Kinetics-Based Structural Requirements of Human Immunoglobulin G Binding Peptides.

Currently, antibodies are widely used not only in research but also in therapy. Hence, peptides that selectively bind to the fragment crystallizable site of an antibody have been extensively utilized in various research efforts such as the preparation of antibody-drug conjugates (ADC). Consequently, appropriate peptides that bind to immunoglobulin G (IgG) with a specific K d value and also k on and k off values will be useful in different applications, and these kinetic parameters have been perhaps overlooked but are key to development of peptide ligands with advantageous binding properties. We prepared structural derivatives of IgG-binding peptide 1 and evaluated the binding affinity and kinetic rates of the products by surface plasmon resonance assay and isothermal titration calorimetry to obtain novel peptides with beneficial antibody binding properties. In this way, 15-Lys8Leu with fast-binding and slow-release features was obtained through a shortened peptide 15-IgBP. On the other hand, we successfully obtained distinctive peptide, 15-Lys8Tle, with a similar K d value but with k on and k off values that were as much as six-fold different from those of 15-IgBP. These new peptides are useful for the elucidation of kinetic effects on the function of IgG-binding peptides and various applications of antibody or antibody-drug interactions, such as immunoliposome, ADC, or half-life extension strategy, by using a peptide with the appropriate kinetic features.

Introduction

Antibodies exhibit high target specificity and produce beneficial functions such as target neutralization and antibody-dependent cellular cytotoxicity in the immune system.[1,2] Consequently, antibodies are utilized not only as research tools in immunology, biology, and chemistry but also as therapeutic agents to fight serious diseases. One of the most promising applications of antibodies is antibody–drug conjugates (ADC) which are expected to be a key to next-generation drugs, particularly for anticancer therapy. The ADC can efficiently deliver a strong antitumor drug to the target tumor tissue with diminished toxicity to noncancerous tissues. Some ADCs have already been clinically approved, and currently, more than fifty kinds of ADCs are in clinical trials worldwide.[3] In connection with this movement, selective conjugation of payloads to antibodies has recently become an important aspect of the development of high-quality ADCs.

One of the versatile approaches to the selective conjugation of the payload is the use of a peptide that selectively binds to the fragment crystallizable (Fc) region of an antibody,[4−6] the tail region of an antibody that interacts with cell surface receptors called Fc receptors. To date, several IgG-binding peptides have been reported (Figure )[7,8] and are represented by the Z33 peptide which was derived in 1996 from the B-domain of protein A by the phage display method.[9] This peptide which consists of 33 amino acid residues shows strong binding affinity (Kd = 43 nM)[9] to the human immunoglobulin G1 (IgG1). In 2000, Wells et al. discovered a shorter monocyclic IgG-binding peptide Fc-III with 13 amino acids (Kd = 185 nM)[10] from peptide screening based on a phage display method.[11] Ito et al. independently reported similar IgG-binding disulfide peptides 1(4) and 2(12) (17 aa) with Kd values of 225 and 10 nM, respectively, and other groups reported two more potent peptide derivatives, FcBP-2[10] (Kd = 2.2) and Fc-III-4C[13] (Kd = 2.45 nM), which were obtained by bicyclization of the aforementioned monocyclic Fc-III peptide to maintain the peptide backbone in the β-sheet-like structure observed in the original Fc-III. A second cyclization of FcBP-2 was accomplished by amide bond formation between both C- and N-terminals with an additional D-Pro-L-Pro sequence.[10] On the other hand, in Fc-III-4C,[13] the similar second cyclization was accomplished by disulfide bond formation between two Cys residues introduced at each terminal. Although these bicyclic peptides showed potent binding affinity, subsequent or detailed structure activity relationship (SAR) studies have not been reported so far, probably due to the difficulties in the chemical synthesis of these peptides with a bicyclic scaffold.

Figure 1

Amino acid sequences of IgG-binding peptides. The lower-case “p” in the sequence of FcBP-2 indicates the d-proline residue.

To date, IgG-binding peptides (Figure ) have been studied for a variety of applications by several research groups. An example is the preparation of ADC with the covalent conjugation reaction. In particular, IgG-binding peptides such as the Z33[14] or Fc-III peptide[5,15] contain a chemically or photochemically reactive group which can form a covalent bond with an antibody to enable the selective modification of the antibody. Based on this concept, our group developed a new method for antibody modification using peptide 1, referred to as the chemical conjugation by the affinity peptide (CCAP) method.[4] In this method, the side chain of Lys8 of peptide 1 is first reacted with a succinimidyl ester of a bivalent cross-linker, disuccinimidyl glutarate, to prepare a peptide which can tether a second succinimidyl ester. The reactive ester on the peptide then makes an amide bond with the amino groups on the surface of the antibody, resulting in covalent bonding of the peptide to the antibody. The conjugation reaction occurs near the peptide binding site (i.e., the Fc-site), which allows the Fab-site to remain intact. Thus, these peptides, particularly if connected to a payload, can be applied to construct a homogeneous ADC.

