Imparting Albumin-Binding Affinity to a Human Protein by Mimicking

Feb 17, 2014 - Imparting Albumin-Binding Affinity to a Human Protein by Mimicking the Contact Surface of a Bacterial Binding Protein. Satoshi Oshiroâ€...
1 downloads 0 Views 3MB Size
Articles pubs.acs.org/acschemicalbiology

Imparting Albumin-Binding Affinity to a Human Protein by Mimicking the Contact Surface of a Bacterial Binding Protein Satoshi Oshiro† and Shinya Honda*,†,‡ †

Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8562, Japan ‡ Biomedical Research Institute, National Institute of Advanced Industrial Science and Technology, Central 6, 1-1-1 Higashi, Tsukuba, Ibaraki 305-8566, Japan S Supporting Information *

ABSTRACT: Attachment of a bacterial albumin-binding protein module is an attractive strategy for extending the plasma residence time of protein therapeutics. However, a protein fused with such a bacterial module could induce unfavorable immune reactions. To address this, we designed an alternative binding protein by imparting albumin-binding affinity to a human protein using molecular surface grafting. The result was a series of human-derived 6 helix-bundle proteins, one of which specifically binds to human serum albumin (HSA) with adequate affinity (KD = 100 nM). The proteins were designed by transferring key binding residues of a bacterial albumin-binding module, Finegoldia magna protein G-related albumin-binding domain (GA) module, onto the human protein scaffold. Despite 13−15 mutations, the designed proteins maintain the original secondary structure by virtue of careful grafting based on structural informatics. Competitive binding assays and thermodynamic analyses of the best binders show that the binding mode resembles that of the GA module, suggesting that the contacting surface of the GA module is mimicked well on the designed protein. These results indicate that the designed protein may act as an alternative low-risk binding module to HSA. Furthermore, molecular surface grafting in combination with structural informatics is an effective approach for avoiding deleterious mutations on a target protein and for imparting the binding function of one protein onto another.

B

peptide has been shown to be antigenic in mice,9 so the induction of an unfavorable immune reaction by ABD-fused protein or peptide is a concern. As an alternative binder, we attempted to impart albuminbinding affinity to a human protein using molecular grafting. Molecular grafting is a structure-based design (SBD) method for mimicking the functional site of a protein onto a different protein. To date, a metal binding site, the HIV-1 gp120 binding site of CD4, and the MDM2 binding site of p53, have been mimicked onto a scorpion toxin scaffold.10−12 Grafting of the functional site of HIV-1 gp41 onto a GCN4 leucine zipper has also been reported.13 Recently, molecular grafting was reported to be applied to a cyclic peptide.14 In general, several residues (for example, between 4−19 residues in the above studies) can be transferred onto the scaffold without significantly altering the scaffold’s structure and still provide the desired mimicked function. In this paper, we conducted molecular grafting onto a human-derived 6-helix bundle protein using available information about the Finegoldia magna protein G-related albuminbinding domain (GA) module, a homologue of ABD (Figure

iologically active proteins such as cytokines, hormones, and monoclonal antibodies have been used as biological therapeutics.1 Recently, smaller molecules such as the Fab region of a monoclonal IgG, single-chain Fv, a single-chain diabody, and non-immunoglobulin scaffold proteins have been and continue to be developed as novel protein therapeutics.2−4 Although these molecules exhibit many advantageous characteristics compared to conventional monoclonal IgGs, their therapeutic effects are limited by their fast clearance from the circulation due to their small molecular weight and lack of an Fc region.2 Serum albumin (SA) is the most abundant protein in plasma (50 mg/mL, 600 μM) and its half-life in human plasma is the longest (19 days) of all plasma proteins.3 The plasma residence time of proteins otherwise cleared quickly from the circulation has been prolonged by noncovalent binding with SA. Recombinant proteins, consisting of otherwise short-lived proteins fused with SA binding peptide,4 anti-SA domain antibody,5 and albumin-binding domain (ABD) from Streptococcal protein G,6,7 all exhibited increased serum half-life. Such fusions have been made with anti-tissue factor Fab, IFN-α2b, soluble complement receptor type 1, and single-chain diabody. Molecular evidence has been reported showing that ABD coupled with indirect targeting to neonatal Fc receptor (FcRn) efficiently extends the half-life of FcRn.8 However, ABD fusion © XXXX American Chemical Society

Received: December 27, 2013 Accepted: February 17, 2014

A

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 1. Schematic illustration of mimicking GA module onto a human scaffold protein. First, human serum albumin (HSA) binding residues of GA module (shown as sticks) were assigned from the crystal structure of the GA module/HSA complex, previously reported.15 Next, HSA binding residues were grafted onto the human scaffold following the selection of mutation sites. Mutation sites on the human scaffold were determined by calculating the surface accessibility of each residues and predicting the folding free energy change due to the mutations prior to design. The figures were generated with Pymol.

