Iterative Optimization of the Cyclic Peptide SFTI-1 Yields Potent Inhibitors of Neutrophil Proteinase 3

Abstract

Neutrophils produce at least four serine proteases that are packaged within azurophilic granules. These enzymes contribute to antimicrobial defense and inflammation but can be destructive if their activities are not properly regulated. Accordingly, they represent therapeutic targets for several diseases, including chronic obstructive pulmonary disease, cystic fibrosis, and rheumatoid arthritis. In this study, we focused on proteinase 3 (PR3), a neutrophil protease with elastase-like specificity, and engineered potent PR3 inhibitors based on the cyclic peptide sunflower trypsin inhibitor-1 (SFTI-1). We used an iterative optimization approach to screen targeted substitutions at the P1, P2, P2′, and P4 positions of SFTI-1, and generated several new inhibitors with K_i values in the low nanomolar range. These SFTI-variants show high stability in human serum and are attractive leads for further optimization.

Neutrophils are widely recognized as critical first responders in the innate immune system. When an injury or infection occurs, neutrophils follow a trail of chemotactic cues to quickly reach the site of damage and eliminate invading pathogens.¹ Neutrophils have three known mechanisms for killing pathogens: phagocytosis, degranulation, and releasing neutrophil extracellular traps (NETs). These processes are assisted by different granules that neutrophils assemble during maturation, including azurophilic granules, specific granules, and gelatinase granules. Each type of granule holds different cytotoxic cargoes, such as oxidative enzymes, proteolytic enzymes, and antimicrobial peptides.²

An important component of azurophilic granules are serine proteases from the chymotrypsin family. To date, four neutrophil serine proteases (NSPs) have been characterized, with different NSPs showing distinct specificities.^3,4 Neutrophil elastase (NE) and proteinase 3 (PR3) have elastase-like specificity and cleave after small aliphatic residues. Cathepsin G has chymotrypsin-like specificity but also cleaves after Lys or Leu to some extent. The most recently identified NSP, NSP4, has trypsin-like specificity and cleaves after Arg. Due to their diverse specificities, NSPs cleave a variety of substrates that modulate diverse physiological functions. NSPs have a direct role in antimicrobial defense by digesting pathogens that have been phagocytosed or captured in NETs.² They also contribute to tissue remodeling by cleaving extracellular matrix proteins (elastin, collagen, and proteoglycans) and can modulate signaling pathways as NE and PR3 act as biased agonists of protease-activated receptors.⁵ NSPs also activate a range of pro-inflammatory cytokines, including IL-1β, IL-8, and, most recently, IL-36,⁶ and inactivate anti-inflammatory proteins, including progranulin.⁷

Although NSPs are critical for immune defense, they can be highly destructive if their activities are not tightly regulated. Control of NSP activity is usually achieved by a range of protease inhibitors, including serpins and Laskowski (standard mechanism) inhibitors, such as elafin and secretory leukocyte proteinase inhibitor.³ However, if the balance between NSP activity and inhibitory control is lost, NSPs are capable of causing tissue damage and chronic inflammation, as proposed in chronic obstructive pulmonary disease, cystic fibrosis, and rheumatoid arthritis.³ Accordingly, NSPs are widely recognized as therapeutic targets and considerable effort has been directed toward developing NSP inhibitors. Of the four known NSPs, NE has received the most attention, and several NE inhibitors have progressed to clinical trials.⁸ However, these trials have had limited success to date,⁸ and there is now growing interest in targeting other NSPs as they also contribute to key disease mechanisms.

One protease of interest is PR3 as it shares several substrates with NE, including extracellular matrix proteins, protease-activated receptor-1, and progranulin.^3,5,7 However, PR3 also displays unique traits that relate to its substrate repertoire and regulation by protease inhibitors. For example, PR3 has been reported to activate a wider range of pro-inflammatory molecules, including pro-IL1β, pro-IL8, and the cathelicidin precursor hCAP18.^3,9 PR3 has also been shown to activate procaspase-3 to promote neutrophil apoptosis via a pathway that is independent of caspase-8 or caspase-9,¹⁰ and cleave the anti-inflammatory protein annexin 1.¹¹ Additionally, whereas NE is potently inhibited by elafin and secretory leukocyte proteinase inhibitor, PR3 is only inhibited by elafin,^12,13 suggesting that it is subject to different regulatory controls to NE. Although PR3 represents an attractive target for inhibitor development, the number of engineered NE inhibitors far outweighs those for PR3. Additionally, most peptide-based inhibitors for PR3 are irreversible, and it has proven challenging to develop potent, reversible PR3 inhibitors.¹⁴

