamide (15 l) at the N-terminus showed slightly decreased potency with K i values of 0.17, 0.13, and 0.48 nM, respectively.Furthermore, removing the alpha-amine of the N-terminal residue led to a compound (15 m) with a K i value of 0.16 nM.Thus, similar to introducing D-amino acids or phenylamine derivatives, modification of the N-terminus had a negative but modest influence on matriptase inhibition, with compound potency kept in the sub-nanomolar range.This tolerance represents a good opportunity to potentially improve the compounds' pharmacokinetic or pharmacodynamic properties, reduce their polarity or enhance their proteolytic stability and selectivity.This is consistent with our previous studies with Nterminal modified compounds that maintained good in vitro potency against TTSPs and showed increased compound stability. [18,30]

Identification of novel pharmacological agents against influenza
To discover potential ketobenzothiazole-based antivirals, we explored the ability of peptidomimetic inhibitors to block multicycle replication of Influenza A/Puerto Rico/8/34 (PR8) H1N1 virus in a cell model.Bronchial human epithelial Calu-3 cells, which endogenously express HA-activating TTSPs, including matriptase, were used for this study. [25]The cells were infected with the PR8 H1N1 virus, and compounds were added at increasing concentrations.Viral titers were then determined in the supernatant of infected Calu- ).This is particularly true for the 15 k-m compounds, which showed a 20-to 60-fold improvement in antiviral potency while exhibiting reduced in vitro activity against matriptase.a] groups contributed to this increased antiviral efficacy in cellulo.
To verify this hypothesis, we compared the stability by incubating compounds 15 m and IN-1 in the presence of Calu-3 cells in a culture media for different times ranging from 0 to 48 h.Our data revealed that after 48 h, compound 15 m remained intact while IN-1 was entirely degraded at 24 h (Figure 2, Table S1), which can explain the 40-fold improvement of 15 m in cellulo IC 50 when compared to IN-1 (Table 4).Altogether, these results suggest that a combination of potent in vitro matriptase inhibition (in the low nanomolar range) and high stability could lead to effective antivirals.Peptidomimetics 15 k-m were the three most potent compounds to efficiently block multicycle PR8 H1 N1 replication in the Calu-3 cell model with IC 50s < 300 nM.Several TTSPs are involved in the cleavage and activation of hemagglutinin, which is essential for viral entry, and Calu-3 cells express several of them. [16,21,22,24,25,46]Designing selective matriptase inhibitors that avoid other TTSPs is known to be challenging since catalytic domains within this family are very similar. [41]Because TMPRSS2 was identified as a H1N1 influenza A virus activating protease, [20] we next tested the inhibition of these 3 compounds against the recombinant protease.To gain further insight into the selectivity of these compounds, we also tested their inhibition on two closely related HA-activating TTSP members, HAT and DESC-1, and also on potential off-targets such as thrombin, a protease of the coagulation cascade, and furin, a ubiquitous protease involved in viral activation, including highly pathogenic avian influenza A viruses [47] (Figure 3, Table S2).The results showed that, although more potent toward matriptase, ketobenzothiazole analogs potently inhibited other TTSPs.This is particularly true for TMPRSS2, with K is � 1 nM and to a lesser extent for HAT, with K is < 10 nM.Compounds were still reasonably potent against DESC1, with K is ranging from 43 to 184 nM.The three compounds were globally less active against thrombin (K is from 338 nM to 3.1 μM), while furin was poorly or not inhibited (K i of 6.3 μM for 15 k and above 10 μM for 15 l and 15 m).This is consistent with our previous studies using this class of inhibitors, being selective for TTSPs when compared to other families of proteases. [18,48]An in-depth analysis of activity against other relevant proteases would enable the completion of this encouraging but succinct selectivity profile to prevent potential adverse effects and provide valuable insights for designing novel compounds.Calu-3 human airway cells are known to endogenously express several TTSPs, [25] but the relative importance of each protease in facilitating influenza entry has not been clearly established.For instance, both matriptase [25] and TMPRSS2 [20] knockdown in those cells reduced H1N1 influenza infection.Moreover, it is possible that functional redundancy among these proteases could compensate for the impairment of a specific TTSP. [42]urrently, we do not know the bona fide in cellulo targets of the compounds developed herein but the observed reduction of viral replication is thought to result at least in part on matriptase inhibition, and most probably through a combined action on both TMPRSS2 and matriptase activity.We cannot rule out that the antiviral activity of the compounds is also due to the inhibition of other proteases, whose implication in infection requires validation, but nonetheless, in the context of influenza, where multiple proteases potentially facilitate viral entry, it could be beneficial to develop broad range inhibitors that can effectively abrogate the activity of not only one but several key TTSPs.One could envision the important deleterious effects of targeting host proteases instead of viral proteins per se.However, in the case of respiratory infectious diseases, temporary and airway-localized exposure to compounds using aerosols could limit those potential side effects.This is in agreement with a recent clinical study using the broad-range serine protease inhibitor aprotinin, which was shown to be effective against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and displayed no major adverse effects. [49]evertheless, identifying precisely both the relative contribution of host proteases to influenza infection and the bona fide target(s) of our compounds in the Calu-3 cell model is crucial to better understanding their antiviral activities.While these efforts are ongoing, our results show that using the prototypical HA-activating TTSP matriptase as a basis for antiviral development is a viable strategy that yielded potent inhibitors against H1N1.Further characterization and optimization will be needed to evaluate the impact of these inhibitors in appropriate in vivo models.

