Structure-activity analysis of truncated albumin-binding domains suggests new lead constructs for potential therapeutic delivery

Rapid clearance by renal filtration is a major impediment to the translation of small bioactive biologics into drugs. To extend serum t 1/2 , a commonly used approach is to attach drug leads to the G-related albumin-binding domain (ABD) to bind albumin and evade clearance. Despite the success of this approach in extending half-lives of a wide range of biologics, it is unclear whether the existing constructs are optimized for binding and size; any improvements along these lines could lead to improved drugs. Characterization of the biophysics of binding of an ABD to albumin in solution could shed light on this question. Here, we examine the binding of an ABD to human serum albumin using isothermal titration calorimetry and assess the structural integrity of the ABD using CD, NMR, and molecular dynamics. A structure-activity analysis of truncations of the ABD suggests that downsized variants could replace the full-length domain. Reducing size could have the benefit of reducing potential immunogenicity problems. We further showed that one of these variants could be used to design a bifunctional molecule with affinity for albumin and a serum protein involved in cholesterol metabolism, PCSK9, demonstrating the potential utility of these fragments in the design of cholesterol-lowering drugs. Future work could extend these in vitro binding studies to other ABD variants to develop therapeutics. Our study presents new under-standing of the solution structural and binding properties of ABDs, which has implications for the design of next-generation long-lasting therapeutics.

Peptides and small proteins have great potential as therapeutic leads because of their high target specificities (1,2). However, they typically have short serum half-lives that range from minutes to a few hours. Even ultrastable peptides that are protease resistant are rapidly eliminated (3). The challenge is that all small biologics (,30 kDa) undergo fast renal clearance by glomerular filtration (4). Increasing dosage levels and drug administration frequency is not a viable counteraction, because it would increase the risk of dose-associated side effects and treatment costs and lead to poor patient adherence and lower quality of life, particularly in the treatment of chronic illnesses.
Several strategies have been proposed for extending in vivo half-lives of peptide and protein drug leads. One approach is to conjugate the lead to PEG chains (5), which increases the hydrodynamic radius and limits glomerular filtration. Its use has led to more than 10 approved therapeutics for the treatment of diseases such as hepatitis C, leukemia, and chronic inflammation (6). However, PEG products have been associated with several safety concerns, including immunogenicity and hypersensitivity (e.g. 25% of healthy individuals were found to already contain anti-PEG antibodies) and tissue accumulation because PEG is not biodegradable (7,8). An alternative is to use a biodegradable peptide or protein tag to smuggle the lead compound past metabolic clearance mechanisms by binding to a carrier protein, such as transthyretin (9) or human serum albumin (10,11). Human serum albumin has attracted considerable interest as a carrier protein because of its long t 1/2 in vivo (;19 days in humans) and abundance in blood (;40 g/l, 600 mM).
A 53-amino-acid fragment from the streptococcal protein G is an example of a naturally occurring albumin-binding domain (ABD; Fig. 1) that has been widely used to extend serum t 1/2 . For example, fusion to this domain or variants of it has successfully extended half-lives of soluble complement receptor type 1 (12), a bispecific antibody (13), Pf155/RESA (14), the bactericidal lysostaphin (15), Affibody scaffold molecules (16)(17)(18), and an avb3-integrin binding protein (19), as well as many other active molecules (20)(21)(22)(23)(24)(25)(26). The interaction of a homologue of this ABD from Finegoldia magna with human serum albumin is the best characterized structurally (Fig. 1a) (27). The crystal structure of it bound to human serum albumin shows that ABD comprises three helices collapsed around a hydrophobic core, presenting the surface-exposed residues of helices 2 and 3 to contact serum albumin (Fig. 1a) (27). Despite the availability of this solid-state structure and the popularity of the ABD as a therapeutic optimization tool, relatively little is known about its structural and albumin-binding properties in solution.
Here, we studied the structural integrity and binding affinity of the ABD from Finegoldia magna using CD, NMR spectroscopy, molecular dynamics (MD) simulations, and isothermal titration calorimetry (ITC). We present a novel strategy to produce the albumin-binding domain by solid-phase peptide synthesis and truncated variants thereof to conduct a structure-activity study. We show that peptide truncations of the domain can have binding affinity comparable to that of the fulllength parent protein and can be attached as tags to bioactive peptides to engender them with albumin affinity.