defined as the query. From the Protein Data Bank (PDB) deposited data, we searched small (less than 100 amino acids) human proteins whose backbone structure was similar to that of the query. Human apoptosis activation factor-1 caspase recruiting domain, APAF-1 CARD (PDB ID: 1cy5, 97 amino acids) was identified from the search. There was only one relevant hit from the small human protein data set (217 entries). The RMSD of the Cα atoms of human APAF-1 CARD (residues 34−64) and GA module (residues 23−53) was 1.00 Å, and the main chain structures were superimposable (Figure 2A). We also confirmed the absence of steric clash between APAF-1 CARD and HSA when the backbone structure of the human protein was aligned on the structure of the GA module/HSA complex. Human APAF-1 CARD (referred to as hAC) was thus chosen as the target scaffold for the design of a new albumin binder. The HSA binding residues were then transferred from the GA module onto hAC. Besides hAC, no molecules were found within the criteria. The second and third best candidates showed large structural deviations (RMSD (Cα) > 2.00 Å) when aligned against the two helixes of GA module (Supporting Information Figure S1). We first designed hAC_18m, which contained all 18 HSAbinding residues of the GA module. The GST-hAC_18m fusion protein was expressed in the soluble fraction after expression in Escherichia coli. GST-hAC_18m was purified by GST-affinity chromatography and anion exchange chromatography. However, the yield of GST-hAC_18m was very low (less than 1 mg/L of culture) compared to GST-hAC (approximately 20 mg/L of culture). After cleavage of the GST-tag, hAC_18m was purified by gel filtration chromatography, but hAC_18m was not eluted as monodispersed form. hAC_18m probably formed soluble aggregates due to destabilization of the tertiary structure of hAC. These results indicated that grafting all 18 residues would be harmful and that the residues required to satisfy function and stability should only be grafted. To

1). By following investigation of the crystal structure of GA module/HSA complex,15 the human serum albumin (HSA)binding residues of the contact surface of the GA module were grafted onto the human protein. We attempted to predict and avoid deleterious mutations prior to constructing the recombinant protein by calculating the accessible surface area (ASA) ratio of the residues and in silico estimates of the change in the stability of the protein due to the mutations. We succeeded in engineering human-based artificial proteins by grafting 13−15 residues. The proteins were shown to maintain the original secondary structure and exhibit adequate affinity for HSA.



RESULTS AND DISCUSSION Design of an Albumin-Binding Protein Based on Human Protein by Grafting the HSA Contact Residues Derived from GA Module. We attempted to design a human serum albumin (HSA) binding protein by transferring the serum albumin-binding residues from the bacterial protein onto a human protein. The design was carried out as follows: (1) determine the serum albumin binding residues of the bacterial albumin-binding module, (2) search for a human protein to use as a target scaffold whose backbone structure is similar to the contacting surface of the bacterial module, (3) select the appropriate residues to be grafted. First, we identified the interfacing residues of the bacterial albumin binding module using the PISA server16 and the crystal structure of the HSA and GA module complex (PDB ID: 1tf0) determined by Lejon et al.15 Eighteen contacting residues located on two helixes of the GA module were identified. Next, we extracted the structural data for the GA module (F. magna derived albumin-binding protein) from the crystal structure of the HSA and GA module complex described above. The region of the GA module which includes the two helices that contribute to HSA binding (residues 23−53) was B

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 2. (A) Structural alignment between GA module (23−53) (yellow) and human APAF-1 CARD (hAC) (PDB ID: 1cy5) (green). HSA contact residues are shown as sticks, with side chains colored magenta. (B) ASA ratio of the candidate mutational residues on hAC and the predicted stability change upon mutation. The bar graph indicates ASA ratio (%) and the line chart shows the value of the predicted stability change (ΔΔG). The candidate mutational residue has the configuration found in the hAC-GA module (e.g., Phe34-Ile23). (C) SPR sensorgrams showing the binding responses of 1 μM hAC and hAC_15m. The measurements were carried out on HSA (approximately 1000 RU) immobilized on a CM5 sensor chip. (D) The CD spectra of hAC and hAC_15m at 293 K. The samples (50 μg/mL) were dissolved in 20 mM sodium phosphate (pH 7.4).

Figure 3. The alignment of amino acid sequences from GA module, Streptococcal Protein G ABD, hAC mutants, and hAC (only the HSA binding region is displayed). HSA binding residues of GA module are shown red and HSA binding residues of Protein G ABD are shown blue. The mutated residues in hAC_mutants are underlined.

C

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

exclude deleterious mutations of hAC_18m, the design was combined with a calculation of the ASA ratio of the side chains of hAC, thus providing an estimation of the stability change upon mutation. To predict the mutable residues of hAC, the ASA ratio of the side chains of hAC was calculated by GETAREA.17 In addition, the structural stability change of the mutations was calculated by the Eris server that calculates the change in the folding free energy (ΔΔG) upon mutation. The Eris server also calculates ΔΔG for the change of small to large residues precisely by introducing main chain flexibility.18,19 Figure 2B shows the result of calculations of the ASA ratio of hAC residues and the predicted ΔΔG of hAC. Residues with a side chain ASA ratio over 50% were Thr36, Ser38, Lys42, Asn45, Pro47, Thr48, Gln50, Gln51, Lys58, and Lys62. Generally, the more exposed a residue is to solvent, the more it can tolerate mutation.20 These surface-exposed residues were therefore chosen for mutation. In addition, the mutation of Met55 to Leu, Met59 to Ile, and Lys63 to His were permitted, even though the ASA ratio of these residues was less than 50%, because these residues were chemically similar to each other. We expected these mutations would have little effect on the stability of hAC. However, Thr36 was not mutated to Ser because the predicted ΔΔG value for T36S was over 1.5 kcal/ mol. According to the study of Streptocaccal protein G ABD (the homologue of GA module) previously reported,21 the mutation of Glu39 to Tyr seemed to be essential to HSAbinding. Thus, we included this mutation in our design, although this mutation was not so favorable for the stability of hAC. Following the process described above, hAC_15m was designed with 15 residues derived from the GA module. The Thr48 and Ala54 of hAC are equivalent to Thr37 and Ala43 of the GA module. The sequences of the mutants are shown in Figure 3. Affinity Analysis of the Designed Protein and Its Secondary Structure. Surface plasmon resonance (SPR) binding assays were conducted to measure the affinity of hAC_15m for HSA, confirming that hAC_15m, but not hAC, bound to HSA (Figure 2C). The sensorgram of hAC_15m fitted well with a 1:1 binding model, providing a dissociation constant (KD) of 1.14 μM (Table 1 and Supporting