Molecular modeling
We conducted modeling studies to understand the interactions of compounds within protease catalytic pockets.The active site of matriptase contains a catalytic triad composed of His656 [57], Asp711 [102], and Ser805 [195] residues, which are essential for its enzymatic activity.We previously reported the matriptase crystal structure with Cpd-8 (see Supporting Information for structure), a ketobenzothiazole-based peptidomimetic TTSP inhibitor (PDB: 6N4T). [48]For molecular modeling studies, we increases the flexibility of the side chain and seems to be beneficial in promoting an additional Π-arene interaction.It should be noted that the steric clash between Phe708 [99] and Phe (P2) of IN-2 is compensated by a slight distortion of the backbone at P4 (Arg) and P3 (Gln).These observations support the experimental findings demonstrating a higher affinity of 15 d towards the target when compared to IN-2.
To highlight the importance of stereochemistry on inhibitory activity, we performed a molecular modeling study on In order to visualize the differences between matriptase and TMPRSS2 within the catalytic pocket and to comprehend the similarity between the K i values, molecular modeling was conducted.Despite having a 40 % identity and a 60 % similarity in the catalytic domain of the two enzymes, we noticed an almost identical secondary structure (Figure 5A).The primary distinction is in the residues near the ketobenzothiazole warhead, where four residues are incorporated into a loop (Y loop of Figure 5A).However, the extension of this loop doesn't appear to obstruct the ligand's approach.Although there are several significant differences (Figure 5B), the docking results with one of our best analogs (15 m) primarily indicate that the TMPRSS2 binding site can accommodate the ligands, and a covalent bond between the warhead and the serine of the catalytic triad is feasible.Compared to matriptase, two conformations, which are very close energetically, are observed (depicted in orange and yellow).Nevertheless, the binding mode is basically the same for both enzymes, however, the orange conformation is not possible in the case of matriptase due to the steric hindrance brought on by Gln783 [175], which is absent for TMPRSS2.This subtlety, along with other mutations, could, therefore, be exploited to optimize selectivity during the design of the next generation of inhibitors.
The compounds demonstrated potency against tested TTSPs but were found to be poor inhibitors of other proteases like thrombin and furin.Inhibiting thrombin should be avoided, as it is an essential part of the clotting process.Thrombin functions as a serine protease, transforming soluble fibrinogen into insoluble fibrin strands and catalyzes numerous other reactions related to coagulation. [50]To explain these discrepancies, modeling studies were conducted.Like other serine proteases, thrombin has an active site composed of a catalytic triad with three residues: His406 [57], Asp462 [102], and Ser568 [195].Unlike TTSPs, thrombin possesses an insertion loop, a structural element positioned above the active site, comprising Tyr410 . The insertion loop is critical in thrombin's function, acting as a lid that helps control access to the catalytic triad (Figure 6A).
Using the thrombin crystal structure (PDB ID: 4AYV), we performed a conformational search for compound 15 m (H)RQAR-Kbt to investigate its selectivity against matriptase and thrombin (Figure 6A).The simulation results clearly show that the binding site is narrower, and the ligand cannot expand.Specifically, the S1' binding subpocket on the warhead side is entirely covered by the insertion loop and Trp413 [À ].The underside of this subpocket is hydrophilic and sterically hindered by the presence of Lys415 [À ], which is unfavorable to the hydrophobic ketobenzothiazole warhead.Altogether, these molecular modeling results explain the selectivity of the ligands towards TTSPs, in agreement with the K i values obtained.
Furin is a protein that belongs to the subtilisin-like proprotein convertase family.It participates in numerous physiological and pathophysiological processes in the cell, including protein maturation, viral pathogenesis, various development processes, tumor progression, immune system functions or infectious diseases. [51]The active site of furin consists of a deep and narrow groove surrounded by several shallow pockets that interact with the side chains of the substrate or inhibitor (Figure 6B).Similar to TTSPs, the active site of furin contains four main subpockets: S1 to S4.The S1 pocket is the primary binding site that accommodates the substrate residue at the cleavage site and largely determines the enzyme's substrate specificity. [52,53]The alpha helix adjacent to the S2 pocket contains two arginine residues (Arg193 [À ] and Arg197 ).These significant unfavorable interactions can lead to a decrease in affinity when the ligand approaches, making it difficult to effectively bind the substrate and form the covalent bond.These findings help explain the kinetic parameters obtained with proteases and demonstrate the tight fitting of IN-1-derived compounds in the matriptase and TMPRSS2 catalytic pocket, which is not the case for other classes of serine proteases.

