Structure-activity relationships across groups

Thiol as a zinc-binding group and thiol masking

All study groups identified the free-thiol zinc-binding group (ZBG) as an important driver of NDM-1 inhibition [38-43]. Among group A, Meng et al. [41] retained the free thiol moiety and demonstrated compounds like compound 14a (IC50 = 0.1 μM) with enhanced potency. This study also provided direct structural support by solving high-resolution NDM-1 complexes with captopril and its optimized derivatives, thereby linking potency changes to specific binding interactions [41]. Similarly, Alfano et al. [42] demonstrated that retaining the free thiol moiety among series 6a-f/7a-d maintained activity, whereas thioester-protected analogues in series 14 were poorly active.

In group B, scaffold optimization strategies consistently preserved the free thiol group, further supporting its essential role in NDM-1 inhibition. Ma et al. [38] synthesized 10 compounds, six of which (3, 4, 6, 7, 8, 11) were more potent than D-captopril. Additionally, among fluorinated analogues designed by Kondratieva et al. [43], preserved activity against NDM-1 was observed when the free thiol moiety was preserved, with only a minority showing loss of inhibition.

In group C, Li et al. [40] observed loss of inhibition upon thiol protection in simplified analogues, supporting a requirement for a free thiol group for activity under their assay conditions. Some analogues were highly potent, like compound 22 (IC50 approximately 1.5 μM), along with others that remained active [40]. Brem et al. [39] further provided structural evidence and additional inhibition data, showing that D-captopril (IC50 ≈20 μM) is more potent than L-captopril due to the orientation of the thiol group in the active site.

Detailed structures of all compounds discussed are presented in Supplementary Table S1.

Overall, the collective evidence across all study groups supports the conclusion that a free thiol is a required moiety for activity, whereas masking the thiol group impairs activity by disrupting zinc coordination within the NDM-1 active site.

Stereochemistry

Stereochemistry was explored from various aspects across all study groups [38-43]. Among group A Meng et al. [41] separated selected racemic mixture of some compounds with promising activity into individual diastereomers and evaluated them independently, demonstrating that (R, S) configuration was more potent than the (S, S) configuration among all tested compounds, some potent compounds like compound 14m and others showed that activity was enhanced with R configuration at the chiral carbon adjacent to thiol moiety. Additionally, Alfano et al. [42] reported that (S, R) configuration was more potent than (S, S) configuration, indicating that potency is increased with the R configuration, but here, instead of the chiral carbon adjacent to thiol, the R configuration is preserved at the carbon adjacent to nitrogen located in the indoline moiety.

In group B, Kondratieva et al. [43] further supported the importance of stereochemistry by identifying the fluorinated stereoisomer (2R,2′R)-5αC as the most important derivative in their series with (IC50 = 0.3 μM). Ma et al. [38] also addressed stereochemical influence indirectly, as they proposed that the most active compound was compound 11 (IC50 = 4.6 μM), which formed a racemate dimer that mimics the binding mode of hydrolysed beta lactam binding geometry, and co-crystal structural analysis highlighted how stereochemical geometry governs active site interaction.

In group C, Brem et al. [39] provided strong stereochemical evidence, demonstrating that D-captopril was the most active captopril stereoisomer against NDM-1 (IC50 = 20.1 μM) and its activity is reduced with epimerization (IC50 = 64.6 μM), whereas L-captopril was substantially weaker (IC50 = 157.4 μM) and epi-L-captopril showed nearly lost of activity (IC50 > 500 μM). Li et al. [40] similarly noted that in simplified free acid and benzylamide analogues, the R configuration at the thiol adjacent stereocenter was related to higher potency. Conversely, more polar derivatives, such as Weinreb amides, N-methoxy amides, the S configuration became more favourable, reflecting that more polar and flexible substituents needed a different geometry to maximize the interactions in the active site [40].

Overall, the collective evidence indicates that stereochemical effects are controlled by the immediate environment of the substituents, in which the best configuration represents a balance among zinc coordination, hydrogen bonding and steric fit, rather than an inherent R/S preference [38-43].

Anchoring group effects

All the study groups reinforce the significant role of the carboxylate moiety as an anchoring group, either directly or indirectly [38-43]. In Group A, evidence was not extensive but favourable. Alfano et al. [42] replaced α-carboxylic acid with 2-carboxamide in their indoline 6a-d series, then substituted it with a glycyl amide moiety in 7a-d series. The findings demonstrated that compounds retaining the original carboxamide arrangement were more active (up to 3-fold more) than their corresponding 2-glycylamide analogues [42]. These findings suggest that the original carboxamide structure is beneficial to inhibition [42]. Within the same group, Meng et al. [41] preserved the carboxylic acid while converting the proline scaffold to other analogues throughout their series and did not vary the carboxylate motif or its orientation. They suggested that the carboxyl group is pivotal for NDM-1 inhibition because of its ability to form a hydrogen bond with Asn220 and should be maintained in the newly designed analogues [41].

