Thiophene-fused γ-lactams inhibit the SARS-CoV-2 main protease via reversible covalent acylation

Enzyme inhibitors working by O-acylation of nucleophilic serine residues are of immense medicinal importance, as exemplified by the β-lactam antibiotics. By contrast, inhibition of nucleophilic cysteine enzymes by S-acylation has not been widely exploited for medicinal applications. The SARS-CoV-2 main protease (Mpro) is a nucleophilic cysteine protease and a validated therapeutic target for COVID-19 treatment using small-molecule inhibitors. The clinically used Mpro inhibitors nirmatrelvir and simnotrelvir work via reversible covalent reaction of their electrophilic nitrile with the Mpro nucleophilic cysteine (Cys145). We report combined structure activity relationship and mass spectrometric studies revealing that appropriately functionalized γ-lactams can potently inhibit Mpro by reversible covalent reaction with Cys145 of Mpro. The results suggest that γ-lactams have potential as electrophilic warheads for development of covalently reacting small-molecule inhibitors of Mpro and, by implication, other nucleophilic cysteine enzymes.


Introduction
g-Lactams are common in bioactive natural products, 1 including e.g. in anantine and derivatives, 2,3 monascuslactams A-D, 4 the proteasome inhibitors lactacystin [5][6][7] and salinosporamide A, 8 and clausenamide. 9They are also present in clinicallyused therapeutics, for example in the antiemetic rolapitant, 10 the respiratory stimulant doxapram, 11 piracetam, which is used to treat cortical myoclonus, 12 the anti-cancer drug ivosidenib, 13 and the antivirals nirmatrelvir (1) 14 and simnotrelvir (2) (Fig. 1). 15][19][20] Small-molecule inhibitors which acylate their target enzyme(s) via covalent reaction of a g-lactam group with a nucleophilic residue have, to our knowledge, not yet been approved for therapeutic use.The lack of therapeutics which employ a g-lactam as an electrophilic warhead for covalent reaction is remarkable, considering (i) the revived interest in the development of covalently reacting small-molecule therapeutics, 25,26 (ii) recent advances in the synthesis of g-lactams, 1,27,28 (iii) the demonstrated safety of active pharmaceutical ingredients which contain a g-lactam group that does not acylate the target enzyme, e.g., nirmatrelvir (1) 14 and simnotrelvir (2), 15 and, in particular, (iv) that many clinically used small-molecules employ a b-lactam as an electrophilic warhead to acylate their target enzyme(s), e.g., penicillin-and cephalosporin-based antibiotics. 29At least in part, this gap may reect the reduced intrinsic reactivity of g-lactams compared to more strained b-lactams based on (non-enzymatic) ring closure rates. 30he covalent reaction of g-lactams with nucleophilic serine enzymes is reported; [31][32][33][34][35][36][37][38][39] however, their analogous reactivity with nucleophilic cysteine enzymes, many of which are contemporary drug targets, 25,26 is, to our knowledge, unknown.SARS-CoV-2 M pro appears to be a suitable target to investigate the reactivity of g-lactams with nucleophilic cysteine enzymes, because many small-molecule inhibitors are reported which employ an electrophilic group for reversible or irreversible covalent reaction with the nucleophilic thiolate of the catalytic cysteine residue of M pro (i.e., Cys145, Fig. 1) and because of the important structural roles of a g-lactam group in many reported substrate-derived M pro inhibitors, 40,41 e.g., nirmatrelvir (1) 14 and simnotrelvir (2). 15The g-lactam group of these inhibitors binds in the S1 subsite of M pro , that is proximate to Cys145 (Fig. 1).