Conclusions
The synthesis of IN-1 (RQAR-Kbt) derivatives with unnatural amino acids or N-terminus modifications led to the discovery of potential antivirals against H1N1 infection.Among these, the most potent inhibitor (15 l) showed a 60-fold improvement with an IC 50 < 100 nM.While ketobenzothiazole analogs are more potent inhibitors for matriptase, they effectively inhibit other TTSPs, specifically TMPRSS2, which is also suspected to be involved in the entry mechanism of the influenza virus.These inhibitors also exhibited a good level of selectivity compared to other families of serine proteases (thrombin, furin) and could serve as a basis for the future development of efficient molecules in vivo against the Influenza A virus, similar to what has been demonstrated against SARS-CoV-2. [18]Importantly, this suggests that the development of inhibitors targeting TTSPs could lead to molecules capable of simultaneously preventing both influenza and SARS-CoV-2 infections.Cell surface TTSPs represent accessible pharmacological therapeutic targets, and we propose that rational drug design based on in vitro TTSP activity and in cellulo compound stability are important steps for the discovery of lead compounds.Molecular modeling has allowed us to rationalize results and explain selectivity, which will undoubtedly facilitate the design of upcoming molecules.Further optimization based on selectivity, stability, toxicity, and in vivo pharmacokinetic properties will be essential for obtaining effective anti-influenza therapeutics.Experimental Section Chemistry General.Amino acids and coupling reagents were obtained from Chem Impex International (USA) and used as received.All reagents and solvents were purchased from Sigma-Aldrich (Canada), Fisher Scientific (USA), or Alfa Aesar (USA).THF, DCM, and MeOH were dried using sodium benzophenone ketyl, P 2 O 5 , and magnesium, respectively.
Thin layer chromatography (TLC) was performed on silica gel 60 F254 plates from Silicyle Inc (Canada).The TLC plates were immersed in binary mixtures with solutions for the mobile phase: Hexane/EtOAc, EtOAc/MeOH, and DCM/MeOH.The ratio of the solutions was adjusted according to the products studied to obtain the correct polarity.UV-visible compounds were visualized by fluorescence (λ ex 254 nm) or by spraying a ninhydrin solution (20 g of ninhydrin in 600 mL of MeOH) to reveal the primary amines of amino acids or by using CAM solution (ammonium pentamolybdate 40 g and 1.6 g of cerium sulfate IV in 800 mL of a mixture of sulfuric acid/water (1 : 9, v/v)) to reveal aromatics, for example, the trityl group after heating.
The 2-chlorotrityl chloride resin, commonly used with a loading of 1.2 mmol/g, was purchased from Matrix Innovation (Canada).Reactions involving this resin were carried out either in 30-60 mL polypropylene cartridges equipped with sintered Teflon taps from Applied Separations (USA) or in 50-100 mL glass reactors from Chemglass Life Sciences (USA).
The characterization of intermediate and final products was performed using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS) from Waters Company (Canada).The UPLC-MS system was equipped with a UV detector set at an excitation wavelength (λ ex ) of 223 nm, a BEH C 18 column (2.1×50 mm, 1.7 μm), and a linear gradient of 5-95 % MeCN/H 2 O containing a fixed percentage of 0.1 % formic acid for 2.5 min.
The final products were purified using preparative HPLC from Waters (Canada) with an ACE C 18 column (250×21.2mm) and a gradient of 10-30 % MeCN/H 2 O + 0.1 % TFA for 30 min at a flow rate of 20 mL/min.All purified final compounds were obtained as the TFA salt after lyophilization.The final tetrapeptides obtained yielded 4-55 %, and all compounds were > 95 % pure by HPLC.
The molecular weights of compounds were confirmed by High-Resolution Mass Spectrometry (HRMS) using an electrospray micromass Maxis 3G from Waters. 1 H NMR spectra were recorded on a Bruker Ascend 400 MHz instrument (Bruker Inc., USA) referenced to internal solvent signals.

