Redirecting Pore Assembly of Staphylococcal α-Hemolysin by Protein Engineering.

α-Hemolysin (αHL), a β-barrel pore-forming toxin (βPFT), is secreted as a water-soluble monomer by Staphylococcus aureus. Upon binding to receptors on target cell membranes, αHL assembles to form heptameric membrane-spanning pores. We have previously engineered αHL to create a protease-activatable toxin that is activated by site-specific proteolysis including by tumor proteases. In this study, we redesigned αHL so that it requires 2-fold activation on target cells through (i) binding to specific receptors, and (ii) extracellular proteolytic cleavage. To assess our strategy, we constructed a fusion protein of αHL with galectin-1 (αHLG1, αHL-Galectin-1 chimera). αHLG1 was cytolytic toward cells that lack a receptor for wild-type αHL. We then constructed protease-activatable mutants of αHLG1 (PAMαHLG1s). PAMαHLG1s were activated by matrix metalloproteinase 2 (MMP-2) and had approximately 50-fold higher cytolytic activity toward MMP-2 overexpressing cells (HT-1080 cells) than toward non-overexpressing cells (HL-60 cells). Our approach provides a novel strategy for tailoring pore-forming toxins for therapeutic applications.

Engineered pore-forming toxins (PFTs) have been extensively studied and applied in biotechnology.[1,2] A few PFTs have been successfully employed in the field of biomolecule sensing, such as DNA sequencing.[2] The structures and assembly mechanisms of PFTs, knowledge of which is essential for efficient protein engineering, have been thoroughly studied: β-barrel pore-forming toxins (βPFTs), such as aerolysin,[3] anthrax toxin,[4]Vibrio cholerae cytolysin (VCC),[5] and α-hemolysin (αHL),[6] which form membrane-spanning β-barrels, bind to specific cell-surface receptors, assemble into inactive prepore intermediates, and transform into membrane-spanning active pores.[7] Among these βPFTs, aerolysin,[8] anthrax toxin[9] and VCC,[10] for example, are synthesized as inactive forms of toxin called “prolysins” and require proteolytic cleavage of redundant a peptide at the C-terminus or N-terminus to generate cytolytic activity toward target cells by receptor binding and pore-formation.[11] For example, furin activates proaerolysin[8] and proanthrax toxin[9] by cleaving peptides from the C-terminus and N-terminus, respectively, and A Disintegrin and Metalloprotease 17 (ADAM-17) activates pro-VCC by cleaving peptides from the N-terminus,[10] presumably after binding to the receptor on target cell membranes.[5] The proteolytic activation of prolysin is an important requirement in the assembly of a toxin, the absence of which prevents the assembly of lytic pores into nontarget membranes. The 2-fold action of receptor binding and protease activation enhances cell specificity. In the present work, we introduce such a 2-fold specificity into αHL by protein engineering.

Protein redesign using genetic modification provides a tuning strategy for the alteration of toxin properties without eliminating cytotoxicity.[1] Several studies have constructed PFT fusion proteins with target-specific ligands such as a colicin–pheromone fusion protein targeted to Staphylococcus aureus.[12] Mechaly et al. constructed an anthrax–EGFR fusion protein, and ablated the native receptor-binding domain of anthrax toxin to improve the receptor specificity toward targeted EGFR bearing cells.[13] In addition to specific receptor binding,  proteolytic cleavage to activate a prolysin provides a highly effective tool for the development of 2-fold specificity. Anthrax toxins have been re-engineered by swapping the furin-recognition domain (i.e., a native protease-recognition domain) with matrix metalloprotease-recognition (MMP-recognition) domain and have shown selective toxicity to MMP overexpressing tumor cells.[14,15] The  requirement for the activation of certain PFTs on the target cell membrane provides a strategy for tailoring cytolytic activity dependent on a specific protease secreted from target cells. In addition to redirecting toxins by the fusion of a heterologous ligand, this ability to regulate cytotoxicity by redesign will greatly contribute to increased cell target specificity.

Pore assembly mechanisms of αHL and its distinct pore properties have been studied using biochemical and genetic approaches.[16−18] αHL lyses rabbit erythrocytes at concentrations 1000-fold lower than those required for human erythrocytes.[19] αHL has been shown to bind to target cell membranes via direct interactions with specific membrane lipids and binding to specific receptors, such as phosphocholine headgroups[20,21] and A Disintegrin and Metalloprotease 10 (ADAM 10).[22] These results account for the high αHL cytolytic activity toward rabbit erythrocytes, which express ADAM 10 highly by comparison with human erythrocytes.[22] To closely explore the mechanism of pore assembly, we previously identified important domains in the αHL polypeptides for binding, oligomerization, and pore formation by using 83 single-cysteine point mutagenesis and truncation mutagenesis.[6,17]

We previously constructed a complementation mutant of αHL consisting of two truncated polypeptide segments encompassing the N-terminal and C-terminal halves of the polypeptide chain. This discontinuous αHL mutant with a nick in the stem domain demonstrated that the integrity of the stem domain of αHL is not required for  pore-formation.[23] Based on this knowledge, we engineered αHL by introducing redundant peptides containing protease recognition sites in discountinuous stem domains. These toxins remained inactive unless the redundant peptides were removed by site-specific proteases[24] such as cathepsin B (i.e., a tumor protease).[25] This was the first reported approach that exogenously introduced a protease-recognition site as a “protease-actuated trigger” in the core region of the pore-formation component of a PFT that does not have a native protease-binding domain.

