Evaluation of Hydra HALT-1 as a toxin moiety for recombinant immunotoxin

Background Immunotoxin is a hybrid protein consisting of a toxin moiety that is linked to a targeting moiety for the purpose of specific elimination of target cells. Toxins used in traditional immunotoxins are practically difficult to be produced in large amount, have poor tissue penetration and a complex internalization process. We hypothesized that the smaller HALT-1, a cytolysin derived from Hydra magnipapillata, can be used as the toxin moiety in construction of a recombinant immunotoxin. Results In this study, pro-inflammatory macrophage was selected as the target cell due to its major roles in numerous inflammatory and autoimmune disorders. We aimed to construct macrophage-targeted recombinant immunotoxins by combining HALT-1 with anti-CD64-scFv in two orientations, and to assess whether their cytotoxic activity and binding capability could be preserved upon molecular fusion. The recombinant immunotoxins, HALT-1-scFv and scFv-HALT-1, were successfully constructed and expressed in Escherichia coli (E. coli). Our data showed that HALT-1 still exhibited significant cytotoxicity against CD64+ and CD64− cell lines upon fusion with anti-CD64 scFv, although it had half cytotoxic activity as compared to HALT-1 alone. As positioning HALT-1 at N- or C-terminus did not affect its potency, the two constructs demonstrated comparable cytotoxic activities with IC50 lower in CD64+ cell line than in CD64− cell line. In contrast, the location of targeting moieties anti-CD64 scFv at C-terminal end was crucial in maintaining the scFv binding capability. Conclusions HALT-1 could be fused with anti-CD64-scFv via a fsexible polypeptide linker. Upon the successful production of this recombinant HALT-1 scFv fusion protein, HALT-1 was proven effective for killing two human cell lines. Hence, this preliminary study strongly suggested that HALT-1 holds potential as the toxin moiety in therapeutic cell targeting.


Background
Immunotoxin, often termed "targeted therapy", is a hybrid protein consisting of a toxin moiety that is linked to a targeting moiety for the purpose of specific elimination of the target cells. The targeting moiety is generally a monoclonal antibody or genetically engineered antibody fragments. The first generation immunotoxins were created by chemically conjugating a monoclonal antibody with a toxin moiety. However, the first generation immunotoxins were large and therefore ineffective in tissue penetration and induced immunogenicity in the host [1]. The latest generation of recombinant immunotoxins were constructed by linking the gene encoding singlechain variable fragment (scFv) with the gene encoding the toxin moiety and expressing them in host cells. Although the latest generation of immunotoxins significantly reduce the molecular weight of the targeting moiety, the toxins commonly used are still too large. Most of these toxins such as RicinA, Pseudomonas exotoxin, and diphtheria toxin are 30-58 kDa and require internalization to the cytosol of target cells to work [1]. These properties lead to disadvantages such as low tissue penetration rate, defect in cytosol delivery and degradation of the immunotoxin in lysosomes before exerting their effect [2,3]. A smaller sized toxin with no internalization process could eliminate these disadvantages.