Synthesis of 5-(2,3-bis(tert-butoxycarbonyl)guanidino)pentanoic acid (3):
To a solution of compound 2 (1.0 equiv, 5.0 g, 13 mmol) in THF/H 2 O (50 mL), LiOH (5.0 equiv, 3 g, 65 mmol) was added.The reaction mixture was stirred for 3 h at room temperature.After completion of the reaction, the reaction mixture was concentrated under reduced pressure, then cooled to 0 °C, and a citric acid solution (1 M) was slowly added dropwise until the pH was 6-7.The solution was stirred for 30 min, filtered, washed with water, then dried in under high vacuum to give a white solid.Yield of 78 %; HPLC purity (UV) 99 %.NMR 1

Introduction of o-nitrobenzenesulfonyl protecting group (o-NBS):
A solution of o-nitrobenzenesulfonyl chloride (4.0 equiv) and collidine (10.0 equiv) in NMP was added to 1 g of peptide-resin (1.2 mmol/g).After mixing for 15 min, the resin was filtered and washed five times with NMP, using approximately 10 mL per gram of resin, and three times with DCM.The resin was dried under a vacuum.A small amount of resin was cleaved with 20 % HFIP in DCM for 10 min, and the o-NBS peptide was analyzed by UPLC/MS.

N-methylation using Mitsunobu reaction:
A solution of PPh 3 (5.0equiv) and MeOH (10.0 equiv) in dry THF was added to the o-NBS peptide fixed on the resin.Then, a solution of DIAD (5.0 equiv) in dry THF was slowly added, and the mixture was mixed for 10 min.The resin was filtered and washed five times with NMP, using approximately 10 mL per gram of resin, and three times with DCM.The resin was dried under a vacuum.A small amount of resin was cleaved with 20 % HFIP in DCM for 10 min, and the peptide was analyzed by UPLC/MS.

