Fetal bovine serum albumin inhibits antimicrobial peptide activity and binds drug only in complex with α1-antitrypsin

Several antimicrobial peptides (AMPs) have been developed for the treatment of infections caused by antibiotic-resistant microbes, but their applications are primarily limited to topical infections because in circulation they are bound and inhibited by serum proteins. Here we have found that some AMPs, such as TP4 from fish tilapia, and drugs, such as antipyretic ibuprofen, were bound by bovine serum albumin only in complex with α1-antitrypsin which is linked by disulfide bond. They existed in dimeric complex (2 albumin -2 α1-antitrypsin) in the bovine serum only at fetal stage, but not after birth. The hydrophobic residues of TP4 were responsible for its binding to the complex. Since bovine serum is a major supplement in most cell culture media, therefore the existence and depletion of active albumin/α1-antitrypsin complex are very important for the assay and production of biomolecules.

ever, that of Polymyxin B was not (Fig. 1b, c bottom panels). To investigate amino acid residues of TP4 that may be responsible for its susceptibility to serum's inhibitory effects, specific hydrophobic and positive-charged residues of TP4 peptide were introduced or depleted and subjected to band shift assays. The results showed that deletion of hydrophobic residues, TP4-F1A,I2A, TP4-I5A,I6A and TP4-F9A,L10A, reduced the FBS-mediated shift, while the introduction of hydrophobic residue, TP4-A12I,A15I, enhanced the shift. In contrast, the deletion of cationic residue, TP4-R18S,R21H and TP4-dC2 (R24 and R25 deletion), did not change the AMP mobility ( Fig. 1d and Supplementary Fig. 1). With respect to the bactericidal activity, the TP4-A12I,A15I became more susceptible to serum's inhibition, while TP4-F1A,I2A was less susceptible (Fig. 1d). This result indicates that the hydrophobic residues of AMPs are responsible for the binding and inhibition caused by serum components.
Purification of serum components responsible for AMP-binding. Two bands, one major and one minor, were pulled down from the crude serum by biotinylated TP4-A12I,A15I peptides which had been immobilized on Streptavidin-conjugated beads (Fig. 2a). The major band consisted of albumin and α1-antitrypsin as determined by liquid chromatography-tandem mass spectrometry (LC/MS/MS), and the minor band was identified to be apolipoprotein AI (Apo-AI) (Supplementary Table 1). The active components responsible for www.nature.com/scientificreports/ TP4-A12I,A15I-binding were mainly located at the 70% ammonia sulfate saturation fraction as analyzed by band shift assay and pulling down experiment (Fig. 2b, c). The active components were further purified by fast performance liquid chromatography (FPLC) Mono Q and S12 column chromatography from the 70% saturated fraction (Fig. 2d). The active fractions of both column eluates ( Supplementary Fig. 2a, e) were collected based on band shift assay ( Supplementary Fig. 2d, h), such as fractions 32-40 of Mono Q and fractions 2-5 of S12 column chromatography. The protein components of each column eluate were analyzed by horizontal native PAGE ( Supplementary Fig. 2b, f) and reduced SDS-PAGE ( Supplementary Fig. 2c, g). The homogeneities of collected  www.nature.com/scientificreports/ protein A (albumin), protein B (α1-antitrypsin), protein C (serotransferrin) and protein D complex (albumin and α1-antitrypsin) as shown on the elution profile of FPLC Mono Q column chromatography (Fig. 2d) were further analyzed by horizontal 8% native PAGE (Fig. 2e) and vertical 15% reduced SDS-PAGE (Fig. 2f). Detailed information for the validations of these proteins A, B, C and D is shown in Supplementary Table 1.

AMP-binding and antitrypsin activities of proteins A, B and D.
As shown, TP4-A12I,A15I was bound by protein D (albumin and α1-antitrypsin complex) only, but not by protein A (albumin) nor protein B (α1-antitrypsin) alone as determined by band shift and pulling down assays (Fig. 3a, b). It is of note that the binding ability of wild type TP4 to protein D was much less than that of TP4-A12I,A15I ( Supplementary Fig. 3a,  b). Similarly, the LL37, SAAP159 as well as Iseganan was also bound by protein D, but not by protein A and protein B. Furthermore, the introduction of hydrophobic residue, TP4-A12I,A15I, increased the binding ability to protein D, deletion of hydrophobic residues, TP4-F1A,I2A, TP4-I5A,I6A and TP4-F9A,L10A, reduced the binding ability. In contrast, the deletion of cationic residue, TP4-R18S,R21H and TP4-dC2, did not change the binding ability ( Supplementary Fig. 3b). For the quantitation of binding affinity, the binding constant of protein D to immobilized biotinylated TP4-A12I,A15I was determined by biolayer interferometry with Kd = 21.7 nM, while that for protein A and B was not detectable (Fig. 3c, d). The bactericidal activity of TP4 (16 μg/ml) against E. coli (10 6 cfu/ml) was inhibited by increasing concentrations of protein D, but not by protein A or B even up to 10 mg/ ml (Fig. 3e). The bactericidal activity of TP4 was inhibited by 5% FBS, but the inhibition markedly decreased if the protein D as well as Apo-AI in the FBS was depleted by TP4-A12I,A15I-immobilized beads (Fig. 3f).