HALT-1 (Hydra actinoporin-like toxin 1), a poreforming toxin derived from Hydra magnipapillata, could be a new candidate for the toxin moiety in the recombinant immunotoxin [4]. HALT-1 resembles actinoporin, a family of α-pore forming toxins (α-PFTs) first identified in sea anemones but also found in other cnidarians [5]. Actinoporins such as equinatoxins and sticholysins bind to sphingomyelins on the cell membrane and create functional pores by oligomerization of four or more than four monomers, leading to an osmotic imbalance in the cell and subsequently cell lysis [6,7]. Early attempts have been made to use equinatoxin II and sticholysin I & II as immunotoxins for anti-parasite and anti-cancer therapy [8][9][10]. One of these early immunotoxins was based on the disulphite linkage between the sticholysin and the monoclonal antibody IOR-T6 that bound directly to the antigen on the surface of immature T-lymphocytes [8]. This immunotoxin was highly toxic for IOR-T6 carrying cells (CEM) and not toxic for non-IOR-T6 cells (K562). Under reducing condition, sticholysin was released from immunotoxin and able to equally kill both cell types [8]. Another "prototype" used avidin and biotinylated secondary antibody to link two separate moieties, anti-Giardia antibody and biotinylated equinatoxin II, in the anti-Giardia assay [9]. The authors demonstrated quite promising results with respect to the specificity of the toxic effect of actinoporins on parasite cells. Although these actinoporin-based immunotoxins belong to the first or second generations of immunotoxin in which the targeting and toxin components are chemically conjugated in vitro, the actinoporins could exert cytolytic activity against targeted cells and were proven as good candidates for constructing immunotoxins. In recent studies, actinoporin is also known to cause cell death in a regulated manner. For example, intracellular ion imbalance that was due to the low-dose exposure of sticholysin II could activate the RIP1-MEK1/2-ERK1/2-pathways and subsequently induce the regulated necrosis-like cell death mechanism [11,12]. Hence, actinoporins including HALT-1 are versatile proteins with multiple modes of action. Moreover, compared to other toxins used for the construction of immunotoxins, actinoporin or HALT-1 is much smaller in size (20.8 kDa) and works by forming pores on cell membrane, which may provide a solution to overcome the disadvantages of other toxins.
Macrophages have been identified as one of the major cellular players in the pathogenesis of numerous chronic inflammatory disorders including vasculitis [13], atherosclerosis [14], rheumatoid arthritis [15], systemic lupus erythematosus [16], making them an attractive target for immunotoxin development. A study by Thepen et al. [17] demonstrated a successful reduction of chronic cutaneous inflammation in a mice model by targeting inflammatory macrophages using CD64 targeted immunotoxin, H22-RicinA. Generally, macrophages are categorized into two distinct phenotypes, which are the M1 (classically activated, proinflammatory) and M2 (alternatively activated, tissue remodelling) macrophages [13,15,18]. It is important to note that the M1/M2 paradigm of macrophage polarization is an oversimplified classification based on in vitro model, which may not directly resemble macrophage behaviour in vivo. Nonetheless, strong activity of pro-inflammatory cytokines and reactive species, which clearly resembles the skew towards M1-like activation, has been associated with the development of persistent chronic inflammation [19]. Therefore, M1-like macrophages may serve as a potent therapeutic target for reducing chronic inflammation. Activated macrophages express a wide variety of cell surface markers. In general, these markers are expressed on both M1 and M2 macrophages. However, some markers are expressed in greater quantity on M1 macrophage and are downregulated on M2 macrophage. An example of such receptor is CD64 (Fc gamma receptor 1), a high-affinity immunoglobulin Fc receptor [20]. The high expression of CD64 on M1-like macrophages makes this receptor an attractive target for specific elimination of M1-like proinflammatory macrophages. Additionally, evidence has shown that CD64 is only expressed on myeloid cell lines including monocytes, macrophages and activated neutrophils [21,22]. Numerous other studies have also shown reduction of inflammation with CD64 targeted immunotoxin further confirming the utility of CD64 as a target for immunotoxin development [23][24][25].
We hypothesized that HALT-1 could be used as a toxin moiety for the construction of recombinant immunotoxin. In this study, we described for the construction of HALT-1-based recombinant immunotoxins by molecular fusion of HALT-1 with anti-CD64 scFv in two different orientations. We then determined the binding potential of the two immunotoxins to CD64 in vitro. Moreover, the efficacy of these recombinant immunotoxins against M1-like macrophages and HeLa cells was evaluated in terms of the cytotoxicity of HALT-1, but not the selective binding affinity of scFv to cells expressing CD64 since the immunotoxins could recognise both M1-like macrophages and HeLa cells via HALT-1. Our findings suggested the potential of using HALT-1 as toxin moiety for construction of recombinant immunotoxins with preferable arrangement of HALT-1 at the N-terminal end. As the binding specificity of scFv to CD64 could have been masked by HALT-1 which recognises the membrane lipids of almost all human cell types, the immunotoxins did not differentiate CD64 + M1-like macrophages from CD64 − HeLa cells. Hence, the future study should replace HALT-1 with a mutant lacking the binding affinity to membrane lipids.