Synthesis of ABD and truncated variants
Previous studies on the ABD have relied solely on recombinant expression to produce the protein. Here, we demonstrate facile assembly of the domain by Fmoc-based solid-phase synthesis. Assembly did not require ligation of separate fragments, typically needed for total chemical synthesis of larger polypeptides. Chemical synthesis of the ABD is significant because it allows for straightforward incorporation of nonstandard amino acids, which could enable site-specific conjugation to an active biomolecule or other alternative forms of late-stage functionalization.
In addition to synthesizing full-length ABD, we made a series of ABD truncations to examine structure-activity relationships (Fig. 1, Table S1, and Fig. S1). These include ABD12, which comprises helices 1 and 2 of ABD; ABD23, which comprises helices 2 and 3; ABD1, which comprises only helix 1; ABD2, which comprises helix 2; ABD3, which comprises helix 3; and ABD29, which is a truncation of ABD2. Two constrained variants of ABD23 were also made, including ABD23ss, which contains a disulfide-bond cross-link, and ABD23ac, which contains a longer acetone cross-link (Fig. 1). ABD23 appeared to be the most promising ABD truncation to be constrained, because the crystal structure pinpoints it to be the minimal region of ABD required for binding to human serum albumin ( Fig. 1).
Structural integrity of ABD peptides in solution-Structural characterization using CD and NMR spectroscopy, as well as MD simulations, showed that overall helicity in solution was correlated to the length of the truncated peptide (Fig. 2). The full-length domain, ABD, exhibited a CD spectral profile typical of a helical protein, with characteristic negative bands at 208 and 222 nm and a positive band at 193 nm (Fig. 2a). This result suggests that synthetic ABD forms a three-helix bundle in solution. In comparison, peptides containing two helices (i.e. ABD12 and ABD23) showed less overall helicity, exhibiting spectral profiles that deviate from that of ABD, with a continual shift away from helicity and toward disorder for single-helix fragments (i.e. ABD1, ABD2, and ABD3). Further truncation resulted in additional loss of structure; for example, as shown by the truncation of ABD2 to ABD29. On the other hand, the constraint of ABD23 increased the overall helicity, albeit modestly, as evidenced by the CD spectra of ABD23ss and ABD23ac.
NMR secondary chemical shift analysis of all peptides in solution confirmed the results from the CD experiments (  . Structure and sequence of ABD and truncation variants. a, crystal structure of a three-helix albumin binding domain (ABD; pink) bound to human serum albumin (green) obtained from the PDB (entry 1TF0). Both proteins are largely composed of helices. The first residue of ABD (T1) is labeled for reference. The inset focuses on the contact region between ABD and albumin, highlighting the three helices (i.e. h1, h2, and h3) of ABD. b, amino acid sequence of ABD. Every tenth residue is labeled above the sequence. The location of the three helices is shown below the sequence. The sequences of ABD truncation peptides are schematically shown underneath. Each gray bar spans the length of the corresponding peptide. ABD23ss and ABD23ac include a disulfide and an acetone cross-link, respectively, as shown. The sequences of all peptides are listed in Table S1. and Fig. S2). For example, many residues of ABD exhibited large negative Ha chemical shift deviations from random coil values, indicative of well-ordered helical structure. In comparison, dual-helix fragments (e.g. ABD23) and single-helix analogues (e.g. ABD2) exhibited progressively smaller chemical shift deviations from random coil, consistent with their reduced helicity compared with ABD as discussed above. MD simulations provided further confirmation of the relationship between peptide size and helicity (Fig. 2c). Thus, the NMR data and MD sim-ulations confirm that overall helicity declines with reducing fragment size.
Solution binding to human serum albumin-Previous studies showing binding of the ABD to serum albumin have predominantly used detection techniques that require surface immobilization of albumin, such as surface plasmon resonance (SPR). SPR is susceptible to mass transport limitations, which can result in artifacts in the fitted association constants. Here, we used ITC to characterize binding to serum albumin in solution. We obtained thermodynamic parameters (e.g. enthalpy change, DH, and entropy change, DS) as well as the dissociation constant (K d ) and stoichiometry (N) of binding for each synthesized peptide to human serum albumin ( Fig. 3 and Fig. S2). Titration of the full-length domain (ABD) into albumin showed that binding was enthalpy driven (DH = 250.4 6 3.0 kcal/mol, DS = 2149 6 10 cal/mol/K; Fig. 3a and Table 1). Negative enthalpy along with negative entropy change (DH , 0, DS , 0) indicates that van der Waals and hydrogen bond interactions govern binding, whereas positive enthalpy and entropy change (DH . 0, DS . 0) is suggestive of a hydrophobic effect and small enthalpy and positive entropy change (DH ; 0, DS . 0) is suggestive of electrostatic/ionic interactions (28). As an illustrative example, binding of bile salts to human serum albumin is largely governed by hydrophobicity and exhibits both positive enthalpy and entropy changes in ITC measurements (29). In the case of ABD binding to albumin, analysis of the structure and dynamics of the reported binding interface of ABD (Fig. 3b) indicates that van der Waals interactions (e.g. involving F27), electrostatic interactions (e.g. involving K34 and K36), and contributions from conformational dy-namics to binding underpin the molecular basis of the observed thermodynamic parameters.
ITC measurements indicate that the affinity of ABD for albumin has a micromolar dissociation constant (45 6 4 mM). Fluorescence polarization measurements of FITC-labeled ABD to human serum albumin confirmed binding affinity to be in the same range (Fig. S3). The binding affinity values and exothermic binding profile of ABD resembles that exhibited by many small-molecule drugs known to bind albumin. For example, warfarin (an anticoagulant) (30), isoflurane (a general anesthetic) (31), and ibrutinib (an antitumor drug) (32) all bind human serum albumin in an exothermic manner and with micromolar dissociation constants. In these cases of low to moderate binding affinities, it is typically difficult to obtain accurate fitted values for the stoichiometry of interaction, because the high concentrations needed to give a profile with a clear inflection point are experimentally infeasible (33). This issue might be a reason for the apparent stoichiometry being below 1:1. Another reason is there are other sites on albumin that interact nonspecifically with ABD. Albumin is known to be a sticky protein reported to have multiple interaction modes