Figure 4. The CD spectra and thermal denaturation measurements of hAC and hAC mutants. (A) CD spectra of hAC and hAC mutants at 293 K. The samples (50 μg/mL) were dissolved in 20 mM sodium phosphate (pH 7.4). (B )Thermal denaturation of hAC and mutants at 278−353 K. The unfolded fraction and melting temperatures (Tm) were calculated from the two state unfolding model, as previously reported.39

Table 1. The Kinetic Parameters (kon, koff) and the Dissociation Constant (KD) for hAC Mutants-HSA Binding and GA Module-HSA Binding kon [(M s)−1] hAC_15m hAC_15m(L35T) hAC_15m(T36S) hAC_15m(E40K) hAC_15m(E41N) hAC_15m(T36S/E41N) GA module a

7.57 × N.D.a 1.02 × N.D. 1.66 × 3.60 × 2.01 ×

104 105 105 105 106

koff [s−1] 8.65 × N.D. 6.33 × N.D. 4.68 × 3.77 × 2.32 ×

10−2 10−2 10−2 10−2 10−3

to HSA with a KD of approximately 1 μM. This was insufficient since a previous study reported that therapeutic proteins exhibited an increased serum residence time if the KD value of the serum albumin-binding protein was less than 600 nM.22 Since grafting 15 residues (including T37T and A43A) onto hAC scarcely affected its secondary structure, the affinity of hAC_15m was improved by further addition of GA modulederived residues onto hAC_15m. Two mutants, one with replacement of Leu35 with Thr (hAC_15m(L35T)) and one with replacement of Thr36 with Ser (hAC_15m(T36S)), were designed by the addition of GA module derived albuminbinding residues (Thr24 and Ser25 of GA module). We also added two binding residues from Streptococcal Protein G ABD, previously identified by mutational analysis.21 hAC_15m(E40K) and hAC_15m(E41N) incorporated these two Streptococcal Protein G ABD derived binding residues (Lys41 and Asn42). The amino acid sequences of these hAC mutants are shown in Figure 3. The affinity of the mutants to bind to HSA was measured by SPR (Table 1, Supporting Information Figure S2). The KD value for hAC_15m(T36S) was 618 nM and for hAC_15m-

KD [M] 1.14 × N.D. 6.18 × N.D. 2.83 × 1.05 × 1.15 ×

10−6 10−7 10−7 10−7 10−9

N.D.: the parameter was not determined due to low solubility.

Information Figure S2A). Circular dichroism (CD) measurements of hAC_15m provided a typical α-helix spectrum very similar to that of hAC (Figure 2D). However, the intensity of spectra in hAC_15m was slightly smaller. As we describe later (Figure 4B), the difference in the intensity of spectra was probably derived from the difference in thermal stability. Affinity Improvement by Incorporation of More Binding Residues. The designed protein, hAC_15m, bound D

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

change were observed upon binding compared with hAC_15m. These results indicate an enthalpy−entropy compensation phenomenon.24 In contrast, hAC_15m(T36S/E41N) showed favorable changes in both enthalpy and entropy (ΔH = −29.43 kJ/mol and −TΔS = −10.41 kJ/mol) upon binding to HSA, and favorable enthalpy and entropic changes were observed in the interaction between GA module and HSA. Analysis of TS thermodynamics using the Eyring equation25 showed that the hAC_15m(T36S/E41N)-HSA interaction involved a small entropic barrier for association and a large entropic barrier for dissociation compared to the other hAC mutants (Supporting Information Figures S4 and S5), suggesting that the interaction mechanism might have changed when the affinity was improved from hAC_15m to hAC_15m(T36S/E41N). A large heat capacity change (ΔCp) for the original bacterial module-HSA binding was observed (Figure 5 and Supporting Information Table S1). The average value of ΔCp for HSA-binding to the hAC mutants was −1.79 ± 0.38 kJ·mol−1·K−1, similar to the equilibrium ΔCp value for typical protein−protein interactions (ΔCp = −1.39 ± 0.85 kJ·mol−1· K−1).26 In comparison, the absolute value of ΔCp for GA module was significantly larger (ΔCp = −4.03 ± 1.30 kJ·mol−1· K−1). It is generally accepted that changes in heat capacity upon protein−protein interaction is associated with the change in hydrophobic hydration. Thus, the large ΔCp may be attributed to dehydration of the contacting surface of GA module. Competition Assay of the Designed Protein against GA Module. FcRn is a key molecule for prolonging the serum residence time of serum albumin and IgG.27 Therefore, it is important that an HSA binder does not interfere with the HSAFcRn interaction. Since GA module binds to the domain II of serum albumin, distant from the FcRn-binding region,28 the attachment of a bacterial albumin-binding module should be an attractive strategy for extending the plasma residence time of protein therapeutics. To confirm that the target site of the designed protein was identical to that of GA module, a competition binding assay was conducted using SPR according to the method of Viht et al. 29 The results demonstrated that hAC_15m(T36S/E41N) inhibited the GA module−HSA interaction in a competitive manner, whereas hAC did not (Figure 6). The inhibition constant (Ki) of hAC_15m(T36S/ E41N) and of GA module was calculated to be 480 and 3.51 nM, respectively. The difference of 2 orders of magnitude in Ki correlated well with the difference in KD between hAC_15m(T36S/E41N) and GA module (Table 1). This result indicates that the designed protein recognizes the same binding site on HSA, and that the contacting surface of GA module is mimicked well on the designed protein. The Efficiency of the Two-Step Approach and the Success Factors in the Protein Design. To obtain a humanbased artificial protein mimicking a bacterial albumin-binding module, we grafted the HSA binding surface from F. magna GA module onto a human protein. There are 18 HSA contacting residues in GA module, so the total number of possible combinations of mutations for this designed protein is in the six digits (∑18Ci = 262 143). The primary object of this study was to identify appropriate mutations to impart sufficient affinity, to maintain the secondary structure, and to achieve acceptable expression yield. According to a previous survey,26 the average interface area in protein−protein complexes is 600−1000 Å2 per protein, and the average number of interface residues is 17 ± 3.5. Hence, several contact residues should be grafted onto a scaffold simultaneously to mimic the binding interface. This

