Substrate derived sequences act as subsite‐blocking motifs in protease inhibitory antibodies

Abstract

Proteases are involved in many physiologic processes, and dysregulated proteolysis is basis of a variety of diseases. Specific inhibition of pathogenetic proteases via monoclonal antibodies therefore holds significant therapeutic promise. Inspired by the competitive mechanism utilized by many naturally occurring and man‐made protease inhibitors, we hypothesized that substrate‐like peptide sequences can act as protease subsite blocking motifs if they occupy only one side of the reaction center. To test this hypothesis, a degenerate codon library representing MMP‐14 substrate profiles at P1–P5' positions was constructed in the context of an anti‐MMP‐14 Fab by replacing its inhibitory motif in CDR‐H3 with MMP‐14 substrate repertoires. After selection for MMP‐14 active‐site binders by phage panning, results indicated that diverse substrate‐like sequences conferring antibodies inhibitory potencies were enriched in the isolated clones. Optimal residues at each of P1–P5' positions were then identified, and the corresponding mutation combinations showed improved characteristics as effective inhibitors of MMP‐14. Insights on efficient library designs for inhibitory peptide motifs were further discussed. Overall, this study proved the concept that substrate‐derived sequences were able to behave as the inhibitory motifs in protease‐specific antibodies. With accumulating data available on protease substrate profiles, we expect the approach described here can be broadly applied to facilitate the generation of antibody inhibitors targeting biomedically important proteases.

1. INTRODUCTION

Proteases are involved in many physiologic processes ranging from embryonic development and autophagy to wound healing and blood coagulation (López‐Otín & Bond, 2008; Turk et al., ²⁰¹²). By hydrolyzing peptide bonds of protein and peptide substrates, proteases trigger irreversible events and thus their activities must be tightly controlled. Dysregulated proteolysis, in consequence, leads to catastrophic implications and is basis of a variety of diseases. Inhibition of the pathogenetic proteases therefore holds significant therapeutic promise (Deu et al., 2012; Drag & Salvesen, ²⁰¹⁰; Vandenbroucke & Libert, ²⁰¹⁴). Despite decades of intensive efforts, conventional drug discovery has only achieved limited successes by targeting a small fraction of all pharmacologically relevant proteases. It is because small‐molecule inhibitors often lack specificity and/or appropriate pharmacokinetic properties required for effective and safe protease‐based therapy. In these aspects, monoclonal antibodies (mAbs) emerge as attractive alternatives with significant advantages such as high selectivity and long serum half‐life. Notably, among over 600 proteases identified in the human genome, around half of them are extracellular and thus druggable by mAbs (Bond, 2019). In 2018, anti‐plasma kallikrein (pKal) DX‐2930 (lanadelumab), as the first FDA‐approved protease inhibitor mAb, entered the market for treatment of hereditary angioedema (HAE) (Banerji et al., 2017). Currently, mAbs inhibiting numerous proteases of biomedical importance have been under investigation as potential therapeutics for indications including cancer (Ager et al., 2015; Chen et al., ²⁰¹⁸; Devy et al., ²⁰⁰⁹; Ling et al., ²⁰¹⁷), thrombosis (David et al., 2016), diabetic neuropathy (Matsuoka et al., 2023), stroke (Ji et al., 2023), and Alzheimer's disease (Atwal et al., 2011), to name a few. This study aims to advance our understanding on the inhibition mechanism of protease mAbs, with the goal to facilitate the development of protease inhibiting therapeutics.

A few distinct mechanisms of macromolecular protease inhibitors have been elucidated (Bode & Huber, 2000; Farady & Craik, ²⁰¹⁰; Ganesan et al., ²⁰¹⁰). Among them, competitive inhibition, that is, by mimicking substrates to recognize the active site of the targeted proteases, is an effective approach. In fact, a large majority of naturally occurring protease inhibitors, including Kunitz‐type serine protease inhibitors and papain‐like cysteine protease inhibitors cystatins, utilize this strategy to achieve highly potent inhibitions. Also called canonical inhibitors, they often insert peptide loops of protruded conformation, which are complementary to the subsite specificity, into the reaction cleft of the targeted proteases. Here, a tendency toward inhibition over proteolysis is needed (Kromann‐Hansen et al., 2016). Taking stefin B as an example, its structure in complex with papain indicates that the N‐terminal segment of stefin B interacts with the non‐prime subsites of papain in a substrate‐like manner, then at the prime side, the N‐terminal loop of stefin B turns away from the catalytic center of papain and thus avoids proteolytic processing (Stubbs et al., 1990). Interestingly, several antibodies isolated from phage display libraries employ the same strategy to behave as protease inhibitors rather than substrates. For instance, a structural study of DX‐2930 suggests that part of its CDR‐H3 binds to the S3–S1 subsites of pKal. However, due to the presence of proteolytically disfavored acidic residues at the P1' and P2' positions, its CDR‐H3 loop abruptly detaches from the prime subsites and thus preventing cleavage by pKal's catalytic serine (Kenniston et al., 2014). Similarly, anti‐matrix metalloprotease (MMP)‐14 mAb 3A2 uses its CDR‐H3 to achieve active‐site inhibition by recognizing one but not both sides of the catalytic zinc (binding to the prime side of MMP‐14 reaction cleft though in the case of 3A2; Nam et al., ²⁰²⁰). More intriguingly, the middle portion of 3A2's CDR‐H3, with a sequence of NLVATP, perfectly matches with MMP‐14 preferred substrate specificity for the P1–P5' positions (Figure 1a; Eckhard et al., ²⁰¹⁶). This inhibition mode is further supported by the observation that incubation of 3A2 Fab with high concentrations of MMP‐14 at low pH generates truncated 3A2 fragments with a cleavage site, identified by Edman sequencing, right between the Asn (P1) and Leu (P1') residues of its CDR‐H3 (Lee et al., 2019).