To further investigate the differential binding of AMP to protein A and protein D, the TP4-A12I,A15I and protein A/D was mixed and subjected to gel filtration chromatography (FPLC S12). There was a front shoulder on the elution profile (Fig. 3g). The albumin showed a broad distribution in the eluates, while the TP4-A12I,A15I coeluted with albumin only at the front shoulder fractions analyzed by reduced SDS-PAGE (Fig. 3h). Interestingly, the albumin co-eluted with TP4-A12I,A15I (fractions 2-7) exhibited very slow mobility on the non-reducing SDS-PAGE with molecular mass close to 130kD which is much higher than that (< 72kD) after reduction with 2-mercaptoethanol (β-ME) (Fig. 3i). This result indicates that the TP4-associated albumin may accompany with other protein(s), such as α1-antitrypsin, through disulfide bond.
To see whether the protein B (α1-antitrypsin) in the protein D complex still possess antitrypsin activity after being bound by protein A (albumin), we found that the proteolytic activity of trypsin toward TP4 and a larger AMP SAAP159 was repressed by protein D as well as free protein B, but only minimally affected by protein A (albumin) as analyzed by 15% reduced SDS-PAGE ( Supplementary Fig. 4).

Disulfide linkage between albumin and α1-antitrypsin is essential for AMP-binding. Protein
D exhibited one major band which moved very slowly and one minor band which moved faster than 72kD on the 8% non-reducing SDS-PAGE (Fig. 4a, left panel). However, it was separated into two bands (D1, D2) after reduction with β-ME and the two bands (D1 and D2) were identified to be albumin and α1-antitrypsin, respectively, by LC/MS/MS analysis (Fig. 4a, right panel and Supplementary Table 1). In addition, protein A (albumin) in non-reducing form moved faster than that in reduced form. To investigate the dynamic of protein D dissociation, increasing concentrations of β-ME were added to protein D and the resulting mixtures were subjected to non-reducing SDS-PAGE and Western blotting analyses using antibodies raised against albumin and α1-antitrypsin, respectively (Fig. 4b). The results demonstrate that protein D was dissociated into protein A and protein B by β-ME. Similarly, the dissociation of protein D by β-ME (0.05-0.8%) was also observed by horizontal native PAGE and high-performance liquid chromatography (HPLC-C4 column) (Fig. 4c, d). To investigate the role of disulfide bond, we found that AMP-binding ability of protein D was destroyed if it was reduced by β-ME as analyzed by band shift (Fig. 4e) and pulling down assay (Fig. 4f). It is of note that some protein D constituents were able to bind the TP4-A12I,A15I after being dissociated by β-ME (0.1-0.4%) (Fig. 4f). These results demonstrate that the disulfide bond between protein A (albumin) and protein B (α1-antitrypsin) is essential for the structural integrity and AMP-binding ability.
Dimerization of protein D is essential for AMP-binding. The non-reduced protein D was recognized by both antibodies raised against albumin and α1-antitrypsin. One major band as well as two minor bands were observed (Fig. 5a). The protein D (A + B complex) existed in monomeric and dimeric forms as shown on the reduced SDS-PAGE after being cross-linked with glutaraldehyde. In contrast, proteins A and B still existed as monomers even after being cross-linked as was done with protein D (Fig. 5b). The monomeric and dimeric forms of protein D were further validated by Western blotting using respective antibodies raised against protein A (albumin) or protein B (α1-antitrypsin) (Fig. 5c). The dimerization of protein D was inhibited by 2 × SDS, equivalent to 0.125% SDS (w/v), as well as 0.05% β-ME (w/v) where protein D was dissociated into protein A and B (Fig. 5d, e). Under the environments (2 × SDS) in which dimerization was inhibited, protein D was unable to bind to TP4-A12I,A15I as analyzed by pulling down assay (Fig. 5f). These results indicate that dimerization of protein D is essential for its AMP-binding activity.