(E41N) was 283 nM. Thus, affinity was improved by the additional transfer of binding residues from the GA module and ABD onto hAC_15m. Furthermore, hAC_15m(T36S/E41N) was designed to incorporate both the T36S and E41N mutations. The KD value for hAC_15m(T36S/E41N) was 105 nM, showing that these additional binding residues enhanced the affinity of the designed protein. However, the expression and purification of hAC_15m(L35T) and hAC_15m(E40K) failed due to their tendency to aggregate. These two mutations may cause conformational destabilization and aggravate aggregation propensity of the proteins. The association rate constant (kon) of hAC_15m(E41N) was 2.2 times larger than that of hAC_15m (Table 1). In contrast, kon of hAC_15m(T36S/E41N) was 3.5 times larger than that of hAC_15m(T36S). Clearly, the effect of the E41N mutation on kon was different for hAC_15m and hAC_15m(T36S), suggesting that a synergetic effect occurred in the double mutation. The CD spectra showed no major difference among the three hAC mutants (hAC_15m, hAC_15m(T36S) and hAC_15m(E41N)), while the double mutant hAC_15m(T36S/E41N) appeared to have slightly different structure content from the other mutants (Figure 4A). The melting temperatures (Tm) of the hAC mutants were 12−16 degrees lower than the wild type (Figure 4B). This indicates that the hAC mutants maintain a secondary structure similar to the wild type at low temperatures. The difference in Tm between the mutants was rather small, suggesting that common mutations among the hAC mutants reduced the thermal stabilities of the mutants by the same mechanism. Thermodynamics of HSA-Binding of the Designed Proteins. To analyze the interactions contributing to the affinity of the hAC mutants, the thermodynamic parameters for HSA-binding of hAC mutants were calculated from the van’t Hoff plot23 (Figure 5 and Supporting Information Figure S3). For the interaction between hAC_15m and HSA, the enthalpy change (ΔH) upon binding was −37.74 kJ/mol and the entropy change (−TΔS) upon binding was 3.82 kJ/mol. This showed that the change in enthalpy was the main contributor to the binding interaction. For hAC_15m(T36S) and hAC_15m(E41N), favorable enthalpy change and unfavorable entropy

Figure 5. Thermodynamic parameters of hAC mutants-HSA binding and GA module-HSA binding. The values of free energy change (ΔG), enthalpy change (ΔH), entropy change (−TΔS) and heat capacity change (ΔCp) at 298 K were calculated by the nonlinear fitting of van’t Hoff plot shown in Supporting Information Figure S4. The Error bars show the error values from the nonlinear regression analyses. E