In addition to covalent ADC preparation, IgG-binding peptides have been applied to a noncovalent method. For example, Ohata and Ball reported an Fc-binding peptide metalated with rhodium, which provided a catalytic chemical modification of an antibody via a rhodium-catalyzed reaction.[6] Moreover, the IgG-binding peptide was utilized as an antibody capturing unit in affinity chromatography replacing protein A, which is commonly used.[12] In the other example, concerning the preparation of immunoliposome, the IgG-binding peptide was loaded onto liposome-recruited antibodies on the surface of liposomes for selective drug delivery.[16] It has also been reported that, based on the interaction of the IgG binding peptide with endogenous antibodies, conjugates of the peptide to a small drug molecule or therapeutic peptide/protein are applicable to the noncovalent-type ADC strategy[17] or an in vivo half-life extension strategy for rapidly biodegradable drugs.[18]

In the development of appropriate IgG-binding peptides, the dissociation constant (Kd) is generally measured in order to evaluate the binding affinity to the antibody. However, peptides with the same Kd value do not always show the same kinetic properties because this value is from a ratio between the association rate constant (kon) and dissociation rate constant (koff). Consequently, appropriate IgG-binding peptides with a specific kinetic property would be required for different applications. For example, when IgG-binding peptides covalently conjugate with antibodies, it may be preferable that the peptides have a higher kon value relative to the koff value, because of the absence of a dissociation step after the covalent-bond forming reaction. Alternatively, when IgG-binding peptides containing catalysts induced the conjugation reaction between payload and antibody as previously cited,[6] it may be preferable for the peptides to have a higher koff value because the catalyst requires prompt release from the antibody after the reaction to interact with another target molecule, resulting in an enhanced catalytic cycle. IgG-binding peptides appropriate for preparing the immune-liposome,[16] noncovalent-type ADC,[17] or half-life extension conjugate[18] are better if they have a slow dissociation rate, that is, a low koff value that leads to the suppression of their undesired dissociation from antibodies in circulating blood.

In recent interpretations of the Kd value in medicinal chemistry, kon and koff values are beginning to be acknowledged as more important factors.[19,20] Though kon and koff values of IgG-binding peptides are essential information for a wide variety of their applications, this question has been rarely studied so far. Hence, in the present study, we performed a series of SAR studies of human IgG-binding peptide 1, which is used in the CCAP method,[4] in order to understand the structural influence on Kd, kon, and koff values, and we found peptides possessing a stronger binding ability than the original peptide (1) and peptides with a similar Kd value but different binding properties, such as fast-binding/fast-release and slow-binding/slow-release.

Results and Discussion

Alanine Scanning Study of IgG-Binding Peptide 1

Derivatives of the IgG-binding peptide (1) were synthesized by a general Fmoc (9-fluorenylmethyloxycarbonyl)-based solid-phase peptide synthesis method[21] and applied to physicochemical evaluation with >95% purity (see the Experimental Section for more details).

Alanine scanning of peptide 1 was first performed to identify the structural requirements for the potent binding activity in each residue, except Ala5 and Cys4/Cys14, which were required for the intramolecular disulfide bridge. We synthesized 14 peptides, and their binding kinetics (Kd, kon, and koff) to an antibody (Herceptin) were measured by surface plasmon resonance (SPR) assay using a 1:1 Langmuir fit model or steady-state analysis (fitting curves of representative peptides are shown in the Supporting Information in Figure S1). As shown in Table , peptides Gly1Ala and Pro2Ala retained the binding affinity (Kd = 248 and 314 nM, respectively) compared to original peptide 1 (Kd = 225 nM). This result suggested that the amino acid residues at the N-terminus (Gly1 and Pro2) were irrelevant to the potent antibody binding. On the other hand, Ala-substitution at Tyr6, His7, Leu11, Val12, and Trp13 led to a complete loss of binding affinity, indicating that these residues are crucial for the antibody binding. All these critical residues are located inside the cyclic structure and are divided into two regions: Tyr6-His7 and Leu11-Trp13, which are designated as minor and major regions, respectively (the residues were highlighted with blue color in the peptide sequence depicted in Table ). While substitutions at Asp3, Gly9, Glu10, Thr15, Phe16, and His17 to Ala resulted in a 2- to 8-fold lower affinity (Kd = 762, 1720, 448, 622, 950, and 696 nM, respectively), Lys8Ala alone showed a slightly stronger affinity (Kd = 171 nM) than the original peptide (1). In contrast to a large change observed in Kd values (171–1720 nM), the kon values changed only minimally from 0.582 to 0.909 s–1·μM–1 except in the case of Gly9Ala. Gly9Ala with kon = 0.252 s–1·μM–1 shows a 3.6-times lower kon value than the original peptide (1), suggesting that the introduction of a small methyl group to Gly9 has a significant negative effect on the peptide binding to the antibody, resulting in a high Kd value of 1720 nM. In comparison to original peptide 1, Lys8Ala showed a slower releasing step (lower koff value) that is enough to compensate for the slower binding step (lower kon value), resulting in the most potent binding affinity of Lys8Ala in this alanine scanning study.