FIGURE 1.

Design and construction of degenerate codon libraries carrying subsite blocking motifs. (a) Diagram showing MMP‐14 subsites S1–S5' and its preferred substrate residues at corresponding P1–P5' positions. The asterisk indicates cleavage site. Residues of mAb 3A2 inhibitory motif, ^100hNLVATP^100m (Nam; Lee et al., 2020), are red‐circled. Degenerate codons are designed to cover majority of MMP‐14 preferred substrate residues. Amino acids introduced by designed codons are grouped as preferred (red), neutral (gray), or disfavored (blue) according to MMP‐14 substrate profiles (Figure S1). (b) Library was constructed by replacing the inhibitory motif of 3A2 (boxed) with degenerate codons representing MMP‐14 substrate specificity at P1–P5'. Same color code is applied as in (a).

The orthosteric mechanism of macromolecular inhibitors inspires us to propose that substrate‐like peptide sequences can act as a protease subsite blocking motif if they occupy only one side of the reaction center. Particularly, we hypothesize that rather than the specific NLVATP sequence discovered in 3A2, a general MMP‐14 substrate sequence covering P1–P5' positions can render antibodies an inhibitory function when it is properly presented in the context of an antibody fold. To test this hypothesis, degenerate codons were designed to represent MMP‐14 substrate profiles, and a 3A2‐derived Fab phage display library carrying substrate‐like sequences in their CDR‐H3s was constructed and subjected to phage panning for the isolation of MMP‐14 specific mAbs. Results showed that identified inhibitory antibodies can indeed use diverse substrate‐like sequences as subsite blocking motifs in their CDR‐H3s. This proof‐of‐concept study paves the way for a rational design of inhibitory motif libraries which shall facilitate the development of therapeutic antibodies targeting biomedically important proteases.

2. RESULTS

2.1. Analysis of MMP‐14 substrate repertoires

Traditionally, the biological roles of MMPs have been considered associated with degradation and remodeling of extracellular matrix components including collagen, gelatin, and fibronectin, and activation of pro‐collagenases such as pro‐MMP2 (Nagase, 2001). However, this paradigm has largely shifted with accumulated evidence from both transgenic knockout models (Holmbeck et al., 1999; Zhou et al., ²⁰⁰⁰) and substrate degradome analysis (Butler et al., 2008). It is now becoming widely accepted that MMPs act as proteolytic processing enzymes to finely tune and tightly control a large variety of signaling molecules (Rodríguez et al., 2010). Indeed, by high‐throughput degradomics analysis using Proteomic Identification of protease Cleavage Sites (PICS) (Schilling & Overall, 2008), over 4300 MMP cleavage sites have been identified (Eckhard et al., 2016). Taking advantage of this rich and accessible dataset, we aimed to design MMP‐14 subsite blocking motifs by broadly representing its substrate repertoires. In this study, profiles of MMP‐14 substrate preference for subsites S1–S4' were analyzed by using WebPICS (Schilling et al., 2011), and data were visualized with iceLogos (Colaert et al., 2009). Using Homo sapiens Swiss‐Prot composition as the reference, the significantly over‐ and under‐represented amino acids relative to their natural abundances were identified for each position (Figure S1A). Results indicated that the P1' position strongly preferred aliphatic residues, especially leucine and isoleucine, and the P1 position was dominated with small and uncharged amino acids, that is, glycine, alanine, and asparagine. In addition, negatively charged residues were under‐represented at the prime subsites in general, and aliphatic and basic residues were disfavored at the P1 position. We also found disparity between GluC‐ and trypsin‐generated human peptide libraries—for example, lysine was over‐represented at P2' and P3' among GluC–derived peptides, while in the trypsin‐generated library, aliphatic (Val, Ile, Leu) and small (Gly, Ala) residues were preferred at these two positions, respectively. Collectively, to maximize the design diversity of subsite blocking motifs, we defined the amino acids over‐represented in either GluC‐ or trypsin‐generated peptide library as preferred, these under‐represented in both libraries as disfavored, and the remaining ones as neutral residues (Figure S1B).