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Streptococcal protein G ABD, a homologue of GA module, to reveal noncontact binding residues. The E41N mutation derived from ABD appeared not in contact with HSA, as judged by GA module/HSA complex structure. This mutation practically enhanced the affinity of hAC_15m and hAC_15m(T36S). We considered that the complementary combination of the structural information of GA module/HSA complex and the mutational information of ABD would be useful to introduce the noncontact binding residue. Future Challenges: Estimations of the Thermal Stability Changes, Synergistic Effect of Double Mutations, and Solvent Effects. Several unexpected results, obtained during the course of this study, highlight challenges associated with improving the current approach for protein design. The T36S mutation was estimated to destabilize the structure of hAC, but this mutation actually stabilized hAC_15m. Also, although destabilizing mutations were excluded, some of the hAC mutants exhibited decreased thermal stability. According to the benchmarking study of the estimation of the protein stability upon mutation, the precision of estimation of the stability of proteins is not so high.30 Specifically, the correlation coefficients between experimental and predicted ΔΔG value were in the range of 0.26−0.59. Thus, the phenomena that the mutation expected to destabilize the structure of protein actually stabilized (e.g., T36S) were not infrequent. To bridge the gap between estimations and experiments, a combination of several stability estimation methods based on different energy functions should help identify destabilizing mutations more precisely. A synergistic effect of multiple binding residues has been widely observed in protein interactions.31,32 Likewise, we observed the synergistic effect of the double mutation in hAC_15m(T36S/E41N): the interaction thermodynamics of this protein was not described by a simple summation of the thermodynamic parameters of hAC_15m(T36S) and hAC_15m(E41N). Since the computational prediction of the synergistic action on protein−protein interactions might be difficult due to its complexity, a stepwise approach would be likely more practical and efficient. Designing a precursor binder with initially moderate affinity for a target molecule would help provide a satisfying solution by limiting the number of mutations on the precursor. Dynamics of the surrounding water molecules, i.e., solvation and desolvation, are usually involved in protein−protein interactions.33 The results of thermodynamic analyses presented here suggested a significant difference in solvation between hAC mutants and GA module. The difference was the likely cause of the 100 times higher affinity of GA module (KD = 1.15 nM) compared to hAC_15m(T36S/E41N) (KD = 105 nM). Designing of a protein that takes into account of solvation effects is a future challenge. The Potential of the Designed Protein as an Alternative to Bacterial Albumin-Binding Protein. Finally, we assessed the potential of hAC_15m(T36S/E41N) as a molecule that could prolong the plasma residence time of protein therapeutics. Its affinity for HSA (KD = 105 nM) appears adequate, as its KD is within the effective KD range (KD < 600 nM).22 Additionally, hAC_15m(T36S/E41N) binds to HSA with high specificity. The results from competitive inhibition studies and the similar thermodynamic profiles of hAC_15m(T36S/E41N) and GA module indicate that hAC_15m(T36S/E41N) will have a binding mode similar to that of GA module. Since FcRn interacts with the opposite side of HSA (domain III) and GA module binds to domain II of

Figure 6. Competition assay of hAC_15m(T36S/E41N) binding against GA module-HSA binding. GA module (approximately 1000 RU) was immobilized onto a CM5 sensor chip. The responses of 1 nM to 4 μM hAC_15m(T36S/E41N) mixed and equilibrated with 10 nM HSA were obtained. The 0.1−400 nM GA module and 0.1 nM to 10 μM hAC were injected as a positive control and as a negative control, respectively. The curve fits for each inhibitor (excluding hAC) were obtained according to eq s6 in Supporting Information and the inhibition constant (Ki) of each inhibitor was calculated. The values are means ± SD (n = 3).

resulted in the design of hAC_15m, which bound to HSA specifically and maintained the original secondary structure of the scaffold protein. Subsequently, hAC_15m(T36S) and hAC_15m(E41N) were designed by addition of a contact residue in GA module or a noncontact residue derived from its homologue. Finally, hAC_15m(T36S/E41N) was designed and shown to bind to HSA with adequate affinity. Only 7 mutants out of the 262 143 possible combinations were designed and tested in this study, demonstrating the efficiency of the twostep approach employed: “imparting specificity by the transfer of contact residues, excluding deleterious mutations” and “affinity maturation by addition of extra residues after imparting specificity onto a scaffold”. Several factors were critical to our success. First, we calculated the surface accessibility of the mutation sites and estimated the effect of the mutations on stability in order to exclude deleterious mutations. Prior to conducting the experiments, the stability of the GA module-HSA complex was calculated by PISA,16 which showed that mutation of partially buried residues was necessary to provide sufficient interface area, although mutation of buried residues would be risky in general. Estimates of the change in stability identified risky mutations and led us to exclude the harmful L35T mutation in the design of hAC_15m. Second, stepwise mutations were conducted to generate 15m, 15m(T36S), and 15m(T36S/E41N). Introduction of single-point mutations onto hAC_15m led to identification of affinity-enhancing mutations (T36S, E41N) and deleterious mutations (L35T, E40K). Consequently, this stepwise mutation approach (i.e., local search around an optimum solution) contributed to the selection of an efficient combination of mutations for affinity enhancement. Third, we introduced a noncontact binding residue by utilizing information regarding a homologue. It is generally difficult to identify noncontact residues that contribute to affinity from the structural information of a protein complex. It needs extensive mutation study. However, the mutation study of GA module has not been reported. Thus, we utilized the information of the mutation study of F