In this study, we redesigned a complementation mutant of αHL by introducing a tumor protease-recognition site in the  stem domain and a receptor-binding domain (i.e., αHL-lectin chimera) at the C-terminus. This mutant required 2-fold activation to produce cytolytic activity toward target cells. The design of this toxin–lectin chimera was based on the characteristics of Vibrio cholera cytolysin (VCC), which has high structural similarity to αHL, but contains two additional C-terminal lectin domains that take part in cell binding and pore-formation: a β-prism domain that interacts with carbohydrate receptors on cell membranes,[26] and a β-trefoil domain that may be involved in oligomerization.[5] Additionally, VCC contains a protease-recognition site that enables the proteolytic cleavage of the proregion that results in conversion of pro-VCC to mature VCC.[5,10] To assess the feasibility of our approach, we constructed a protease-activatable mutant of αHLG1 (PAMαHLG1). We fused galectin-1 to the C-terminus of αHL (αHLG1). We then introduced a protease-recognition site in the stem loop of αHLG1 flanked by a peptide extension so as to inactivate the toxin (i.e., form a prolysin). Our approach provides a template for engineering PFTs for therapeutic applications.

Results

αHLG1 Has Increased Hemolytic Activity toward Human RBCs Compared with αHL

Engineered αHLs with C-terminal extensions have been previously reported to form functional pores. For example, αHL fused to the 94 amino acid residues (289–382) of the C-terminal tail of hemolysin II from Bacillus cereus(27) and subunit dimers of αHL (αHL-linker-αHL)[28] are hemolytically active. In the present study, we fused galectin-1 (14 kDa) to the C-terminus of αHL (34 kDa) through a glycine–serine linker (TSSGSS) to form αHLG1 (48 kDa) (Figure A–C). The fused lectin domain (galectin-1) was introduced by cassette mutagenesis, and is readily switchable to other binding molecules, such as mAbs and cancer-specific ligands. Galectin-1 is a β-galactoside-binding protein with high affinity toward human erythrocytes.[29] αHLG1 (29 nM) exhibited the same hemolytic activity toward 0.5% rabbit red blood cells (rRBCs) as αHL (wild-type αHL, 29 nM) (Figure A). By contrast, αHLG1 bound more extensively to human red blood cells (hRBCs) and had a higher hemolytic activity toward hRBCs compared with αHL (wild-type αHL) (Figure B,D). We assessed the differential hemolytic activity between αHLG1 and αHL toward hRBCs through long-duration hemolysis assays. αHLG1 exhibited a 172-fold higher lysis rate (% cell lysiss –1) (Figure C and Table , % cell lysis = ((ODinitial – OD80)/(ODinitial – ODfinal)) × 100, where ODinitial, ODfinal, and OD80 are OD values (595 nm) at initial, final, and 80% cell lysis timepoints, respectively). The initial lag period (time to 10% lysis) of αHLG1 was 21 times shorter than that of αHL on hRBC (Figure C and Table ). αHLG1 formed SDS-stable oligomers on both rRBCs and hRBCs, whereas αHL did so only on rRBCs, as analyzed by SDS-polyacrylamide gel electrophoresis (Figure D). The extents of binding of αHLG1 to both rRBCs and hRBCs were higher than those of αHL: 2-fold higher for rRBCs and 1000-fold higher for hRBCs (Table S1). Galectin-1 has been reported to increase the osmofragility of hRBCs by cross-linking membrane constituents.[29,30] We found that galectin-1 itself had no such effect (Figure S2). We concluded that the binding of galectin-1 to hRBC membranes improved the hemolytic activity  of αHLG1.

Figure 1

Construction of αHLG1. (A) αHLG1 comprises human galectin-1 (Gal1, 14 kDa) fused to the C-terminus of staphylococcal α-hemolysin (αHL, 34 kDa) through a linker (TSSGSS). (B) Representation of the αHLG1 monomer. (C) Expression of monomeric forms of galectin-1, αHL, and αHLG1. The proteins were synthesized by IVTT (In Vitro Transcription and Translation) in the presence of [35S]methionine and subjected to electrophoresis in a 10% SDS-polyacrylamide gel followed by autoradiography. The molecular masses of galectin-1, αHL, and αHLG1 are 14, 34, and 48 kDa, respectively. M: protein molecular mass markers ([14C]methylated protein, Amersham Bioscience).

Figure 2

Hemolytic activities of αHL and αHLG1 toward rabbit and human erythrocytes. IVTT proteins (29 nM) were added to the first well of each row in a microtiter plate and subjected to 2-fold serial dilution in MBSA across each row leaving 50 μL in each well. 1.0% rRBCs (A) or hRBCs (B) suspended in MBSA (50 μL) were then added to each well (0.5% final concentration of RBCs), and the light scattering at 595 nm was recorded over 2 h. (C) Long-duration hemolytic activity assays for αHL and αHLG1 toward hRBCs. The activities of αHL (◆) and αHLG1 (■) (48 nM) toward hRBC were monitored for 72 h. The results plotted are from three independent assays (mean ± SD, n = 3). (D) Extents of binding of αHL and αHLG1 to RBC as demonstrated by SDS-polyacrylamide gel electrophoresis. IVTT proteins (29 nM) were incubated with 0.5% rRBC (left) or 0.5% hRBC (right) for 20 min at room temperature. * membrane-bound monomer; **, membrane-bound heptamer. M: Protein molecular mass markers, 14C-methylated protein (Amersham Bioscience). H1, αHLG1 monomer; α1, αHL monomer; H7, αHLG1 heptamer; α7, αHL heptamer.