In a recent study, we designed a potent and selective inhibitor for cathepsin G based on sunflower trypsin inhibitor-1 (SFTI-1).¹⁵ SFTI-1 is a 14-amino acid cyclic peptide (Figure 1A) that is produced in sunflower seeds and is a potent, reversible inhibitor of trypsin (K_i < 0.1 nM).¹⁶ To produce cathepsin G inhibitors, we identified preferred P1–P4 sequences to substitute into SFTI-1 by screening a sparse matrix substrate library¹⁵ that was designed using the positional scanning data for cathepsin G.⁴ The P1–P4 specificity of PR3 has also been characterized using high-throughput libraries (positional scanning⁴ and a hybrid combinatorial substrate library¹⁷) and individual synthetic peptides with defined sequences.¹⁸⁻²⁸ However, many of the residues preferred by PR3 are hydrophobic, which presents a challenge for synthesizing and screening a sparse matrix library.

Figure 1.

PR3 inhibitor design strategy. (A) Structure of SFTI-1 (PDB ID 1SFI). (B) Model of the SFTI-variant used as a design template showing the P1, P2, P2′, and P4 residues (for clarity, side chains not displayed). Inhibitor sequences are shown below, together with amino acids substituted at each position.

Therefore, we considered an alternative strategy that involved screening residues directly in the SFTI-scaffold. The first residue to be screened was the most important binding contact (P1). After identifying the optimal P1 residue, our plan was to incorporate this residue into the lead variant and proceed to screening the next most important position. The scaffold that we selected for iterative optimization was an engineered SFTI-variant (Figure 1B) that we previously reported²⁹ and used to produce inhibitor libraries.^30,31

PR3 has elastase-like P1 specificity and prefers small aliphatic amino acids, including both proteinogenic (Ala and Val)¹⁸ and nonproteinogenic (norvaline [Nva] and α-aminobutyric acid [Abu]) amino acids.^18,21 Therefore, we synthesized four head-to-tail cyclic peptides (compounds 1–4, Table 1, Supplementary Figure 1) that each contained a separate P1 residue favored by PR3. Each compound was assessed by performing competitive inhibition assays and determining the inhibition constant (K_i). Assays were also performed using an endogenous PR3 inhibitor, elafin, and the K_i value we determined (0.74 ± 0.06 nM, Table 2) is similar to the reported value.¹³ Among the four SFTI-variants, 3 (P1 Abu) was the most potent inhibitor (K_i = 9.8 ± 1.2 nM, Table 2), which was 2.3-fold more potent than 4 (P1 Nva) and 5.2-fold more potent than 1 (P1 Ala). Therefore, we selected Abu as the P1 residue and proceeded with screening the P2 position.

Table 1. Sequences and Validation Data for SFTI-Variants.

compound	series	sequencea	theoretical [M + H]+	determined [M + H]+	purity (%)
1	I	c[GTCTAlaSIPPICNPN]	1367.6	1367.7	99.4
2	I	c[GTCTValSIPPICNPN]	1395.6	1395.6	99.2
3	I	c[GTCTAbuSIPPICNPN]	1381.6	1381.7	97.6
4	I	c[GTCTNvaSIPPICNPN]	1395.6	1395.7	97.9
5	II	c[GTCDAbuSYPPICNPN]	1445.6	1445.6	96.8
6	II	c[GTCDAbuSDPPICNPN]	1397.5	1397.5	97.6
7	II	c[GTCYAbuSYPPICNPN]	1493.7	1493.6	99.6
8	II	c[GTCYAbuSDPPICNPN]	1445.6	1445.6	98.9
9	II	c[GTCYAbuSEPPICNPN]	1459.6	1459.5	99.6
10	II	c[GTCYAbuSWPPICNPN]	1516.7	1516.6	99.2
11	III	c[GNleCYAbuSYPPICNPN]	1505.7	1505.6	99.7
12	III	c[GTrpCYAbuSYPPICNPN]	1578.8	1578.6	99.7
13	III	c[GBipCYAbuSYPPICNPN]	1615.9	1615.6	99.9
14	III	c[GKZ*CYAbuSYPPICNPN]	1689.3	1688.6	99.9
15		c[GBipCYNvaSYPPICNPN]	1629.9	1629.6	99.9

^aCyclized Gly1–Asn14 with a disulfide bond between Cys3–Cys11. New substitutions are shown in bold. Abbreviations: Abu, α-aminobutyric acid; Bip, 4,4′-biphenyl-l-alanine; KZ*, Nε-(2-chlorobenzyloxycarbonyl)-l-lysine; Nle, norleucine; Nva, norvaline.