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Protein Expression and Purification. Human APAF-1 CARD (termed hAC in this paper) was expressed and purified as previously described.37 In brief, hAC was expressed in the soluble fraction and then purified by ammonium sulfate precipitation, anion-exchange chromatography, and gel filtration chromatography (GFC). hAC mutants were expressed as glutathione-S-transferase (GST)-fusion proteins and purified by GST-affinity chromatography and anionexchange chromatography. After cleavage of the GST-tag, monomeric hAC mutants were purified by GFC. As a control, wild-type hAC was prepared in the GST-fusion protein format. GA module was expressed and purified as previously reported.38 In brief, GA module was expressed with a His-tag and purified by Ni affinity chromatography and GFC. Recombinant proteins were identified using SDS−PAGE and MALDI-TOF mass spectrometry (Supporting Information Figure S7). For further details, see Supporting Information. Binding Analysis Using Surface Plasmon Resonance (SPR). A Biacore T100 instrument (GE Healthcare) was used for kinetic and affinity analyses of the hAC mutants or GA module to HSA. HSA (approximately 1000 RU) was immobilized on a CM5 sensor chip by amine coupling chemistry. All experiments were fitted to a 1:1 binding model and kinetic constants (kon and koff) were determined using BIAevaluation 3.2 software. The dissociation constant (KD) was calculated by the equation KD = koff/kon. For further details, see Supporting Information. Circular Dichroism (CD) Spectroscopy. Far-UV CD spectra were acquired on a J-805 spectropolarimeter (JASCO) using a 0.2 mm path length quartz cell at 293 K. Protein concentrations were 50 μg/ mL in 20 mM sodium phosphate, pH 7.4. For the thermal stability analysis, ellipticity was measured at 222 nm between 278 and 353 K (scanning at 1 K/min). The unfolded fraction was calculated using the two-state unfolding model as previously reported.39 Curve fitting was executed by Igor pro 5.03J (WaveMetrics). Thermodynamic Study of the Interaction between hAC Mutants and HSA. The parameters KD, kon, koff for the hAC mutants and the GA module were determined from SPR measurements at various temperatures ranging from 283 to 308 K using the procedure described above. Then, the changes in Gibbs free energy (ΔG), enthalpy (ΔH), entropy (ΔS), and heat capacity (ΔCp) for the interactions were calculated by nonlinear fitting of the van’t Hoff plot. The transition state (TS) free energy (ΔG⧧on, ΔG⧧off), TS enthalpy (ΔH⧧on, ΔH⧧off), TS entropy (ΔS⧧on, ΔS⧧off), and TS heat capacity (ΔC⧧p,on, ΔC⧧p,off) of association (or dissociation) were calculated from kon, koff as a function of temperature through an Eyring analysis. The equations for nonlinear fitting of the van’t Hoff plot and the Eyring plot are described in Supporting Information. Competitive Binding Assay Using SPR. Inhibitor (0.1−400 nM GA module or 1 nM to 4 μM hAC_15m(T36S/E41N)) and 10 nM HSA were mixed and equilibrated. The mixture was injected onto a GA module (1000 RU) immobilized CM5 sensor chip for 180 s. Binding responses were recorded at the association state of the mixtures according to the method previously reported.29 Responses were plotted and the inhibition constant (Ki) was calculated as described in Supporting Information. As a negative control, 10 nM HSA mixed with 0.1 nM to 10 μM hAC was also examined.

HSA without inhibiting the HSA-FcRn interaction (Supporting Information Figure S6), hAC_15m(T36S/E41N), which has a similar binding surface to GA module, would not interfere with the HSA−FcRn interaction. On the basis of the results obtained in this study, however, we are not able to exclude the possibility of any allosteric competitive inhibition in HSA-hAC_15m(T36S/E41N) binding. The number of mutations incorporated into hAC_15m(T36S/E41N) is similar to that of engineered 10th type III domain of human fibronectin, which did not elicit neutralizing antibodies in phase I clinical trials,34,35 so we expect that hAC_15m(T36S/E41N) will also elicit low antigenicity. However, there are still several problems to be solved or investigated regarding hAC_15m(T36S/E41N). First, the thermal stability of hAC_15m(T36S/E41N) is too low: its melting temperature is close to physiological temperature (Tm = 311.5 K). Second, unexpected toxicity may be a concern because the intrinsic function of hAC to associate with procaspase-9 may be affected in serum. To address these problems, hAC_15m(T36S/E41N) needs to be further engineered. Conclusions. In summary, the HSA-binding surface of bacterial albumin-binding module was mimicked onto a human protein. Careful selection of mutation sites based on structural informatics and stepwise addition of extra residues enabled us to impart sufficient affinity and specificity in just two steps. The designed protein is a prototype for novel low-risk albuminbinders. The surface grafting strategy shown here holds promise for the engineering of other human proteins.



METHODS

Assignment of Binding Residues. The HSA binding residues of the GA module were assigned by the PISA server (http://www.ebi.ac. uk/msd-srv/prot_int/cgi-bin/piserver)16 using the crystal structure of the GA module/HSA complex (PDB ID: 1tf0).15 The HSA binding residues were defined as follows: (1) buried surface area is over 0 Å2 and (2) side chain atoms are in contact with HSA. Computational Search for Target Scaffolds in Human Protein Resources. Structural similarity searches were carried out using the PDB (http://www.rcsb.org/pdb/home/home.do) and PDBefold (http://www.ebi.ac.uk/msd-srv/ssm/)36 services. The coordinate data of the 2−3 helix (residues 23−53) of F. magna GA module were used as the query and were retrieved from the crystal structure of GA module/HSA complex (PDB ID: 1tf0). A human protein data set was prepared using the following conditions: Organism ‘Homo sapiens’, Number of chains (Biological assembly) ‘1’, Resolution ‘0−3.0 Å’, Chain length ‘20−100’, Macromolecule type ‘Protein’, and Identity cutoff ‘70%’. Structural comparison calculations between the query and every protein in the data set were executed using these parameters: Lowest acceptable match ‘0%’, Precision ‘Highest’, and Align length ‘29−31’. Human proteins satisfying these criteria were selected as target scaffolds for the following molecular grafting: (1) The backbone structure of the human protein is superimposable on the GA module 2−3 helix. (2) There is no steric clash between the human protein and HSA when the backbone structure of the human protein is aligned on the structure of the GA module/HSA complex. The steric clash was checked with Pymol. Calculation of Relative Accessible Surface Area (ASA) and Prediction of Change in Structural Stability. The ASA ratio of each residue was calculated by GETAREA (http://curie.utmb.edu/ getarea.html) using static coordinate data.17 The radius of the water probe was 1.4 Å. Other parameters were set as default values. The change in stability upon a single mutation at the grafting site was estimated using Eris server (http://troll.med.unc.edu/eris/login. php).19 To improve the precision of the calculation, the ‘backbone flexibility’ option was introduced into the calculation of ΔΔG. In the prediction of change in structural stability, cross-correlation effects of multiple mutations were not considered.