Of M pro inhibitors that react covalently, those that react reversibly may be preferred over those that react irreversibly, as the latter may also react with 'off-targets' in an irreversible manner.Indeed Cys145 of M pro reacts reversibly with the nitrile group of the clinically-used drugs 1 and 2. 14,15,42 Many investigational M pro inhibitors, however, employ highly reactive electrophiles for covalent reaction with Cys145, including e.g., aldehydes, a-ketoamides, and Michael acceptors, 19,43,44 which may potentially compromise safety, as reported for some clinically-used small-molecules bearing reactive electrophiles; 26,45 The use of electrophilic groups with relatively low intrinsic reactivity is thus desirable.The observation that the glactam of both 1 and 2 is stable in cells 14,15 likely reects its reduced reactivity compared to more reactive electrophiles, indicating that covalently reacting g-lactams may have potential for development of safe COVID-19 therapeutics.However, by contrast with b-lactams, 21,22 g-lactams have, to our knowledge, not yet been considered as electrophilic warheads for covalent reaction with M pro Cys145.
During the course of investigations aimed at developing penicillin-based M pro inhibitors which acylate Cys145 via blactam ring opening (e.g., 3 and 4), 21,22 we synthesized the glactam analogue 5 to probe the effect of altering the lactam group on potency, including with respect to reversibility of acylation.Consistent with studies revealing that acylation of nucleophilic serine residues is more reversible with a g-lactam compared to an analogous b-lactam, 46,47 g-lactam 5 inhibits isolated recombinant SARS-CoV-2 M pro ∼4-fold less efficiently than the structurally-related b-lactam 3. 22 Mass spectrometric analyses indicated that, by contrast with b-lactam 3, g-lactam 5 did not react to form a stable acyl-enzyme complex, suggesting that it may bind to M pro principally via non-covalent interactions. 22ere we report the synthesis of thiophene-fused g-lactams which inhibit isolated recombinant SARS-CoV-2 M pro more efficiently than b-lactam 3 and g-lactam 5 (Fig. 1).Mass spectrometric evidence supports a mechanism involving reversible covalent reaction of the g-lactam group with Cys145.The results reveal bicyclic g-lactams are useful scaffolds for the inhibition of nucleophilic cysteine enzymes.

Results
Thiophene-fused g-lactams inhibit SARS-CoV-2 M pro Natural product inspired trans-ring-fused g-lactams can inhibit serine proteases via acylating their nucleophilic serine residue, 31,32 as is also the case for thiophene-fused g-lactams. 33n the latter case, it is proposed that, following acylation, the presence of the thiophene ring hinders deacylation by sequestering electron density on the g-lactam-derived amine. 33Based on the proposal that a hydrophobic thiophene ring may bind in the hydrophobic S2 pocket of M pro that is in proximity of Cys145 (Fig. 1), we synthesized an initial set of thiophene-fused g-lactams (7-9) following modications of reported procedures. 33][50][51][52][53][54][55][56] Analysis of the half-maximum inhibitory concentrations (IC 50 -values) revealed that both the regioisomeric thiophenefused g-lactams 7 and 8 moderately inhibited isolated recombinant SARS-CoV-2 M pro with similar potencies (IC 50 ∼ 8.5 and 8.4 mM, respectively; Table 1, entries i and ii).By contrast, the regioisomeric g-lactam 9 did not inhibit M pro over the tested concentration range (Table 1, entry iii), showing that the position of the thiophene sulfur atom with respect to the g-  23 (G) view of the surface from the reported SARS-CoV-2 M pro :1 complex structure active site (PDB ID: 7TE0 24 ), revealing that the g-lactam group of 1 binds in the S1 subsite, whereas its bicyclic leucine mimic binds in the S2 subsite, its tert-butyl group is solvent exposed, and its trifluoroacetamide group binds in the S4 subsite.g-Lactam, b-lactam, and pyrazolidinone groups are in green, blue, and ochre, respectively.Bn: benzyl; Ph: phenyl.lactam nitrogen atom affects inhibition potency.The substitution of the methylene group of 9 with an NSO 2 Me group to give 10 did not result in inhibition (Table 1, entry iv).To investigate the effect of the thiophene ring of 7 and 8, we synthesized the corresponding benzene-fused g-lactam 11 using the route employed for synthesis of 7 and 8. 11 did not inhibit M pro over the tested concentration range (Table 1, entry v), indicating that the size of the ring fused to the g-lactam, nature of delocalization, and/or the presence of a sulfur atom in that ring are important for the inhibition manifested by 7 and 8.