Table 1

Hemolytic Activity Comparison Chart

	initial lag period (s)	lysis rate (% cell lyseds –1)
αHL	2500	6.2 × 10–6
αHLG1	120	1 × 10–3

Inhibition of αHLG1 Binding to hRBCs by Carbohydrates

To verify the role of the galectin-1 domain in the binding of αHLG1 to hRBCs, we performed an inhibition assay. Lactose is a well-known inhibitor of galectin-1,[30] and it has been reported that more than 10 mM lactose is required to efficiently inhibit galectin-1 activity.[31] We performed an inhibition assay with various concentrations of lactose and found that 10 mM lactose completely inhibited the hemolytic activity of αHLG1 (20 nM) (Figure A) and diminished the binding of αHLG1 toward hRBCs (Figure B) while 3 and 6 mM lactose partially inhibited the hemolytic activity (Figure S3). By contrast, galactose and sucrose failed to completely inhibit both the binding and hemolysis of αHLG1 toward hRBCs (Figure A,B). These results suggest that the galectin-1 domain of αHLG1 binds to the surfaces of hRBCs and facilitates the hemolysis by the αHL domain of αHLG1.

Figure 3

Inhibition of αHLG1 activity toward hRBCs by carbohydrates. (A) Inhibition of the hemolytic activity of αHLG1 toward hRBCs. αHLG1 preincubated (20 min) with carbohydrates (lactose, galactose, sucrose: 10 mM) was added to the first well of each row and 2-fold serially diluted across the row (50 μL final volume in each well). The hemolysis assay was initiated by adding 1% hRBCs (50 μL to each well, 0.5% RBCs final). (B) Inhibition of binding and oligomerization. αHLG1 (20 nM), which had been preincubated with a carbohydrate (10 mM), was incubated with 0.5% hRBCs for 20 min at room temperature. A membrane pellet was recovered by centrifugation and resuspended in MBSA. The pellets and supernatants (αHLG1, αHLG1/lactose, αHLG1/galactose, αHLG1/sucrose) were examined by electrophoresis in a 10% SDS-polyacrylamide gel, followed by autoradiography. M, protein molecular mass markers; Lac, lactose; Gal, galactose; Suc, sucrose.

Cytotoxicity of αHLG1 toward Cancer Cells Is Higher than That of αHL

We assessed whether the galectin-1 domain of αHLG1 mediated binding to human cancer cells (HL-60, promyelocytic leukemia) expressing galectin-1 receptors as it does toward RBCs. Both monomeric and dimeric galectin-1 have been reported to bind to HL-60 cells through poly-N-acetyllactosamine (Galβ1–4GlcNAc).[30] We evaluated the cytolytic activity of αHLG1 toward HL-60 cells using flow cytometry (Figure A and Figure S5). In brief, lysed cells and healthy cells (10 000 cells in total) were sorted and quantified using two parameters: forward-scattered light (FSC, proportional to cell-surface size) and side-scattered light (SCC, proportional to cell granularity). The cytolytic activity of αHLG1 (19 nM) was at least 2-fold higher than that of αHL (19 nM) toward HL-60 cells (1 × 107 cells mL –1, Figure A). αHLG1 also had a shorter lag period (time to 5% cell death): αHLG1, 25 min; αHL, 35 min. The percentages of cell death at 45 min were 10.0 ± 0.5% for αHL and 24.6 ± 2.6% for αHLG1 (mean ± SD, n = 3), and the lysis rates (% cell death min –1) were 0.38 ± 0.01 for αHL and 0.78 ± 0.04 for αHLG1 (mean ± SD, n = 3). These findings were consistent with the results of quantitative binding assays (αHL, 9.6% of αHL bound; αHLG1, 43% of αHL bound) (Figure B and Table S1).

Figure 4

Cytotoxicity of αHLG1 toward human cancer cells. (A) Lysis of HL-60 cells as detected by flow cytometry. Lysis was monitored every 5 min for 75 min (10 000 cells at each time point) at room temperature: αHL (◆) and αHLG1 (■). Briefly, HL-60 cells were washed with PBS and medium (IMDM) containing 3% FBS. The assay was started by adding IVTT proteins (19 nM) to the cells (1 × 107 cells mL–1). % Lysis = (number of toxin-treated dead cells – number of untreated dead cells)/(10 000 – number of untreated dead cells) × 100. The detailed experimental procedures are described in Methods. (B) Extent of binding of αHL and αHLG1 to HL-60 cells in IMDM as determined by electrophoresis in a 10% SDS-polyacrylamide gel. Proteins (19 nM) radiolabeled with [35S]methionine were incubated with cells (1 × 107 cells mL–1) for 40 min at room temperature. After centrifugation, the pellets were treated with DNase, and samples were subjected to 10% SDS-PAGE, followed by autoradiography: α1 and H1, monomers of αHL and αHLG1, respectively; α7 and H7, heptamers of αHL and αHLG1, respectively. (C) Cytotoxicity of αHL and αHLG1 toward HT-1080 cells. Protein (48 nM) was incubated with HT-1080 cells (1 × 107 cells mL–1) in assay medium (DMEM, 3% FBS) for 2 h with 5% CO2 at 37 °C. The CytoTox 96 assay (Promega) was used to measure cytotoxicity. (D) Extent of binding of αHL and αHLG1 to HT-1080 cells. The toxins (48 nM) were incubated with HT-1080 cells in DMEM for 40 min at room temperature. Extents of binding were determined by electrophoresis in a 10% SDS-polyacrylamide gel, followed by autoradiography: α1 and H1, monomers of αHL and αHLG1, respectively; α7 and H7, heptamers of αHL and αHLG1, respectively.