Table 1
Thermodynamic parameters for binding of ABD peptides to human serum albumin obtained from ITC for binding to a partner (34), and it can bind a range of stressed proteins (35). It is possible that the crystal structure of the ABD bound to albumin (27) has one of these binding modes.
The binding promiscuity of ABD might explain the lower binding affinity in solution measured in this study by ITC compared with that measured using serum albumin immobilized on a carboxymethylated dextran chip surface obtained previously by SPR (36). The ability of ABD to bind to multiple sites on albumin, as well as possibly interacting with the chip surface in the SPR flow cell, will reduce flow cell diffusion and introduce mass transport limitations, which is a common source of error in SPR surface binding kinetics that will result in higher apparent affinities (37). In support of this possibility, the thermodynamic profile derived from SPR measurements (36) differs from that obtained here from ITC and is characterized by a negative enthalpy and positive entropy change instead, which might come about from ionic interactions between positively charged surface residues of ABD (Fig. 3b) and the negatively charged chip surface. Another common reason for deviation in binding kinetics between SPR and solution techniques is surface heterogeneity, which would have affected previous SPR measurements because a nonspecific coupling method was used to immobilize albumin (36). Altogether, the results support the value of obtaining solution binding kinetics from a technique such as ITC to complement surface binding experiments.
ABD truncations (ABD12, ABD23, ABD1, ABD2, and ABD3) exhibited binding to human serum albumin as measured by ITC (Table 1 and Fig. S4). This result is consistent with albumin being a sticky protein and able to bind nonspecifically with hydrophobic residues that are solvent exposed in the truncated variants but hidden in the core of parent protein. Of the truncated variants, ABD23 is the most likely to bind to the same interface on albumin as the full-length domain indicated by the crystal structure, because it incorporates all residues at the contact interface (Fig. 1a). Indeed, these residues can be transferred to a functionally unrelated protein to impart albumin-binding affinity, providing further evidence that they mediate binding (36). Unexpectedly, ABD and ABD23 have similar binding affinities (Table 1 and Fig. S3) despite ABD23 being less helical and rigid than ABD ( Fig. 2 and 3b). The thermodynamic parameters reveal that this result is due to enthalpyentropy compensation ( Table 1). The shift in enthalpy and entropy change (DDH . 0, DDS . 0) suggests increased contributions from the hydrophobic effect, consistent with the exposed hydrophobic residues of ABD23 (that are hidden in ABD) creating new contacts with albumin. Interestingly, conformational constraint of ABD23 with either a disulfide bond (ABD23ss) or an acetone linker (ABD23ac) did not significantly alter binding affinity (Table 1). This result apparently contradicts the often-assumed paradigm that limiting conformational entropy loss would improve binding affinity. However, exceptions to that paradigm are common and attributed to the complexity of the enthalpy-entropy compensation phenomenon (38).
ABD23 could be further truncated to become ABD2, which comprises helix 2 of the full-length domain and retains binding to serum albumin (Table 1 and Fig. S3). However, its affinity to albumin could not be accurately determined by ITC, because it had reduced solubility and saturation could not be reached. Accordingly, the determined binding values for ABD2 should be considered estimates and be used only in a qualitative manner. Furthermore, the reduced solubility of ABD2 might limit applications for this peptide. Nevertheless, ABD2 is interesting because it possibly binds to the same contact region of human serum albumin as ABD; a competitive ITC experiment in which ABD2 was titrated into a binding-saturated ABD-albumin mixture resulted in no binding being detected (Fig. S5). Our study of ABD truncations identified several peptides, such as ABD23 and ABD2 (because of their existing supportive crystallographic data), that could be used as tags that bind human serum albumin.
Potential as tags that bind albumin-To explore their utility, we confirmed that ABD truncations could be conjugated to peptidic drug leads to confer to them affinity for albumin. The drug leads we used are peptidic inhibitors of the interaction between low-density lipoprotein receptor (LDL-R) and proprotein convertase substilisin/kexin type 9 (PCSK9), which is a validated drug target for the treatment of hypercholesterolemia (39,40), a condition that would benefit from drugs with long in vivo half-lives. A recent study showed that fusion of a small biologic inhibitor of PCSK9 to an albumin-binding domain resulted in extended t 1/2 and a concomitant reduced serum cholesterol up to 21 days after a single administration to monkeys (26). We selected two peptides that bind PCSK9: a peptide discovered from phage display (41) (which we call PBP01) and a fragment of LDL-R, which we recently showed could restore LDL-R recycling (42, 43) (which we call PBP02 here). These peptides were conjugated to ABD truncations to make four peptides, ABD2-PBP01, PBP01-ABD2, PBP01-ABD3, and PBP02-ABD2, which were characterized using CD and NMR spectroscopy, demonstrating that various combinations of tagpeptide drug can be made chemically (Fig. 4, a-c). Although we attached albumin-binding peptides to a functional cargo using backbone amide bond linkage, other methods of conjugation are also possible, such as using the ketone within the acetone bridge of ABD23ac as a chemoselective ligation handle.
We then demonstrated that a conjugated peptide can be designed so that it has affinity for both human serum albumin and its intended drug target (in this case, PCSK9). Specifically, PBP02-ABD2, which comprises PBP02 conjugated to ABD2, binds human serum albumin with affinity comparable to that of ABD2, as measured using ITC (Fig. 4d). Consistent with these data, a computational model of PBP02-ABD2 bound to human serum albumin (Fig. 5a) shows that the PBP02 component of the fusion peptide can be located away from the albumin binding interface so as not to interfere with potential interactions between ABD2 and albumin.
To determine whether fusion of ABD2 to PBP02 affects its interaction with PCSK9, we measured binding of PBP02-ABD2 to PCSK9 using SPR. To minimize surface heterogeneity-related artifacts that might arise when nonspecific immobilization methods are used, as mentioned above, we used PCSK9 that had been site-specifically labeled with biotin for capture onto streptavidin-coated sensor chips. The SPR data show that PBP02-ABD2 binds PCSK9 with micromolar affinity (Fig. 4e), which is in agreement with previous reports using SPR, fluorescence resonance transfer binding experiments, and functional cell-based assays (42,43). Addition of human serum albumin to the PBP02-ABD2 sample did not affect its binding to PCSK9, demonstrating that its binding affinity for human serum albumin does not prevent it from binding PCSK9 (Fig. 4f). This result can be structurally rationalized by a computational model of the PBP02-ABD2:PCSK9 complex (Fig. 5d). Ideally, ITC experiments could be used to further refine the precision of the observed affinity between PBP02-ABD2 and PCSK9; however, the important conclusion from the SPR and modeling experiments is that addition of ABD2 to PBP02 did not affect the therapeutically relevant activity of PBP02. Thus, we have shown that the peptide fragments studied here have potential for enhancing the albumin affinity of peptidic drugs without the loss of the therapeutic activity.