Results
In vitro assessment of α-CD64-scFv (or scFv) binding to CD64 CD64 has been shown to be a good choice of target for the development of therapies against many kinds of monocyte/macrophages related inflammatory diseases such as rheumatoid arthritis, inflammatory skin diseases and acute myeloid leukemia (AML). α-CD64-scFv used in our study is based on the amino acid sequence of H22(scFv) that was reported to show specific binding towards CD64 [26][27][28]. α-CD64-scFv was expressed as a recombinant protein in BL21(DE3) E. coli cells and its solubility was assessed before the purification. Soluble lysate of expressed culture was compared with the insoluble cell debris on SDS-PAGE (Fig. 1a). Our results showed α-CD64-scFv present as an insoluble 32 kD protein. Hence it was purified under denaturing condition and then followed by refolding. The final yield of recombinant α-CD64-scFv was 144 μg/mL (Fig. 1b) with the recovery of slightly less than 40%.
The binding of α-CD64-scFv to CD64 at various concentrations was assessed by ELISA. In Fig. 1c, there was a sharp increase of absorbance from 0 to 1.0 ng/mL and then the absorbance remained constant even though the concentration of CD64 was increased to 10 ng/mL. This suggested that there was a direct proportional  Figure S1; Additional file 2 relationship between CD64 concentration and binding to α-CD64-scFv. As such, α-CD64-scFv is a potential vehicle for immunotoxin to be targeted to M1 macrophages and could be used in the production of immunotoxin.
CD64-binding and cytotoxicity of HALT-1-scFv and scFv-HALT-1 Two recombinant immunotoxins with opposite orientations of α-CD64-scFv and HALT-1 were constructed, one with HALT-1 at the N-terminus and α-CD64-scFv at the C-terminus, and the other having HALT-1 at the C-terminus and α-CD64-scFv at the N-terminus. Figure 2 shows the schematic drawing of the recombinant immunotoxins in pET22b expression vector. Orientation of α-CD64-scFv and HALT-1 would determine whether or not the recombinant immunotoxin can be produced in E. coli and bind on the cell membrane to form the oligomeric pores. Many known immunotoxins have their own preference of moiety orientation. For instance, Pseudomonas exotoxin A is often positioned at the Cterminal end of immunotoxin [29,30] whereas Diphtheria toxin tends to be placed at the N-terminal end [31,32]. In this study, we prepared the recombinant immunotoxin in two different orientations in such that one might work better than the other.
Both recombinant immunotoxins were successfully expressed in BL21(DE3) E. coli cells in the presence of 1 mM IPTG ( Fig. 3a and b). However, both recombinant immunotoxins were present in insoluble inclusion bodies ( Fig. 3c and d). The insoluble inclusion bodies were isolated, denatured, and purified with Ni-NTA column before the refolding procedure. Stepwise dialysis refolding was processed over a long period of time to achieve high refolding efficiency and recovery of bioactive immunotoxin. Then we confirmed their purity by SDS-PAGE ( Fig. 3e and f). The total yield after refolding was below 40%, indicating that less than 40% of recombinant immunotoxins were refolded properly, soluble and biologically active (Supplementary Table 3). Those that left unfolded or misfolded would remain as aggregates.
To evaluate whether α-CD64-scFv is still able to bind CD64 after it has coupled with HALT-1 in the immunotoxin, we did an ELISA assay. A CD64-coated 96-well immunoplate was treated with various concentrations of the recombinant immunotoxins, HALT-1-scFv and scFv-HALT-1. Our results showed concentrationdependent binding of recombinant HALT-1-scFv to CD64 from 0 to 10 μg/mL (Fig. 4a). Although HALT-1-scFv also showed weaker non-specific binding towards 3% BSA in CD64(−) control wells, its binding to CD64 was significantly one-fold higher as compared with CD64(−) control (One Way Repeated Measures ANOVA p = 0.0381) (Fig. 4a). scFv-HALT-1, on the other hand, showed no binding to CD64 (Fig. 4b). This might be due to the incorrect folding of α-CD64-scFv during the refolding step and subsequent loss of its binding activity towards CD64. Figure 4c demonstrated that the HALT-1 moiety did not contribute to the CD64 binding of the recombinant immunotoxins.