We evaluated the binding and cytolytic activity of αHLG1 on different cell types (other than RBCs and HL-60 cells), e.g., human fibrosarcoma cells (HT-1080 cells, 1 × 107 cells mL–1). HT-1080 cells display N-acetyllactosamine, a receptor for galectin-1,[32] and express matrix metalloproteinases (MMPs), which are required for activation of PAMαHLG1.[33] αHLG1 (48 nM) had cytolytic activity toward HT-1080 cells (69%), which was higher than that of αHL (28%) (Figure C). αHLG1 formed SDS-stable oligomers in the presence of HT-1080 cells, whereas αHL monomers were detected, but oligomers were not (Figure D). Based on a band intensity comparison, the extents of binding of αHLG1 and αHL (monomers and oligomers) to HT-1080 cell membranes were 42% and 13% of αHL bound, respectively (Table S1).

The cytotoxicities of αHLG1 toward cells from both a hematological malignancy (HL-60 cells) and a solid tumor (HT-1080 cells) were greater than that of αHL (Figure A,C). Our results suggest that αHL fusion protein with interchangeable receptor-binding domains may be readily modified and redirected to bind to targeted cells.

Construction of PAMαHLG1, a Protease-Activatable Mutant of αHLG1

PAMαHL, the two-chain overlap mutant of αHL, was prepared as previously described.[25] Two-chain overlap mutants of αHL are activated when redundant amino acids in the stem domain (β-barrels) are removed by proteolysis.[24] In other words, additional overlapping amino acid residues in the β-barrels of two-chain mutants of αHL inhibit cytolytic pore-formation. We have also demonstrated that few PAMαHLs selected from a library of candidates are activated by tumor proteases.[25] These results prompted us to construct PAMαHLG1s (protease-activatable mutants of αHLG1) by converting αHLG1 to a two-chain overlap mutant (Figure A). PAMαHLG1s are “prolysins” that require activation and incorporate a targeting mechanism conferring 2-fold specificity: (1) a galectin-1 domain to bind to receptors on the target-cell surface; and (2) a protease-recognition site for activation by removal of residues 119–132 (Figure A and Table S2). In this study, we constructed two mutants of αHL, PAMαHL (as a control) and PAMαHLG1: PAMαHL is composed of two truncated fragments of αHL, amino acids 1–131, 16 kDa, and amino acids 119–293, 22 kDa. PAMαHLG1 is composed of two truncated fragments of αHLG1: amino acids 1–131 and amino acids 119–293-galectin-1 of αHLG1, 36 kDa (Figure A,B).

Figure 5

Protease-activatable mutants (PAMs) of α-hemolysin. (A) Schematic diagram of PAMs. The proteins are composed of two polypeptide chains: PAM-A contains amino acids 1–131 of αHL (16 kDa) and is present in both PAMαHL and PAMαHLG1. PAMαHL-B is present in PAMαHL and contains amino acids 119–293 of αHL (22 kDa). PAMαHLG1-B is present in PAMαHLG1 and contains amino acids 119–293 of αHL followed by a linker (TSSGSS) and galectin-1 (14 kDa). A protease recognition site is present in amino acids 127–134 of αHL in both PAMαHL-B and PAMαHLG1-B. (B) Monomers of the PAMs were subjected to electrophoresis in a 10% SDS-polyacrylamide gel, followed by autoradiography. A, PAM-A; B, PAMαHL; B′, PAMαHLG1-B. (C) Hemolysis of hRBCs by PAMαHLG1s after protease-activation. Activated PAMαHLG1 was added to a well of a microtiter plate and subjected to 2-fold serial dilution in MBSA down a column of the plate to give a final volume of 50 μL in each well. Hemolysis assays were initiated by adding 1% hRBCs (50 μL) to the wells and monitored for 3 h by observing the decrease in light scattering at 595 nm. +, protease-treated hRBC; −, untreated hRBC.

Cell-Specific Lytic Activity of PAMαHLG1

The proteases, cathepsin B and matrix metalloproteinases (MMPs), are involved in tumor invasion and metastasis.[34,35] We constructed four different PAMs (protease-activatable mutants): PLHL and IGHL for PAMαHL, and PLHLG1 and IGHLG1 for PAMαHLG1. These proteins have two different MMP-recognition sequences: ···PLGLAGGG··· and ···TRRIGGLG···, which are prefixed PL- and IG-, respectively (Table S2). PXX↓H (H, hydrophobic residue) is a good recognition motif for most MMPs, whereas L/IXX↓H, HSX↓L, and HXX↓H are specifically recognized by MMP-2.[36] Therefore, PLHLG1 (PLG↓H) and IGHLG1 (L/IGG↓H) were evaluated for their ability to lyse hRBC after treatment with MMP-2 (0.3 μg μL–1) (Figure C). Both constructs were activated for hemolysis by MMP-2, although the activity of IGHLG1 was higher than that of PLHLG1. PLHLG1 (36 nM) was also activated by cathepsin B (0.8 μg μL–1) although its activity was limited, but IGHLG1 (36 nM) was not activated (Figure C). These findings were consistent with our previous report showing that αHL-RR (a two-chain mutant of αHL with the same protease-recognition site as IGHL/IGHLG1, Table S2) was not activated by cathepsin B.[25]