Table 2. Inhibition Data for Selected PR3 Inhibitors.

compounda	protease	Ki (nM) ± SEM
Elafin	PR3	0.74 ± 0.06
	NE	0.15 ± 0.03
1	PR3	51 ± 2.6
2	PR3	245 ± 20
3	PR3	9.8 ± 1.2
4	PR3	23 ± 4.1
7	PR3	7.0 ± 1.1
	NE	3.2 ± 0.3
	Cathepsin G	IC50 > 10,000
	Chymase	IC50 > 10,000
	Trypsin	IC50 > 10,000
	Chymotrypsin	IC50 > 10,000
	Thrombin	IC50 > 10,000
13	PR3	17 ± 1.8
	NE	24 ± 6.5
	Cathepsin G	IC50 > 10,000
	Chymase	IC50 > 10,000
	Trypsin	IC50 > 10,000
	Chymotrypsin	IC50 > 10,000
	Thrombin	IC50 > 10,000
15	PR3	6.1 ± 0.6
	NE	16 ± 2.8
	Cathepsin G	IC50 > 5,000
	Chymase	IC50 > 10,000
	Trypsin	IC50 > 10,000
	Chymotrypsin	463 ± 36
	Thrombin	IC50 > 10,000

^aCompounds 1–15 are cyclized Gly1–Asn14 with a disulfide bond between Cys3–Cys11 (sequences listed in Table 1).

The P2 specificity of PR3 has been characterized using peptide substrates or inhibitors, which revealed that Asp^19,25,26 and Tyr²² were favored. These findings are in agreement with data from high-throughput library screens.^4,17 We also elected to combine P2 optimization with substitutions at P2′. Here, synthetic substrates containing P2′ Asp have been shown to be highly preferred by PR3,^20,21,26 with modeling analyses indicating that P2′ Asp forms a salt bridge with Arg143.^20,26 In the PR3 crystal structure, Arg143 occupies a prominent position in the S2′ subsite.³² PR3 has also been modeled in complex with the Kazal-type inhibitor OMTKY3,²⁰ where Tyr is present at P2′ and could engage Arg143 via a cation−π interaction. Therefore, to test different interactions with Arg143, we selected Asp and Tyr for substitution at P2′ and synthesized four cyclic peptides (5–8) where each P2′ replacement was combined with either Asp or Tyr at P2.

Compound 7 (P2 Tyr, P2′ Tyr) was the most potent PR3 inhibitor in series II (Figure 2). Indeed, 7 showed much higher activity at 40 nM than each of the other peptides at 160 nM. This finding indicated that replacing Tyr with Asp at either P2 (5) or P2′ (8) was detrimental to the activity of the SFTI-based inhibitors. The weak activity of inhibitors with P2′ Asp was unexpected; hence, we synthesized two additional inhibitors (9–10) to test a second acidic (Glu) or aromatic (Trp) residue at P2′. Compound 10 (P2′ Trp) showed similar activity to 7 (P2′ Tyr), whereas 9 (P2′ Glu) was a weak inhibitor of PR3. Based on these findings, we selected 7 for further optimization and tested substitutions at the P4 position.

Figure 2.

Inhibitory activity of 5–14 against PR3. Each variant was tested in competitive inhibition assays at two concentrations (x-axis). Compound numbers and unique substitutions are shown below. Structures of the nonproteinogenic amino acids KZ* and Bip are shown on the right.

After examining the reported P4 specificity of PR3, we designed four additional SFTI-variants (11–14, Table 1). The residues selected to substitute at P4 were Trp and Nle, which were the two most preferred residues in a positional scanning screen,⁴ together with 4,4′-biphenyl-l-alanine (Bip) and Nε-(2-chlorobenzyloxycarbonyl)-l-lysine (Lys-[2-Cl-Z], shortened to KZ*), which were favored in a hybrid combinatorial substrate library.¹⁷ These findings are largely consistent with a recent study that assessed P4 substitutions in peptide-based phosphonate inhibitors and found that hydrophobic residues were preferred.²⁸ Overall, the four inhibitors in the P4 screen showed relatively similar activity (Figure 2), indicating that each substitution was well-tolerated. Compounds 11–14 were also tested against NE, which revealed that 14 (P4 KZ*) was the most potent NE inhibitor, whereas 13 (P4 Bip) had the weakest activity against this enzyme (Supplementary Figure 4).