The M pro inhibition potency of g-lactams 7 and 8 is in the range of that reported for penicillin V sulfone benzyl ester 3 (IC 50 ∼ 6.6 mM; 22 Fig. 1), which inhibits M pro via covalent reaction of its b-lactam with the active site Cys145. 22Notably, glactams 7 and 8 inhibit M pro ∼3-fold more efficiently than the reported g-lactam 5 (IC 50 ∼ 26.1 mM; 22 Fig. 1), which inhibits M pro apparently via non-covalent interactions, 22 and ∼5-fold more efficiently than the reported pyrazolidinone 6 (IC 50 ∼ 45 mM; 23 Fig. 1).

Substitution affects the inhibition potency of thiophene-fused g-lactams
Structure activity relationship studies were performed to investigate whether the lactam nitrogen substituent and substituents a to the lactam carbonyl affect inhibition potency.Derivatives of g-lactam 7 with a single a substituent were synthesized from commercially-sourced 2-nitrothiophene ( 12) in 5 steps following modication of reported procedures (Scheme 1A). 33Initially, 12 was efficiently alkylated with ethyl chloroacetate to give thiophene 13 as a single regioisomer.13 was alkylated using an alkyl halide with Cs 2 CO 3 as a base to give thiophenes 14a-d and 14f; 14e was synthesized from 13 using catalytic amounts of 1,1,3,3-tetramethylguanidine as a base and acrylonitrile as a Michael acceptor, as reported for related nitriles. 57Nitrothiophenes 14a-f were converted to the corresponding sulfonamides 15a-f following nitro-reduction, sulfonylation, and saponication.g-Lactams 16a-f were obtained from 15a-f via HATU 58 -mediated amide bond formation.Derivatives of g-lactam 7 which bear two identical a substituents, i.e., 17a and 17b, were directly synthesized from 7 via an alkylation reaction (Scheme 1B).
The M pro inhibition results reveal that the addition of a methyl group a to the carbonyl of g-lactam 7 increases inhibition potency by ∼2-fold, whereas the addition of a second methyl group ablates inhibition (Table 2, entries i and ii).The length of the alkyl substituent a to the g-lactam carbonyl did not appear to substantially affect potency: 16b, which bears a propyl a Assays were performed as reported using SPE-MS, employing SARS-CoV-2 M pro (0.05 mM) and substrate peptide (2.0 mM). 42Results are means of two independent runs each composed of technical duplicates (n = 2; mean ± standard deviation, SD).Representative dose-response curves of selected g-lactams are shown in Fig. 2. substituent a to the lactam carbonyl, inhibited isolated recombinant SARS-CoV-2 M pro with similar potency as 16a which bears a methyl group at the same position (Table 2, entries i and iii).By contrast, isomeric isopropyl-substituted glactam 16c inhibited M pro ∼5-fold less efficiently than propylsubstituted 16b (IC 50 ∼ 16 mM; Table 2, entry iv).However, sterically bulky substituents are not necessarily detrimental for efficient inhibition, since benzyl substituted g-lactam 16d inhibited with similar potency as 16a and 16b (IC 50 ∼ 4.0 mM; Table 2, entry v).g-Lactams 16e and 16f which are both derived from propylsubstituted g-lactam 16b, but which contain a nitrile or olen, respectively, in their alkyl substituent, inhibited M pro with similar potency to 16b (IC 50 ∼ 2.7 and 3.5 mM, respectively; Table 2, entries vi and vii).Similar to a-disubstituted g-lactam 17a, 17b which bears two allyl substituents a to its lactam carbonyl did not inhibit M pro (Table 2, entry viii).Both the spiro g-lactam 18, which was synthesized from 17b via a ring-closing metathesis, 60 and the isomeric g-lactam 19, which was synthesized from g-lactam 8 following an analogous synthesis route (ESI †), did not efficiently inhibit M pro (Table 2, entries ix and x).