2.2. Design of a degenerate library encoding MMP‐14 substrate sequences

To accommodate most of the preferred residues while minimizing incorporation of the disfavored ones, degenerate codons were carefully designed for each of P1–P5' positions (Figure 1a). Stop and rare codons were excluded to prevent unwanted truncation during Fab expression. For P1 position, codon RVC was chosen to encode Gly, Asn, and Ala, which covered >95% of MMP‐14 cleaved peptide sequences, while inevitably introduced neutral residues Asp, Ser, and Thr. For P1' position, preferred Leu, Ile, Val, Met, and Trp were given by codon WKS, nevertheless disfavored Arg and Ser were also included. Furthermore, for P2'–P4' positions, degenerate codons were designed to encompass preferred residues identified from both GluC‐ and trypsin‐derived libraries. Overall, for each of S1–S4' subsites, degenerate codon designs can accommodate 50%–75% preferred, 37.5%–50% neutral, and up to 25% disfavored residues of MMP‐14 substrates. Special for P5' position, as degenerate designs will introduce at least 67% undesired residues, the preferred design (Lys, Pro, and Asp) was achieved by three corresponding codons (AAA, CCG, and GAT) using individual oligonucleotides. Combining P1–P5' six positions, the designed amino acid diversity was 2.3 × 10⁴.

2.3. Construction of 3A2 Fab CDR‐H3 library carrying MMP‐14 substrate repertoires

To test our hypothesis that substrate‐like peptides can provide inhibitory functions, a synthetic Fab phage library was constructed in which the NLVATP motif in CDR‐H3 of 3A2 was replaced by degenerate codons representing MMP‐14 substrate specificity at P1–P5' (Figure 1b). Notably, in our design, other CDRs of 3A2 and its CDR‐H3 regions flanking the inhibitory motif were unchanged, allowing a proper display of the substrate‐like peptides for effective subsite blocking. To avoid library bias introduced by PCR, an amplification free approach was utilized to assemble the diversified CDR‐H3 fragments (Ge et al., 2010; Nam et al., ²⁰¹⁶). More specifically, chemically synthesized oligonucleotides (Table S1) were annealed and subjected to gap‐filling by using T4 DNA polymerase and T4 DNA ligase (Figure S2). Cloning the resultant fragments into a phagemid encoding 3A2 Fab generated 9.2 × 10⁵ transformants, sufficient to cover the designed library diversity. Sequencing of 12 randomly picked colonies verified the incorporation of 50%–75% preferred residues at P1–P4' positions and 100% at P5', all consistent with the codon design.

2.4. Isolation of MMP‐14 active site binders by phage panning

To select MMP‐14 specific Fabs from the constructed library, two rounds of solution‐based phage panning were conducted with biotinylated cdMMP‐14 (Figure S3) (Fellouse & Sidhu, 2006). N‐terminal domain of tissue inhibitor of metalloproteinase‐2 (nTIMP‐2) was used as the eluent to isolate Fab clones that bound at the active site or the surrounding cleft. Results of polyclonal phage ELISA clearly indicated that cdMMP‐14 specific binders were enriched through the selection (Figure 2a). Furthermore, competitive ELISA showed a nTIMP‐2 concentration‐dependent response (Figure 2b), suggesting that isolated Fabs and nTIMP‐2 at least partially shared their epitopes. Followed monoclonal ELISA indicated that majority of the isolated clones (184 out of 288) were positive (>2‐fold over background) with dozens showed high signals (>5 folds) (Figure S4). Sequencing 53 isolated clones identified 34 unique Fabs, among which 11 were produced for further characterizations (Figure 2c).

FIGURE 2.

ELISA and sequence analysis of post‐panning phage clones. (a) Polyclonal phage ELISA. Prior‐panning library and no MMP‐14 coating were used as controls. (b) Competitive ELISA of post Round 2 polyclonal phages in the presence of increasing concentrations of nTIMP‐2. (c) Summary of monoclonal phage ELISA results and distributions of clones for sequencing and Fab production.