Cytotoxicity of the recombinant immunotoxins were assessed in vitro by measuring viability of CD64 + M1like macrophages and CD64 − HeLa cells treated with various concentrations of immunotoxins. Before proceeding with the cytotoxicity assay, PCR was utilized to validate the expression of CD64 in activated M1-like macrophages as well as the lack of CD64 expression in CD64 − HeLa cells. The results clearly demonstrated the expression of CD64 in M1-like macrophages and the lack of CD64 expression in HeLa cells (Fig. 5a). Figure 5b, c and d display the viability of CD64 + M1-like macrophages and CD64 − HeLa cells following treatment with increasing concentrations of HALT-1, HALT-1-scFv and scFv-HALT-1. CD64 + M1-like macrophages are slightly more susceptible than CD64 − HeLa cells to the cytotoxicity of HALT-1, either alone or in  conjugation with scFv (Fig. 5b, c and d). HALT-1 toxin alone has an IC 50 of 5.05 μg/mL against CD64 + M1like macrophages and 12.55 μg/mL against CD64 − HeLa cells (Fig. 5b). HALT-1-scFv was shown to have cytolytic activity with IC 50 of approximately 10.05 μg/ mL on CD64 + M1-like macrophages while its IC 50 towards CD64 − HeLa cells was found at 17.95 μg/mL (Fig. 5c). Despite the absence of CD64 on the cell membrane of CD64 − HeLa cells, HALT-1-scFv still exerted cytotoxicity as the HALT-1 moiety could recognise its target sites on the cell membrane. Both CD64 + and CD64 − cells were also killed by scFv-HALT-1 at IC 50 of 9.95 μg/mL and 18.8 μg/mL, respectively (Fig. 5d). Since scFv-HALT-1 immunotoxin did not bind CD64 (Fig. 4b), this cytotoxicity must be due to the activity of the HALT-1 moiety. Therefore, the immunotoxins that we constructed in this study reduced the cell viability of CD64 + and CD64 − cells, but whether they selectively target CD64 + cells would μg/mL) were used in the assay with or without CD64. Each assay was performed in triplicate and error bars represent standard deviations from the mean of triplicate. Recombinant HALT-1 alone was used as a control to indicate that nonspecific binding did not occur between HALT-1 and CD64. Fluorescence intensity was obtained by subtracting the fluorescence value with 0 μg/ mL of either HALT-1-scFv, scFv-HALT-1 or HALT-1 require more study using mutant HALT-1 lacking receptor-binding activity.
Taken together these results showed that there were only small differences in the cell viability between cells treated with the two immunotoxins and HALT-1. This supports two conclusions: (1) the toxicity of HALT-1 was not significantly altered by the presence of scFv moiety in the recombinant immunotoxin and (2) the toxicity of immunotoxins was due primarily to HALT-1 binding to cells and was independent of the presence (M1-like macrophage) or absence (HeLa cells) of CD64 on the cell membrane.

