Phosphinic Peptides as Tool Compounds for the Study of Pharmacologically Relevant Zn-Metalloproteases

Abstract

Phosphinic peptides constitute an important class of bioactive compounds that have found a wide range of applications in the field of biology and pharmacology of Zn-metalloproteases, the largest family of proteases in humans. They are designed to mimic the structure of natural substrates during their proteolysis, thus acting as mechanism-based, transition state analogue inhibitors. A combination of electrostatic interactions between the phosphinic acid group and the Zn cation as well as optimal noncovalent enzyme–ligand interactions can result in both high binding affinity for the desired target and selectivity against other proteases. Due to these unique properties, phosphinic peptides have been mainly employed as tool compounds for (a) the purposes of rational drug design by serving as ligands in X-ray crystal structures of target enzymes and allowing the identification of crucial interactions that govern optimal molecular recognition, and (b) the delineation of biological pathways where Zn-metalloproteases are key regulators. For the latter objective, inhibitors of the phosphinopeptidic type have been used either unmodified or after being transformed to probes of various types, thus expanding the arsenal of functional tools available to researchers. The aim of this review is to summarize all recent research achievements in which phosphinic peptides have played a central role as tool compounds in the understanding of the mechanism and biological functions of Zn-metalloproteases in both health and disease.

Zn-metalloproteases constitute the most populated superfamily of proteolytic enzymes in humans (194 members versus 176 serine proteases, 150 cysteine proteases, 28 threonine proteases, and 21 aspartic proteases)¹ with several of its members being validated pharmaceutical targets for severe pathologies such as cancer, inflammatory, and cardiovascular diseases.²⁻⁶ Therapeutic manipulation of pathological conditions associated with deregulation of Zn-metalloproteases’ activity requires detailed understanding of their role in complex biological processes and availability of pharmaceutical agents with optimized selectivity, bioavailability, and safety profiles.^4,5 Reaching these goals can be a long and arduous process that strongly depends on the development of tool compounds that provide insights into the function of target proteases on a molecular level. In most cases, these tool compounds are, or are based on, active-site-directed small-molecule inhibitors initially designed to probe intrinsic properties of target proteases, such as subsite occupancy preferences or substrate-processing mechanisms.⁴

Among the various types of zinc-binding chemical scaffolds (Figure 2) that have been employed as tool compounds for the study of Zn-metalloproteases,⁷ phosphinic peptide inhibitors have been extensively explored.⁸⁻¹³ Phosphinic peptides are hydrolytically stable peptide analogues which contain a phosphinic acid moiety in lieu of a scissile peptide bond, thus mimicking the tetrahedral geometry of transient gem-diol species formed during the transition state of proteolysis (Figure 1).¹⁰ Acting as transition state analogues, phosphinic peptides are characterized by high binding affinity for their target which is mainly attributed to the optimized network of interactions in the transition state of an enzymatic reaction rather than strong zinc chelation. The first literature report about phosphinic peptide inhibitors dates back to 1984, when Paul Bartlett and his colleagues described a pseudodipeptide inhibitor of aspartyl protease pepsin.^14,15 During that time, researchers from the pharmaceutical company Squibb were working on phosphinic inhibitors of the Zn-metalloprotease angiotensin-converting enzyme (ACE), leading a few years later to the well-known antihypertensive drug fosinopril.¹⁶ However, phosphinic peptides were established as transition state analogue inhibitors of Zn-metalloproteases after the independent studies of Grobelny, Kollman, and Bartlett on thermolysin.¹⁷⁻¹⁹ Their mechanism of action was later verified by crystallographic analysis of an astacin complex bound to a phosphinic peptide.²⁰ Notably, phosphinic peptide inhibitors of thermolysin were developed much later than their phosphonamidic isosteres (Figure 2b)^21,22 as a result of the incorrect initial assumption that a NH group adjacent to phosphorus is essential for accurate reproduction of the transition state by participating in a critical hydrogen bond with the carbonyl group of Ala113.^11,23,24 Interestingly, although phosphinic peptides are unable to form this important hydrogen bond, they were found equipotent to their phosphonamidic isosteres. This is a result of a decrease in their solvation energy, as compared to phosphonic and phosphonamidic isosteres, that counterbalances the loss of the hydrogen bond observed in phosphonamidates.^18,19 Taken together, the above observations and the fact that phosphinic peptides are completely stable against hydrolysis unlike phosphonamidates and phosphonates, phosphinic peptides were quickly qualified as potential drug candidates and stimulated pharmaceutical research toward most of the known medicinally relevant Zn-metalloproteases.¹⁰ In the following years, synthetic advances pioneered by Dive and Yiotakis concerning the wide production of diverse phosphinic peptides by means of combinatorial chemistry²⁵⁻²⁷ played pivotal role to the expansion of the field and allowed the identification of potent Zn-protease inhibitors with remarkable selectivity profiles. Due to their unique target specificity, these phosphinic peptide inhibitors have been extensively used as tool compounds in biological and pharmacological studies whereas crystallographic analysis of their complexes with Zn-metalloproteases has accelerated rational, structure-based drug discovery by shedding light on the crucial enzyme–ligand interactions which determine binding affinity. In the new era of phosphinic peptides, privileged inhibitors have been transformed into innovative active-site-directed probes for Zn-metalloproteases, extending the range of applications to functional proteomics and diagnostics. In this review, we attempt to present in a systematic manner the most important advancements in the field, focusing our attention on the contribution of phosphinopeptidic tool compounds in structural and biological studies of pharmacologically important Zn-metalloproteases.

Figure 2.

(a) Representative examples of zinc-binding moieties used in Zn-metalloprotease inhibitors. (b) Generic structures of phosphonic, phosphonamidic, and phosphinic peptides.

Figure 1.

Generic chemical structure of a phosphinic peptide and schematic representation of its mode of action as a transition state analogue inhibitor of a Zn-metalloprotease.

1. Structural Studies with Phosphinic Peptide Inhibitors of Zn-Metalloproteases

The wealth of structural information acquired from the crystallographic analysis of Zn-metalloproteases’ complexes with phosphinic inhibitors has revealed certain common binding patterns. Phosphinic peptides occupy the active site of proteases usually by adopting an extended conformation which resembles the orientation of a natural substrate within the catalytic site of the protease. The phosphinopeptidic backbone is centered around the zinc ion which interacts with the negatively charged phosphinic group in a bidentate manner. In most cases, specificity subsites, as these have been described by Schechter and Berger (S₃, S₂, S₁, S₁′, S₂′, S₃′, etc.),²⁸ are accommodating the respective side chains of the inhibitor (P₃, P₂, P₁, P₁′, P₂′, P₃′, etc.), allowing additional interactions that enhance binding affinity. Substrate-binding grooves of Zn-metalloproteases are shaped according to the exact position of the scissile peptide bond which is replaced by the phosphinic acid group in the case of inhibitors. Therefore, phosphinic inhibitors of endopeptidases, such as astacin (Figure 3A), occupy both primed and unprimed specificity pockets. In the case of exopeptidases, aminopeptidases, or carboxypeptidases (APN and CPB, respectively, Figure 3B,C), inhibitors occupy the specificity pocket S₁ and S₁′, respectively, and they also extend toward the opposite side of the zinc, forming additional interactions with one or more pockets. Moreover, the shape and charge distributions of the subsites of the catalytic groove determine substrate specificity, e.g. the S₁ pocket of CPB is strongly negative to optimally accommodate Lys or Arg (Figure 3C). Obviously, detailed analysis of the exact network of enzyme–ligand interactions which are highly variable among different Zn-proteases is crucial for the rational design of selective inhibitors, as discussed below.

Figure 3.

Surface representation of the specificity pockets of the (A) endopeptidase astacin in complex with a phosphinic peptide (PDB ID: 1QJI), (B) aminopeptidase hAPN in complex with phosphinic inhibitor PL250 (PDB: 2DQM and inhibitor from 2ZXG), and (C) carboxypeptidase CPB in complex with phosphinic inhibitor EF6525 (PDB ID: 3WC7). The surfaces are colored according to the electrostatic potential (red: −5 k_BT/e; 0: white; +5 k_BT/e: blue), calculated using PDB 2PQR (AMBER FF) and APBS in PyMOL.²⁹ The catalytic Zn(II) is shown as a yellow sphere, and the phosphorus atom of the ligand is shown in orange.