Conclusions
We have demonstrated the chemical synthesis of an albumin-binding domain and characterized its binding to serum albumin in solution. Our study highlights the importance of using solution binding techniques such as ITC to obtain crucial information on albumin binding. Structure-activity relationship studies with truncated peptides suggested that downsized variants are useful alternatives to the full-length domain for extending drug half-lives. These fragments might be easier to handle and less likely to induce an immune response because they are smaller than their parent domain. More broadly, the strategy we have employed here to downsize an original target domain and make constrained analogues could be applied to other albumin-binding domains, including those that have been engineered to span an 11,000-fold range in affinity and others with high affinity that have resulted in long terminal half-lives of 60 h in mice (44). A more expansive toolkit of albumin-binding peptides with different affinities for albumin is important, because apparent efficacy comes as a trade-off between affinity for the carrier protein and affinity for the target (45). Moreover, long half-lives would be beneficial in chronic conditions with persistent symptoms, such as dyslipidemia, whereas intermediated half-lives would be preferred in applications that . Structure and binding of peptide conjugates. a, schematic illustration of the sequences of ABD peptide conjugates, showing albumin binding peptides (ABD2 or ABD3) linked using a peptide linker to a protein binding peptide (PBP01 or PBP02) that bind to protein convertase subtilisin/kexin type 9 (PCSK9). b, CD spectra of peptide conjugates, showing molar ellipticity (u) as a function of wavelength. c, 1D 1 H NMR spectra of PBP02 and PBP02-ABD2. Similarity in the two spectra indicates the attachment of ABD2 to PBP02 has no significant effect on the structure of PBP02. d, binding isotherm of PBP02-ABD2 to human serum albumin, measured using ITC. e, binding of PBP02-ABD2 to immobilized PCSK9 monitored by SPR. Sensorgrams for varying concentrations of (ranging from 0.31-10.00 mM) are overlaid. f, binding of peptide conjugates to PCSK9 and the effect of human serum albumin (HSA) on binding. The response level at 10 mM peptide concentration is shown for various peptide samples with and without HSA.
benefit from a short treatment window, such as diagnostics. Therefore, continual investigation into new albumin-binding sequences and their binding affinities for serum albumin has broad implications in drug design.