Deprotection of o-nitrobenzenesulfonyl protecting group:
The resin was treated with a solution of mercaptoethanol (10.0 equiv) and DBU (5.0 equiv) in NMP and mixed for 15 min.This deprotection condition was repeated one more time; then the resin was filtered and washed five times with NMP, using approximately 10 mL per gram of resin, and three times with DCM.The resin was dried under a vacuum.A small amount of resin was cleaved with 20 % HFIP in DCM for 10 min, and the peptide was analyzed by UPLC/MS.

Coupling reaction with HATU/HOAt:
A solution of Boc-AA-OH or Fmoc-AA-OH (3.0 equiv), HOAt (3.0 equiv), and DIPEA (6.0 equiv) in NMP was added to the resin and mixed for 3 h.The resin was filtered and washed five times with NMP, using approximately 10 mL per gram of resin, and three times with DCM.The resin was dried under a vacuum.A small amount of resin was cleaved with 20 % HFIP in DCM for 10 min, and the peptide was analyzed by UPLC/MS.

Acetylation of the N-terminus:
A solution of acetic acid (5.0 equiv) HATU (2.0 equiv), and DIPEA (5.0 equiv) in DMF was added to the resin.After 1 h, the resin was filtered and washed three times with DMF, three times with iPrOH, and three times with DCM using approximately 10 mL per gram of resin.The resin was then dried under a vacuum.

Sulfonylation of the N-terminus:
A solution of DIPEA (5.0 equiv), methyl sulfonyl chloride (2.0 equiv), or phenylsulfonyl chloride (5.0 equiv) in DCM (10 mL/g of resin) was added to the resin.After 1 h, the resin was filtered and washed three times with DMF, three times with iPrOH, and three times with DCM using approximately 10 mL per gram of resin.The resin was then dried under a vacuum.

General cleavage conditions of the chlorotrityl resin:
A solution of a 20 % HFIP solution in DCM (10 mL) was added to the resin and mixed for 30 min.The resin was then filtered and washed once with 20 % HFIP in DCM, followed by three washes with DCM using approximately 10 mL/g of resin.The filtrate was evaporated using a rotavapor and dried under a vacuum to yield the tripeptide.

Protein alignment and chymotrypsinogen Numbering System:
Human matriptase (Uniprot Q9Y5Y6), thrombin (Uniprot P0074) and furin (Uniprot P09958) amino acid sequences were aligned to the bovine chymotrypsinogen A sequence (Uniprot P00766) using the SnapGene version 7.0.2software.Corresponding chymotrypsinogen amino acid (or a dash (À ) when no correspondence is found) is indicated in brackets throughout the text.
In vitro inhibition assay: Active recombinant human matriptase, DESC1, and HAT were purified as previously described. [43]Active recombinant human TMPRSS2 (Cusabio), furin (Bio-Techne) and thrombin (MilliporeSigma) were obtained from commercial sources.Inhibitor dissociation constants (K i ) were determined as reported previously. [18,30,31]Briefly, proteases were incubated with different concentrations of the inhibitor in the presence of a fluorogenic substrate (Boc-RVRR-AMC for furin and Boc-QAR-AMC for the others), and proteolytic activity was monitored by measuring the release of fluorescence (λ ex 360 nm, λ em 460 nm) in a FLX800 TBE microplate reader (Bio-Tek Instruments, USA).Compounds were considered tight-binding inhibitors if substantial inhibition occurred at a ratio I/E = 10.In this case, for the K i determination, plots of enzyme velocity as a function of inhibitor concentration were fitted by non-linear regression analysis to a Morrison K i equation.If substantial inhibition occurred only at I/E > 10, compounds were treated as classical reversible inhibitors.Because the concentration of active TMPRSS2 could not be determined in the commercial preparation, K is were derived from the IC 50 dose-response curves using the Cheng-Prusoff equation for competitive reversible inhibitors, as published previously. [36,54]Non-linear regression and statistical analysis were performed using GraphPad Prism 9.5.1 for Windows.