HT-1080 cells constitutively secrete MMP-2[33,37] and cathepsin B[38] and display N-acetyllactosamine, a galectin-1 receptor,[32] whereas HL-60 cells display N-acetyllactosamine and secrete cathepsin B[39] but do not overexpress MMP-2.[33,40] Accordingly, the susceptibility of HT-1080 cells to PAMs (PLHL, 9.8%; PLHLG1, 25%; IGHL, 8.0%; and IGHLG1, 19%) was higher than that of HL-60 cells (PLHL, 0.20%; PLHLG1, 1.9%; IGHL, 0.47%; and IGHLG1, 0.44%), suggesting that MMP-2 activated cytotoxicity (Figure A). Additionally, the cytotoxicity of PAMαHLG1 (PLHLG1, 25%; and IGHLG1, 19%) was higher than that of PAMαHL (PLHL, 9.8%; and IGHL, 8.1%) toward HT-1080 cells, suggesting that the galectin-1 domain of PAMαHLG1 enhances activity (Figure A). Oligomers of PAMαHLG1 associated with HT-1080 cells were visible on SDS-polyacrylamide gels, but PAMαHL oligomers were not detected, while PAMs were cytotoxic (Figure B). No oligomers were detected on HL-60 cells (Figure B). To assess MMP-2 activity in our experimental model, we performed gelatin zymography using a 10% SDS-PAGE gelatin zymogram gel (Bio-Rad). Pro-MMP-2 (74 kDa) and active-MMP-2 (63 kDa) were detected in HT-1080 cells, but not in HL-60 cells (Figure C). This result clearly indicates that MMP-2 secreted from HT-1080 cells activates PAMαHLG1s. Further, the dual regulatory domains of cytotoxicity in PAMαHLG1, i.e., the fused galectin-1 domain and the MMP-2-recognition domain, did not eliminate the native cytolytic activity of αHL. This result provides  promising evidence of 2-fold specificity of engineered αHL as we previously proposed.[41]

Figure 6

Cytotoxicity of PAMs toward cultured cancer cell lines, HT-1080 and HL-60 cells. (A) Cytotoxicity of PAMs toward HT-1080 (solid bars) or HL-60 (lined bars) cells. The cytotoxicity was evaluated by using the CytoTox 96 kit (Promega) after 8 h of incubation with PAMαHL or PAMαHLG1; the absorbance at 490 nm representing the concentration of released lactate dehydrogenase was recorded with a microtiter plate reader. Cytotoxicity (%) = ((As – Aspon)/(Amax – Aspon)) × 100 where As is the absorbance of toxin-treated cells, Aspon is the absorbance of untreated cells (spontaneously dead cells), and Amax is the absorbance of cells treated with lysis buffer. (B) Binding assay of PAMs to HT-1080 and HL-60 cells determined by electrophoresis in a 10% SDS-polyacrylamide gel, followed by autoradiography. PAMs (PLHL, PLHLG1, IGHL, IGHLG1; 62 nM) were incubated with HT-1080 cells or HL-60 cells (1 × 107 cells mL–1) in 5% CO2 for 40 min at 37 °C. After centrifugation, the resuspended pellets and the supernatants were analyzed. ▲, oligomers; P, pellet; S, supernatant. (C) Extent of MMP-2 expression from HT-1080 and HL-60 cells analyzed by SDS-PAGE gelatin zymography. HT-1080 cells and HL-60 cells were incubated in serum-free DMEM and IMDM, respectively, in 5% CO2 at 37 °C overnight. The media were collected, centrifuged to remove debris, and mixed with zymogram sample buffer before loading into the zymogram gel (2 μg final concentration of total protein).

Discussion

In the present study, our approach was to modulate the cytotoxicity of αHL by protein engineering. We re-engineered α-HL into a protoxin chimera (PAMαHLG1): Full length αHL was fused with galectin-1, and divided into two complementary fragments, which are cotranslated in vitro. PAMαHLG1 consists of three domains: the toxin domain (αHL), a proteinase-recognition domain in the stem  forming region of αHL (MMP-2), and a receptor-binding domain at the C-terminus of αHL (galectin-1) (Figure ). Our future goal is to expand our approach to generate a tailored targeted toxin called “proimmunolysin” for cancer therapy with 2-fold specificity provided by specific ligand–receptor binding and protease activation. Tumor-specific ligands, such as cancer-specific monoclonal antibodies (mAb) and growth factors might be used rather than galectin-1.

It has been reported that the binding of αHL to hRBC is weak due to a lack of high affinity binding sites, including clustered sphingo-cholesterol domains[20] and ADAM 10.[22] The binding of αHL to hRBC is enhanced by the fusion of galectin1 (Figure ). We believe that the galectin-1 thereby increases the concentration of αHL on the cell surface and thus enhances the hemolytic activity. We assumed the following pore assembly mechanism of PAMαHLG1, which is based on the pore assembly mechanism of VCC proposed by Olson et al.[5] A monomer of PAMαHLG1 binds to the galectin-1 receptor. MMP-2 on the proximity of  the cells cleaves redundant peptides at the central loop of PAMαHLG1 (Figure ), and PAMαHLG1 forms functional pores containing a nick at the central loop. However, bettter understanding of the mechanism of pore-formation by PAMαHLG1 will be necessary to engineer αHL for cancer therapeutics. For example, we need to know whether proteolytic activation occurs before PAMαHLG1 binding to membranes. We believe that the two fragments of PAMαHLG1 come together in solution because they can be ligated to each other.[24] Once they are together, the redundant peptides are removed by proteolysis, which allows assembly into a functional pore. Proteolysis might also occur when the fragments are apart, and then the complementary fragments would become active after association.