1.1. Aminopeptidases

Aminopeptidase N (APN) is a membrane-bound ectoenzyme which is overexpressed in several cancer types and is considered an attractive target for the development of anticancer therapies.^30,31 It belongs to the M1 family of metalloproteases that preferentially removes neutral amino acids, most notably alanine, from the N-terminus of peptides. Its active site cleft is characterized by two conserved signature sequences which are common in all M1 aminopeptidases: the GXMEN exopeptidase loop and a consensus zinc-binding motif HExxHx18E. Among the first crystal structures of APN which provided valuable structural information on key enzyme–ligand interactions was that of E. coli APN bound to phosphinic inhibitor PL250 (Figure 4A).³² PL250 was found tightly bound to the active site of APN in an extended conformation antiparallel to the GAMEN motif (Figure 4B). The nonprimed region of the active site is characterized by a negatively charged cavity typical for aminopeptidases with the amino group of PL250 being hydrogen bonded with Glu121 and Glu264. The bottom of the S₁ pocket is delimited by Met260 (replaced by an alanine in mammalian APN) which is highly flexible, and its displacement can create additional space for the accommodation of larger P₁ side chains. The enzyme–inhibitor interactions around the zinc ion were found almost identical with those of the APN/bestatin complex, with the only exception being Glu298 which is slightly displaced to form a hydrogen bond with one of the two phosphinic oxygens (Figure 4B). The other oxygen of the phosphinic group forms a hydrogen bond with Tyr381 which highlights the involvement of this tyrosine for the stabilization of the transition state of the enzymatic reaction, as established in the past with the astacin paradigm.²⁰ In a more recent report, Mucha and colleagues succeeded in solving the crystal structure of a bacterial APN isoform (N. meningitides APN, NmAPN) bound to phosphinic peptide 1 (Figure 4A), a potent aminopeptidase inhibitor previously developed for dinuclear leucine aminopeptidase (LAP),³³ en route to selective APN inhibitors that could spare LAP (i.e., compound 2, Figure 4A).³⁴ Crystallographic data of NmAPN complexed with derivative 2 revealed that its high potency toward APN isoforms is attributed to a salt bridge between Asp323 and the aminomethylene group of the P₁′ substituent in combination with optimal π–π stacking interactions observed between the pyridyl group and the S₁ subsite.³⁴

Figure 4.

(A) Structure and inhibition profile of PL250,³²1, and 2.³⁴ (B) Close-up view of the active site of E. coli APN (green C atoms) in complex with phosphinic inhibitor PL250 (yellow C atoms, PDB ID: 2ZXG). The conformation of Glu298 in the complex of human APN with bestatin (purple C atoms, PDB ID: 2DQM) is also shown. Green dashed lines indicate APN–inhibitor hydrogen-bonding interactions. Zinc atom is shown as a gray sphere.

In a collaboration of Australian and Polish researchers, the utility of compound 1 (Figure 4A) as a tool for the study of aminopeptidases was further expanded to PfA-M1 and PfA-M17, two enzymes from Plasmodium falciparum parasites that share many common features with APN and LAP, respectively.³⁵PfA-M1 and PfA-M17 degrade hemoglobin of infected organisms to produce essential nutrients for parasite growth and development within the erythrocytes, and therefore they are considered pharmaceutical targets for antimalarial therapy.³⁶ After establishing the efficacy of phosphinic dipeptide 1 to potently inhibit PfA-M1 and PfA-M17,³⁵ the researchers solved the X-ray crystal structures of the two enzymes complexed with 1 and identified critical interactions that account for the enzymes’ binding preferences.^37,38 In contrast to other aminopeptidases, both enzymes do not undergo major conformational changes upon inhibitor binding, as concluded by comparison of the unbound and ligand-bound structures, which implies alternative mechanisms for substrate entry into the active site. In this regard, it was proposed that PfA-M1 substrates or inhibitors use a large channel at the C-terminal domain to access the buried active site whereas amino acid products are released by a small channel at the N-terminal domain.³⁸ The inhibitor possesses a similarly extended conformation on both crystal structures with small variations in the orientation of side chains (Figure 5). However, the higher binding affinity of 1 for PfA-M17 was attributed to the more extensive interactions of the inhibitor with PfA-M17’s active site as compared to PfA-M1. Careful inspection of PfA-M1 and PfA-M17 active sites also provided a reasonable explanation for their distinct substrate specificity. In particular, the S₁ cavity of PfA-M17 is shaped by residues Met392, Met396, Phe398, Thr486, Gly489, Leu492, and Phe583, forming a narrow hydrophobic pocket that cannot accommodate substrates with polar groups in their P₁ position (Figure 5B). On the contrary, the respective pocket of PfA-M1 is much larger (Figure 5A) and the polar carboxylate of Glu572 located at the base of this pocket allows the formation of ionic interactions with polar substrates such as arginine. Of notable interest is the flexibility observed in the side chain of Met1034 which controls the size of S₁ hydrophobic pocket of PfA-M1 and may explain the broader substate specificity of this enzyme.

Figure 5.

(A) Active site of PfA-M1 (green C atoms) in complex with phosphinic inhibitor 1 (yellow C atoms, PDB ID: 3EBI). (B) Active site of PfA-M17 (cyan C atoms) in complex with phosphinic inhibitor 1 (yellow C atoms, PDB ID: 3KR5). Zinc atom is shown as a gray sphere.

During the past decade, a series of crystallographic studies concerning antigen processing endoplasmic reticulum M1 aminopeptidases 1 and 2 (ERAP1 and ERAP2) and insulin-regulated aminopeptidase (IRAP) with bound phosphinic peptide inhibitors have shed light not only on salient mechanistic features of the proteolytic reactions catalyzed by these enzymes but also on crucial binding interactions that have guided the rational design of improved inhibitors with potential clinical value.^39,40 DG013A, the first inhibitor designed for ERAPs/IRAP, is a phosphinic tripeptide able to bind to all three enzymes with nanomolar affinity (Figure 6).^41,42 Its discovery was accompanied by the structural elucidation of its complex with ERAP2⁴¹ while very recently the X-ray crystal structure of ERAP1 with DG013A was also solved,⁴³ allowing a direct comparison of binding modes between the two enzymes (Figure 7). In both structures, phosphinic oxygens form stabilizing hydrogen bonds with catalytic residues Glu354/Glu371 and Tyr438/Tyr455, for ERAP2/ERAP1, respectively, unveiling the key interactions which are established in the transition state of proteolysis. Additional explanations for the high binding affinity of DG013A were found in the tight polar interactions of its terminal amino group with three glutamates located at the S₁ cavity of both active sites and the hydrophobic interactions between the phenyl group of the P₁ position and Phe433/Phe455 which is almost identically oriented in both structures. The importance of hydrophobic interactions between DG013A’s indole ring and Tyr438/Tyr455 is also highlighted in these reports; however, additional π-stacking between indole ring and Tyr892 which is unique in ERAP2 (Figure 7B) causes a large conformational change in the P₂′ position of DG013A that differentiates the binding poses of the inhibitor in the active sites of the two enzymes.

Figure 6.

Structure and inhibition profile of DG013A, DG026, DG046, 3, and 4.⁴²

Figure 7.

Close-up view of the active site of (A) ERAP1 (green C atoms, PDB ID: 6M8P) and (B) ERAP2 (cyan C atoms, PDB ID: 4JBS), in complex with DG013A (yellow C atoms). Dashed lines indicate aminopeptidase–inhibitor hydrogen-bonding interactions. A yellow dashed line is also used to indicate the π–π interactions between Tyr892 and indole ring of DG013A. Zinc atom is shown as a gray sphere.

Further SAR studies based on the ERAP2/DG013A structure led to a number of selective inhibitors of ERAPs/IRAP that served as valuable tools for investigating the molecular basis of targeted inhibition.⁴² For example, an X-ray crystal structure of IRAP bound to the IRAP-selective inhibitor DG026 (Figure 6) revealed a ligand-induced conformational closing of the enzyme which is believed to be triggered by interactions between the inhibitor and the GAMEN loop of IRAP that reorientates in order to approach the bound inhibitor.⁴⁴ This unusual plasticity of IRAP’s GAMEN loop was considered responsible for the formation of a network of interactions that induce a near-optimal fitting of DG026 into the active site and result in high potency and selectivity. Comparison with the closed conformations of ERAP1 and ERAP2 provide additional explanations for the observed selectivity for IRAP, such as the loss of an important π-stacking interaction between the P₁ phenyl ring and Tyr961 (replaced by Ser869 in ERAP1) and a steric clash between bulky P₁′ side chain and the indole ring of Trp363 of ERAP2 that replaces a smaller isoleucine in the case of IRAP (Figure 8A). A few years ago, the first crystal structure of ERAP1 bound to a highly potent, nanomolar inhibitor (phosphinic inhibitor DG046, Figure 6) was resolved in an attempt to map enzyme–inhibitor interactions that will guide optimization efforts.⁴⁵ Besides the expected interactions, an unusual extensive network of T-shaped π-stacking interactions between Phe433 and the phenyl rings of the P₁ and P₂′ positions in combination with a favorable hydrophobic stacking of the alkyne group of P₁′ position with His353 were identified as the source of DG046’s high binding affinity whereas selectivity for IRAP was attributed to similar factors as in the case of DG026 mentioned above (Figure 8B). Recently, two more crystal structures of ERAP2 complexed with small molecule phosphinopeptidic inhibitors (3 and 4, Figure 6) became available that revealed an ill-defined S₁′ specificity pocket which can accommodate P₁′ side chains of distinct spatial characteristics by making only opportunistic interactions with nearby residues.⁴⁶ Interestingly, cocrystallization of 3 with ERAP2 led to the identification of two different orientations for the long P₁′ isoxazole side chain occupying different S₁′ subsites (Figure 9A). This unexpected finding implies that long or bulky side chains are able to exploit the additional space of the S₁′ cavity by probing distant specificity pockets, as clearly demonstrated in the ERAP2/4 crystal structure (Figure 9B), thus optimizing P₁′/S₁′ interactions which have been proposed as one of the main determinants for the potent inhibition of ERAP2.⁴⁷

Figure 8.