Discussion
Recombinant immunotoxins were successfully constructed by molecular fusion of HALT-1 and α-CD64-scFv via a flexible linker (Fig. 2). The serine and glycinerich linker was selected due to several reasons; (1) flexible glycine-rich regions have been observed as natural linkers in multidomain proteins; (2) glycine and serine provide good flexibility due to their small sizes; (3) serine and glycine help maintain stability of the linker structure in the aqueous solvent by forming hydrogen bonds with water; (4) linker length is within the optimal length of 6 or 10 ± 5.8 residues [33,34]. As stated above, pET22b expression vector was chosen because it contains pelB leader sequence to bring the protein to periplasmic space for disulfide bonds formation [35]. Both recombinant immunotoxins were expected to have the molecular weight of 50.6 kDa. However, the SDS-PAGE results indicated that the molecular weights of HALT-1-scFv was larger than expected (approximately 52.8 kDa, Fig. 3a) while the molecular weight of scFv-HALT-1 had the approximately correct size of 50.6 kDa (Fig. 3b). The increase in molecular weight of HALT-1-scFv is likely caused by the failure of the pelB leader sequence (approximately 2.2 kDa) to be cleaved off during expression. This result was corroborated with the subsequent solubility test (Fig. 3c) showing these immunotoxins were expressed as insoluble inclusion bodies which reflected the failure of periplasmic translocation and cleavage of pelB leader sequence. However, despite the successful cleavage of pelB leader sequence, scFv-HALT-1 was also present in the inclusion body (Fig. 3d). In future studies, eukaryotic expression system such as green alga could be used to improve the production and solubility of the recombinant immunotoxins [36].
Refolded recombinant HALT-1-scFv, but not scFv-HALT-1, showed in vitro binding to CD64. This suggested that HALT-1-scFv where HALT-1 and α-CD64-scFv were placed at the N-terminus and C-terminus, respectively, could be selected for further development of immunotoxin. However, HALT-1-scFv also showed significant degree of non-specific binding towards BSA. Reports have shown that different antibodies and proteins may exert cross-reactivity with BSA [37,38]. The nonspecific cross-reactivity towards BSA in our results might be one of the rare examples. Additionally, previous report has also shown that glycerol may differentially affect antibody-antigen interaction depending on the antibody clone and antigen [39]. The high concentration of glycerol used in our desalting buffer during the protein refolding may contribute to the signal in CD64(−) ELISA. Expression system that eliminates the need to refold the recombinant immunotoxin or an improved redox refolding system could eliminate the need of glycerol in the refolding protocol.
Our study aims to demonstrate whether HALT-1 could maintain its cytotoxic function after it was conjugated to the scFv if the conjugated scFv preserves its target binding capability, and what orientation of HALT-1 in recombinant immunotoxin that gives the highest efficacy in killing the target cells. A number of recombinant immunotoxins targeting CD64 have been developed and reported in several studies as listed in Table 1. Compared to H22(scFv)-ETA, H22(scFv)2-ETA, granzymeB-H22(scFv), and H22(scFv)-MAP, the IC 50 of HALT-1-scFv was significantly higher at approximately 189 nM (equivalent to 10.05 μg/mL). This significantly higher working concentration can be explained by the requirement of four or more monomers of HALT-1 to oligomerize in close proximity to form a functional pore and induce cell lysis. Albeit the higher working concentration, HALT-1 based recombinant immunotoxin can be beneficial in selective toxicity. Since at least four monomers of HALT-1 are needed to oligomerize in close proximity, the recombinant immunotoxin could presumably have low cytolytic activity when cells have low and sparse expression of the target surface receptor. As such, HALT-1 based immunotoxin targeting CD64 can selectively eliminate M1-like macrophage that express high amount of the surface receptor. When comparing the cytotoxic activity of HALT-1 before and after conjugating with α-CD64-scFv, either HALT-1-scFv or scFv-HALT-1 displayed a two times lower cytotoxic activity than that observed for HALT-1 (Fig. 5). This is not surprising because the reduction of activity was also observed in GFP conjugated equinatoxin II (GFP-EqtII). The fusion of GFP to the C-terminus of equinatoxin II caused the toxin