The effect of varying the g-lactam nitrogen substituent of 7 on inhibition potency was investigated (Table 3).Derivatives of 7, i.e., 23a and 23c-f, were synthesized from commerciallysourced 2-(thiophen-3-yl)acetic acid (20) in 3 steps, employing copper-catalysed reaction of thiophene bromide 21 with activated amines (Scheme 2). 61,62g-Lactam 23b was synthesized from thiophene 13 in a similar manner to which 7 was prepared (ESI †).Substituting the methylsulfonyl group of g-lactam 7 for an acetyl group ablated M pro inhibition (Table 3, entry ii), whereas use of benzylsulfonyl or phenylsulfonyl groups apparently increased potency by ∼2-fold (Table 3, entries iii and iv).In some cases, the addition of substituents on the phenyl ring of 23c para to the sulfonyl group further increased potency (Table 3, entries v-vii); the CF 3 -substituted g-lactam 23f appeared to be a particularly potent M pro inhibitor, being ∼3and ∼6-fold more potent than g-lactams 23c and 7, respectively (IC 50 ∼ 1.3 mM; Table 3, entry vii).Notably, g-lactam 23f inhibits M pro ∼20-fold more efficiently than our reported non-covalently reacting glactam M pro inhibitor 5 and ∼6-fold more efficiently than our reported covalently reacting b-lactam M pro inhibitor 3 (Fig. 1).a Assays were performed as reported using SPE-MS, employing SARS-CoV-2 M pro (0.05 mM) and substrate peptide (2.0 mM). 42Results are means of two independent runs each composed of technical duplicates (n = 2; mean ± SD).Representative dose-response curves of selected g-lactams are shown in Fig. 2.
Thiophene-fused g-lactams inhibit M pro via reversible covalent reaction Protein-observed MS studies under denaturing conditions were performed with selected synthetic g-lactams to investigate whether they inhibit isolated recombinant SARS-CoV-2 M pro via non-covalent interactions, as for g-lactam 5, 22 or via covalent reaction.The results reveal that some of the tested g-lactams covalently react with M pro during the tested time period (i.e., 4 h), as shown by the anticipated mass shis (Fig. 3); the stoichiometry of the covalent reaction appears to be, at least predominantly, 1 : 1, suggesting that g-lactams react selectively with a single nucleophilic M pro residue, likely Cys145.Nonetheless, the MS studies imply that some of the g-lactams, i.e., 16a, 16b, and 23d, may have capacity to covalently react with M pro residues other than Cys145, albeit at substantially lower levels even when being used in excess; note that M pro has eleven cysteine residues in addition to Cys145, all of which can covalently react with non-specic inhibitors such as ebselen. 49ariable levels of M pro acylation were observed depending on the g-lactam employed, suggesting that initial binding constants, reaction rates, and/or stabilities of the acyl-enzyme complex differ depending on the substitution pattern.Notably, complete M pro acylation was not observed under the tested conditions, an observation which may reect the reversibility of the reaction and the comparatively low enzyme to g-lactam ratio employed in the assay (i.e., 1 : 5); this ratio was chosen to avoid g-lactam-induced ionization suppression of M pro observed at higher g-lactam concentrations, thus perturbing data analysis.
To localize the site of covalent modication and to probe whether covalent reaction occurs with Cys145, 16a was incubated with M pro that had been previously reacted with a smallmolecule inhibitor 42 that selectively and irreversibly reacts with Cys145 (Fig. 4); 16a was used for this study because its levels of M pro acylation were apparently higher than those for the unsubstituted 7 and 8, and because it has the least bulky substituent amongst those g-lactams that covalently react with M pro , a property which may favour more efficient covalent reaction.We have reported that an alkyne derivative of nirmatrelvir (24; Fig. 4), in which the electrophilic nitrile is substituted for an alkyne, reacts selectively and irreversibly with Cys145. 42Thus, isolated recombinant SARS-CoV-2 M pro was incubated with a ∼16-fold excess of the alkyne derivative of 1 (i.e., 24) to block the thiol of Cys145 by stoichiometric thioenol ether formation (Fig. 4A). 42The excess of 24 was removed by washing and the resultant covalent M pro :24 complex was incubated with 16a.The results reveal that 16a does not covalently react with the covalent M pro :nirmatrelvir alkyne (24) 42 complex within the tested time (i.e., 2 h).g-Lactam 16a thus likely reacts selectively under the tested conditions with Cys145, but not, at least substantially, with other surface-exposed cysteine residues of M pro .