(A) Active site of IRAP (gray C atoms, PDB ID: 5MJ6) in complex with DG026 (yellow C atoms) illustrating the residue differences with ERAP1 (green C atoms, PDB ID: 6Q4R) and ERAP2 (cyan C atoms, PDB ID: 5AB0). Yellow dashed lines indicate hydrogen bonds, π–π and hydrophobic interactions between DG026 and IRAP. (B) Active site of ERAP1 (green C atoms, PDB ID: 6Q4R) in complex with DG046 (yellow C atoms) illustrating the residue differences with ERAP2 (cyan C atoms, PDB ID: 5AB0) and IRAP (gray C atoms, PDB ID: 5MJ6). The distance (Å) between the propyne moiety of DG046 and Trp363 of ERAP2 is indicated. Zinc atom is shown as a gray sphere.

Figure 9.

(A) Surface representation of ERAP2 active site in complex with phosphinic inhibitor 3 (PDB ID: 7P7P) that has been resolved in two alternative conformations (white and yellow C atoms). (B) Surface representation of ERAP2 active site in complex with phosphinic inhibitor 4 (PDB ID: 7PFS).

Apart from the aforementioned structural studies on ERAPs/IRAP complexes with phosphinic tripeptides that have contributed to the detailed mapping of inhibitor–active site interactions, longer phosphinic peptides have also been designed and employed as tool compounds in important mechanistic studies. In particular, a 10-mer phosphinic peptide (DG025, Figure 10A) was designed based on the sequence of a known antigenic epitope from the aggrecan protein and cocrystallized with both ERAP2 and IRAP in order to investigate how these enzymes recognize their substrates.^48,49 Comparative analysis of the obtained structures showed that the two enzymes adopt distinct overall conformations: a closed conformational state was found in the case of the ERAP2/DG025 complex whereas IRAP was captured in an intermediate, semiclosed configuration (Figure 10B). In these two configurations, DG025 adopts completely different orientations in the active site, with the exception of the first two residues that are restricted by the presence of the zinc atom and the adjacent conserved catalytic residues. This conformational flexibility of DG025’s peptide chain becomes more evident in its C-terminus, suggesting that these two enzymes allow substrates to scan their active site cleft and optimize binding by forming the most efficient yet opportunistic interactions with their environment. On the contrary, it has been proposed that ERAP1 follows a different processing mechanism which involves recognition of substrate’s C-terminus by a specific regulatory pocket lying ∼30 Å far from the catalytic site. This structural feature enables the selection of substrates of specific length, because shorter or longer peptides cannot simultaneously recognize both the regulatory site with their C-terminal “tail” and the catalytic site with their N-terminal “head”, leading to lower catalytic efficiency. The structural basis of this “molecular ruler” mechanism, as proposed previously by Goldberg and colleagues,⁵⁰ was recently delineated by cocrystallizing ERAP1 with suitably designed phosphinic peptides of various lengths.^43,51 In one of these structures, a 15-mer phosphinic peptide (DG055, Figure 11A) was found to interact with the catalytic center through its N-terminal end and to extend toward a distal site that displays “carboxypeptidase” characteristics that anchors the C-terminal end of the peptide (Figure 11B).⁵¹ Interestingly, the central region of DG055 was found disordered which suggests a conformational freedom that may have important implications in the catalytic mechanism. Notably, this interaction between the peptides’ C-terminal moieties and the regulatory (or allosteric) site of ERAP1 revealed an extremely powerful alternative for the development of small-molecule allosteric inhibitors that may prove advantageous drug candidates as compared to their active-site counterparts.⁵²⁻⁵⁴

Figure 10.

(A) Structure and inhibition profile of DG025.^48,49 (B) Superposition of DG025 from its X-ray structures in complex with ERAP2 (white, PDB ID: 5AB0) and IRAP (yellow, PDB ID: 4Z7I).

Figure 11.

(A) Structure and inhibition profile of DG055.⁵¹ (B) Surface representation of ERAP1 in complex with the 15-mer phosphinic pseudopeptide inhibitor DG055 (PDB ID: 6RYF). The C-terminal residues of DG055 were resolved in two alternative conformations (yellow and white C atoms) whereas residues 6–10 (gray conformation) or 6–9 (yellow conformation) of the peptide were not resolved due to high mobility. Zinc atom is shown as a purple sphere.

Evidently, the wealth of structural information acquired from the analysis of aminopeptidases/phosphinic peptides complexes mainly involves mononuclear aminopeptidases; however, the first aminopeptidase that was cocrystallized with a phosphinic peptide was the dinuclear bacterial aminopeptidase (PepV) isolated from Lactobacillus delbrueckii.⁵⁵ Phosphinic dipeptide 5 (Figure 11A), previously developed as inhibitor of aminopeptidase A (APA),⁵⁶ was employed to explore the preference of PepV for dipeptides and the role of the two zinc ions in the catalytic mechanism. Structural analysis of the PepV/5 complex revealed a “catalytic domain” and a “lid domain” that traps the substrate in a tunnel-like active site with two exits toward water. The amino group of 5 forms hydrogen bonds with Asp177 and a water molecule located at the “entrance” of the tunnel whereas the C-terminal carboxylate of 5 is tightly fixed in a network of hydrogen bonds with Arg350, Gly415, and Asn217 residues forming a recognition site known as the “carboxylate groove” (Figure 12B). This latter structural feature explains the propensity of PepV to act as a dipeptidase rather than a real aminopeptidase. The phosphinic group is located between the two Zn atoms with one of the phosphinic oxygens bridging the two ions and the other oxygen binding to Zn1 in a bidentate manner. This binding pose of the phosphinic group led researchers to the suggestion that Zn1 mainly participates in the polarization of the scissile peptide bond while Zn2 mainly activates a water molecule to promote the nucleophilic attack. Further confirmation of this statement came by the crystallographic analysis of another dinuclear dipeptidase from E. coli (isoaspartyl dipeptidase) bound to a phosphinic dipeptide which revealed a similar arrangement of the phosphinic group between the two Zn ions.⁵⁷

Figure 12.

(A) Chemical structure of inhibitor 5. (B) Close-up view of the active site of PepV (green C atoms) in complex with phosphinic inhibitor 5 (yellow C atoms, PDB ID: 1LFW). The two metal ions are indicated as Zn1 and Zn2, including two structural solvent molecules designated as Wat1 and Wat2. Hydrogen bonds are indicated by yellow dashed lines. Zinc atoms are shown as gray spheres.

1.2. Endopeptidases

A landmark discovery in the history of structural biology of Zn-metalloproteases was achieved in 1996 with the elucidation of the X-ray crystal structure of astacin complexed with phosphinic peptide 6 (Figure 13A).^20,58 Astacin is the prototype of the astacin family of metallopeptidases and the metzincin superfamily⁵⁹ that also includes adamalysins, matrixins, and serralysins.⁶⁰ It was isolated from crayfish Astacus astacus L., and since then several members of this family have been found in humans, such as meprins α and β which contribute to extracellular matrix (ECM) remodeling; they are overexpressed in chronic inflammation, certain cancers, and fibrosis.⁶¹ Unlike other metzincins, the zinc atom of astacin is pentacoordinated using a Tyr149 as the fifth ligand, as revealed by the first X-ray crystal structure of this enzyme in 1992.⁶² This unusual feature of astacin raised the question of whether Tyr149 is essential for catalysis; therefore, structural elucidation of astacin bound to phosphinic peptide 6 (Figure 13A) was pursued. Indeed, crystallographic analysis revealed that coordination of the phosphinic oxygen with the zinc atom is accompanied by simultaneous shifting of the phenolic oxygen of Tyr149–5 Å away from the metallic center and formation of a stabilizing hydrogen bond with the phosphinic oxygen (Figure 13B).²⁰ This movement of Tyr149 upon substrate binding was called the “tyrosine switch”, and it was suggested to play a central role in transition state stabilization. Moreover, an overall hinge-like movement was observed by which the distance between the lower and upper subdomains of astacin is decreased by ∼1 Å. Apart from that, the short distance between Glu93 and a phosphinic oxygen of 6 implies that Glu93 may act as a proton shuttle between an activated molecule of water and the nitrogen of the scissile peptide bond during hydrolysis. The importance of Glu93 as a general base in the catalytic reaction was later confirmed by the complete loss of activity observed in a Glu93Ala astacin mutant.⁶³

Figure 13.

(A) Structure and inhibition constant of compound 6.⁵⁸ (B) Superimposed X-ray structures of ligand-free astacin (cyan C atoms, PDB ID: 1IAC) and inhibitor 6-bound astacin (pink C atoms, PDB ID: 1QJI) illustrating the movement of the Tyr149 side chain. Dashed lines indicate the distances (Å) between the phenolic oxygen of Tyr149 and either the metal (ligand-free) or the phosphinic O (inhibitor-bound). Inhibitor 6 is shown in yellow C atoms, and the zinc atom is shown as a gray sphere.