Engineered PFTs that primarily kill target cells by breaching cell membranes offer potential improvements over current targeted toxin therapies where cell entry is required. Bacterial PFTs, such as Listeria monocytogenes cholesterol-dependent pore-forming cytolysin (listeriolysin O),[42] and a sea anemone cytolysin,[43] have been exploited to create anticancer agents. Collier and colleagues reported that mPA-ZHER2, in which the cell-binding domain of the pore-forming component (PA) of anthrax toxin was swapped with an affibody[44] targeting HER2, showed high specificity and potency against HER2-positive cancer cells. Interestingly, mPA-ZHER2 showed cytotoxicity against a trastuzumab-resistant cell line.[13,45] Recently, bacteria that deliver pore-forming toxins, i.e., cytolysin A and α-hemolysin, near tumors have been shown to produce therapeutic benefits.[46−48] Further, the antitumor activity of αHL has been demonstrated in xenograft mice by killing tumor cells by necrosis.[48]

In this report, we demonstrated a potential for using engineered αHL in cancer therapeutics. However, there is one drawback: PAMαHL chimera bind to the native receptor of αHL, ADAM10.[22] ADAM10 is expressed on HL-60 and HT-1080 cells, which were our target cell lines.[49,50] This explains why αHL possessed limited cytotoxicity toward HT-1080 and HL-60 cells, but not toward hRBCs that lack ADAM 10[22] (Figures and 4). To enhance the binding specificity of PAMαHL chimera toward target cell membranes, it will be necessary to locate and ablate the ADAM10-binding site on αHL. Additionally, a tumor-associated protease can serve an important role, as a “biological switch”, in targeted cancer therapeutics such as in nanoparticles,[51] cell-penetrating peptides,[52] liposomal drugs,[53] and prodrugs.[54] Torchilin and colleagues have claimed that a single targeting factor, such as proteinase-triggered mechanism, may not be effective due to regulated tumor proteinases in noncancerous cells.[55] Therefore, our 2-fold specificity  approach utilizing dual targeting is an efficient means to attenuate toxic effects toward noncancerous cells.

We previously demonstrated that pore-formation by αHL increased the transport of trehalose into fibroblasts, which significantly increased the survival of cells during cryopreservation.[56] Recently, engineered pore-forming toxins have been developed for the delivery of therapeutic agents.[11] Pirie et al. reported that engineered cholesterol-dependent cytolysins, listeriolysin and perfringolysin, improved the delivery of immunotoxins. These toxins disrupted endosomes releasing immunotoxins into the cytoplasm, and thereby enhanced therapeutic efficacy, whereas  traditional endosomal delivery systems result in the lysosomal degradation of immunotoxins.[57] These results indicate that our  2-fold specificity approach might be applied for the targeted delivery of impermeant anticancer agents.

Methods

Cell Lines

Human red blood cells were purchased from Stanford Blood Center (Palo Alto, CA). HL-60 cells were grown under 5% CO2 at 37 °C in Iscove’s modified Dulbecco’s medium (IMDM, ThermoFisher). HT-1080 cells at 80% confluence were treated with trypsin-EDTA [0.25% (w/v) trypsin and 0.38 g L–1 EDTA with phenol red, GIBCO] at 37 °C, under 5% CO2, for 5 min, followed by the addition of FBS to inhibit trypsinization.

Construction of αHLG1 and PAMαHLG1s

To construct αHLG1, a modified αHL gene in a T7 vector (pT7-αHL-RL3-CTEx) was used.[18] The αHL-RL3-CTEx gene encodes the entire wild-type αHL protein (WTαHL amino acids 1–293)[18] but the gene was modified to contain six silent restriction sites (from 5′ to 3′: SacII, HpaI, BsiWI, StuI, AflII, XhoI) within the central region that encodes the transmembrane β-barrel of the assembled heptameric pore. Additionally, the αHL-RL3-CTEx gene carries a 3′-extension containing inverted EarI sites that allow for a seamless addition of coding sequences. Codon-293 of WTαHL is in bold.

pT7-αHL-RL3-CTEx was used to construct another extended αHL gene in the T7 vector, pT7-αHL-ESA, which encodes a flexible linker (Thr–Ser–Ser–Gly–Ser–Ser) that extends from the last codon of the WTαHL gene. Two EarI sites follow this linker. The 3′-extension in pT7-αHL-ESA contains restriction sites (from 5′ to 3′: SpeI, EarI, ApaI, EarI, HindIII) and includes a His tag-encoding sequence for metal chelate affinity purification of expressed protein (Figure S6). To construct pT7-αHL-ESA, pT7-αHL-RL3-CTEx was digested with EarI and HindIII and ligated with the following three DNA cassettes to yield pT7-αHL-ESA: cassette 1, 5′-AATACTAGTAGCGGATCGTCCAGAAGAG (sense), 5′-ACCTCTTCTGGACGATCCGCTACTAGT (antisense); cassette 2, phosphorylated 5′-GTGGGCCCGCAGCGGCTCTTCCAGTTC (sense), phosphorylated 5′-CCGAACTGGAAGAGCCGCTGCGGGCCC (antisense); and cassette 3, 5′-GGGACACCACCATCACCACCATTGATA (sense) and 5′-AGCTTATCAATGGTGGTGATGGTGGTGTC (antisense). To construct pT7-αHLG1, the human galectin-1 gene was amplified by PCR from a plasmid (pOTB7) obtained from the American Type Culture Collection (ATTC, Manassas, VA) with the following primers: 5′-AGTTACTCTTCGTCCATGGCTTGTGGTCTG GTCGC (forward) and 5′-AGTGGCGCCTACTCTTCGACTGTCAAAGGCCACACATT (reverse). These primers place EarI sites (underlined) at the 5′ and 3′ ends of the gene, which are inverted with respect to each other. The resulting DNA fragment was digested with EarI and ligated into pT7-αHL-ESA that had been cut with the same enzyme to yield pT7-αHLG1. In this paper, we did not use the His tag-encoding version of the fused galectin-1 gene (i.e., αHLG1-H6). Therefore, a stop codon was included in the reverse primer (bold italics).