Experimental procedures
General Human serum albumin (fatty acid free, globulin free, 99%) was purchased from Sigma-Aldrich (product code A3782).

Peptide synthesis
Peptides were synthesized using solid-phase peptide synthesis (46). Briefly, all peptides were assembled on 2-chlorotrityl-chloride resin, with the exception of ABD29 and PBP01-ABD3, which were assembled on rink-amide resin, at a 0.125 or 0.25 mmol scale using Fmoc solid-phase peptide synthesis on a Symphony Multiplex Synthesizer. Deprotection of the Fmoc group was done with 30%, v/v, piperidine in dimethylformamide (DMF). Side-chain-protected amino acids were activated using 2-(6-chloro-1-H-benzotriazole-1-yl)-1,1,3,3-tetramethylaminium hexafluorophosphate (4 eq) and N,N-diisopropylethylamine (4 eq) in DMF and were coupled twice. The assembled peptides were washed with dichloromethane before drying under nitrogen. The peptides were cleaved from the resin with a 94:3:3 mixture of TFA, triisopropylsilane, and water for 2 h. The peptides were then precipitated with ice-cold diethyl ether, followed by extraction with 50%, v/v, acetonitrile, before lyophilization. Formation of the disulfide bond in ABD23ss was achieved by dissolving reduced peptide in 0.1 M NH 4 HCO 3 , pH 8.1, and leaving the mixture for 24 h. Formation of the acetone-linked bridge in ABD23ac was done as previously described (47). Briefly, peptide was dissolved at 0.1 mM in 50 mM NH 4 HCO 3 , pH 8.1, with tris(2-carboxyethyl)phosphine) (1.1 eq) and dichloroacetone (1.5 eq). The list of peptides synthesized for this study is given in Table S1.

CD
Peptides were dissolved in water or 10 mM Tris-HCl, pH 7.5, 100 mM NaF at 25-50 mM. Spectra were recorded at room temperature with a 0.1-cm-path-length quartz cell by accumulating three scans, from 190 to 260 nm, using a CD spectropolarimeter (Jasco J-810). Molar ellipticity [u] was determined as previously described (48).