Table 1

Alanine Scanning of IgG Binding Peptide 1a

peptide	Kd (nM)	kon (s–1·μM–1)	koff (s–1)
peptide 1	225 ± 2	0.909 ± 0.002	0.204 ± 0.000
Gly1Ala	248 ± 3	0.768 ± 0.002	0.191 ± 0.000
Pro2Ala	314 ± 10	0.635 ± 0.002	0.200 ± 0.002
Asp3Ala	762 ± 5	0.649 ± 0.003	0.495 ± 0.001
Tyr6Ala	n.b.b	−	−
His7Ala	n.b.b	−	−
Lys8Ala	171 ± 4	0.622 ± 0.002	0.106 ± 0.000
Gly9Ala	1720 ± 5	0.252 ± 0.001	0.434 ± 0.001
Glu10Ala	448 ± 20c	−	−
Leu11Ala	n.b.b	−	−
Val12Ala	n.b.b	−	−
Trp13Ala	n.b.b	−	−
Thr15Ala	622 ± 5	0.851 ± 0.004	0.530 ± 0.001
Phe16Ala	950 ± 4	0.582 ± 0.002	0.552 ± 0.000
His17Ala	696 ± 2	0.714 ± 0.001	0.497 ± 0.001

Binding affinity was measured by using Herceptin with an immobilized amount of 2000 RU.

n.b.: no detectable binding.

Steady-state analysis was applied because of the low reliability of a fitting curve (U-value > 14). Data are presented as mean ± SE (n = 5). The important residues are indicated with blue color in the peptide sequence (see the second paragraph of Results and Discussion).

Next, the effects of a series of Ala substitutions on the secondary structure were evaluated by circular dichroism (CD) spectrometry. This is a first report of the CD spectrum analysis of the series of peptides, but the phenomenon has been already observed by NMR[10] and X-ray cocrystal structure[11] analysis showing that the peptides show a β-sheet-like structure involving a β-bulge, which is a twisted secondary structure in the β-strand sometimes observed in the protein. However, as shown in Figure , CD analysis indicated that peptide 1 displays a random coil-like structure with a random coil content of 62% determined on the basis of Reed’s ref (22). This discrepancy could probably be attributed to the lack of spectral information of β-bulge because of the nonrepetitive feature of the β-bulge structure[23] and low frequency in peptides or proteins. In addition, there is a twisted β-bulge structure between Val12 and Trp13, residues which are involved in the major region covered in the Ala scanning. For these reasons, the secondary structure of peptide 1 could be detected as a random coil in the CD spectral analysis. CD spectra of other Ala-substituted peptides also indicated the similar random coil-like structure with the random coil content ranging from 55 to 62%, regardless of the binding affinity (Figure for four representative peptides, Figure S2 for other peptides). In the CD spectra analysis (Figure ), we found that three peptides (peptide 1, Gly1Ala, and Pro2Ala) possessing a good binding affinity showed a small maximal signal around 230 nm, while two other peptides (His7Ala and Val12Ala) possessing no significant affinity did not. A similar correlation between the maximal signal and binding affinity was also observed in the other peptides with the exception of Leu11Ala and Gly9Ala (Figure S2). The maximal signal around 230 nm, that is not considered in the secondary structure calculation of Reed’s reference, is reported as a signal derived from a disulfide bond[24,25] or π–π stacking.[26] Although it is unclear what structure in the peptides is truly related to the maximal signal observed around 230 nm, the signal in the CD spectrum could indicate a conformation which is intimately related to the binding affinity.