Assay inhibition of replication of H1 N1 in bronchial epithelial cells Calu-3:
In 24-well plates, adenocarcinoma bronchial epithelial Calu-3 cells grown to confluence were infected with A/Puerto Rico/ 8/34 (H1N1; PR8) at a multiplicity of infection (MOI) of 0.003.Subsequently, the cells were incubated with the compounds at different concentrations.Viral titers were determined 48 h postinfection as detailed previously. [25]ability assay: Calu-3 cells were grown to confluence in 96-well plates.The cells were washed twice with PBS and then incubated in triplicate with 100 μM of the inhibitor in EMEM without FBS, supplemented with 0.1 % BSA.Cell media were collected at six time points (0, 30 min, 1 h, 6 h, 24 h, and 48 h).Acetonitrile (75 % in water) was added to the samples at a ratio of (1.11 : 1).Samples were vortexed for 10 seconds and then centrifuged at 13,000 × g for 20 min at 4 °C.The supernatant was collected and stored at À 80 °C.UV spectra of the samples were analyzed using a PDA eλ UV-Vis detector on a UPLC from Waters Company (Canada) equipped with a BEH SHIELD RP18 (2.1×100 mm, 1.7 μm) column.UPLC elution gradients are available in Supporting Information Table S1.The data represent one experiment performed in triplicates.
Molecular Modeling: Molecular modeling studies were conducted using the Molecular Operating Environment (MOE) software, version MOE2022.02.The 3D coordinates of matriptase, thrombin, and furin were obtained from the Protein Data Bank (PDB ID: 3NCL, 6 N4T, 4AYV, 4RYD) and loaded into MOE.Polar hydrogens and partial charges were added.For the protonation process, a temperature of 300 K, a concentration of 0.1 mol/L salt in the solvent, and a pH of seven were specified.Any missing atoms, alternate geometries, or other crystallographic artifacts were corrected using the QuickPrep function.Subsequently, the structures were energyminimized in the Amber12:EHT force field to achieve a root mean square (RMS) gradient of 0.1 kcal/mol.3D models were built for all synthesized compounds, and their energies were minimized to an RMS gradient of 0.1 kcal/mol using the Amber12:EHT force field.The structures were protonated, and partial charges were calculated to assign ionization states and position hydrogens in the macromolecular structure based on its 3D coordinates.Before conducting the conformational search, a covalent bond was created between the serine of the catalytic triad and the ketone of the ketobenzothiazole.Compounds The conformational search was performed on the ligand-receptor complex using the LowModeMD Search method of MOE. [55]This method involved a short molecular dynamics simulation using velocities with some kinetic energy on the high-frequency vibrational modes, which was appropriate for peptide conformations.The guanidine group of the arginine in P1 was constrained through key interactions with Asp at the bottom of the S1 pocket.
The conformational search using LowModeMD was conducted with AMBER12:EHT as the molecular mechanics force field, using default parameters (rejection limit: 100; RMS gradient: 0.05; conformation limit: 10,000; iteration limit: 10,000).Subsequently, a second round of energy minimization was performed around the ligand-binding site.
The low-energy conformations of the inhibitor-protein complexes were analyzed for their binding interactions.The generated conformation exhibiting the lowest potential energy values was selected for further analysis.The interactions of all molecules in the binding cavity and their geometry and orientation were analyzed.

Supporting Information Summary
The Supporting Information contains stability parameters, K i values, chemistry data (weight, yield, purity, HRMS data), structure of compound 8, and supplementary figures ( 1 H NMR, UV, and MS spectra) of compounds described herein.

Figure 2 .
Figure 2. Stability of analogues IN-1 and 15 m in the presence of Calu-3 cells.Compounds were incubated for the indicated time on cells and extracellular media was analyzed by UPLC coupled with a UV/Vis detector.

Figure 3 .
Figure 3. Specificity of selected compounds toward serine proteases.Data are the mean of log(K i ) with n � 3 and are represented as a heat map.Related K is are found in S2 table.