ASSOCIATED CONTENT

S Supporting Information *

Supplementary methods; structural alignments of the second and third best candidates against GA module; SPR sensograms and kinetic analyses; van’t Hoff plots, Eyring plots, and transition state analyses of hAC_mutants-HSA binding and GA module-HSA binding; structural comparison between GA module/HSA complex and FcRn/HSA complex; characterization of recombinant proteins; table of ΔCp, ΔC⧧p,on, ΔC⧧p,off. This material is available free of charge via the Internet at http://pubs.acs.org. G

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology



Articles

(16) Krissinel, E., and Henrick, K. (2007) Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774−797. (17) Fraczkiewicz, R., and Braun, W. (1998) Exact and efficient analytical calculation of the accessible surface areas and their gradients for macromolecules. J. Comput. Chem. 19, 319−333. (18) Yin, S., Ding, F., and Dokholyan, N. V. (2007) Modeling backbone flexibility improves protein stability estimation. Structure 15, 1567−1576. (19) Yin, S., Ding, F., and Dokholyan, N. V. (2007) Eris: An automated estimator of protein stability. Nat. Methods 4, 466−467. (20) Go, M., and Miyazawa, S. (1980) Relationship between mutability, polarity and exteriority of amino-acid-residues in protein evolution. Int. J. Pept. Protein Res. 15, 211−224. (21) Linhult, M., Binz, H. K., Uhlen, M., and Hober, S. (2002) Mutational analysis of the interaction between albumin-binding domain from streptococcal protein G and human serum albumin. Protein Sci. 11, 206−213. (22) Kontermann, R. E. (2011) Strategies for extended serum halflife of protein therapeutics. Curr. Opin. Biotechnol. 22, 868−876. (23) Zeder-Lutz, G., Zuber, E., Witz, J., and Van Regenmortel, M. H. (1997) Thermodynamic analysis of antigen-antibody binding using biosensor measurements at different temperatures. Anal. Biochem. 246, 123−132. (24) Dunitz, J. D. (1995) Win some, lose someEnthalpy-entropy compensation in weak intermolecular interactions. Chem. Biol. 2, 709− 712. (25) Roos, H., Karlsson, R., Nilshans, H., and Persson, A. (1998) Thermodynamic analysis of protein interactions with biosensor technology. J. Mol. Recognit. 11, 204−210. (26) Stites, W. E. (1997) Protein-protein interactions: Interface structure, binding thermodynamics, and mutational analysis. Chem. Rev. 97, 1233−1250. (27) Chaudhury, C., Mehnaz, S., Robinson, J. M., Hayton, W. L., Pearl, D. K., Roopenian, D. C., and Anderson, C. L. (2003) The major histocompatibility complex-related Fc receptor for IgG (FcRn) binds albumin and prolongs its lifespan. J. Exp. Med. 197, 315−322. (28) Andersen, J. T., Dalhus, B., Cameron, J., Daba, M. B., Plumridge, A., Evans, L., Brennan, S. O., Gunnarsen, K. S., Bjoras, M., Sleep, D., and Sandlie, I. (2012) Structure-based mutagenesis reveals the albumin-binding site of the neonatal Fc receptor. Nat. Commun. 3, 610. (29) Viht, K., Schweinsberg, S., Lust, M., Vaasa, A., Raidaru, G., Lavogina, D., Uri, A., and Herberg, F. W. (2007) Surface-plasmonresonance-based biosensor with immobilized bisubstrate analog inhibitor for the determination of affinities of ATP- and proteincompetitive ligands of cAMP-dependent protein kinase. Anal. Biochem. 362, 268−277. (30) Potapov, V., Cohen, M., and Schreiber, G. (2009) Assessing computational methods for predicting protein stability upon mutation: good on average but not in the details. Protein Eng., Des. Sel. 22, 553− 560. (31) Han, J., Kim, H. J., Lee, S. C., Hong, S., Park, K., Jeon, Y. H., Kim, D., Cheong, H. K., and Kim, H. S. (2012) Structure-based rational design of a Toll-like receptor 4 (TLR4) decoy receptor with high binding affinity for a target protein. PLoS One 7, e30929. (32) Zhang, Z., and Palzkill, T. (2003) Determinants of binding affinity and specificity for the interaction of TEM-1 and SME-1 betalactamase with beta-lactamase inhibitory protein. J. Biol. Chem. 278, 45706−45712. (33) Bhat, T. N., Bentley, G. A., Boulot, G., Greene, M. I., Tello, D., Dall’Acqua, W., Souchon, H., Schwarz, F. P., Mariuzza, R. A., and Poljak, R. J. (1994) Bound water molecules and conformational stabilization help mediate an antigen-antibody association. Proc. Natl. Acad. Sci. U.S.A. 91, 1089−1093. (34) Getmanova, E. V., Chen, Y., Bloom, L., Gokemeijer, J., Shamah, S., Warikoo, V., Wang, J., Ling, V., and Sun, L. (2006) Antagonists to human and mouse vascular endothelial growth factor receptor 2 generated by directed protein evolution in vitro. Chem. Biol. 13, 549− 556.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank H. Watanabe, N. Fukuda, and T. Shimizu for critical reading of the manuscript.