More direct evidence is provided by Ma et al. [38], who clearly noted that the carboxylate position in D-captopril derivatives affect NDM-1 inhibition. For example, potency decreased after shifting the carboxylate moiety from the ortho position, as shown in compound 3 (IC50 = 6.9 μM) to the meta position and compound 5 (IC50 =28.4 μM), while relocation to the para position and compound 6 (IC50 4.9 μM) improved activity [38]. These effects are supported by structural analysis, as the para carboxylate position strengthened hydrophobic interactions with F70 and formed two water-mediated hydrogen bonds stabilized by N220, K211, and H250, whereas the meta isomer weakened both hydrophobic and hydrogen-bonding interactions [38]. Moreover, substitution of the para-carboxylate with an amide decreased potency, as observed in compound 7 (IC50=11.1 μM), because hydrogen bonds to water molecules were weakened either by a longer distance or by a lack of stabilization [38]. Kondratieva et al. [43] also explored the significance of the carboxylic acid moiety by deleting it or changing its position and orientation, but its influence seems more context-dependent within the fluorinated series than uniformly dominant. In some analogues, carboxylate removal was tolerated, whereas in others it caused significant loss of activity, indicating context-specific contributions rather than absolute dominance [43].

In group C, Li et al. [40] also provided indirect support for the anchoring role of the carboxylate group in simplified captopril analogues. Compounds with a free carboxylic acid have been demonstrated to exhibit measurable NDM-1 inhibition, while masking the acid by methyl ester blocked the activity, as noted in compound 9 (IC50 > 200 μM) [40]. Notably, highly potent simplified analogues, such as compounds 21 and 22 (IC50 5 ± 0.4 and 1.5 ± 0.2 μM, respectively), also retained the free carboxylate moiety, further reinforcing its compatibility with strong inhibition [40]. Also, Brem et al. [39] demonstrated that stereochemistry regulates the presentation of the carboxylate to the conserved basic residues within the active site, thereby providing a structural explanation for the superior potency of D-captopril over the L enantiomer

In general, all the study groups combined support that the carboxylate moiety is a major anchoring moiety in NDM-1 inhibitors and primarily through its role in hydrogen bonding, position optimization, and stabilization within the active site. Its specific contribution, however, can vary with the context of scaffolds, the orientation of substituents, and the overall structural environment.

Pyrrolidine ring modification

Group A studies redesigned the pyrrolidine scaffold to assess the influence of ring geometry on activity, rather than directly modifying the pyrrolidine ring. Alfano et al. [42] replaced the pyrrolidine ring of captopril with an indoline bicyclic system, and Meng et al. [41] replaced the pyrrolidine ring with variant scaffolds including L-tryptophan, L-phenylalanine, L-homophenylalanine, biphenyl, indole, or 2-benzo[b] thiophene-derived frameworks. Both studies showed that replacement of the pyrrolidine ring was well tolerated [41,42].

Clearer evidence can be derived from group B studies for the role of the pyrrolidine ring scaffold [37,42]. Ma et al. [38] showed that ring expansion from a five-membered pyrrolidine ring of D-captopril to a six-membered ring improved NDM-1 inhibition as shown in compound 3 (IC50 = 6.9 μM) compared to D-captopril (IC50=21.8 μM), whereas further expansion to a seven-membered ring reduced potency as shown in compound 4 (IC50=20.4 μM). These findings indicate that over-expansion interferes with optimal positioning in the active site [38]. Kondratieva et al. [43] further emphasized the importance of ring geometry, stereochemical arrangement and positioning of substituents through pyrrolidine to piperidine expansion in fluorinated captopril analogues. They found that some compounds retained high activity as compound 5αE (IC50 = 3.2 μM), whereas other analogues were much less active as compound **5βE (**IC50 >145 μM) [43].

In Group C, Li et al. [40] simplified the pyrrolidine ring with less polar amide substituents, such as linear aliphatic amides and benzyl amides. Substitution with aliphatic amide produced moderately active compounds (IC50 ≈13 to 20 μM) and bulky heteroaromatic analogs were usually weak. On the other hand, benzyl amide substitution as in compound 21 (IC50 = 5 μM) and compound 22 (IC50 = 1.5 μM) resulted in potent compounds [40].

These findings suggest that while the original pyrrolidine ring is not a structural requirement, scaffold modifications must preserve the spatial orientation of the thiol zinc-binding group and the anchoring functionality to maintain effective inhibition. Collectively, evidence across all groups supports that excessive expansion or improperly oriented substituents reduce potency, whereas para-oriented expansion of a six-membered ring may enhance activity.