Figure 2

CD spectra of representative IgG binding peptides for the comparison of the maximal signal at 230 nm that is detected in the spectrum of peptide 1, Gly1Ala, and Pro2Ala but not in the spectrum of His7Ala and Val12Ala.

Truncation Study of IgG-Binding Peptide 1

We next performed a study of truncation from both N- and C-termini of peptide 1 in an effort to identify the minimum sequence required for IgG binding (Table ). As suggested in Ala-scanning, peptide 1 (3–17) with Gly1 and Pro2 deletion retained the beneficial binding affinity (Kd = 267 nM) compared to original peptide 1, suggesting again that the two N-terminal residues are not necessary for binding. Contrary to this, peptide 1 (4–17) with deletion of Gly1–Asp3 residues showed a dramatically higher Kd value of 1470 nM. The koff value of the peptide was markedly decreased to 1.05 s–1 from 0.204 s–1 in peptide 1, while the kon value (0.716 s–1·μM–1) retained that observed in peptide 1 (0.909 s–1·μM–1), suggesting that Asp3, which is located at the outside of the cyclic structure, contributes significantly to the dissociation step. Therefore, the minimum sequence for the activity required at the N-terminus begins at Asp3.

Table 2

Antibody Binding Affinities of Truncated Peptidesa

peptide	Kd (nM)	kon (s–1·μM–1)	koff (s–1)
peptide 1 (from Table 1)	225 ± 2	0.909 ± 0.002	0.204 ± 0.000
2–17	256 ± 11	0.732 ± 0.001	0.187 ± 0.002
3–17 (15-IgBP)b	267 ± 4	0.741 ± 0.003	0.198 ± 0.001
4–17	1470 ± 10	0.716 ± 0.003	1.05 ± 0.00
1–16	703 ± 4	0.778 ± 0.003	0.546 ± 0.001
1–15	1640 ± 10	0.567 ± 0.003	0.929 ± 0.002
1–14	n.b.c	−	−
2–16	690 ± 8	0.687 ± 0.005	0.473 ± 0.002
3–15	>2000	−	−
4–14	n.b.c	−	−

Binding affinity was measured by using Herceptin with an immobilized amount of 2000 RU.

Data are presented as mean ± SE (n = 5).

n.b.: no detectable binding.

Otherwise, C-terminal deletion led a stepwise decrease of the IgG binding affinity The Kd values of peptide 1, peptide 1 (1–16), and peptide 1 (1–15) were 225, 703, and 1640 nM, respectively, and peptide 1 (1–14) showed no detectable binding, and there was a similar gradual decrease in both the kon and koff values. These results suggest that C-terminal residues are valuable in maintaining the potent binding activity. The previously reported IgG-binding peptides Fc-III and the bicyclized peptides[10,13] (Figure ) lack the residues corresponding to Phe16 and His17 in peptide 1 as an interaction site to the antibody, and the effect of these residues on the affinity was not previously discussed. Our finding is therefore the first report of the importance of the two C-terminal residues. Taking these results into account, we concluded that peptide 1 (3–17) designated as 15-IgBP, and comprised 15 amino acid residues from the C-terminus of peptide 1, is a minimum sequence required for strong IgG-binding affinity. Therefore, 15-IgBP was used as a lead peptide with the monocyclic scaffold in further SAR studies.

Additionally, the peptide 1 (3–15) which involves almost the same amino acid sequence as that of Fc-III (13 aa, Kd = 185 nM[10]), except for N-terminal acetylation and/or substitutions of Lys8Leu and Tyr13Trp, did not show any significant binding affinity to the antibody (Kd > 2000 nM, Table ). This indicates that the structural differences between Fc-III and peptide 1 (3–15) significantly influence the binding affinity.

Amino Acid Substitution at Position 8 of 15-IgBP

For further structural optimization starting with 15-IgBP, we focused on the Lys residue at position 8 because substitution at only this position improved the binding affinity in the Ala-scanning study, and this residue is used for the cross-linking reaction with IgG in the CCAP method.[4] First, peptides possessing a basic amino acid with a shortened lysine side chain, that is, ornithine (Orn) and 2,4-diaminobutyric acid (Dab), 15-Lys8Orn and 15-Lys8Dab, respectively, were synthesized, and their antibody binding affinity was analyzed by SPR (Table ). 15-Lys8Dab with a two carbon shorter side chain than Lys exhibited about 4-fold increased binding affinity (Kd = 68.9 nM) compared with 15-IgBP (Kd = 267 nM), while 15-Lys8Orn exhibited lower affinity (Kd = 487 nM). Thus, Dab at position 8 seemed to be the most attractive residue for the CCAP method,[4] because it showed the highest binding affinity among the amino acid residues with a primary amino group which can be utilized for the reaction with a cross-linker reagent in the method. We prepared their acetylated forms, that is, 15-Lys8Lys(Ac), 15-Lys8Orn(Ac), and 15-Lys8Dab(Ac) to assess the effect of basicity on the antibody binding. Interestingly, the affinity of 15-Lys8Orn(Ac) was increased by the acetylation [15-Lys8Orn vs 15-Lys8Orn(Ac): Kd = 487 vs 138 nM], in contrary to Dab [15-Lys8Dab vs. 15-Lys8Dab(Ac): Kd = 68.9 vs 116 nM].