2.5. Broad spectrum substrate‐like sequences were enriched

Sequence analysis of the 34 identified clones revealed the following features among selected Fabs (Figure 3): (1) Diverse sequences were used as subsite blocking motifs—for example, at P1 position, all three preferred residues in design were preserved, though at altered abundancies in comparison to the PICS degradome repertoires; and at P1' position, the proportions of preferred Ile, Leu, and Met all increased after selection. (2) Decreased shares of disfavored residues—for example, charged residues Arg (at P1') and Glu (at P2'), introduced undesirably by degenerate codons, were not found in any sequenced clones. (3) Disappear of certain preferred residues specific to GluC‐derived library—for example, although over‐represented at P2', P4', and P5' positions in GluC‐generated substrate repertoires, lysine was not found at these positions after selection. (4) Increased shares of neutral residues at distant prime positions such as Arg at P3' and Thr at P4', and (5) Pro was selected at P5'. Collectively, after selection for MMP‐14 active‐site binders, broad spectrum substrate‐like sequences were enriched in isolated clones with overall representations of disfavored and neutral residues reduced.

FIGURE 3.

Amino acid distribution at P1–P5' subsite positions. (a) Design. (b) After selection. MMP‐14 preferred, neutral, and disfavored residues (categorized in Figure S1) are colored in red, gray, and blue, respectively.

2.6. Substrate‐mimicking CDR‐H3 sequences conferred antibodies inhibitory potencies

Eleven isolated Fabs were produced, and their binding affinities and inhibitory potencies were measured. Results indicated that all tested Fabs exhibited nanomolar affinities (EC₅₀ = 23–180 nM) and potencies (IC₅₀ = 51–240 nM) (Table 1), strongly supporting the notion that diverse substrate‐derived sequences can act as inhibitory motifs. In addition, clone 2A8, which carried the combination of the most enriched AA at each position, was one of the strongest binders and inhibitors. Six Fabs with inhibition potencies greater than 100 nM were further tested by competitive ELISA (Figure S5). The gradually decreased signals responding to increasing nTIMP‐2 concentrations mirrored the result obtained with polyclonal phages (Figure 2b), indicating that isolated Fabs bound most likely to MMP‐14 reaction cleft and/or its vicinity—a sign of active‐site inhibitors. Important to standard mechanism inhibitors, the proteolytic stabilities of produced Fabs were also examined. Results showed that majority of tested Fabs sustained 4 h‐incubation with 2 μM cdMMP‐14 at 37°C, having their stabilities comparable to that of 3A2 (Figure S6). However, clones 1E9 (ASIRPP) and 3G5 (AMVRTP) exhibited significantly compromised stabilities. Interestingly, these two clones had improved binding affinities while weekended inhibition potencies, suggesting that they presumably behaved more like substrates. Overall, Fab characterizations confirmed that in general substrate‐mimicking CDR‐H3 sequences were effective as active‐site inhibitory motifs.

TABLE 1.

Characterizations of isolated and designed Fabs.

Clone	P1–P5' sequence a	ELISA EC50 (nM)	Inhibition IC50 (nM) b	Stability (%) c
3A2	NLVA T P	77 ± 18	76 ± 3	59 ± 1
2C11	ALI RT P	23 ± 4	54 ± 4	47 ± 6
2C10	T MI RT P	33 ± 0	71 ± 1	37 ± 0
1E9	A S I R PP	35 ± 1	130 ± 6	10 ± 1
3G5	AMV RT P	38 ± 1	110 ± 8	8 ± 5
2A8 d	AMI RT P	40 ± 1	60 ± 8	21 ± 6
2H8	NMV RT P	42 ± 2	140 ± 11	65 ± 8
3H7	S MI RT P	44 ± 4	51 ± 2	30 ± 6
2B3	A S I RT P	48 ± 6	140 ± 3	39 ± 1
3A3	ALVA T P	97 ± 4	74 ± 6	40 ± 4
3C2	AMIA T P	110 ± 6	69 ± 2	23 ± 1
3C11	AMIG T P	180 ± 24	240 ± 30	56 ± 16
Design 1	NLI RT P	49 ± 3	88 ± 3	71 ± 0
Design 2	S LI RT P	34 ± 1	78 ± 2	49 ± 9
Design 3	T LI RT P	40 ± 11	27 ± 0	38 ± 0

^a In the context of CDR‐H3 sequence VKLQKDKSHQWIRXXXXXXYGRYVMDY. MMP‐14 preferred, neutral, and disfavored residues are shown in red, gray, and blue, respectively.

^b Inhibition assays were performed with 10 nM cdMMP‐14, 1 μM M‐2350, and 1 nM‐1 μM Fab in 50 mM Tris–HCl, 150 mM NaCl, 5 mM CaCl₂, pH 7.5.

^c Stability was tested by incubation of 2 μM Fab with 2 μM cdMMP‐14 at 37°C for 4 h. The amounts of intact Fab remained were determined by non‐reducing SDS‐PAGE as relative stability.

^d Clone 2A8 carries the combination of the most enriched AA at each substrate position.