It was of interest to investigate whether the synthetic g-lactams react reversibly with Cys145.Thus, g-lactam 16a was  a Assays were performed as reported using SPE-MS, employing SARS-CoV-2 M pro (0.05 mM) and substrate peptide (2.0 mM). 42Results are means of two independent runs each composed of technical duplicates (n = 2; mean ± SD).Representative dose response curves of selected g-lactams are shown in Fig. 2.

Discussion
Penicillins and related antibiotics inhibit bacterial cell wall biosynthesis via covalent reaction of their electrophilic b-lactam ring with the nucleophilic serine residue of transpeptidases to give stable acyl-enzyme complexes. 680][71][72][73][74][75] Subsequently, 1,6-diazabicyclo [3.2.1]octane-based compounds (DBOs), [76][77][78][79] including Fig. 4 Evidence that g-lactam 16a inhibits M pro by selective reversible covalent reaction with Cys145.(A) g-Lactam 16a does not covalently react with the covalent M pro :nirmatrelvir alkyne derivative 24 42 complex obtained via irreversible covalent reaction of M pro Cys145 with 24, 42 indicating that g-lactams selectively react with the nucleophilic thiolate of Cys145 under the tested conditions; (B) addition of an excess of 24 42 to a mixture containing the covalent M pro :16a complex results in stoichiometric formation of the corresponding covalent M pro :24 42 complex, substantial levels of the M pro :16a complex were not detected using SPE-MS implying that the reaction of g-lactams with M pro is reversible and/or that the acyl-enzyme complex is not stable towards hydrolysis.Assays were performed using SPE-MS, as described in the Experimental section, employing SARS-CoV-2 M pro (3.0 mM), and, if appropriate, g-lactam 16a (15 mM) and/or nirmatrelvir alkyne 24 42 (50 mM) in buffer (20 mM HEPES, pH 7.5).Representative spectra of technical duplicates are shown.
Our combined results imply that g-lactams have potential to be useful covalently reacting inhibitors of nucleophilic cysteine enzymes, in particular SARS-CoV-2 M pro .They reveal that thiophene-fused g-lactams can efficiently inhibit M pro in vitro (Tables 1-3), in accord with the proposal that the thiophene ring helps to sequester electron density of the g-lactam-derived amine following acyl-enzyme complex formation. 33The g-lactam thiophene ring appears to be important for efficient M pro inhibition, since its substitution by a benzene ring ablates inhibition and, interestingly, the regioisomeric positioning of the sulfur atom within the thiophene ring also affects inhibition potency (Table 1).The reasons for this observation, including precisely how sequestration of the lone pair(s) on the g-lactamderived amine N atom affects the extent of inhibition, are under investigation.The results also show that substitution both a to the g-lactam carbonyl and on the g-lactam N atom affect inhibition potency (Tables 2 and 3).Thus, there is considerable scope for further optimization of the identied g-lactam M pro inhibitors, in particular with respect to optimal binding in the S1 or S2 pocket.The knowledge that g-lactams can bind in the S1 pocket of M pro via non-covalent interactions 14,24,42 suggests that derivatives of 7 and 8 possessing a second g-lactam binding in the S2 or S1 pocket are of interest.
Our MS studies reveal that g-lactams acylate the nucleophilic thiolate of Cys145 in a reversible manner (Fig. 3-5).The results thus indicate that lactam rings other than b-lactams have potential for development as covalently reacting inhibitors of M pro and, by implication, of other nucleophilic cysteine enzymes.However, it should be noted that g-lactams do not necessarily have to covalently react with M pro for efficient inhibition, because e.g., g-lactam 5 (Fig. 1) is reported to likely inhibit M pro without covalently reacting. 22The formation of a glactam from an ester, including an acyl-enzyme complex, is intrinsically more favourable than that of a b-lactam, as indicated by studies on the reactivity of g-lactams with a nucleophilic serine enzyme. 30,46The more reversible nature of g-lactam versus b-lactam reaction with nucleophilic residues may, in some circumstances, be an advantage with respect to limiting (irreversible) off-target reactivity.Reversibility may, however, be an unfavourable property with respect to the relative potency of inhibition for analogous gand b-lactams.