During the past two decades, Dive and colleagues employed phosphinic peptides extensively as tool compounds for the investigation of biochemical and structural aspects of matrix metalloproteinase (MMP) inhibition and for exploring the potential of phosphinic inhibitors in various disease models.⁶⁴⁻⁶⁷ MMPs form a set of 23 endopeptidases mainly involved in extracellular matrix (ECM) degradation which is an important biological process for the pathophysiology of several diseases such as cancer metastasis, arthritis, and cardiovascular diseases.^68,69 Numerous studies have highlighted the complex biology of MMPs and the need for developing selective inhibitors after the failure of first-generation inhibitors to reach the clinic due to negligible therapeutic efficacy and untoward side effects, such as musculosceletal syndromes (MSS).⁷⁰ In this direction, intense screening efforts of phosphinic pseudopeptide libraries led to compound RXP470.1, a highly selective nanomolar inhibitor of MMP-12 that fully discriminates between different MMPs (Figure 14A).⁷¹ Its unique properties prompted researchers to explore the molecular determinants of the observed selectivity by using crystallographic and calorimetric studies.⁷² As shown in Figure 14B, the P₁′ isoxazolyl side chain of RXP470.1 extends into the S₁′ hydrophobic pocket of MMP-12 and is engaged in multiple interactions with residues forming the wall of the deep S₁′ channel. The P₂′-P₃′ Glu-Glu motif of RXP470.1 does not contribute to the MMP-12 binding affinity, however it impairs binding to other MMPs, as it was verified by comparison of RXPA470.1 with an analogue bearing alanine residues at these positions. It was also proposed that binding of RXPA470.1 to MMP-12 is mostly entropy driven with the m-Cl phenyl group of the P₁′ side chain significantly contributing to this behavior. Moreover, in a follow-up study it was demonstrated that replacement of the phosphinic moiety by a carboxylic acid zinc-binding group did not affect potency for MMP-12; however, selectivity was dramatically diminished.⁷³ Interestingly, starting from RXP470.1 as a structural template, several MMP-12 selective inhibitors were identified bearing no ZBG or a P₃′ glutamate, implying that occupancy of the S₁′ specificity pocket by a hydrophobic moiety and the presence of a P₂′ glutamate are the two key determinants for RXP470.1′s potency and selectivity.^74,75

Figure 14.

(A) Structure and inhibition profile of RXP470.1.⁷¹ (B) Active site of MMP-12 (cyan C atoms) in complex with RXP470.1 (yellow C atoms, PDB ID: 4GQL). Yellow dashed lines indicate the interactions of the phosphinic group with the enzyme. Zinc atom is shown as a gray sphere.

A few years after the elucidation of astacin’s structure in complex with a phosphinic peptide, the first snapshot of an MMP interacting with a transition state analogue was acquired.²⁰ Stromelysin 3 (MMP-11), the first MMP whose proteolytic action was linked to invasive cancer and considered as a first-line pharmaceutical target, was cocrystallized with one of the first phosphinopeptidic MMP inhibitors reported,⁷⁶ namely RXP03 (Figure 15A).⁷⁷ In this structure, besides the expected zinc chelation by the phosphinic acid moiety and a hydrogen bond between the catalytic Glu220 and one phosphinic oxygen, an unusual hydrogen bond was observed between the other phosphinic oxygen and a water molecule which is polarized by the carbonyl group of Pro239 (Figure 15B). This bond was found to replace an important stabilizing interaction between Tyr149 of astacin²⁰ or His231 of thermolysin⁷⁸ with an identically positioned phosphinic oxygen in their crystal structures with phosphinic peptides, revealing significant differences in the catalytic mechanisms of these three enzymes. The P₁′ phenylpropyl side chain of RXP03 fills a deep S₁′ hydrophobic channel reaching toward the side chain of Gln215, a residue that is unique in MMP-11 and is considered to delimit the size of the S₁′ pocket. Moreover, although the hydrophobic indole ring of the P₂′ position of RXP03 is almost entirely exposed to the solvent, its role in RXP03’s inhibitory potency is crucial, as implied by the dramatic drop of binding affinity displayed by an analogue bearing an alanine at the same position.⁷⁶ Interestingly, the orientation of this indole ring in the MMP11/RXP03 complex is completely different from that observed in the crystal structure of MMP9 bound to another Trp-containing phosphinopeptidic inhibitor (RXP409, Figure 15A) solved a few years later (Figure 15C).⁷⁹ In the latter structure, a noticeable feature is the reorientation of Arg424’s side chain to create a deeper S₁′ cavity able to accommodate the long isoxazolyl chain of RXP409. This flexibility of residues located in the S₁′ loop, a highly variable segment among MMPs in shape and length, was recognized in the following years as the most challenging obstacle in the discovery of selective inhibitors of MMPs.⁸⁰

Figure 15.

(A) Structure and inhibition profile of MMP inhibitors RXP03⁷⁶ and RXP409.⁷⁷ (B) Close-up view of the active site of MMP-11 (cyan C atoms) in complex with phosphinic inhibitor RXP03 (magenta C atoms, PDB ID: 1HV5). Green dashed lines indicate hydrogen bonds of a water molecule between the carbonyl group of Pro239 and a phosphinic oxygen. (C) Close-up view of the active site of MMP-9 (green C atoms) in complex with phosphinic inhibitor RXP409 (yellow C atoms, PDB ID: 2OVZ). Distances are indicated in angstroms (Å). Zinc atom is shown as a gray sphere.

In 2010, phosphinic peptide PPI-2 (Figure 16A) bearing collagenous sequences of Gly-Pro type was cocrystallized with Pz-peptidase A,⁸¹ a bacterial endopeptidase that degrades oligopeptidic collagen fragments after collagenolysis and shares many functional characteristics with mammalian endopeptidase 24–15 (or thimet oligopeptidase, TOP) which is involved in neuropeptide metabolism and antigen presentation.⁸² The inhibitor was found trapped in a buried, tunnel-like cavity which contains the catalytic site and connects two gateways of different sizes serving as the entrance of collagenous peptides and exit of hydrolytic products. The tight binding of the inhibitor was mainly attributed to three tyrosine residues (Tyr486, Tyr487, and Tyr490) that form multiple hydrogen bonds with the inhibitor backbone and, in addition, create a hydrophobic nest which accommodates ideally the P₁′/P₂′ Gly-Pro sequence (Figure 16B). Moreover, the indole ring of Trp377 defines the length of substrates being accepted in the primed region of the active site by rotating ∼90° to prevent larger peptides from entering the active site channel. This movement was not observed in the structure of the enzyme bound to a phosphinic inhibitor with a noncollagenous P₁/P₂ sequence,⁸¹ implying that Trp377 controls the specificity of Pz-peptidase A for collagenous peptides of appropriate length.

Figure 16.

(A) Structure and inhibition profile of PPI-2.⁸³ (B) Close-up view of the active site of Pz-peptidase A (cyan C atoms) in complex with phosphinic inhibitor PPI-2 (magenta C atoms, PDB ID: 3AHO). Yellow dashed lines indicate hydrogen-bonding interactions. Zinc atom is shown as a gray sphere.

1.3. Carboxypeptidases

Pharmacological inhibition of angiotensin-converting enzyme (ACE) is a well-established and highly successful approach for the regulation of arterial vasoconstriction and the treatment of several cardiovascular and renal diseases.⁸⁴⁻⁸⁶ ACE is a membrane-bound zinc dipeptidyl carboxydipeptidase that is responsible for the cleavage of a variety of peptide substrates, and it is widely known for its role in the conversion of AngI to the vasoconstrictor hormone AngII. After the introduction of the first ACE inhibitors in the market during the late 70s, it was realized that somatic ACE (sACE) consists of two homologous domains, the N- and C-domain, which are both enzymatically active against the same substrates, albeit with different specificity.^87,88 Interestingly, all inhibitors known at that time were essentially mixed inhibitors of both N- and C-domain active sites of ACE.⁸⁵ In the following years, researchers proposed that the development of C-domain-selective inhibitors could alleviate severe adverse side effects associated with commercial mixed ACE inhibitors, such as cough and angioedema.⁸⁵ These side effects were attributed to the accumulation of bradykinin, an important physiological substrate of ACE which is cleaved equipotently by both domain active sites, whereas AngI is preferentially cleaved in vivo by the C-domain. On the other hand, N-domain-selective ACE inhibitors could increase the levels of antifibrotic peptide AcSDKP, an N-domain specific physiological substrate of ACE, in cases of treatment of diseases involving fibrosis without affecting blood pressure.⁸⁹

During the early 2000s, two milestone achievements paved the way toward the detailed understanding of structural determinants governing domain selectivity in ACE inhibition: the discovery of two highly domain-selective phosphinopeptidic inhibitors, namely RXP407 (Figure 18A) and RXPA380 (Figure 17A) reported by Dive’s research group in 1999⁹⁰ and 2003,⁹¹ respectively, and the determination of the first crystal structure of testicular ACE (tACE), an ACE isoform that is identical to the C-domain of sACE by Sturrock, Acharya, and colleagues in 2003.⁹² A few years later, the crystal structure of tACE bound to the C-domain-selective inhibitor RXPA380 became available, allowing the identification of all key interactions formed between RXPA380 and the C-domain active site of ACE which are absent in the N-domain active site.⁹³ Being the longest inhibitor cocrystallized with tACE, RXPA380 was found to adopt an elongated conformation extending from the S₂ to the S₂′ cavity of the active site (Figure 17B). One important parameter for the observed C-domain selectivity of RXPA380 is that the Trp residue of the inhibitor is better accommodated in the more unpolar S₂′ subsite of tACE by interacting with Val379 and Val380 (Val955 and Val956 of sACE, respectively). These residues are replaced by Ser357 and Thr358 in the S₂′ subsite of the N-domain which results in a more polar environment and reduced binding affinity (Figure 17B). Another observation that supports this hypothesis is the complete loss of inhibitory activity of RXPA380 toward AnCE, an ACE homologue from Drosophila melanogaster that resembles the C-domain of sACE but has Phe363 and Thr364 residues in lieu of Val379 and Val380 residues of tACE.⁹⁴ This was later verified by analysis of the crystal structure of AnCE/RXPA380 which showed that these differences in S₂′/P₂′ hydrophobic interactions between AnCE and tACE have a crucial impact in binding affinity (Figure 17B).⁹⁵

Figure 18.