Figure 4 .
Figure 4. Docking of IN-1 to matriptase A) Interaction between IN-1(purple) and matriptase (grey and cyan) B) Comparison of the docking mode of IN-1 (purple) with Cpd-8 (yellow) in the binding pocket of matriptase.C) Comparison of the orientation of IN-2 (white) and 15 d (red) into the binding site of matriptase (cyan).D) Comparison of the orientation of compound IN-1 (purple) and IN-4 (green) in the binding site of matriptase (cyan).Distance between Ser805 [195] of the catalytic triad and the ketone of the warhead is 2.74 Å.
compound IN-4 (RQAr-Kbt), where the corresponding D-amino acid replaced the L-amino acid at the P1 position.Substitution of the P1 Arg residue with its dextrogyre form had a deleterious effect on the inhibitory activity of IN-4 (K i 4.6 nM), which is 400fold less potent than IN-1 (K i 0.011 nM).According to our molecular docking of IN-4 in the matriptase active site, the guanidine group of D-Arg at P1 could locate itself properly in the S1 pocket.However, the drastic conformational change interferes with the formation of a covalent bond between the warhead and Ser805 [195] of the catalytic triad, which could explain this major loss of inhibition (Figure 4D).IN-4 warhead's electrophilic ketone is distanced from Ser805 [195] (2.74 Å for IN-4 compared to 1.5 Å for IN-1), hindering the potential Ser805 [195] nucleophilic attack, which is essential to obtain the reversible-covalent form of inhibition.

Figure 5 .
Figure 5.Comparison between matriptase and TMPRSS2 A) Overlay of the secondary structures of matriptase (blue) and TMPRSS2 (gray) reveals only two significant differences near the binding site (red square) and another distinct region farther from the binding site ( yellow square).B) Matriptase (cyan) and TMPRSS2 (purple) exhibit 15 unique amino acids differences near the binding site.Residues 661 [À ] to 664 [60] are not found on TMPRSS2 and represent the Y loop in panel A. The molecules in panel A and B represent the two lowest energy conformations obtained for the docking of 15 m to TMPRSS2, the yellow being the lowest.
[À ]), contributing to a positively charged and highly hydrophilic region.Upon examining the placement of the low-energy conformation of peptidomimetic 15 m in the furin active site, we observed that the arginine at the P1 position was indeed situated within the S1 pocket.However, as depicted in Figure 6B, the hydrophobic ketobenzothiazole warhead in compound 15 m is located in a very hydrophilic pocket surrounded by three basic amino acids (Arg193 [À ], Arg197 [À ], His364 [140]), and two other polar residues (Thr365 [141] and Ser363 [139]

Figure 6 .
Figure 6.Docking simulation between 15 m (orange) and A) thrombin (PDB ID: 4AYV, of the thrombin insertion loop in white) or B) furin (PDB ID: 4RYD, residues in the S1' pocket in purple and catalytic triad in cyan).
15 m, 15 d, IN-1, IN-2, IN-3, and IN-4 were selected for molecular modeling studies based on the experimental assay results.

Table 1 .
SAR of tetrapeptides with Phe derivatives at the P2 position.
i values are the means � standard deviations of at least three independent experiments.[b] K i previously published in Colombo et al.

Table 4
3 cells by viral plaque assays at 48 h post-infection, and IC 50 values were determined ( ).Our data revealed that phenylalanine derivatives 15 a-c inhibit H1 N1 virus replication with IC 50 values above 1 μM, while 15 d showed better potency with an IC 50 value of 292 nM, a substantial improvement compared to the reference compound IN-1 (5.6 μM).Diastereoisomers 15 e and 15 f, with D-Gln and D-Arg at the P3 and P4 positions, displayed IC 50 values of 1093 nM and 391 nM, respectively, and the combination of both substitutions led to a slightly reduced IC 50 of 2100 nM (15 g).N-methylated derivatives 15 h and 15 i were

Table 3 .
SAR of tetrapeptides with N-terminal modifications at the P4 position.K i values are the means � standard deviations of at least three independent experiments.[b] K i previously published in Colombo et al.