Table 3

Antibody Binding Affinity of the Peptide Derivatives with the Substitution at the Lys8 Positiona

Binding affinity was measured by using Herceptin with an immobilized amount of 2000 RU.

Data are presented as mean ± SE (n = 5).

In an attempt to understand these results, molecular modeling was performed on the basis of the X-ray crystal structure of the Fc-III peptide (PDB: 1DN2)[11] (Figures and S3). This model shows that the antibody has two Glu residues (Glu380 and Glu382) in the proximity of the binding region of Lys8 in 15-IgBP, suggesting that the amino group interacts electrostatically with these Glu residues. We speculate that the observed high affinity of 15-Lys8Dab can be attributed to the electrostatic interaction with Glu382. This is consistent with the observation that acetylation of the amino group decreases the affinity. On the other hand, the amino group of Orn cannot form a favorable electrostatic interaction with the Glu residues because they are too far apart. The acetylated Orn could however gain a new hydrophobic interaction with Pro387 which is between Glu380 and Glu382 (Figure ), and this would result in a higher affinity. This finding suggests that the structural derivatization focused on hydrophobicity at position 8, which can interact with the hydrophobic region around Pro387, could be a promising way to obtain a higher binding affinity, as discussed in the next paragraph. Additionally, the peptide 15-Lys8Arg shows a strong binding affinity (Kd = 29.1 nM), which might be attributed to the guanidino group interacting with both Glu380 and Glu382 of the antibody. This peptide shows a fast-binding feature with the highest kon value of 1.60 s–1·μM–1.

Figure 3

Model of the antibody binding mode of peptide 1 around the Lys8 position based on the X-ray cocrystal structure of human IgG1 and Fc-III peptides (PDB: 1DN2). Peptide 1 and antibody are shown as magenta and green ribbons, respectively. The antibody surface is shown in atom colors (blue: nitrogen, red: oxygen, and gray: carbon). Lys8 of peptide 1 and Glu380, Glu382, and Pro387 of antibody are shown.

As mentioned above, we had focused on the hydrophobicity at position 8 in 15-IgBP but actually, the Fc-III peptide (Figure ) has a hydrophobic Leu residue at this position. First, the peptide with substitution of Lys8 for norleucine (Nle) at position 8 (15-Lys8Nle) that corresponds to the des-ε-amino-Lys peptide of 15-IgBP was synthesized and evaluated. The substitution drastically increased the binding affinity (Kd = 36.7 nM) by 7 times with a slower dissociation rate (koff = 0.0261 s–1) than 15-IgBP, indicating that the hydrophobic alkyl group was possibly better than an amino group at this position, probably by associating with the hydrophobic side chain of Pro387 in the antibody in spite of the existence of anionic groups Glu380 and Glu382 in the proximity of the binding site. With a shorter side chain, 15-Lys8Nva also showed similar binding affinity (Kd = 44.6 nM).