(A) Structure and inhibition profile of the N-domain-selective ACE inhibitors RXP407⁹⁰ and 33RE.¹⁰¹ (B) Active site of N-ACE (cyan C atoms) in complex with RXP407 (yellow C atoms, PDB ID: 3NXQ) superimposed with active site residues of tACE (gray C atoms, PDB ID: 2OC2) and AnCE (purple C atoms, PDB ID: 2X97). Only residues that differ are indicated. Zinc atom is shown as a gray sphere.

Figure 17.

(A) Structure and inhibition profile of the C-domain selective ACE inhibitor RXPA380.⁹⁶ (B) Active site of tACE (gray C atoms) in complex with RXPA380 (yellow C atoms, PDB ID: 2OC2) superimposed with active site residues of the N-domain of ACE (cyan C atoms, PDB: 3NXQ) and AnCE (purple C atoms, PDB ID: 2X96). Only residues that differ are indicated. Zinc atom is shown as a gray sphere.

Although early docking studies with RXPA380 succeeded in predicting the contribution of S₂′ subsite’s hydrophobicity in the C-domain selectivity,⁹⁶ they failed to detect the significance of S₂/P₂ interactions which was later identified by analysis of the tACE/RXPA380 crystal structure.⁹³ In particular, a beneficial aromatic interaction between Phe391 of tACE and the terminal benzyl group of RXPA380 is lost in the N-domain where a Tyr369 is located at the same position (Figure 17B). Interestingly, kinetic experiments involving various tACE mutants showed that a F391Y mutation leads to a dramatic 34-fold decrease in binding affinity for RXPA380, whereas a marginal 2.6-fold decrease is observed for the S₂′ double mutant VV379,380ST which highlights the principal role of optimized S₂/P₂ interactions in domain selectivity.⁹⁷ Regarding the pseudoproline residue in the P₁′ position of RXPA380, its importance for the C-domain selectivity seems to be related to entropic factors favoring the lining of Trp indole within the S₂′ cavity of tACE. Indeed, replacement of pseudoPro with a pseudoAla caused a drop in C/N selectivity from 3300 to 90. Apparently, the loss of rigidity allows the pseudoAla inhibitor to adopt a conformation that can also be accepted by the N-domain which is evident by the 222-fold increase of binding affinity for the N-domain as compared to the 6-fold increase for the C-domain of sACE.⁹⁶

Eleven years after the discovery of the highly N-domain-selective inhibitor of ACE RXP407, two crystal structures appeared in the literature that provided insights on the origin of N/C selectivity displayed by this phosphinic pseudotetrapeptide: the first structure involved an ACE N-domain mutant⁹⁵ and the second the ACE homologue from Drosophila melanogaster AnCE.⁹⁸ As observed in the case of RXPA380, RXP407 also adopts an extended conformation interacting with the active site of ACE N-domain mutant through a network of twelve stabilizing hydrogen bonds, five more than those found in AnCE. Residues Arg381 and Tyr369 positioned at the S₂ subsite of the N-domain of ACE that are replaced by Glu403 and Phe391 in the C-domain, and by Phe375 and Thr387 in the C-domain-like isoform AnCE, were found in close proximity with the carboxylic acid functionality of the P₂ position of RXP407 (Figure 18B). In accordance with previous computational predictions^99,100 and kinetic analysis of N-domain mutants,⁹⁷ these two interactions were identified as the key interactions for the outstanding N/C selectivity of RXP407. In addition, the crystal structure of the N-domain of ACE bound to inhibitor 33RE (Figure 18A), an analogue of RXP407 incorporating the bioisosteric tetrazole ring in lieu of the P₂ position carboxylic acid functionality, revealed a loss of the critical salt bridge with Arg381 whereas hydrogen bonding with Tyr369 was maintained.¹⁰¹ Given that 33RE displays a ∼1000-fold selectivity for the N-domain, this observation implies that Tyr369 contributes more to the observed N/C selectivity than Arg381, which was further supported by assessing N/C selectivity against N-domain analogues mutated at these specific positions.^97,102

The discovery of FII (Figure 19A), a phosphinic tripeptide inhibitor that simultaneously targets the C-domain of ACE and endothelin 1-converting enzyme (ECE-1)-sparing ACE’s N-domain and neutral endopeptidase (NEP),¹⁰³ prompted a series of crystallographic studies in an attempt to shed light on the origin of this unique inhibition profile.^104,105 This type of inhibition aims to lower plasma concentrations of the two most potent vasoconstrictive peptides, angiotensin I and endothelin 1, without affecting bradykinin levels through NEP inactivation. Unlike all known clinical ACE inhibitors, FII is characterized by an unusual R stereochemical configuration at its P₁′ position which proved to be crucial for its unprecedented selectivity, as revealed by analysis of the crystal structures of FII bound to tACE, N-ACE, and AnCE and comparison with the corresponding complexes with FI, a diastereoisomer of FII with the opposite configuration at the P₁′ position.^104,105 In particular, it was demonstrated that the P₁′ isoxazole side chain of FII is oriented toward the S₂′ cleft of both the C-ACE and the N-ACE domain; however, the more hydrophobic nature of the tACE primed site can accommodate better aromatic moieties as compared to the N-domain active site (Figure 19B).¹⁰⁴ Strikingly, in the case of the C-ACE/FII complex, an additional inhibitor molecule was identified in close proximity to the inhibitor molecule bound in the expected canonical mode (not shown in Figure 19B). This was the first time that an inhibitor molecule was observed at a secondary binding site in ACE, opening the way for the exploration of allosteric domain-specific inhibition.

Figure 19.

(A) Structure and inhibition profile of ACE/ECE dual inhibitor FII.¹⁰³ (B) Active site of N-ACE (cyan C atoms) in complex with inhibitor FII (magenta C atoms, PDB ID: 2XYD) superimposed with the active site of tACE (gray C atoms) in complex with the same inhibitor (yellow C atoms, PDB ID: 2XY9). Zinc atom is shown as a gray sphere.

Carboxypeptidase A (CPA) is the prototype enzyme of the M14 family of carboxypeptidases involved in several biological functions ranging from simple breakdown of dietary proteins or peptides in the digestive track to the regulation of innate immune responses or the processing of neuropeptides.^106,107 CPA, together with thermolysin, was extensively used during the 1980s in pioneering biochemical and crystallographic studies that have shaped our understanding of how Zn-metalloproteases act on their substrates.¹⁰⁸ In these studies, phosphonamidic and phosphonic peptides were employed as transition state analogue ligands of CPA;^109−112 however, it was only a few years ago that the first crystal structures of CPA/phosphinopeptidic inhibitor complexes were reported.¹¹³ As illustrated in Figure 20, the CPA-selective inhibitor 7, which also shows some preference for CPA1 over CPA2 and CPA3 isoforms, is interacting through its phosphinic moiety with the catalytic machinery of CPA, that is the Zn ion, the “general base” Glu270 which is responsible for the deprotonation of the nucleophilic water molecule, and Arg127 which polarizes the carbonyl group of the scissile peptide bond during proteolysis. The C-terminal carboxylate of 7 is anchored by the triad Asn144, Arg145, and Arg127 that serves as the carboxypeptidase recognition site. Interestingly, the origin of compound 7’s selectivity for CPA over CPB lies mainly in the length of the P₁′ side chain because shorter groups can be easily accommodated by CPB as well. Presumably the hydrophobic pocket of CPA shaped by Met203, Ile243, Ile247, Ile255, and Thr268 is more spacious than in the case of CPB where Met203 is replaced by a bulkier isoleucine.

Figure 20.

(A) Structure and inhibition profile of hCPA1 inhibitor 7.¹¹³ (B) Close-up view of the active site of hCPA1 (yellow C atoms) in complex with phosphinic inhibitor 7 (green C atoms, PDB ID: 6I6Z). Gray dashed lines indicate hydrogen bonding and electrostatic interactions of the carboxy terminus of 7. Zinc atom is shown as a gray sphere.

Two more X-ray crystal structures of medicinally relevant carboxypeptidases have been published using phosphinic peptides as tool compounds for their structural characterization (Figure 21). The first one concerns the complex between phosphinic dipeptide EF6525 and plasma CPB or TAFIa (activated thrombin-activatable fibrinolysis inhibitor),¹¹⁴ a Zn-metalloprotease that plays an important role in the regulation of fibrinolysis and is considered a promising pharmacological target for novel antithrombotic therapies.¹¹⁵ The second structure is the inhibitor-bound M14 carboxypeptidase Csd4 from Helicobacter pylori,¹¹⁶ an enzyme implicated in the trimming of un-cross-linked peptidoglycan peptide chains during bacterial wall biosynthesis and emerges as a novel alternative target for antibiotic development.¹¹⁷ In both structures, a similar arrangement is observed in the catalytic zinc environment as described above for CPA, with the only exception of Gln47 of Csd4 that replaces a typical zinc-coordinating histidine, which also interacts with the C-terminal carboxylate of inhibitor 8 (Figure 22). Finally, interactions of the amino group of EF6525’s P₁′ side chain with Asp255, Ser207, and two water molecules polarized by Thr268 and Ala250 clearly support the established preference of CPB for removing basic residues from its substrates.