REFERENCES

(1) Walsh, G. (2010) Biopharmaceutical benchmarks 2010. Nat. Biotechnol. 28, 917−924. (2) Kontermann, R. E. (2009) Strategies to extend plasma half-lives of recombinant antibodies. BioDrugs 23, 93−109. (3) Peters, T., Jr. (1995) All About AlbuminBiochemistry, Genetics, and Medical Applications, Academic Press, San Diego, CA. (4) Dennis, M. S., Zhang, M., Meng, Y. G., Kadkhodayan, M., Kirchhofer, D., Combs, D., and Damico, L. A. (2002) Albumin binding as a general strategy for improving the pharmacokinetics of proteins. J. Biol. Chem. 277, 35035−35043. (5) Walker, A., Dunlevy, G., Rycroft, D., Topley, P., Holt, L. J., Herbert, T., Davies, M., Cook, F., Holmes, S., Jespers, L., and Herring, C. (2010) Anti-serum albumin domain antibodies in the development of highly potent, efficacious and long-acting interferon. Protein Eng., Des. Sel. 23, 271−278. (6) Makrides, S. C., Nygren, P. A., Andrews, B., Ford, P. J., Evans, K. S., Hayman, E. G., Adari, H., Uhlen, M., and Toth, C. A. (1996) Extended in vivo half-life of human soluble complement receptor type 1 fused to a serum albumin-binding receptor. J. Pharmacol. Exp. Ther. 277, 534−542. (7) Stork, R., Muller, D., and Kontermann, R. E. (2007) A novel trifunctional antibody fusion protein with improved pharmacokinetic properties generated by fusing a bispecific single-chain diabody with an albumin-binding domain from streptococcal protein G. Protein Eng., Des. Sel. 20, 569−576. (8) Andersen, J. T., Pehrson, R., Tolmachev, V., Daba, M. B., Abrahmsen, L., and Ekblad, C. (2011) Extending half-life by indirect targeting of the neonatal Fc receptor (FcRn) using a minimal albumin binding domain. J. Biol. Chem. 286, 5234−5241. (9) Sjolander, A., Nygren, P. A., Stahl, S., Berzins, K., Uhlen, M., Perlmann, P., and Andersson, R. (1997) The serum albumin-binding region of streptococcal protein G: a bacterial fusion partner with carrier-related properties. J. Immunol. Methods 201, 115−123. (10) Vita, C., Roumestand, C., Toma, F., and Menez, A. (1995) Scorpion toxins as natural scaffolds for protein engineering. Proc. Natl. Acad. Sci. U.S.A. 92, 6404−6408. (11) Vita, C., Drakopoulou, E., Vizzavona, J., Rochette, S., Martin, L., Menez, A., Roumestand, C., Yang, Y. S., Ylisastigui, L., Benjouad, A., and Gluckman, J. C. (1999) Rational engineering of a miniprotein that reproduces the core of the CD4 site interacting with HIV-1 envelope glycoprotein. Proc. Natl. Acad. Sci. U.S.A. 96, 13091−13096. (12) Li, C., Liu, M., Monbo, J., Zou, G. Z., Li, C. Q., Yuan, W. R., Zella, D., and Lu, W. Y. (2008) Turning a scorpion toxin into an antitumor miniprotein. J. Am. Chem. Soc. 130, 13546−13548. (13) Sia, S. K., and Kim, P. S. (2003) Protein grafting of an HIV-1inhibiting epitope. Proc. Natl. Acad. Sci. U.S.A. 100, 9756−9761. (14) Wang, C. K., Gruber, C. W., Cemazar, M., Siatskas, C., Tagore, P., Payne, N., Sun, G., Wang, S., Bernard, C. C., and Craik, D. J. (2013) Molecular grafting onto a stable framework yields novel cyclic peptides for the treatment of multiple sclerosis. ACS Chem. Biol. 9, 156−163. (15) Lejon, S., Frick, I. M., Bjorck, L., Wikstrom, M., and Svensson, S. (2004) Crystal structure and biological implications of a bacterial albumin binding module in complex with human serum albumin. J. Biol. Chem. 279, 42924−42928. H

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

(35) Tolcher, A. W., Sweeney, C. J., Papadopoulos, K., Patnaik, A., Chiorean, E. G., Mita, A. C., Sankhala, K., Furfine, E., Gokemeijer, J., Iacono, L., Eaton, C., Silver, B. A., and Mita, M. (2011) Phase I and pharmacokinetic study of CT-322 (BMS-844203), a targeted Adnectin inhibitor of VEGFR-2 based on a domain of human fibronectin. Clin. Cancer Res. 17, 363−371. (36) Krissinel, E., and Henrick, K. (2004) Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2256− 2268. (37) Vaughn, D. E., Rodriguez, J., Lazebnik, Y., and Joshua-Tor, L. (1999) Crystal structure of Apaf-1 caspase recruitment domain: an alpha-helical Greek key fold for apoptotic signaling. J. Mol. Biol. 293, 439−447. (38) Jonsson, A., Dogan, J., Herne, N., Abrahmsen, L., and Nygren, P. A. (2008) Engineering of a femtomolar affinity binding protein to human serum albumin. Protein Eng., Des. Sel. 21, 515−527. (39) Honda, S., Kobayashi, N., and Munekata, E. (2000) Thermodynamics of a beta-hairpin structure: Evidence for cooperative formation of folding nucleus. J. Mol. Biol. 295, 269−278.

I

dx.doi.org/10.1021/cb400946m | ACS Chem. Biol. XXXX, XXX, XXX−XXX