TP4 peptide binding sites on protein D. To elucidate the interactions between TP4-A12I,A15I and protein D (albumin and α1-antitrypsin), the AMP/protein D complex was cross-linked with glutaraldehyde, separated by reduced SDS-PAGE and analyzed by Western blotting using antibodies raised against albumin, α1-antitrypsin and TP4 (Fig. 6a). Interestingly, TP4-A12I,A15I was associated with protein A, protein B as well as monomeric and dimeric protein D (Fig. 6a). However, only protein B was still tightly associated with the immobilized TP4-A12I,A15I if the pulled down-protein D was dissociated by β-ME (Fig. 6b). These results suggest that TP4-A12I,A15I is close to protein B in the albumin and α1-antitrypsin complex (Fig. 6c). www.nature.com/scientificreports/ Binding site of AMP and drug in protein D. The blue shift of fluorescence emission spectrum of protein D was observed following the addition of TP4-A12I,A15I or antipyretic agent ibuprofen, but these reagents did not cause blue shift in protein A (albumin) 13 . As a positive control, the anionic surfactant, sarkosyl, caused blue shifts for both protein A and protein D (Fig. 7a). In contrast, no blue shift was seen in protein A and protein D by the anticonvulsant phenytoin which is weekly bound by albumin 14 . Furthermore, only protein D, but not protein A, was bound by TP4-A12I,A15I analyzed by band-shifted assay. Interestingly, the binding of protein D www.nature.com/scientificreports/ by TP4-A12I,A15I was competitively inhibited by ibuprofen, but not by warfarin or phenytoin (Fig. 7b). These results suggest that TP4-A12I,A15I and ibuprofen may share the same binding site probably at the drug-binding site II of albumin in the protein D complex.
Variation of protein D content in bovine serum. The bovine serum is the major supplement of cell culture media which is generally used in modern medical researches for the assay and production of biomol- (e, f) Decrease of AMP-binding ability of protein D by β-ME. The binding ability of proteins D, A and B to TP4-A12I,A15I peptide were determined by band shift and pulling down assays after reduction with β-ME. 1X represents 0.05% β-ME. Bio, biotinylated TP4-A12I,A15I. www.nature.com/scientificreports/ ecules. We are interested in the amount of protein D in the bovine serum at different developmental stages or sources of vender. The protein D was seen only in the bovine serum at the fetal stage but not after birth analyzed by non-reducing 15% and 8% SDS-PAGE and pulling down experiments (Fig. 8a, b). The homologous protein D was not seen in the serum of adult human. Interestingly, the Apo-AI was minimally seen in FBS, but it markedly increased in calf serum as that in human serum (Fig. 8a). Furthermore, the ratio of protein D to A in FBS varied with serum batch or source of vender (Gibco, CORNIING and Biological industries) as analyzed by non- www.nature.com/scientificreports/ reducing SDS-PAGE (Fig. 8c). It is suggested that the AMP-or drug-binding capacity of serum may vary with the source of serum. In PBS, the antimicrobial activity of TP4 against E. coli was markedly repressed by 10% FBS with 16-folds increase in MBC. Interestingly, the TP4 activity was also markedly inhibited with 32-folds increase in MBC if the assay is performed in DMEM medium alone without FBS addition (Fig. 8d). Similarly, the bactericidal activity of TP4 was further repressed by the addition of 10% FBS in DMEM which is generally used for cell culture (Fig. 8d).