Figure 21.

(A) Structure and inhibition profile of inhibitor EF6525.⁸³ (B) Close-up view of the active site of plasma CPB in complex with phosphinic inhibitor EF6525 (PDB ID: 3WC7). The inhibitor has been resolved in two alternative conformations that are shown with different C atom colors (yellow and pink) whereas yellow dashed lines indicate hydrogen bonding interactions. Zinc atom is shown as a gray sphere.

Figure 22.

(A) Structure and inhibition profile of inhibitor 8.¹¹⁶ (B) Close-up view of the active site of carboxypeptidase Csd4 in complex with phosphinic inhibitor 8 (PDB ID: 5D2R). Yellow dashed lines indicate hydrogen bonds. Zinc atom is shown as a gray sphere.

2. Biological Studies with Phosphinic Peptide Inhibitors of Zn-Metalloproteases

Phosphinic peptides have been valuable tools in several studies aiming to interrogate biological questions and biochemical mechanisms. Below, these studies are presented based on the target enzyme investigated:

Angiotensin converting enzyme is a key component of the renin–angiotensin system, with important roles in controlling blood pressure. As described above, ACE has two catalytic domains with distinct biological functions¹¹⁸ which can be selectively targeted by inhibitors RXP407 and RXPA380 (Figures 17A and 18A).^90,91 Phosphinic peptide RXP 407 is a potent inhibitor of the N-domain active site of ACE and is also effective in in vivo systems.¹¹⁹ In particular, RXP407 has been shown to block in vivo the degradation of the hemoregulatory peptide AcSDKP¹²⁰ and along with the phosphinic inhibitor RXPA380 has been shown to prevent angiotensin I-induced vasoconstriction.¹²¹ Interestingly, these two inhibitors were proposed to exert their biological effects through distinct contributions on the two active sites of ACE.^91,122 RXP407 has also been shown to increase plasma concentrations of the immunomodulatory peptide AcSDKP in mice and thus may be useful for protecting haematopoietic cells from the toxicity of cancer chemotherapy.⁸⁹ Prodrug versions of phosphinic peptides have been developed,such as the drug fosinopril (Figure 23) used for the treatment of hypertension and some types of chronic heart failure.^123−125 Finally, selective phosphinic inhibitors have also been developed for a homologue of ACE that degrades AngII and has been recently recognized as the major viral entry site for SARS-CoV-2, namely angiotensin-converting enzyme 2 (ACE-2).¹²⁶ A representative example is compound 9 which displays excellent selectivity for ACE-2 over ACE and CPA (Figure 23). Although these inhibitors have not been employed in cardiovascular therapeutics because of the cardioprotective role of ACE2, the renewed scientific interest for ACE-2 is expected to lead to an overall re-evaluation of ACE-2 pharmacological inhibition for potential applications in SARS-CoV-2 infection.¹²⁷

Figure 23.

Structure and inhibition profile of inhibitors PL-265 (currently in clinical phase 1a),¹²⁸ fosinopril (approved antihypertensive drug),¹²⁴ compound 9,¹²⁶ phosphodiepryl 21,²⁷ phosphodiepryl 33,¹³⁶ DG011A,⁴² DG002,⁴¹ RXP1001,¹³⁷ and ΨTRH.¹³⁸ In the case of ΨΤPH, ED₅₀ value corresponds to the inhibition of rat brain membrane-bound PPII activity.

Aminopeptidase N (APN) is a broad specificity aminopeptidase involved in the processing of various peptide hormones and plays central role in the migration, invasion, angiogenesis, and metastasis of tumor cells. APN is also involved in the degradation of endogenous enkephalin opioid peptides. As mentioned above, phosphinic peptide inhibitors have been developed for APN and used to explore several of its biological functions. Long-lasting oral analgesic effects in peripherally controlled pain have been reported using phosphinic dual inhibitors of APN and NEP.¹²⁸ Indeed, several studies have reported the antinociceptive properties of various phosphinic inhibitors targeting APN and NEP, suggesting that they may constitute a promising class of central analgesics.^129−132 In this regard, the effects of compound PL265 (Figure 23), a prodrug form of a dual enkephalinase phosphinopeptidic inhibitor (DENKI) currently undergoing clinical trials, were explored in a murine model of cancer-induced bone pain and revealed synergistic efficacy with various analgesic compounds.¹³³ More recently, PL265 proved effective as a topical treatment for the alleviation of ocular pain and inflammation.¹³⁴ Furthermore, labeled APN inhibitors based on a phosphinic peptide scaffold have also been used for radioautography and positron-emission tomography applications as potential tools for tumor diagnosis and prognosis.^132,135

Phosphinic peptide inhibitors have also been utilized for the exploration of the biology of carboxypeptidases. Compound EF6265 (Figure 21A), a potent inhibitor of thrombin-activatable fibrinolysis inhibitor (TAFI), was demonstrated to protect against sepsis-induced organ dysfunction in rats and helped establish that TAFI plays an important role in the deterioration of organ dysfunction in sepsis through the regulation of fibrinolysis and inflammation.¹³⁹ It was also shown that EF6265 is able to protect against renal dysfunction in rat thrombotic glomerulonephritis through enhancing fibrinolysis.¹⁴⁰ Moreover, EF6265 specifically inhibited plasma carboxypeptidase B activity and enhanced tPA-mediated clot lysis, decreased detectable thrombi factor in a rat microthrombosis model, and enhanced exogenous tissue plasminogen activator-mediated thrombolysis in an arteriovenous shunt model.¹⁴¹ These results validated the use of TAFI inhibitors to enhance physiological fibrinolysis and as novel adjunctive agents to tissue plasminogen activator for thrombo-embolic diseases. Finally, glutamate carboxypeptidase II (GCPII, also known as N-acetyl-l-aspartyl-l-glutamate peptidase I, NAALADase I) inhibition by phosphinic acid-based inhibitors significantly prevented neurodegeneration in a middle cerebral artery occlusion model of cerebral ischemia, suggesting that NAALADase inhibition could be useful for the treatment of both neurodegenerative disorders and peripheral neuropathies.¹⁴²

Neurotensin (NT) and neuromedin N (NN) are two neuropeptides that are efficiently inactivated by peptidases in vitro. Preliminary studies on neurotensin-degrading zinc metallopeptidases (endopeptidase 3.4.24.15 and 3.4.24.16) by using a phosphonamidic peptide inhibitor have shown that their mixed inhibition can prevent neurotensin degradation, in vitro and in vivo, in central and peripheral assays. In particular, it was demonstrated that endopeptidase inhibition was able to potentiate neurotensin-induced antinociception in the mouse hot plate test and the neurotensin-induced contraction of a smooth muscle preparation.¹⁴³ Nevertheless, the hydrolytic sensitivity of the phosphonamide bond to the pH of the medium forced researchers to develop stable phosphinic peptide inhibitors, a challenging task that was accomplished by synthesizing and screening, for the first time, combinatorial libraries of phosphinic peptides.^26,27 Indeed, phosphodiepryl 21 (Figure 23), a phosphinic inhibitor of endopeptidase 3.4.24.15 that spares endopeptidase 3.4.24.16,²⁷ was shown to drastically prolong the forepaw licking latency of mice tested on the hot plate that were injected with neurotensin.¹⁴⁴ Phosphodiepryl 21 was also used to investigate whether endopeptidase 3.4.24.15 contributes to the secretion of amyloid-beta peptide by human transfected cells, leading to the conclusion that this enzyme does not possess significant secretase activity.¹⁴⁵ After the discovery of phosphodiepryl 21, the selective endopeptidase 3.4.24.16 inhibitor phosphodiepryl 33 was identified (Figure 23)²⁶ that was used to establish the in vivo contribution of endopeptidase 24-16 in the central physiological inactivation of neurotensin.¹³⁶ Furthermore, phosphodiepryl 33 also served as a tool compound in experiments that demonstrated that endopeptidase 3.4.24.16 activity in vascular endothelial cells may play an important role in the degradation of bradykinin and/or other peptides in the circulation.¹⁴⁶

Antigen-processing aminopeptidases that include ER aminopeptidase 1 and 2 (ERAP1 and ERAP2) as well as insulin-regulated aminopeptidase (IRAP) process antigenic peptides and can regulate adaptive immune responses, although roles in innate immune responses have also been proposed.¹⁴⁷ Phosphinic peptide inhibitors have been used to interrogate different aspects of the biology of these enzymes. DG013A (Figure 6), a rationally designed phosphinic inhibitor, was shown to affect antigen presentation and T-cell responses in cellular models.⁴¹ The same compound was later used to probe changes induced in the global immunopeptidome of cancer cells due to ERAP1 activity.¹⁴⁸ DG013A and its analogue DG002 (Figure 23) were also used to demonstrate the role of ERAP1 in IL-1β, IL-6, and TNFα production by human peripheral blood monocytes¹⁴⁹ as well as its role in free heavy chain expression and Th17 responses in ankylosing spondylitis (AS), IL-2 production by KIR3DL2CD3ε- reporter cells, and Th17 expansion and IL-17A secretion by AS CD4⁺ T cells, establishing that ERAP1 inhibition may have potential in AS treatment.¹⁵⁰ Phosphinic peptide inhibitor DG026 (Figure 6) was also used to explore cross-presentation, demonstrating modulation of the response of CD8⁺ T-cells isolated from the lymph nodes of Rag-1^–/– C57Bl/6 OT-I mice to FACS-sorted splenic CD8α⁺ and CD11b⁺ conventional DCs exposed to soluble OVA.⁴² More recently, the ERAP2-selective phosphinic peptide DG011A (Figure 23), was used to interrogate ERAP2-dependent changes in the immunopeptidome of MOLT-4 leukemia cancer cells. Inhibitor treatment resulted in the presentation of many novel and potentially antigenic peptides, suggesting that ERAP2 inhibition could prove beneficial in cancer immunotherapy.¹⁵¹