To further understand the delicate balance between proteolysis and inhibition, mutants at each of P1–P4' positions were separately grouped for comparisons (Figure 4). As expected, substitutions of preferred Met (2A8) or Leu (2C11) at P1' with disfavored Ser (2B3) impaired both affinity and potency (2.6‐fold decrease), and presence of a neutral residue Thr (2B3) instead of preferred Pro (1E9) at P4' improved stability. However, exceptions also existed—a neutral residue Arg (2A8) at P3' but not preferred residues Ala (3C2) or Gly (3C11) gave better affinity and potency, and for the non‐prime P1 position, replacing preferred Ala (3A3 and 3G5) with another preferred residue Asn (3A2 and 2H8) improved stability without significant changes in affinity or potency (Figure S7).

FIGURE 4.

Characterizations of representative Fabs isolated from the subsite motif library. Clones are grouped to mutations for P1–P4' each position. (a) Binding affinity and inhibition potency toward cdMMP‐14, determined by ELISA and FRET assay, respectively. Error bars represent standard deviation of duplicate measurement. (b) Proteolytic stability. Reactions were performed in 50 mM HEPES, pH 7.5, 150 mM NaCl with 2 μM Fab and 2 μM cdMMP‐14 at 37°C for 4 h incubation. The amounts of intact Fab were determined by SDS‐PAGE under non‐reducing condition.

2.7. Mutation designs showed balanced characteristics

Above comparisons performed among site‐specific mutants clearly demonstrated a trade‐off between inhibition and stability. Yet some clones such as 2C11 able to well balance these two properties encouraged us to design and test additional mutation combinations. Three characteristics—binding EC₅₀, inhibition IC₅₀, and stability—were considered collectively to identify desired residue(s) for each position (Table 2). Clearly, Leu was optimal for P1'. Ile and Thr were desired for P2' and P4', respectively, as other options dramatically impaired stability. Arg was chosen for P3' since it was more beneficial overall. For P1 position, residues Ala, Asn, Ser, and Thr remained to be considered. Collectively, four motif sequences ALIRTP, NLIRTP, SLIRTP, and TLIRTP passed our filters.

TABLE 2.

Selection of desired residues at each P position.

Position	P1			P1'	P2'	P3'	P4'
CDR‐H3 (P1–P5')	(N/A)LVATP	(N/A)MVRTP	(A/T/S)MIRTP	A(M/L/S)IRTP	AM(I/V)RTP	AMI(R/A/G)TP	ASIR(P/T)P
EC50 (nM)	77 /97	42/38	40/ 33 /44	40/ 23 /48	40/38	40 /110/184	35 /48
IC50 (nM)	76/74	138/ 108	60/71/ 51	60/ 54 /140	60 /108	60 /69/243	129/140
Stability (%)	59 /40	65 /8	21/ 37 /30	21/ 47 /39	21 /8	21/23/ 56	10/ 39
Desired residues	N	N/A	T/S	L	I	R	T

Note: Desired properties are shown in red, excluding the ones with less than 10% difference.

Notably, ALIRTP has been identified as clone 2C11. The other three motif designs were then produced for tests. As results shown in Figure 5 and Table 1, all three additional designs had balanced characteristics. Particularly, NLIRTP (Design 1) and TLIRTP (Design 3) exhibited the highest proteolytic stability (more stable than the parent 3A2) and inhibitory strength (IC₅₀ = 27 nM), respectively, among all tested Fabs. As selectivity is another desired feature for protease inhibitors, cross‐activities were further tested for Designs 1 and 3 on MMP‐9 and MMP‐12 D4 (Lee et al., 2020), which catalytic domains share high sequence and structural homologies with that of MMP‐14. Results indicated that both Designs showed no inhibitory functions up to a concentration of 1 μM, suggesting they were highly selective toward MMP‐14 over other MMPs (Figure S8).

FIGURE 5.

Biochemical characterizations of designed Fab mutants. (a) Binding affinity and inhibition potency toward cdMMP‐14, determined by ELISA and FRET assays, respectively. Error bars represent standard deviation of duplicate measurement. (b) Proteolytic stability. Reactions were performed in 50 mM HEPES, pH 7.5, 150 mM NaCl with 2 μM Fab and 2 μM cdMMP‐14 at 37°C for 4 h incubation. The amounts of intact Fab were determined by SDS‐PAGE under non‐reducing condition.

3. DISCUSSION

To develop protease inhibitory therapeutics, the generation of mAbs not only specifically bind but also efficiently inhibit often presents as a bottleneck. To address this challenge, several engineering strategies have been implemented such as design of molecular mimicry for immunization (Sela‐Passwell et al., 2011), synthesis of long CDR libraries (Nam et al., 2016), and development of screening/selection methods including competitive elution (Devy et al., 2009; Kenniston et al., ²⁰¹⁴), epitope‐specific FACS (Nam et al., 2017), deep sequencing (Lopez et al., 2017), and functional selection (Lopez et al., 2019). Notably, grafting polypeptide sequences derived from natural or synthetic inhibitors into CDR‐H3 has been proven as a valuable approach (Liu et al., 2015; Nam et al., ²⁰¹⁷), and the similar strategy has been further extended to pro‐domains of cathepsins (Dai et al., 2020). Demonstrating substrate‐like sequences as effective inhibitory motifs, this study provides an alternative approach and enriches our toolbox for the generation of inhibitory antibodies.