NMR spectroscopy
Peptides were dissolved in 550 ml of 90%, v/v, H 2 O-10%, v/v, D 2 O. The pH was adjusted to 5.3-5.5 using concentrated NaOH. NMR spectra were recorded on a Bruker 500-MHz or 600-MHz spectrometer. One-and two-dimensional NMR spectra ( 1 H, 1 H TOCSY, NOESY) were acquired at 298 K and referenced to 4,4-dimethyl-4-silapentane-1-sulfonic acid at 0 ppm. Spectra were analyzed using CcpNmr analysis to obtain chemical shift assignments. Secondary shifts were calculated from random coil shifts reported by Wishart et al. (49).

Isothermal titration calorimetry
Binding isotherms were measured using a MicroCal VP-ITC instrument at room temperature. A buffer of 50 mM Hepes, pH Figure 5. Structure of a peptide conjugate bound to human serum albumin or protein convertase subtilisin/kexin type 9. a, structure of PBP02-ABD2 (which comprises PBP02 in gray and ABD2 in pink) bound to human serum albumin (green) predicted using homology modeling. b and c, focus on selected interactions between residues of PBP02-ABD2 and human serum albumin. d, structure of PBP02-ABD2 bound to protein convertase subtilisin/ kexin type 9 (orange) predicted using homology modeling. 7.4, 100 mM NaCl was used in all cases. Typically, 70 mM human serum albumin was prepared for the reaction cell and 790 mM the peptide for the injection syringe. Water was used in the reference cell. The injection protocol started with a single injection of 0.5 ml titrant (over 1 s) followed by 12 injections of 3.22 ml of titrant (each over 6.44 s). Each injection was followed by a wait time of 180 s. A constant stirring speed of 1000 rpm was used. Isotherms were analyzed using MicroCal Origin software and the one-set-of-sites model. Experiments were performed in duplicate.

SPR
Peptides were dissolved in running buffer and the concentrations determined by NanoDrop A 280 measurements using molar extinction coefficients determined using CyBase (50,51). All surface plasmon resonance (SPR) experiments were performed on a Biacore 3000 (GE Healthcare Lifesciences, Fairfield, CT, USA) instrument at 25°C, as previously described (43). Biotinylated PCSK9 (Acro Biosystems) was captured on a streptavidin sensor chip using a flow rate of 5 ml/min and a 100 mg/ml protein solution in running buffer, typically achieving a 2000-3000 response unit (RU) change. SPR binding curves were generated from a flow rate of 20 ml/min, with a buffer composed of 20 mM Hepes and 100 mM NaCl at pH 7.5, 5 mM CaCl 2 . SPR binding curves were generated from 0.625, 1.25, 2.50, 5.00, and 10.00 mM peptide concentrations, with 60-s injection and 120-s dissociation cycles. Data were analyzed using GE BIAevaluation software (version 4.1.1) and kinetic analysis of reference-subtracted sensorgrams.

Molecular dynamics
Molecular dynamics simulations were carried out as previously described (52). Briefly, starting coordinates for ABD, ABD23, and ABD2 were extracted from the crystal structure of ABD bound to human serum albumin (PDB entry 1TF0), and each was fully solvated in a cubic box of TIP3P water (with sizes of 70 3 70 3 70 Å, 60 3 60 3 60 Å, and 50 3 50 3 50 Å, respectively) using the program VMD and the CHARMM36 force field. Each system was minimized with 250 steps of conjugate gradient energy minimization and equilibrated at 315 K. Production runs were conducted using ACEMD 1.6.0 using an NVT ensemble with a cutoff of 9 Å, hydrogen mass repartitioning, and a time step of 4 fs, and coordinates were saved every 200 ps. Assignment of secondary structure at each time point was done using the program STRIDE.

Computational modeling
A structural model of PBP02-ABD2 bound to human serum albumin was constructed using MODELLER v9. 16. Template coordinates were taken from the crystal structure of PBP02 (tEGF-A; PDB entry 4NE9) and that of ABD bound to human serum albumin (PDB entry 1TF0). A model of PBP02-ABD2 was first constructed and then used as a template to construct a model of it bound to albumin.
Author contributions-C. K. W., and D. J. C. conceptualization; C. K. W. formal analysis; C. K. W., A. S. A., and J. W. investigation; C. K. W. writing-original draft; A. S. A. and D. J. C. writing-review and editing; D. J. C., and C. K. W. funding acquisition; D. J. C. project administration.
Funding and additional information-D. J. C. is an Australian Research Council (ARC) Australian Laureate (FL150100146). Work in our laboratory on peptide scaffolds is supported by grants from the ARC (DP150100443) and the National Health & Medical Research Council (APP1107403).
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.