Inhibition of MMPs by phosphinic peptides has been extensively explored. RXP470.1 (Figure 14A), a selective MMP-12 inhibitor, reduced atherosclerotic progression and improved plaque phenotype in apolipoprotein E (apoE) null mice.¹⁵² The same compound was found to worsen postmyocardial infarction cardiac dysfunction by delaying inflammation resolution.¹⁵³ RXP470.1 was also used to unravel a novel role for MMP-12 in antiviral immunity, mediated by clearing systemic IFN-a. Accordingly, treatment of coxsackievirus type B3-infected wild-type mice with RXP470.1 elevated systemic IFN-α levels and reduced viral replication in pancreas, suggesting that inhibition of extracellular MMP-12 could lead to novel antiviral treatments.¹⁵⁴ RXP470.1 has been also used to demonstrate the role of MMP-12 in the C-terminal truncation of IFN-γ as part of a negative feedback mechanism for the proinflammatory IFN-γ activation of macrophages.¹⁵⁵ Another phosphinic peptide inhibitor of MMPs, RXP03 (Figure 15A), prevented rat liver injury after induction of normothermic ischemia, suggesting that MMP inhibitors could have clinical relevance in liver-associated ischemic diseases.¹⁵⁶ Similarly, RXP409 (Figure 15A) protected rat livers from prolonged cold ischemia–warm reperfusion injury.¹⁵⁷ RXP03 also showed promising results in a C26 colon carcinoma mouse model, retarding tumor growth, albeit its effects were varied depending on the treatment schedule, suggesting complex spatiotemporal functions of MMPs in cancer growth.¹⁵⁸ Furthermore, triple-helical peptide inhibitors of MMPs were examined for potency in a mouse model of multiple sclerosis leading to a marked reduction in severity and weight loss. The same compounds were used in a mouse model of sepsis where they minimized lung damage and increased production of the anti-inflammatory cytokine IL-10.¹⁵⁹ Finally, phosphinic peptides have been used in the development of FRET probes able to detect MMP-12 and MMP-13 activities in vivo in mouse models of inflammatory arthritis.¹⁶⁰

In addition to the above systems, phosphinic peptide inhibitor 8 (Figure 22A) of H. pylori metalloprotease Csd4 caused significant cell straightening when incubated with H. pylori, presumably by inhibiting cell-shape determining proteases, thus highlighting a potential novel antimicrobial target.¹¹⁶ Moreover, RXP1001 (Figure 23), a broad-spectrum inhibitor of astacins, was found to be able to regulate the procollagen processing by BMP-1/tolloid-like proteinases in human keratocytes and HT1080 cells at μM concentrations.¹³⁷ Finally, intramedial septum administration of a phosphinic inhibitor (ΨTRH, Figure 23) of pyroglutamyl peptidase II (PPII), a TRH (thyrotropin-releasing hormone)-degrading Zn-peptidase, in male Wistar rats subjected to ethanol-induced narcosis enhanced the analeptic effect of TRH, thus strengthening the rationale of developing PPII inhibitors in order to explore new CNS disorder therapies.¹⁶¹

3. Molecular Probes for the Detection of Zn-Metalloproteases Based on a Phosphinic Peptide Scaffold

Besides the extensive use of phosphinic peptides as tool compounds in structural and biological studies of pharmacologically relevant Zn-metalloproteases, their transformation into molecular probes for applications in diagnostics and proteomics has also been the topic of several research efforts during the last 15 years. Given the great recent advances in the field of chemical biology and the high binding affinity and target selectivity of certain phosphinic peptide inhibitors, these compounds have been used for the detection of functional Zn-proteases (mainly MMPs) by acting as the recognition warhead of the probe during its interaction with the enzyme. In general, the transformation of inhibitors to molecular probes involves the attachment of a reporter tag (fluorescent, radioactive, etc.) separated from the phosphinic ligand by a flexible linker. Variations from this original designing theme have also been reported (noncovalent probes, traceless covalent probes) and will be described below in greater detail.

In the field of MMPs, the multifaceted role of these enzymes in severe pathologies such as cancer or atherosclerosis prompted researchers to develop probes able to target the active site of these enzymes and record their functional state in complex proteomes.⁶⁹ Unlike serine or cysteine proteases which can be covalently modified by electrophilic probes, covalent modification of zinc metalloproteinases, such as MMPs, is more challenging due to the lack of a conserved active-site nucleophile. Unspecific photo-cross-linking methods have been employed in this regard, but the low sensitivity of early photoreactive activity-based probes of Zn-metalloproteases^162,163 initiated research efforts toward the development of improved tools that could react with higher cross-linking yields. In 2007, on the basis of a novel chemical strategy leading to potent isoxazole/isoxazoline-containing phosphinopeptidic inhibitors,¹⁶⁴ Dive and colleagues described a highly efficient probe (Figure 24, probe 10) bearing a photoactivatable azido group at the distal end of the P₁′ side chain and radioactive tritium as a reporter tag.¹⁶⁵ Probe 10 is able to target several MMPs, reaching a detection limit of 2.5 fmol for hMMP-12 which can be further improved by a factor of 2 when the pH before photoactivation is raised to 12.¹⁶⁶ In addition, adjustment of the pH to higher values was also found to be beneficial for MMP-3 which supported prior experimental evidence that the reactive intermediate formed by the photoexcitation of azide preferentially targets basic residues in position 241 (Lys for MMP-12, His for MMP-3) located at their S₁′ loop.^166,167 The high variability of this position explains the low efficacy of probe 10 against other MMPs; therefore, a next-generation pan-MMP probe was developed bearing a trifluoromethyldiazirine in its P₁′ position (probe 11, Figure 24).¹⁶⁸ This isoxazole-containing phosphinopeptidic probe was able to efficiently label human MMP-2, -9, -12, and -13 in complex proteomes and also murine MMP-12 which is not sufficiently detected by probe 10. Indeed, in the first attempt to detect an MMP in animal fluids without sample preconcentration, probe 10 succeeded in labeling the catalytic domain of murine MMP-12 in bronchoalveolar lavage fluids (BALf) from mice exposed to ultrafine carbon black nanoparticles, an animal model of chronic obstructive pulmonary disease (COPD).¹⁶⁹

Figure 24.

(a) Chemical structures of photoreactive MMP probes 10 and 11. (b) Schematic representation of the detection of active forms of MMPs by photoactivatable radiolabeled activity-based probes 10 and 11.

The high inhibitory potency that can be achieved by phosphinic peptides in combination with their slow-binding properties⁵⁸ triggered research efforts toward affinity, noncovalent active-site-directed probes for MMPs, as an alternative to the photo-cross-linking approach which is difficult to implement in vivo. Indeed, biotinylated probe 12 (Figure 25), that displays low picomolar activity for six MMPs, allowed the capture of active forms of three hMMPs (2, 8, and 12) from a tumor extract and their unambiguous detection by MALDI-MS analysis in a single affinity experiment.¹⁷⁰ Furthermore, circumventing photoexcitation for MMP labeling was also the main objective of Devel’s research team who recently introduced an innovative “traceless” approach for the covalent modification of hMMP-12 without any external trigger.¹⁷¹ Probe 13 (Figure 25), which is based on the structure of the selective MMP-12 inhibitor RXP470.1 (Figure 14A), contains a reactive cleavable acyl imidazole linker as a leaving group between the recognition element (RXP470.1) and a fluorescent analytical tag. Upon binding of the probe, attack of a nearby nucleophilic residue transfers the tag on the protein whereas the inhibitor part is released and the enzymatic activity is partially restored. The authors unambiguously identified Lys177 of hMMP-12 as the reactive nucleophilic residue and demonstrated that this probe can specifically label hMMP-12 in a nanogram scale in complex proteomes.

Figure 25.

(a) Chemical structure of affinity MMP probe 12. (b) Schematic representation of the interaction of probe 12 with the active site of MMPs. (c) Chemical structure of activity-based MMP probe with no photolabile group 13. (d) Schematic representation of the interaction of probe 13 with the active site of MMPs illustrating the cleavage of the probe by an active-site nucleophile.

The selective MMP-12 phosphinic inhibitor RXP407.1 has also been incorporated into optical and radiolabeled probes for vascular imaging molecular studies.^172−174 With RXP470.1 as a starting template, several probes where evaluated as candidates for the detection of active forms of hMMP-12 in preclinical models, leading to probe 134 (Figure 26) which displayed high in vivo stability, rapid renal clearance, and slow dissociation of the hMMP-12/probe complex.¹⁷² After further optimization, researchers developed probe 15 bearing the near-infrared fluorophore ZW800-1 (Figure 26) which was successfully used in the specific detection of hMMP-12 in murine models of sterile inflammation and carotid aneurysm.¹⁷³ Finally, due to the inherent limitations of optical imaging that preclude the application of fluorescent probes for noninvasive imaging, the development of a radiolabeled homologue was pursued based on the RXP470.1-based optical probes.¹⁷⁴ In this regard, radiotracer 16 (Figure 26) was evaluated in a murine model of abdominal aortic aneurysm exhibiting high specificity for MMP-12, optimal pharmacokinetic and biodistribution properties and high in vivo uptake compared to nonaneurysmal aorta.