To characterize 13 in further detail, we determined the inhibitor’s K_i against PR3 and NE. For comparison, we also included 7 as it was the lead inhibitor in the previous series. Compound 7 was a potent inhibitor of both PR3 (K_i = 7.0 ± 1.1 nM) and NE (K_i = 3.2 ± 0.3 nM), but displayed no significant activity against cathepsin G, chymase, chymotrypsin, trypsin, and thrombin (Table 2). Replacing P4 Thr with Bip to generate 13 decreased the inhibitor’s activity against NE by 7.5-fold but also decreased its activity against PR3 by 2.4-fold (K_i = 17 ± 1.8 nM, Table 2). As both 7 and 13 displayed potent activity toward PR3 and NE, we synthesized an additional peptide to test whether substituting P1 Abu with Nva would alter the inhibitor’s selectivity. P1 Nva is found at the P1 position of reversible inhibitors^24,33 and irreversible inhibitors^27,28 that target PR3 and was well tolerated by PR3 in the SFTI-screen (4, Table 2). Substituting P1 Abu with Nva (15) led to an improvement in activity against PR3 (K_i = 6.1 ± 0.6 nM) and modest selectivity over NE (2.6-fold). Interestingly, increasing the length of the P1 side chain by one carbon (Abu to Nva) also increased the inhibitor’s activity against two enzymes with chymotrypsin-like specificity, particularly chymotrypsin (K_i = 463 ± 36 nM) and, to a lesser extent, cathepsin G (IC₅₀ > 5 μM, Table 2).

Having characterized the activity of 7, 13, and 15, we next analyzed each peptide by NMR spectroscopy to examine whether they retained well-defined structures. In the ¹H 1D spectrum for each peptide, peaks in the amide spectral region (7–9 ppm) were sharp and well-dispersed, indicating that the SFTI-variants had well-defined structures (Supplementary Figures 2–3). TOCSY and NOESY experiments were also performed to assign spin systems for each residue and calculate secondary Hα chemical shifts (Figure 3). Generally, these values were similar to previously reported data for SFTI-1.^34,35 Local differences were observed at several positions, including the segment spanning residues 2–5, which contains three substituted amino acids. Additionally, an upfield shift of 0.7 ppm was identified for Pro8 Hα. Both of these effects have been observed in previous studies, either when the P4, P2, and P1 residues were also substituted in the SFTI-scaffold³¹ or Pro8 was preceded by an aromatic residue (likely due to a ring current effect).^31,34

Figure 3.

Secondary Hα chemical shifts for 7, 13, and 15. SFTI-1 is also shown (black) and its sequence is illustrated below. Modified residues are enclosed by a square, and the new substitutions are listed below.

We also performed molecular dynamics simulations to study the interaction between PR3 and 15, which contains preferred substitutions at P4, P2, P1, and P2′ (Figure 4). The modeled complex indicated that P2′ Tyr was well placed to engage Arg143 via a cation−π interaction. In the S2 subsite, Lys99 was in close proximity to the P2 Tyr residue, as well as Asn12 of the inhibitor, where a hydrogen bond was present in one-third of frames. At P4, Bip extended into the S4 subsite with each ring contributing to separate interactions, including a π–π interaction with Phe215 and a cation−π interaction with Arg177.

Figure 4.

Model of 15 bound to PR3. Compound 15 is shown in stick representation, and PR3 is shown as a (A) ribbon diagram (transparent surface) or (B) surface colored by electrostatic potential. Inhibitor residues are labeled in three letter codes (green) and PR3 residues are labeled in single letter codes (black).

Finally, we examined the stability of 7, 13, and 15 in human serum to assess their potential for further optimization. These assays were also performed using an acyclic, disulfide-deficient mutant of SFTI-1 that was rapidly degraded and essentially undetectable after 2 h incubation in human serum at 37 °C (Figure 5). By contrast, each of the cyclic SFTI-variants was highly resistant to degradation, with 90–95% of peptide remaining after 24 h. This result highlights one of key advantages of macrocyclic peptides as they are widely recognized as being stable engineering frameworks.