All protease-activatable mutants of αHLG1 (PAMαHLG1s) were constructed from pT7-αHLG1 and pT7-RR (a two-chain mutant of αHL containing a protease cleavage site, amino acid sequence: TRRI)[25] by in vitro recombination of PCR-amplified fragments[59] (Figure S1). Two plasmids were used to encode PAMαHLG1: pT7-αHL-A and pT7-αHLG1-B, which produced PAM-A (residues 1–131 of αHL) and PAMαHLG1-B (residues 119–293 of αHL, followed by a TSSGSS linker and residues 1–134 of galectin-1), respectively. pT7-RR (which contains two HindIII sites and two NdeI sites) was digested with restriction enzymes: HindIII to make pT7-αHL-A and with NdeI to make pT7-αHLG1-B (Figure S1). pT7-RR was solely digested with HindIII, which resulted in ‘self-ligation’ and form pT7-αHL-A (Figure S1). pT7-RR/NdeI (pT7-αHLG1-B) was further digested with MfeI and SacI, and was ligated to the double digested fragment from pT7-αHLG1/MfeI and SacI to make pT7-αHLG1-B (Figure S1). An MMP-2-recognition domain was introduced into pT7-αHLG1-B by PCR mutagenesis with the following primers: 5′-ACTGGTGATGATCCACTAGGTCTAGCCGGTGGAGGCCTTATT-3′ (sense); 5′-AATAAGGCCTCCACCGGCTAGACCTAGTGGATCATCACCAGT-3′ (antisense).

In Vitro Transcription and Translation (IVTT)

As we previously demonstrated, proteins were expressed in a cell-free coupled IVTT system by following the instruction  provided by the company (Promega).[27]

The concentration of αHLG1 from an IVTT mix was calculated based on analysis of radioactive band intensities after synthesis in the presence of [35S]methionine and electrophoresis in 10% SDS-polyacrylamide gels as previously described.[27] After electrophoresis, the gels were dried and exposed to a phosphor-imager screen  (Molecular Imager Pharos FX Plus, Bio-Rad). For quantitation, the numbers of methionine residues in the polypeptides (αHL, 7; αHLG1, 9; PAM-A, 3; PAMαHL-B, 5; PAMαHLG1-B, 7) were taken into consideration as well as the known αHL concentration derived from hemolysis assay.

The ratio of the number of methionines is 1.3 (αHL:αHLG1 = 7:9): [αHL] (amount (μL) of αHL for experiment) = [αHLG1] (amount (μL) of αHLG1 for experiment) × 1.3 (band intensity of αHLG1/band intensity of αHL). To evaluate the band intensity, an equal volume of radiolabeled  IVTTs were loaded onto a 10% SDS-polyacrylamide gel. After electrophoresis, the gel was dried and exposed to the phosphor-imager screen (Kodak) for 30 min total (proper exposure duration is critical to prevent saturation). The specific radioactivity of αHLG1 was determined by using the specific activity of αHL from a hemolysis assay; HC50 of αHL is 25 ng mL–1.[18]

No unexpected or unusually high safety hazards were encountered. All experiments were performed at BSL2 level. However, overexpression and purification of αHLG1 may be considered a potential biohazard because αHLG1 is active toward human cells. However, we used IVTT for safety reasons, so that a human-directed pore-forming protein was not expressed in Escherichia coli.

Lysis Assays

Whole blood was washed to yield RBCs (hRBC or rRBC), which were suspended at 1% in MBSA [10 mM 3-(N-morpholino) propanesulfonic acid (MOPS), 150 mM NaCl, 1 mg mL–1 bovine serum albumin (BSA), pH 7.4]. Proteins synthesized by IVTT were diluted into MBSA and subjected to 2-fold serial dilution in the same buffer in microtiter wells (50 μL final volume). RBCs (1%, 50 μL) were pipetted into each well containing the serially diluted protein to produce a final assay volume of 100 μL. Hemolysis was recorded by monitoring the decreases in light scattering at 595 nm. For long-duration hemolysis assays, the microtiter plate was covered with an adhesive film to prevent evaporation between measurements and stored at room temperature. The optical density (OD) at 595 nm was monitored at 10 different time points up to 72 h (TECAN). The presented results are the average of three experimental sets (n = 3). The percent of cells lysed at a time (t) was calculated: % cell lysis = ((ODinitial – OD)/(ODinitial – ODfinal)) × 100, where ODinitial, ODfinal, and OD are the initial and final ODs, and the OD at time t, respectively. The initial lag period was taken to be the time at which 10% lysis had occurred. The hemolysis rate (% cells lysed s–1) was derived from the time required to reach ∼80% lysis. HL-60 cells were washed twice with PBS and resuspended in IMDM containing 3% FBS. The cells were then divided into three tubes (polystyrene round-bottom tube, Falcon): control, αHL, and αHLG1. The lysis of HL-60 cells (1 × 107 cells mL–1) was monitored by flow cytometry (FACSCalibur, Becton Dickinson) every 5 min for 75 min after the addition of the toxin. The sample tubes were gently agitated during the incubation. Forward scatter (FSC) and side scatter (SSC) were monitored to identify damaged cells and cell debris. Data analysis was performed by Cellquest (BectonDickinson, Franklin Lakes, NJ) (Figure S5). Percent lysis (%) = ((Ns – Nc)/(Nt – Nc)) × 100, where Ns is the number of lysed cells in the treated sample, Nc is the number of lysed cells in the untreated control sample, and Nt is total number of cells in the assay (Nt = 10 000). The results are presented as the means of three independent assays (n = 3, with triplicate reads in each).