Subsequently, branched alkyl group-containing amino acids were introduced at the Lys8 position, and it was found that the peptides with β-branched amino acids, Val or tert-leucine (Tle), showed a lower affinity (Kd = 77.8 and 326 nM, respectively) with slow binding properties (kon = 0.335 and 0.118 s–1·μM–1, respectively), compared to those of 15-Lys8Nle (Kd = 36.7 nM, kon = 0.714 s–1·μM–1). Interestingly, the 15-Lys8Tle peptide has a similar Kd value to 15-IgBP (Kd = 267 nM), while the kon and koff values of 15-Lys8Tle were approximately six times slower. These peptides are representative examples of IgG binding peptides with similar affinities and different kinetic features, which can support a variety of applications of the antibody based on the kinetic effects of IgG binding peptides. On the other hand, γ-branched structures highly enhance antibody binding affinity, particularly with 15-Lys8Leu showing the lowest Kd value of 8.19 nM with fast-binding (1.47 s–1·μM–1) and slow-release (0.0116 s–1) qualities. The residence time of 15-Lys8Leu calculated from the koff value is 86 s, which is 17 times longer than original peptide 1 and 15-IgBP which are resident each for 5 seconds. In addition, bulkier amino acids such as tert-butylalanine [Ala(t-Bu), Kd = 15.6 nM] or cyclohexylalanine (Cha, Kd = 17.0 nM) are also acceptable for their high binding affinity. In particular, 15-Lys8Leu and 15-Lys8Cha containing residues with a secondary carbon atom at the γ-position showed faster binding properties (kon = 1.41–1.47 s–1·μM–1) than peptides with the tertiary carbon structure [15-Lys8Ala(] (kon = 0.975 s–1·μM–1). The affinity of 15-Lys8Phe (Kd = 40.0 nM) was similar to that of 15-Lys8Nle, indicating that an aromatic structure is also tolerated. In view of the overall results, the γ-branched amino acids (e.g., Leu) at position 8 were the most suited to the strong antibody binding affinity.

In the analysis of the CD spectra, peptides with hydrophobic amino acid residues at the 8-position with a few exceptions (15-Lys8Val and 15-Lys8Tle) showed the maximum signal at 230 nm (Supporting Information, Figure S2). This is consistent with the aforementioned findings that the maximum signal can be related to the binding affinity. However, the exceptional β-branched peptides 15-Lys8Val and 15-Lys8Tle show no maximum signal at 230 nm in spite of their binding affinities (Kd = 77.8 and 326 nM, respectively). Judging from the maximum signal in the CD spectra, the peptides might not adopt the conformation preferred for the binding because β-branched amino acids, which form a type-I′ β-turn only with difficulty,[27] exist at the turn structure, that is, at Lys8 and Gly9, and this disturbed turn structure might exhibit the slow binding properties (kon = 0.335 and 0.118 s–1·μM–1) of the peptides. This inference about the involvement of the turn structure on the kon value is supported by the fact that substitution of Gly9 by Ala in the turn region also shows a low kon value in the Ala scanning (Table ). On the other hand, 15-Lys8Val and 15-Lys8Tle show some measure of binding affinity, probably because the slow binding property might be compensated by a very slow releasing property (koff = 0.0261 and 0.0386 s–1, respectively) which derives from the hydrophobic residue at position 8. This slowing effect on the koff value by hydrophobic residues at position 8 is supported by the results of the series of peptides with hydrophobic residues [e.g., 15-Lys8Leu or 15-Lys8Ala( vs 15-IgBP, Table ]. In contrast to 15-Lys8Val and 15-Lys8Tle, the other peptides such as His7Ala or Val12Ala with no maximal signal in the CD spectra show no affinities in SPR (Figures and S2), probably due to their fast releasing features, although their koff values were not determined. These findings suggest that the secondary structure indicated by the CD spectra of the peptides is important for the strong antibody binding affinities and potent kon values.

Isothermal Titration Calorimetry Measurements

Isothermal titration calorimetry (ITC) measurements of representative derivatives, such as peptide 1, 15-IgBP, 15-Lys8Tle and 15-Lys8Leu, were performed in order to compare the calorimetric parameters (Table and Figure ). It was found that the interaction of each peptide with human IgG1 (Herceptin) was exothermic (Supporting Information, Figure S4). The number of binding sites of the four peptides is almost two what was expected from the C2 symmetric structure of the antibody. The binding of these peptides is site-selective, suggesting that these peptides would be highly applicable to the site-specific conjugation method for the homogenous ADC preparation. Compared to the other three peptides (peptide 1, 15-IgBP, and 15-Lys8Tle, Kd = 310, 210, and 200 nM, respectively), the Kd value of 15-Lys8Leu is notably low (12 nM) similar to those judging by the results of SPR assay. In comparison with 1 and 15-IgBP, the deletion of N-terminal residues increased the entropy (−TΔS = 0.16 vs 3.5 kcal/mol, respectively) while decreasing the enthalpy (ΔH = −9.1 vs −13 kcal/mol). This result suggests that an enthalpy-driven feature of the binding is raised as a result of the deletion of flexible N-terminal residues, especially the secondary structure disruptors, Gly and Pro.[28,29] On the other hand, 15-IgBP and 15-Lys8Tle show no major difference in the calorimetric parameters. Although they have different kon and koff values as shown in Table , such kinetic rates cannot be assessed in ITC analysis in the absence of a special protocol.[30] The comparison with 15-IgBP and 15-Lys8Leu indicated that the hydrophobic substitution of Lys to Leu gained enthalpy (ΔH = −13 vs −15 kcal/mol), although generally, hydrophobic interactions are considered to be entropy-driven. This suggests again that the ε-amino group of Lys8 is energetically disfavored in spite of its possible electrostatic interactions with Glu380 and Glu382, when compared to the interaction provided by the hydrophobic alkyl group. The peptide 15-Lys8Leu was obtained and found to have the best enthalpy-driven feature with a high antibody binding affinity (Kd = 12 ± 1 nM).