Active‐site binders employ a variety of mechanisms to act as inhibitors rather than substrates, for example, self‐resynthesis (Zakharova et al., 2009), reversely orientated peptide loops (Schneider et al., 2012), unintimate contact (Nam et al., 2020), and occupying one side of the cleft (Kenniston et al., 2014). In this study, many peptide sequences mimicking P1–P5' positions clearly show a trade‐off between inhibition and stability—potent inhibitors are often less stable (e.g., 2A8), while highly stable clones are of weak inhibition (e.g., 3C11). It implies that an improved library design encoding only non‐prime or prime positions (e.g., P1'–P5') might be able to break the trade‐off. It is worth mentioning that to facilitate cleavage in the stability assays, cdMMP‐14 was used at 2 μM, a concentration two to three orders of magnitude higher than its physiological concentrations at normal and tumorous tissues (Jonsson et al., 2016). Therefore, the in vitro stability results shown in this study should not reflect their in vivo half‐lives (Chen et al., 2018). Another limitation of our study is the convenient use of a synthetic peptide substrate in inhibition assays. As substrate‐dependency has been documented for proteolysis (Atwal et al., 2011), it is possible that the isolated Fabs could behave differently with macromolecular substrates.

This study also finds unexpected/unpredictable characteristics of isolated peptide motifs. For instance, PICS neutral residues rather than preferred ones are optimal at P3' and P4' (Table 2). And comparisons of Asn/Ala at P1, Ile/Val at P2', and Pro/Thr at P4' indicate that one preferred residue can greatly out‐perform another preferred residue (Figure 4 and Figure S7). These results also imply that loop flexibility without intimate contacts to the catalytic zinc can be important for stability (Nam et al., 2020). In addition, some selection results can be degradome repertoire specific (e.g., no Lys, preferred among GluC‐generated peptides, were found at P2' or P4' post selection) or restricted to the parent clone 3A2 (e.g., Pro at P5'). Collectively, these observations suggest that precise rational designs will be challenging, and a combinatorial approach could be more applicable for the optimal inhibitory motifs.

In conclusion, we prove that substrate derived sequences can act as protease subsite blocking motifs. This study also provides insights on efficient designs for inhibitory peptide libraries. With both physiological substrate specificity databases (e.g., MEROPS) and high‐throughput protease cleavage repertoires (e.g., PICS) available, we expect the approach described here can be broadly applied for the generation of antibody inhibitors targeting many biomedically important proteases.

4. MATERIALS AND METHODS

4.1. Recombinant production of MMPs, nTIMP‐2, and BirA

DNA fragment encoding AviTag‐His‐tag (GSGS‐GLNDIFEAQKIEWHE‐H₆) with a short N‐terminal GS linker was assembled from polynucleotides and cloned into NdeI/EcoRI sites on pMopac16‐cdMMP14 (Nam & Ge, 2016) to give pMopac16‐cdMMP14‐Avi for MMP‐14 catalytic domain (cdMMP14)‐AviTag periplasmic expression. Escherichia coli BL21 harboring pDsbC (Rodriguez et al., 2017) was transformed with pMopac16‐cdMMP14‐Avi and grown in 2 × YT supplemented with 100 μg/mL ampicillin and 20 μg/mL chloramphenicol at 37°C and 200 rpm until an OD₆₀₀ of 0.6–0.8 was reached. The expression was induced by the addition of 0.1 mM IPTG at 30°C and 200 rpm for 14–16 h. Produced cdMMP14‐AviTag was purified from periplasmic solutions by Ni‐NTA affinity chromatography. cdMMP‐14 without AviTag, cdMMP‐9, cdMMP‐12 D4 (Lee et al., 2017), and the N‐terminal domain of tissue inhibitors of metalloproteinase (nTIMP‐2) were prepared by periplasmic expression similarly and/or as described (Lee et al., 2017; Nam et al., ²⁰¹⁸). Biotin ligase gene BirA (O'callaghan et al., ¹⁹⁹⁹) was PCR amplified from pET21a‐BirA (Addgene) and cloned into BamHI/ XhoI sites on pGEX‐4 T‐1 to give pGEX‐GST‐BirA. E. coli BL21 was transformed and cultured in 0.8 L LB/Amp (100 μg/mL) containing 0.7% glucose at 37°C and 200 rpm until an OD₆₀₀ of 0.6–0.8 was reached, and the protein expression was induced with 0.4 mM IPTG at room temperature overnight. The cultured cells were lysed by sonication and GST‐BirA was purified from the supernatant using glutathione agarose (Pierce). In vitro biotinylation was performed in 50 mM Tris–HCl, pH 7.5 containing 150 mM NaCl, 5 mM CaCl₂, 5 mM MgCl₂, 2 mM ATP, 0.15 mM D‐biotin, 0.5 μM GST‐BirA, 260 μM cdMMP14‐AviTag with rotation at 30°C for 1 h (Fairhead & Howarth, 2015). Biotinylated cdMMP‐14 was purified using Ni‐NTA chromatography at 4°C. Purities of prepared recombinant proteins were verified by SDS‐PAGE and their concentrations were measured by UV spectrophotometer. Enzyme kinetics measurement of cdMMP‐14 demonstrated expected activities (Figure S3A), and streptavidin pulldown assays showed complete biotinylation of produced biotin‐cdMMP‐14 (Figure S3B).