Hydrophobic cap and steric ceiling

Hydrophobic substitution was often associated with enhanced binding to lipophilic regions close to the NDM-1 active site; however, this effect diminished with steric bulk substituents or unfavourable orientation [38-43]. This effect was especially noticeable in Group A. Meng et al. [41] developed mercaptopropionamide analogues by adding various hydrophobic motifs on both sides of the mercaptopropionamide core and demonstrated substantial potency improvement with various aromatic hydrophobic groups, as illustrated by biphenyl-containing compound 14a (IC50 = 0.10 μM), and indole analogue compound 14m (IC50 = 0.12 μM) [41]. Conversely, extending the hydrophobic system a notch further was not always advantageous, as observed with larger polycyclic compounds such as phenanthrene, dibenzo[b,d]thiophene, and trifluoromethylbenzene. These results indicate that the steric capacity of the binding pocket is already reached [41]. Similarly, Alfano et al. [42] demonstrated that the moderate hydrophobic expansion in their indoline-based series significantly improved potency and increasing alkyl bulk at the 3-position enhanced activity through hydrophobic capping, with the 3,3-diethyl analogue 6d (IC50=3.5 μM) representing one of the more active compounds in this series.

In group B, Ma et al. [38] investigated hydrophobic substitution by expanding the D-captopril scaffold weather by ring expansion or aryl extension and noticed that moderate hydrophobic enlargement improved activity as shown in their six-membered ring analogue (compound 3, IC50 = 6.8 μM), which was more effective than D-captopril, while excessive enlargement or poor orientation reduced potency, as seen in compound 10 (IC50 > 300 μM) [38]. Kondratieva et al. [43] further explored hydrophobic substitution by varying the hydrophobic cap in trifluoro-methylated mercaptopropionamide analogues, demonstrating that pyrrolidine and piperidine-based hydrophobic motifs, such as 5αC (IC50 = 0.3 to 4.5 μM), were more favourable than poorly oriented β-CF3 analogues, which were often weak or inactive (IC50 > 300). This suggests that hydrophobic substitution tends to improve potency only when paired with α-CF3 configuration [43].

Group C provided more indirect evidence. Li et al. [40] found that some benzyl amide analogues represented some of the most effective simplified analogues, such as compound 22 (IC50 = 1.5 μM), which is about a five-fold better inhibitor than D-captopril (IC50 = 7.9 μM). However, alteration of the benzyl ring was not tolerated; for example, meta-hydroxybenzyl analogue compound 23 retained moderate activity (IC50 ≈5 μM), whereas an ortho-hydroxy substitution, methyl substitution, or fluorine substitution resulted in inactive compounds (IC50 > 200 μM) [40]. The significance of hydrophobic interactions is supported by structural information presented by Brem et al. [39], rather than by investigating the potential of hydrophobic substitution in captopril-based inhibition.

Overall, the collective evidence suggests the following SAR rule: hydrophobic capping can enhance potency by improving pocket occupancy, but the benefit is limited by a steric ceiling and poor orientation [38-43].

Fluorinated analogues

Various fluorination strategies were explored across study groups [40,41,43]. In group A, Meng et al. [41] introduced a trifluoromethyl-substituted phenyl moiety, generating compound 14q, which exhibited moderate activity (IC50 = 3.90 μM).

In group B, Kondratieva et al. [43] reported a systematic fluorination approach in which they used trifluoromethyl (CF3)-substituted mercaptopropionamide scaffolds while maintaining the thiol group. Activity in their series was highly dependent on the location of fluorination: αCF3 analogues were more active than βCF3 analogues. As an example, 5αC, 5αE, and 5αF showed IC50 values of 0.3 to 4.5 μM, while numerous βCF3 analogues were weak or inactive [43]. These findings suggest that fluorination may enhance potency by improving hydrophobic interactions and conformational stability, but this enhancement relies on the location of the fluorinated group and the scaffold architecture [43].

From group C, Li et al. [40] explored fluorination less extensively, and used the peripheral aryl fluorination instead of core scaffold modification; however, ortho-fluorobenzyl compound 26 was not active (IC50 > 200 μM), which suggested that the simple fluorine substitution of the benzyl cap was not conducive in this simplified scaffold.

Scaffold simplification

Li et al. [40] from group C investigated simplified captopril analogues. They found that the pyrrolidine residue could be substituted with less reactive amide fragments of 3-mercapto-2-methylpropanoic acid without significantly affecting activity, indicating that a considerable reduction in scaffold complexity could be achieved without loss of activity. In their simplified series, a simple hydrophobic benzyl amide was particularly favourable, such as analogue 22, which was the strongest compound (IC50 = 1.5 μM) in the whole series [40]. Substitution on the benzyl ring was, however, poorly tolerated, with the meta-hydroxybenzyl analogue 23 remaining active (IC50 = 5 μM) while the ortho-hydroxy, methyl and fluor substituted analogues of the benzyl ring were not active. Further simplification of the structure generated compounds 31 and 32 that inhibit NDM-1 (IC50 = 15 and 10.4 μM, respectively), indicating that the thiol-containing moiety comprises the core of the essential inhibitory unit, though the proline moiety and the hydrophobic cap could also be simplified, although not with a wide structural range [40].