Discussion
Albumin is the most abundant protein in human serum (HSA, 45 mg/ml), followed in descending order by α1-antitrypsin (A1AT, 3.5 mg/ml), transferrin (2.5 mg/ml), α2-macroglobulin (2.5 mg/ml), haptoglobulin (1.8 mg/ml), low-density lipoprotein (LDL, 1.2 mg/ml), α1-acid glycoprotein (AGP, 0.8 mg/ml) and high-density lipoprotein (HDL, 0.6 mg/ml) 11 . These proteins are involved in the transportation of fatty acids, cholesterols, toxins and drugs in the blood stream. More than 90% of anticoagulation drug warfarin and antipyretic drug ibuprofen were shown to bind to site I and site II of human albumin, respectively [8][9][10] . However, the binding ability of anticonvulsant drug phenytoin to albumin is low although 90% of the drug is bound to serum protein 14 . The Figure 6. Binding of TP4-A12I,A15I peptides to the albumin/α1-antitrypsin complex. (a) Binding of TP4-A12I,A15I peptides to protein D. The AMP-protein D complexes were cross-linked with glutaraldehyde (GA), separated by 8% or 15% reduced SDS-PAGE, stained by Coomassie blue and detected by Western blotting using antibodies raised against albumin, α1-antitrypsin and TP4, respectively. 1X represents 0.00625% GA. (b) Binding of TP4-A12I,A15I peptides to the α1-antitrypsin subunit. The protein D was pulled down by immobilized biotinylated TP4-A12I,A15I on Streptavidin beads. The retained proteins after β-mercaptoethanol (β-ME) wash were analyzed by 8% reduced SDS-PAGE and Coomassie blue staining. 1X represents 0.25% β-ME. (c) Model for the interaction between AMP and protein D. The active complex is composed of two identical protein D and each subunit is composed of protein A (albumin) and protein B (α1-antitrypsin) linked by disulfide bond. www.nature.com/scientificreports/ lipopeptide antibiotics daptomycin is highly bound to human serum components (90 to 94%) 15 in the following order, albumin > > α1-antitrypsin > low-density lipoprotein > high-density lipoprotein > transferrin 11 . In addition, some antibiotics, such as ceftriaxone and moxifloxacin, and short cationic AMPs, such as CAP1-CAP9, are also bound and inhibited by human and bovine serum albumin 12,16,17 . In this study we found that the fetal bovine serum albumin was active in the binding and inhibition of AMP activity only in complex with α1-antitrypsin, whereas the free form albumin or free α1-antitrypsin alone was almost unable to exert these binding properties when analyzed by all the following detection methods: band shift, pulling down, biolayer interferometry, gel filtration chromatography, bactericidal activity and blue shifts of fluorescence emission spectrum. The existence of active albumin and α1-antitrypsin complex has not been reported in human or bovine serum yet. Our results also show that the relative amounts of free and complex albumin in the serum varied www.nature.com/scientificreports/ with the developmental stage of bovine, batches of serum sources (Fig. 8). Therefore, the amount of albumin/ α1-antitrypsin complex is very important because bovine serum is a major supplement of culture media which is used for the assay and production of biomolecules. In addition to the protein D (albumin/α1-antitrypsin complex), the culture medium (DMEM) which contains salts and divalent cations is also inhibitory to the antimicrobial activity of TP4 (Fig. 8). To prevent the inhibitory effects of serum component on the antimicrobial activity of AMP, the protein D in the FBS may be depleted by the AMP-immobilized beads before adding to the culture media, DMEM or RPMI. The band shift of TP4-A12I,A15I on native PAGE occurred with bovine protein D, but not with protein A (Fig. 3a). Interestingly, the binding between protein D and TP4-A12I,A15I was blocked if the protein D-containing mixture was pre-incubated with ibuprofen, but not with warfarin or phenytoin, before binding to TP4-A12I,A15I (Fig. 7b). It is known that warfarin and ibuprofen bind to human albumin at the binding site I and II, respectively 10 . Furthermore, the conformation changes of albumin could be monitored by the blue shift of fluorescence emission spectrum exerted by tryptophan residues. The movement of tryptophan residues, W131 and W214 located in sites I and II of BSA, induced by drug treatment could be detected by the assay 13,18 . For example, the anionic surfactant sarkosyl drove the blue shifts in both proteins A and D (Fig. 7a). The TP4-A12I,A15I and ibuprofen drove blue shift only in bovine protein D, but not protein A, whereas phenytoin did not exert blue shift in both proteins. These results suggest that the TP4-A12I,A15I and ibuprofen share the same binding site probably at the binding site II of albumin in protein D. www.nature.com/scientificreports/ The α1-antitrypsin belongs to the serpin family being inhibitors of serine proteinases. Both protein B (free form α1-antitrypsin) and protein D (complex form) exerted anti-proteolytic activity (supplementary Fig. 4). It is a glycoprotein with 394 amino acid residues and composed of 3 β-sheets and 9 α-helices having only one cysteine at Cys232 for disulfide linkage with albumin which contains 17 pairs of disulfide bond and one free cysteine 19 . The TP4 signals were seen on the A and B subunits as well as on monomeric and dimeric forms of protein D as detected by Western blotting after being cross-linked by glutaraldehyde (Fig. 6a). However, protein B still bound to TP4-A12I,A15I if protein A was removed from the AMP-protein D complex by β-ME (Fig. 6b). It indicates that AMP lies close to protein B (α1-antitrypsin) in the complex. Furthermore, the dimerization of two identical protein D molecules was important for AMP-binding since the pulling down of protein D by immobilized AMP was inhibited by anionic surfactant SDS which disrupts hydrophobic interaction between these two subunits (Fig. 5e, f). The model of AMP binding in the complex is proposed as shown in Fig. 6c. The complex is composed of two identical protein D molecules tethered by hydrophobic interaction, each protein D subunit consisting of protein A and B linked by disulfide bond.