Table 4

Thermodynamic Parameters for the Interaction between Herceptin and Peptides at 25 °C Measured by ITCa

	N	Kd (nM)	ΔG (kcal/mol)	ΔH (kcal/mol)	–TΔS (kcal/mol)
peptide 1	2.0 ± 0.1	310 ± 30	–8.9 ± 0.1	–9.1 ± 0.7	0.16 ± 0.73
15-IgBP	2.0 ± 0.0	210 ± 10	–9.1 ± 0.0	–13 ± 0	3.5 ± 0.1
15-Lys8Tle	2.0 ± 0.0	200 ± 20	–9.2 ± 0.0	–12 ± 0	2.4 ± 0.3
15-Lys8Leu	2.0 ± 0.0	12 ± 1	–11 ± 0	–15 ± 0	3.9 ± 0.3

ITC was measured three times for each sample.

Figure 4

Thermodynamic parameters of peptide 1, 15-IgBP, 15-Lys8Tle, and 15-Lys8Leu shown in Table .

Conclusions

We synthesized 37 kinds of derivatives of IgG-binding peptide 1 (17 residues, Kd = 225 nM) and evaluated their binding affinities against human IgG1 (Herceptin). The results of Ala-scanning and truncation of peptide 1 revealed that two N-terminal residues (Gly1 and Pro2) are irrelevant to the binding, but two C-terminal residues (Phe16 and His17) have an impact on binding affinity. We obtained the shortened peptide (15-IgBP, 15 residues) which retains the affinity (Kd = 267 nM). During further structure derivatization of 15-IgBP, we found that the Lys8 position is preferential for binding to the hydrophobic structure, particularly γ-branched structures. This led to the development of 15-Lys8Leu (Kd = 8.19 nM) which shows about 25-times higher affinity than 1 and 15-IgBP. We successfully obtained peptides with the similar Kd values but with kon and koff values that are approximately six-fold different (15-IgBP vs 15-Lys8Tle). These peptides could be useful in elucidation of the kinetic effect for the function of IgG-binding peptides and development of a potent covalent- or noncovalent-type conjugate with the antibody by using the peptide with the appropriate kinetic feature. In addition, we found that the conformational changes which were detected at 230–235 nm in the CD analysis affect the kon value. This knowledge will be very helpful for the derivatization of more potent IgG-binding peptides. Further SAR studies of 15-IgBP focused on other residues are currently in progress.

Experimental Section

General Procedures

Reagents and solvents were purchased from Wako Pure Chemical Industries (Osaka, Japan), Sigma-Aldrich (St. Louis, MO), Watanabe Chemical Industries (Hiroshima, Japan), and Tokyo Chemical Industries (Tokyo, Japan). All were used as received. Herceptin was purchased from Chugai Pharmaceutical Co., Ltd. Mass spectra were obtained on a Waters MICRO MASS LCT-premier mass spectrometer.