Figure 26.

Chemical structures of RXP470.1-derived optical probes 14 and 15 and radiotracer 16.

Phosphinic peptides have been extensively used as recognition elements in different types of MMP-targeting probes; however, only a few probes have been developed for other Zn-metalloproteases. In such a case, the selective carboxypeptidase A inhibitor 7 (Figure 20A) was transformed to a high affinity, optical probe by attaching a 6S-IDCC fluorescent tag to its N-terminal end, but no further evaluation of this probe was reported. In addition, an ERAP1 photoactivatable probe (DG023, Figure 27) was recently described as a molecular tool for the covalent modification of ERAP1.⁴³ In particular, the photolabile benzophenone group attached to the C-terminus of probe DG023 was cross-linked to the protein after UV irradiation, revealing the exact position of the C-terminal regulatory site of ERAP1 and verifying parallel crystallographic studies of ERAP1 bound to phosphinic peptides (vide supra, Figure 11).

Figure 27.

Chemical structure of ERAP1 photoactivatable probe DG023.

4. Future Perspectives and Conclusions

During the last decades, phosphinic peptides have been extensively used as tool compounds for structural, biological, and pharmacological studies of Zn-metalloproteases.⁸⁻¹³ Regarding structural studies, phosphinic peptides have been an ideal choice for cocrystallization ligands of Zn-proteases for two main reasons. The first concerns their ability to inhibit target enzymes by acting as transition-state analogues which offers a powerful tool for the delineation of catalytic mechanisms with important implications to drug design. Starting from the first crystallographic analyses of phosphinic peptides with astacin²⁰ and MMP-11⁷⁷ that uncovered fundamental features of catalytic processes mediated by Zn-proteases up until the very recent use of a phosphinic peptide in the investigation of the structural basis of antigen selection by aminopeptidase ERAP1,⁵¹ this class of pseudopeptides have provided mechanistic information for most of the currently known medicinally important Zn-metalloproteases. The second reason relates to the peptide-like structure of these compounds which facilitates rational inhibitor design and allows direct probing of proteases’ specificity sites and exploration of structural determinants that govern inhibitor selectivity. Prominent examples in this regard are the crystallographic studies with ACE domain-selective inhibitors RXPA380⁹³ and RXP407⁹⁸ or the more recent structural analysis of the IRAP/DG026 complex.⁴⁴ Obviously, the availability of selective tool compounds is of paramount importance not only for structural but also for biological studies, and phosphinic peptides present certain advantages as compared to other classes of Zn-metalloprotease inhibitors. In particular, unlike hydroxamates or thiolates, phosphinic peptides are able to interrogate both primed and unprimed subsites of a Zn-protease, a property that in the past has allowed the discovery of RXPA380⁹¹ and RXP407,⁹⁰ whereas previously discovered inhibitors were unable to probe both sides of ACE’s active sites. In addition, the phosphinic acid group is a weaker zinc-binding ligand which means that binding affinity is determined by noncovalent interactions at the specificity pockets and not by the binding capacity of the zinc-binding group. Consequently, proper structural optimization of the inhibitor may bring out subtle differences between similar active sites and lead to increased selectivity. This combination of favorable properties (weak ZBG, extended secondary interactions) has resulted of several cases of phosphinopeptidic inhibitors with a remarkable ability to differentiate structurally and functionally similar Zn-proteases such as the endopeptidases 24.15 and 24.16 (phosphodiepryl 21 and 33, Figure 23),^26,27 the two domain active sites of ACE (RXP407 and RXPA380, Figures 17 and 18),^90,91 ACE-2 from ACE and CPA (inhibitor 9, Figure 23),¹²⁶ or MMP-12 from eight other members of the matrixin family (RXP470.1, Figure 12A).⁷¹ In the latter case, a recent study showed that selectivity is severely impaired when the phosphinic group of RXP470.1 is replaced by other ZBGs such as hydroxamic or carboxylic moieties;⁷³ however, the lack of similar studies in the literature do not allow generalization of this conclusion. Another major concern that must be taken under consideration is that optimization opportunities strongly depend on synthetic advances in the field,^10,11,13,175 as highlighted by the application of combinatorial chemistry principles in the discovery of selective inhibitors of several Zn-metalloproteases^26,27,71,90 or the successful preparation of capricious sequences which allowed the synthesis of RXPA380^96,176 or inhibitors of aminopeptidase A.^56,177 Finally, the transformation of phosphinic inhibitors of MMPs to molecular probes has set the basis for similar studies with other pharmacologically important Zn-metalloproteases, expanding the range of applications of this privileged class of tool compounds. Future development of novel, improved tool compounds or probes thereof will be largely accelerated by addressing unmet challenges such as stereoselective synthetic strategies, late-stage diversification approaches, or synthesis of unusual, conformationally constrained analogues.^178−180

From a biological point of view, it has been demonstrated that phosphinic peptides are devoid of any toxicity in in vitro cell assays^{137,148,151,158} or in mice animal models under chronic intravenous administration for over 1 month.¹⁵² Furthermore, they are characterized by increased metabolic stability despite the presence of peptide bonds in their structure, as previously established in the cases of pseudotripeptides RXP407⁹⁰ and RXP03¹⁵⁸ and pseudotetrapeptide RXPA380.⁹¹ Presumably, the absence of any detectable metabolites may be attributed to the negatively charged phosphinic group under physiological conditions which prevents their interaction with nonspecific proteases of plasma and/or their uptake by hepatocytes that could lead to their modification by cytochrome P450 enzymes. This low propensity of phosphinic peptides to penetrate cell membranes makes them ideal tool compounds when plasma-soluble or membrane-bound Zn-metalloproteases are targeted. A noticeable example is the case of RXP470.1, a selective MMP-12 inhibitor, which displays no detectable cell permeability and thus is able to differentiate between the beneficial intracellular MMP-12 that increases the secretion of IFN-α after acute viral infection and the harmful extracellular MMP-12 which degrades IFN-α and weakens antiviral immune response.¹⁵⁴ For intracellular Zn-metalloproteases, partial membrane penetration has been reported in certain cases^145,148 which implies that cell permeability may be inhibitor-dependent or even cell-dependent. Adoption of prodrug strategies, as in the case of known FDA-approved organophosphorus drugs for the treatment of viral infections,^181,182 could be a viable approach to ameliorate this potential weakness. Interestingly, in a recent comparative study of the pharmacokinetic parameters of different phosphine oxides, esterification of a phosphinic acid group massively increased Caco-2 permeability.¹⁸³ Although prodrug ester technologies have been used in the field, until now they mainly aim for the improvement of oral availability of inhibitors targeting extracellular Zn-metalloproteases.^125,128,131 Future advances in that direction are expected to significantly increase the versatility offered by the toolbox of phosphinic peptides.

Glossary

Abbreviations

CCR2

CC chemokine receptor 2

CCL2

CC chemokine ligand 2

CCR5

CC chemokine receptor 5

TLC

thin layer chromatography

ACE

angiotensin-converting enzyme

AnCE

ACE from Drosophila melanogaster

AngI

angiotensin I

AngII

angiotensin II

APA

aminopeptidase A

APN

aminopeptidase N

APN

aminopeptidase N

AS

ankylosing spondylitis

BALf

bronchoalveolar lavage fluids

CD

cluster of differentiation

CNS

central nervous system

COPD

chronic obstructive pulmonary disease

CPA

carboxypeptidase A

CPB

carboxypeptidase B

DC

dentritic cells

DENKI

dual enkephalinase inhibitor

ECE-1

endothelin-1-converting enzyme

ECM

extracellular matrix

ER

endoplasmic reticulum

ERAP

endoplasmic reticulum aminopeptidase

FACS

fluorescence-activated cell sorting

FRET

fluorescence resonance energy transfer

GCPII

glutamate carboxypeptidase II

IFN

interferon

IRAP

insulin-regulated aminopeptidase

LAP

leucine aminopeptidase

MALDI-MS

matrix-assisted laser desorption/ionization mass spectrometry

MMP

matrix metalloproteinase

MSS

musculosceletal syndrome

NAALADase

N-acetyl-l-aspartyl-l-glutamate peptidase

NEP

neutral endopeptidase

Nm

Neisseria meningitides

NN

neuromedin N

NT

neurotensin

OVA

ovalbumin

Pf

Plasmodium falciparum

PPII

pyroglutamyl peptidase II

SAR

structure–activity relationships

SARS

severe acute respiratory syndrome

TAFI

thrombin-activatable fibrinolysis inhibitor

TNFα

tumor necrosis factor alpha

TOP

thimet oligopeptidase

tPA

tissue-type plasminogen activator

TRH

thyrotropin-releasing hormone

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

D.G. acknowledges support by funds from the Special Account for Research Grants of NKUA (account nos. 10504 and 18638).

The authors declare no competing financial interest.

Dedication

This work is dedicated to Dr. Vincent Dive for his outstanding contribution to the field.