4.2. Construction of synthetic CDR‐H3 Fab phage library

Polynucleotides were synthesized by IDT with 5' phosphorylation on the internal primers (Table S1). Two terminal primers (600 pmol each), five internal constant primers (400 pmol each), and three internal degenerate primers (133 pmol each) were mixed with 40 μL annealing buffer (100 mM Tris–HCl, pH 8.0, 500 mM NaCl) to a total volume of 400 μL. The mixture was incubated using a thermal cycler at 95°C for 2 min, cooled from 95 to 25°C by 1.5°C per min, and held at 4°C; 320 μL annealed product was added with 20 μL dNTP (10 mM), 40 μL T4 DNA ligase buffer (10×), 10 μL T4 DNA ligase (400 U/μL, NEB), and 10 μL T4 DNA polymerase (3 U/μL, NEB), and incubated at 37°C for 1 h followed by enzyme inactivation at 75°C for 25 min in the presence of 10 mM EDTA. The assembled fragment was gel‐purified, digested with AflII/BsmBI (NEB), and gel‐purified. The resultant fragment was ligated into the same restriction sites on Fab 3A2 phagemid and electroporated to E. coli SS320 complement cells at 2.5 kV (12.5 kV/cm). After recovery in SOC media at 37°C and 200 rpm for 1 h, the cells were spread on 2 × YT/Tet (10 μg/mL)/Amp (60 μg/mL) agar plates for incubation at 37°C for 8–9 h. The obtained colonies were collected and stored in 20% glycerol at −80°C till use. Dozens of library clones were randomly picked for miniprep and sequencing.

4.3. Phage panning

E. coli SS320 cells carrying the constructed Fab library were incubated with M13KO7 helper phages at a phage/cell ratio of 20 in 2 × YT at 37°C for 30 min without shaking. Cells were pelleted by centrifugation, resuspended in fresh 100 mL 2 × YT/Amp (100 μg/mL)/Kan (35 μg/mL), and incubated at 30°C and 200 rpm overnight for phage propagation. Generated library phages were purified by 20% PEG‐8000/2.5 M NaCl and prepared in 50 mM Tris–HCl, pH 7.4, 150 mM NaCl. During each round of solution‐based phage panning, ~10¹² phage particles were first depleted with 20 μL MyOne streptavidin T1 Dynabeads (Invitrogen) three times by gentle inversion at room temperature for 20 min. The remaining phages (~1 mL) were added with biotinylated cdMMP‐14 (100 nM in Round 1 and 10 nM in Round 2), incubated for 30 min with gentle inversion, and then 20 μL MyOne beads were added to capture Fab phage clones that bound with biotinylated cMMP‐14. After extensive washing (3 and 10 times for the R1 and R2, respectively), selected phages were eluted by incubation with 200 μL of 30 μM nTIMP‐2 at room temperature for 1 h. Eluted phages were added to 2 mL E. coli SS320 cells of 0.45 OD₆₀₀ and incubated at 37°C without shaking for 30 min; 5 × 10¹⁰ M13KO7 helper phages were then added for another 30‐min incubation, and the cells were cultured in fresh 50 mL 2 × YT/Amp (100 μg/mL)/Kan (35 μg/mL) at 30°C and 200 rpm overnight for phage propagation.