In addition to protein D, the serum protein apolipoprotein AI (Apo-AI) was also pulled down from crude fetal bovine serum by immobilized TP4-A12I,A15I, but in a less amount (Fig. 2a). The Apo-AI with Mr = 28,080 Da is the major constituent of high-density lipoprotein (HDL), which is involved in the removal of cholesterol from peripheral tissues and its catabolism. The protein has been shown to bind human AMP, LL37, in an 1:1 complex and inhibits its antimicrobial and cytotoxic activities 20,21 . The amount of Apo-A1 in fetal bovine serum was much less than that of protein D (Fig. 8a, b). Interestingly, the amount of protein D decreased and Apo-Al increased in the bovine serum after birth. It is of note that the absence of protein D and presence of Apo-Al were also seen in adult human serum (Fig. 8a). It is suggested that protein D may play important roles in the development of fetus in bovine and probably in human, while the Apo-AI in HDL may play roles in the neutralization and transportation of AMPs and drugs in the sera of bovine and human after birth.
In conclusion, the albumin is active in AMP-binding only in complex with α1-antitrypsin in bovine serum at fetal stage, but not after birth. It is composed of two identical albumin-α1-antitrypsin subunits linked by disulfide bond. The amount of albumin/α1-antitrypsin complex in bovine serum is very important and may be depleted by AMP-immobilized beads because FBS is a major supplement for cell culture media which is used for the assay and production of biomolecules. www.nature.com/scientificreports/ Identification of TP4-binding proteins in FBS. Streptavidin-conjugated beads were incubated with 10 μg of biotinylated TP4-A12I,A15I peptides in 600 μl PC buffer for one hour on a rolling wheel. The immobilized biotinylated AMPs were further mixed with serum components in the indicated 600 μl buffer(s) at room temperature for 30 min on a rolling wheel, then washed twice with 600 μl respective buffer and subjected to SDS-PAGE and Coomassie Blue staining. Specific proteins were excised from SDS-PAGE gel and subjected to in-gel trypsin digestion and liquid chromatography-tandem mass spectrometry 26 .
Binding kinetics of protein to biotinylated TP4-A12I,A15I. The biolayer interferometry were measured by an Octet RED384 instrument (ForteBio). Biotinylated TP4-A12I,A15I (10 μM) were immobilized on a Streptavidin biosensor. Serial dilutions of serum components in 200 μl PC buffer were used as analyte. Affinity (Kd) and kinetic parameters (k on and k off ) were calculated from a global fit (1:1) of the data using the Octet RED384 software 27 .
Antitrypsin activity analysis. The inhibitions of trypsin's proteolytic activity by serum components on peptides TP4 and SAAP159 were determined as follows. Freshly diluted trypsin (2 ng; 0.2 μg/ml) was pre-incubated with various concentrations of serum components in PC buffer for 30 min before the mixture was added to TP4 (2 μg; 0.2 mg/ml) or SAAP159 (4 μg/; 0.4 mg/ml) peptides in 10 μl PC buffer at 37 °C for 30 min. The digested peptide fragments were analyzed by 15% reduced SDS-PAGE and Coomassie Blue staining.
Cross-linking assay. Small aliquots of serum components (4 μg) in the presence/absence of AMPs (2 μg) were incubated with various concentrations of glutaraldehyde in 10 μl PC buffer at 37 °C for 30 min. The crosslinked complexes were analyzed by reduced SDS-PAGE and Coomassie Blue staining 28 .
High-performance liquid chromatography. The 2-mercaptoethanol-treated samples in 50 μl was uploaded into a C4 column (250 mm × 4.6 mm, 5 μm, Vydac) at 37 °C, washed with A buffer (0.1% trifluoroacetic acid and 2% acetonitrile, v/v), eluted with a gradient of B buffer (0.1% trifluoroacetic acid and 98% acetonitrile, v/v) at a flow rate of 1 ml/min and detected at the wavelength of 220 nm. The relative amount of serum components was quantified by Agilent ChemStation program.
Fluorescence emission spectrum measurement. The serum components (protein A or D, 0.75 μM each) were incubated with various concentrations of reagent, such as sarkosyl, TP4-A12I,A15I, ibuprofen and phenytoin, in 200 μl PC buffer at 25 °C for 30 min. The mixtures were excited by 285 nm and its emission spectra were recorded between 300 and 450 nm using a temperature-controlled fluorimeter (FP8500, Jasco, Japan) 13 .
Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request.