Inhibition of αHLG1 Activity by Carbohydrates

To examine hemolysis inhibition, αHLG1 was mixed with MBSA containing various carbohydrates and incubated for 20 min at room temperature before 2-fold serial dilution across a row of a 96-well microtiter plate. Hemolysis of 0.5% RBCs was monitored by observing the decrease in light scattering at 595 nm for 1 h. The inhibition of membrane binding by carbohydrates was performed by incubating 35S-labeled αHLG1 made by IVTT with the assay buffer (MBSA and carbohydrates) for 20 min before the addition of washed RBCs in MBSA to a final concentration of 0.5%. The assay mix (αHLG1, 20 nM final) was incubated for 30 min and centrifuged. The supernatant and resuspended pellet were loaded onto 10% SDS-polyacrylamide gels. After electrophoresis, dried gels were subjected to autoradiography.

Binding Assay

IVTT proteins were added to the cells (20 μL of 0.5% RBCs, 30 μL of HL-60 cells or HT-1080 cells at 1 × 107 cells mL–1) and incubated for 20 min (RBCs) or 40 min (HL-60 and HT-1080 cells) at room temperature. The cells were recovered by centrifugation for 3 min at 12000g. For the HL-60 or HT-1080 cells, cell pellets (18 μL) were resuspended in MBSA and treated with 2 μL of DNase (2000 units mL–1, New England Biolabs) for 30 min at 37 °C. The cells were then pelleted and washed with MBSA twice. After centrifugation, the pellets and supernatants were dissolved in 2× loading buffer without heating, and separated in a 10% SDS-polyacrylamide gel. Autoradiography was carried out as detailed above. To examine the binding of PAM (protease-activatable mutants), PAMs were incubated with HL-60 or HT-1080 cells (1 × 107 cells mL–1) for 40 min at 37 °C under 5% CO2.

Activation of PAMs by Proteolysis

Cathepsin B (Calbiochem) was first activated in 50 mM sodium acetate, 15 mM EDTA, and 30 mM dithiothreitol at pH 5.9, for 30 min at 37 °C. PAMs were then treated with the activated cathepsin B (0.8 μg μL–1) for 40 min at 37 °C, followed by inactivation with leupeptin (1 mM). Procedures for proenzyme activation and MMP-2 storage were as described by the manufacturer (Calbiochem). The proenzyme (as supplied) was mixed 10:1 (v/v) with 30 mM aminophenylmercuric acetate (APMA) in 0.1 M NaOH (Sigma-Aldrich) and incubated for 2 h at 37 °C prior to storage in 50 mM Tris-HCl, pH 7.6, 5 mM CaCl2, and 20% glycerol at −80 °C. Activated MMP-2 (0.3 μg μL–1 final) was incubated with PAMs for 40 min at 37 °C and inhibited by adding 1.25 mM MMP-2 inhibitor I (cis-9-octadecenoyl-N-hydroxylamide, Calbiochem). PAMαHLG1 (36 nM) was incubated at 37 °C for 30 min with MMP-2 (0.3 μg μL–1) or cathepsin B (0.8 μg μL–1).

Cytotoxicity Assay

HT-1080 cells were cultured to 80% confluence and trypsinized as detailed previously. The harvested HT-1080 cells were resuspended in DMEM (10% FBS), and HL-60 cell pellets were resuspended with IMDM (20% FBS). The cells were then seeded onto uncoated 96-well tissue-culture plates (Falcon) and incubated for 16 h (80% confluence, HT-1080 cells). The medium was then removed from the wells, and the cells were washed twice with PBS. Toxins in medium including 3% FBS were added to each well and incubated for different time periods (HL-60 cells, 75 min for αHL and αHLG1, 8 h for PAMs; HT-1080 cells, 2 h for αHL and αHLG1, 8 h for PAMs) under 5% CO2 at 37 °C. The cytotoxicities toward HL-60 and HT-1080 cells were measured by determining lactate dehydrogenase (LDH) activity released upon cell lysis (CytoTox 96 Assay, Promega), by measuring absorbance at 490 nm with a microplate reader and the Magellan management program (both from TECAN). Percent cytotoxicity is defined as cytotoxicity (%) = ((As – Aspon)/(Amax – Aspon)) × 100, where As is absorbance from the LDH assay (cells treated with toxin), Aspon is absorbance from untreated cells, and Amax is absorbance from fully lysed cells.

Gelatin Zymography

MMP-2 enzymatic activity in tissue culture media was determined by 10% SDS-PAGE gelatin zymography (Ready Gel Zymogram Gel, Bio-Rad). HT-1080 cells and HL-60 cells in serum-free media were incubated at 5% CO2 at 37 °C for 18 h. Media were gently collected by pipetting and centrifuged for 5 min to remove debris and floating cells. Total protein concentration in the clear media was measured and  equal amounts of proteins (2 μg, final) were loaded in each lane of  the SDS-PAGE gelatin zymogram gel. Before electrophoresis media were mixed with the zymogram sample buffer (62.5 mM Tris-HCl, 4% SDS, 25% glycerol, 0.01% Bromophenol Blue, Bio-Rad).  After electrophoresis, the gel was incubated in renaturing solution (2.5% Triton X-100) for 30 min with gentle agitation, incubated in development solution (50 mM Tris, 200 mM NaCl, 5 mM CaCl2, 0.02% Brij-35, pH 7.5, Bio-Rad) at 37 °C overnight, stained with 0.5% Coomassie blue R-250 for 1 h, and destained with destaining solution (40% methanol and 10% acetic acid).