6][67] We did not observe evidence for g-lactam hydrolysis in our MS studies; this potential problem can be limited by steric extrusion of hydrolytic water from the active site, as precedented with work on inhibiting nucleophilic serine enzymes. 33,46,47The potential g-Fig.5 Proposed scheme for reaction of thiophene-fused g-lactams with nucleophilic cysteine enzymes.Reaction steps include: (a) the reversible non-covalent binding of the thiophene-fused g-lactam to the enzyme active site; (b) reversible nucleophilic attack of the cysteine thiolate to the proximate g-lactam carbonyl forming a tetrahedral intermediate; (c) reversible g-lactam fission and formation of an acyl-enzyme complex; (d) (reversible) protonation of the resultant amine, which may or may not be associated with a conformational change to form a hydrolytically more stable acyl-enzyme complex; (e) irreversible hydrolysis of the acyl-enzyme complex.Note there is variation in the general acid base machinery and oxy-anion stabilising mechanisms employed by nucleophilic cysteine enzymes.
lactam liability concerning reversibility in acylation can be overcome if the amine lone pair derived from the g-lactam is sequestered in the acyl-enzyme complex; indeed, this concept enabled the initial development of thiophene-fused g-lactams as serine protease inhibitors. 33It should be noted that the acylation ability of 5,5-trans-fused bicyclic pyrrolidine lactams 31 and related g-lactams is likely not, at least principally, a result of their ring strained structure. 90Indeed, the normally efficient natural substrates of proteases are themselves unstrained amides.Hence, empirical optimization remains of importance in inhibitor development.
We have not yet obtained a crystal structure of the acylenzyme complex formed by covalent reaction of Cys145 with a glactam, something that may reect reversible binding and/or labile nature of this intermediate.In addition to inhibiting by covalent reaction with Cys145, our combined mass spectrometric evidence imply that g-lactams can also inhibit M pro via non-covalent binding. 22It is thus possible that g-lactams can bind to the M pro active site in different conformations, i.e., one that enables covalent reaction of the g-lactam group with Cys145 or one that enables non-covalent binding of the g-lactam group, for example in the S1 pocket, as crystallographically observed for the g-lactam of nirmatrelvir (1). 14,24Our current work is thus focused on substituents a to the g-lactam carbonyl which engage with residues at the M pro active site to promote formation of a stable acyl-enzyme complex.

Conclusions
Our results expand the repertoire of covalently reacting groups for efficient SARS-CoV-2 M pro inhibition to g-lactams and suggest that g-lactams may also covalently react with other disease-relevant nucleophilic cysteine enzymes.2][93] Considering that both b-lactams 21,22 and, as we now report, g-lactams can covalently react with Cys145, other related ring systems also have potential for M pro inhibition.5][96][97] The corresponding band gsultam derivatives and d-lactams, which are reported to inhibit human neutrophil elastase, [98][99][100] may also be suited to covalent reaction with nucleophilic cysteine enzymes, including M pro .

Experimental section g-Lactam synthesis
Thiophene-fused g-lactams were synthesized following modications of reported procedures. 33

Fig. 3 g
Fig. 3 g-Lactams react covalently with isolated recombinant M pro .Analysis of a reaction mixture of M pro and g-lactams (A) 7, (B) 8, (C) 16a, (D) 16b, (E) 23b, and (F) 23d prior (bottom) and 4 h post (top) incubation with the respective g-lactam.Assays were performed using SPE-MS as described in the Experimental section employing SARS-CoV-2 M pro (3.0 mM) and, if appropriate, a g-lactam (15 mM) in buffer (20 mM HEPES, pH 7.5).Representative spectra of technical duplicates are shown.

Table 2
Effect of substituents a to the g-lactam carbonyl on M pro inhibition a IC 50 [mM] g-Lactam a IC 50 [mM] i 3.9 AE 0.1 vi 2.7 AE 0.1

Table 3
The g-lactam nitrogen substituent affects M pro inhibition a IC 50[mM]