Solid-Phase Peptide Synthesis

The IgG-binding peptides were synthesized by the Fmoc-based peptide synthetic method[21] using an automatic peptide synthesizer (Prelude). Using the Fmoc-NH-SAL resin (40 μmol), the Fmoc group was deprotected with 20% piperidine in dimethylformamide (DMF) for 10 min twice. The peptide chains were elongated with the Fmoc-amino acid (5 equiv), 1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate (5 equiv), 1-hydroxy-7-azabenzotriazole (5 equiv), and N,N-diisopropyl-ethylamine (DIPEA, 10 equiv) for 30 min. These reactions were repeated to lengthen the desired peptide. The N-terminal amino group was acetylated with acetic anhydride (3 equiv) and DIPEA (3 equiv) in DMF for 15 min. For the solid-phase disulfide formation, the synthesized resins were treated with I2/DMF for 4 h. The resin was washed with DMF until the color disappeared and washed with Et2O and MeOH and then dried in vacuo. Cleavage from resin and final deprotection was performed by the treatment with trifluoroacetyl (TFA)/triisopropylsilane (TIS)/H2O/1,2-ethanedithiol (40:1:1:0.4) for 3 h. The crudes were precipitated with Et2O and washed twice. After drying, the residual solids were purified by reverse-phase (RP) high-performance liquid chromatography (HPLC) [SunFire PrepC18 OBD 19 × 150 mm (5 μm)] to give the desired peptides. For the liquid-phase disulfide formation, the pure peptides containing free thiol groups were obtained by the same methods as described above regarding cleavage from the peptidyl resins and HPLC purification. The peptides were dissolved in 20% dimethyl sulfoxide (DMSO)/sodium phosphate buffer (pH 7.0, 100 mM) and stirred for 48 h at rt. The solution was directly purified by RP-HPLC [SunFire PrepC18 OBD 19 × 150 mm (5 μm)] to give the desired peptides. The purity of synthesized peptides was analyzed by RP chromatography (COSMOSIL 5C18 AR-II, 4.6 i.d. × 150 mm) using a binary solvent system with a linear gradient starting from 10% MeCN in 0.1% TFA aq to 50% MeCN in 0.1% TFA aq at a flow rate of 0.9 mL/min with detection at UV 230 nm. The peptides that were used for kinetic analysis had an HPLC purity of >95%. The yields and analytical data are provided in the Supporting Information.

Acetylation of Peptide Side Chains

After disulfide bridging, the purified peptides were dissolved in DMSO at a concentration of 2.5 μM. N-Succinimidyl acetate (10 equiv) and N-methylmorphiline (20 equiv) were added to the reaction solution. After 30 min, the desired peptides were directly purified by RP-HPLC [SunFire PrepC18 OBD 19 × 150 mm (5 μm)].

CD Spectra

The CD spectra of the peptides were measured under previously reported conditions[17] using a Jasco J-1500CD spectrometer (JASCO, Japan) in a quartz cell with a 0.5 cm path length. Spectra were collected between 190 and 250 nm with a scan speed of 100 nm/min, a response time of 1 s, and a bandwidth of 1 nm at 25 °C. The peptides were dissolved in a 10 mM phosphate buffer (pH7.4) including 10% 2,2,2-trifluoroethanol at a concentration of 2.5 μM. The secondary structure composition of the peptide was determined using software based on published methods.[22]

SPR Assay

The binding kinetics were determined by a Biacore T-200 system using the previous conditions.[17] Herceptin (human IgG1) was dissolved in an acetate buffer (pH 5.5) and immobilized by premixed N-hydroxysuccimimide and EDC·HCl (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride) on a CM5 sensor chip. The analytes (peptide derivatives) were adjusted to the desired concentration by a serial dilution in a running buffer (HBS-EP; 0.01 M N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid, 0.15 M NaCl, 3 mM ethylenediaminetetraacetic acid, 0.005% Tween 20, pH 7.4). The sensorgrams were obtained with an association time of 180 s, a dissociation time of 600 s, and a flow rate of 50 μL/min. To determine the binding kinetics (kon, koff and Kd), the sensorgrams obtained in this way were analyzed by the Biacore T200 Evaluation software Ver. 1.0 using a 1:1 binding model. When the curve fitting was unreliable (U-value > 14), steady-state analysis was applied for the determination of the Kd value. As a negative control, we applied the other species antibody, rat IgG2a. Because the representative peptide derivatives, 15-IgBP and 15-Lys8Leu, did not show detectable affinity (kon and koff also) against the rat IgG2a, this series of peptides specifically bind to human IgG (Herceptin) and affinity data of which are directly comparable.

Molecular Modeling

The crystal structure of the Fc region of human IgG1 in a complex with the Fc-III peptide (PDB code: 1DN2)[11] was obtained from the Protein Data Bank. To model the complex between peptide 1 and the Fc region, the peptides were constructed on the basis of the secondary structure of Fc-III by a molecular operating environment (MOE ver. 2018.0101) software. The minimization process was performed using the Amber10:EHT force field.

ITC Measurement

The antibody bindings of peptide 1, 15-IgBP, 15-Lys8Leu, and 15-Lys8Tle were estimated by ITC measurements, and the thermodynamic parameters are shown in Table . The concentration of peptide and antibody was normalized by the absorbance at 280 nm. The experiment was performed under 25 °C condition utilizing PEAQ-ITC (Microcal Inc.). The final concentration of the antibody and peptides was 10–20 and 150–300 μM, respectively. The data were fitted with the single-site binding model utilizing analysis software-implemented I PEAQ-ITC software. The measurement was performed three times for each sample.