4.4. Phage ELISA

Ninety‐six‐well microtiter plates were coated with 60 μL of 5 μg/mL streptavidin at 4°C overnight and blocked with 0.5% BSA. After washing with a washing buffer (50 mM Tris–HCl buffer, pH 7.5, 150 mM NaCl, 5 mM CaCl₂, 0.05% Tween 20), 50 μL of 100 nM biotinylated cdMMP‐14 was added into each well, and the plates were incubated at room temperature for 30 min and washed. For polyclonal phage ELISA, freshly purified phages in 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 5 mM CaCl₂, containing 0.5% BSA were added (~10⁹ per well) and incubated at room temperature for 1 h. In competitive ELISA, 10‐fold serially diluted nTIMP‐2 starting at 30 μM was also added. After washing, the presence of bound phages was detected by the addition of 1:5000 horseradish peroxidase‐conjugated anti‐M13 antibody (Cytiva) and 50 μL of 3,3′,5,5′‐tetramethylbenzidine (Thermo Scientific). The chromogenic reaction was stopped by the addition of 50 μL of 1 M sulfuric acid, and the absorbance at 450 nm was measured. Wells without coating with biotinylated cdMMP‐14 were used as negative controls. All ELISA was conducted in triplicates.

For monoclonal phage screening, E. coli SS320 cells infected with phages post two rounds of panning were grown on agar plates of 2 × YT/Amp (100 μg/mL). Colonies were randomly picked and inoculated to 450 μL of 2 × YT/Amp (100 μg/mL)/Tet (5 μg/mL) in 96‐well plates for culture at 37°C and 200 rpm for 3–4 h. The cultured cells (10 μL) were transferred to new plates containing 200 μL 2 × YT/Amp (100 μg/mL) containing ~10⁵/μL M13KO7 helper phages, and incubated at 37°C without shaking for 30 min. After the addition of 200 μL of 2 × YT/Kan (50 μg/mL) to each well, the plates were incubated at 30°C and 200 rpm overnight. Cell culture supernatants were collected by centrifugation and 10 μL samples were used for monoclonal phage ELISA.

4.5. Cloning, expression, and purification of Fabs

Isolated clones were cultured in 2 × YT/Amp (100 μg/mL) and Fab phagemids were miniprepped. After sequencing, identified heavy chain fragments were cloned via BglII/SalI sites into a Fab expression vector carrying the light chain gene of 3A2 (Nam et al., 2016). Obtained plasmids were used for transformation of E. coli BL21, and cells were cultured in 600 mL 2 × YT/Amp (100 μg/mL) at 30°C and 200 rpm for 16–18 h. The periplasmic fractions were prepared by osmotic shocks (Nam & Ge, 2013). Cells were centrifuged and the pellets were resuspended in 60 mL periplasmic buffer (200 mM Tris–HCl, pH 7.5, 20% sucrose, 75 μg/mL lysozyme) by vortex. After incubation for 10 min at room temperature with gentle shaking, 60 mL ice‐cold water was added for osmotic shock followed by incubation on ice for 10 min with gentle shaking. After removal of cell debris by centrifugation, Fabs were purified from the supernatant by Ni‐NTA affinity chromatography. Purified Fabs were dialyzed against 50 mM HEPES and 150 mM NaCl (pH 7.5), and their purities were verified by SDS‐PAGE and concentrations were measured by UV spectrophotometer.

4.6. Fab ELISA

Ninety‐six‐well plates were coated with streptavidin and biotinylated cdMMP‐14 as described above. Purified Fabs were added as two‐fold serial dilutions with a starting concentration of 1 μM. The presence of bound Fabs was detected with 1:5000 HRP‐conjugated anti‐human Fab (Sigma) and the signals were developed with TMB. Competitive ELISA was conducted with 50 nM Fab in the presence of four‐fold serially diluted nTIMP‐2 starting from 50 μM. Reactions were stopped by acidification using 1 M sulfuric acid, and absorbances at 450 nm were measured. Error bars represent the standard deviation of duplicate measurements.

4.7. Inhibitory potency measurement by FRET assays

Two‐fold serially diluted Fabs (starting at 1 μM) were incubated with purified cdMMPs (10 nM cdMMP‐14, 10 nM cdMMP‐9, or 30 nM cdMMP‐12 D4) in 100 μL assay buffer (50 mM Tris–HCl, 150 mM NaCl, 5 mM CaCl₂, pH 7.5) at room temperature for 1 h. The reactions were initiated by adding 1 μL of 100 μM M‐2350 (Bachem) in 20% DMSO, and the assays were performed at 37°C in 96‐well plates with continuous shaking. The generated FRET signals were monitored for 15 min with excitation at 328 nm and emission at 393 nm. Values of inhibition IC₅₀ were determined accordingly. Error bars represent the standard deviation of duplicate measurements.

4.8. Proteolytic stability

Fabs at 2.0 μM were incubated with 2.0 μM cdMMP‐14 in 30 μL of HEPES buffer (50 mM HEPES, 150 mM NaCl, pH 7.5) at 37°C for 4 h. The resulting solutions were analyzed by 12% non‐reducing SDS‐PAGE.

Supporting information

Data S1: Supporting Information.

Choe H, Antee T, Ge X. Substrate derived sequences act as subsite‐blocking motifs in protease inhibitory antibodies. Protein Science. 2023;32(7):e4691. 10.1002/pro.4691