Figure 5.

Stability of 7, 13, and 15 in human serum at 37 °C. An acyclic, disulfide-deficient mutant of SFTI-1 was also tested (open circles), and a 2 h time point was included for this peptide as it was rapidly degraded. Data represent the mean ± SEM from three experiments.

Overall, this study has identified several cyclic peptide-based PR3 inhibitors that show potent activity, are stable in human serum, and are promising leads for further optimization. Our design strategy was based on an iterative approach where we screened targeted substitutions at SFTI’s key binding contacts, starting with P1. This approach utilized the existing specificity data for PR3 in a different way to our previous engineering studies, including our study on cathepsin G.¹⁵ Guided by this approach, we produced PR3 inhibitors with K_i values in the low nanomolar range, and it is likely that this design strategy will be applicable to other proteases of interest.

Compared to existing reversible PR3 inhibitors, our SFTI-variants show higher potency by several orders of magnitude. Previous reversible inhibitors were produced by chemically modifying the P1–P1′ segment of selective PR3 substrates, either by introducing an aza-amino acid at the P1 position or a ketomethylene dipeptide isostere at P1–P1′.^24,33 In these peptidomimetics, the binding kinetics of the parent substrate show a close correlation with the activity of the corresponding inhibitor.^24,33 Accordingly, the PR3 selectivity of the parent substrate is also preserved, although both inhibitors have K_i values for PR3 in the low micromolar range. By contrast, SFTI-1 is a Laskowski (standard mechanism) inhibitor, and its mode of action involves cleavage and re-ligation of the scissile bond. This unusual mechanism might explain why certain substitutions preferred by PR3 according to its substrate specificity did not generate potent SFTI-based inhibitors. For example, PR3 has been shown to favor substrates containing Asp at P2,^4,17,19,26 but neither of the SFTI-variants with this substitution (5–6) were potent PR3 inhibitors. Similarly, substrates containing Asp at P2′ are highly preferred by PR3,^20,21,26 but this substitution did not yield potent SFTI-variants for PR3 (6, 8). These results impacted the selectivity of the SFTI-based inhibitors, as P2 Asp and P2′ Asp are key residues in synthetic substrates and inhibitors that are selective for PR3 over NE.^{19−21,24,26,33}

Another class of PR3 inhibitors that has received considerable attention is irreversible peptide-based inhibitors. These compounds contain a peptide sequence that spans P1–P4, followed by a C-terminal warhead such as a di(chlorophenyl)-phosphonate^25,28 or diphenyl-phosphonate.²⁷ This approach has successfully generated inhibitors that are potent and selective for PR3.^27,28 Although these compounds have different mechanisms and kinetics to the SFTI-variants reported here, there are some similarities in the specificity trends at several positions. For example, P1 Nva or Abu generate more potent phosphonate inhibitors than P1 Ala,^27,28 and these two residues outperformed Ala or Val in the SFTI-variant screen (1–4). Additionally, proteinogenic and nonproteinogenic hydrophobic amino acids shown to be preferred by PR3 at P4^17,28 were well tolerated when substituted into the SFTI-scaffold. However, a critical difference was observed at the P2 position, where P2 Asp was compatible with potent phosphonate inhibitors^25,28 but not SFTI-based inhibitors. Nonetheless, by incorporating another residue at this position, we were able to produce potent SFTI-based PR3 inhibitors, and improving their selectivity is the focus of ongoing work. The extent of this challenge is perhaps apparent in endogenous inhibitors of PR3, including serpins and elafin, which are also not selective for PR3 and inhibit NE with similar efficacy.^3,14

Experimental Section

Detailed methods are provided in the Supporting Information.

Glossary

ABBREVIATIONS

NE

neutrophil elastase

NSP

neutrophil serine protease

PR3

proteinase 3

SFTI-1

sunflower trypsin inhibitor-1.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.9b00253.

• Methods, peptide characterization data, and NMR spectra (PDF)

This study was funded by a grant from the Australian Research Council (ARC) [DP150100443]. S.J.D. and J.E.S. are National Health and Medical Research Council Early Career Fellows [GNT1120066 and GNT1069819], and D.J.C. is an ARC Laureate Fellow [FL150100146].

The authors declare no competing financial interest.