becoming less haemolytic than the native toxin [41]. Despite the opposite orientations, HALT-1 in HALT-1-scFv and scFv-HALT-1 demonstrated similar cytotoxic activity in either CD64 + M1-like macrophages (IC 50 = 10.05 and 9.95 μg/mL, respectively) or CD64 − HeLa cells (IC 50 = 17.95 and 18.80 μg/mL, respectively). One may expect scFv-HALT-1 to have much less activity than HALT-1-scFv since the N-terminus of HALT-1 is functionally important. The role of Nterminal α-helix of actinoporin has been previously reported for equinatoxin II. Gutierréz-Aguirre et al. [42] has shown that equinatoxin II mutant having substitution of valine at position 22 to tryptophan (V22W) could establish interactions with the interface of membrane and prevent it from insertion into the lipid bilayer. Similarly, the N-terminal α-helix of double cysteine mutant 8-69 (V8C and K69C) of equinatoxin II was immobilised in the oxidised form but regained lipid penetration in the reduced form [43]. The initial binding of mutant 8-69 on lipid membrane was not affected under the oxidative condition, indicating that the N-terminal α-helix is only involved in the second stage of membrane insertion [43]. Thus, the addition of bulky scFv to the Nterminus of HALT-1 might hinder the detachment of Nterminal helix from the core protein and subsequently its insertion into the lipid bilayer. However, the cytolytic activity of scFv-HALT-1 did not seem to be affected by the fusion of scFv. Having said that, scFv-HALT-1 may not be the choice of immunotoxin because it failed to bind CD64 in vitro in this study. Furthermore, HALT-1 exerted different cytotoxic activities against different human cell lines regardless whether it was connected to a single-chain scFv or not [44]. In our case, HALT-1 worked more effectively in M1-like macrophages than in HeLa cells (Fig. 5). In the next course of development, we will introduce mutation(s) in the receptor binding domain of HALT-1 so that the immunotoxin would be directed by its targeting moiety to CD64 expressing cells. Two residues of HALT-1, tryptophan at position 113 and tyrosine at position 129, have been previously substituted into alanine respectively and these mutants, which did not bind membrane lipid, are appropriate candidates as the toxin component of immunotoxin [45]. Moreover, flow cytometry can be introduced to evaluate the binding of HALT-1-scFv or scFv-HALT-1 to CD64 expressing cells in comparison with the commercial anti-CD64 antibody. Lastly, that efficacy of HALT-1 based recombinant immunotoxin can be improved by modification of the scFv moiety to bivalent tandem scFv. Bivalent tandem H22(scFv)2-ETA showed ten folds reduction in IC 50 compared to H22(scFv)-ETA [25]. Almost all toxins currently used in developing recombinant immunotoxins require internalization to cytosol to exert their toxicity, which could lead to degradation of the immunotoxin in lysosome. By comparison, HALT-1 works on the cell surface without the needs for internalization into the cytosol, which avoids the complexity of entry mechanism and the degradation in lysosome. It has been argued that the necrotic action of HALT-1 could cause intracellular components to induce an inflammatory response in the neighbouring cells [46]. Recently, it became clear that HALT-1 at IC 50 (0.51 μM or 10.61 μg/mL) could also induce an apoptotic pathway in HeLa cells and that the same apoptotic effect could be induced in other cell lines with similar IC 50 values [44]. Hence, apoptosis could occur in HALT-1 treated cells when sub-lytic concentration of HALT-1 was used [44]. Interestingly, the apoptotic pathway induced by HALT-1 might not require the internalization of toxin. Recent studies of sticholysin II, a member of actinoporin family as mentioned above, have proven that the pore formation, if not for cell lysis, would lead to the ion efflux which subsequently activates the apoptotic signalling pathway [11,12]. This feature differs HALT-1 from other bacterial α-PFT and thus HALT-1 can be an alternative candidate for the construction of immunotoxins.

Conclusions
We fused HALT-1 to α-CD64-scFv via a flexible linker peptide and demonstrated that HALT-1 could be utilized as a toxin moiety in recombinant immunotoxins. Our preliminary data suggested that the positioning of α-CD64-scFv in immunotoxin is crucial for its binding to CD64 and HALT-1 has reduced half of its cytotoxicity following the conjugation with α-CD64-scFv. Despite the limitations, the small molecular size of HALT-1 and the ability to exert its toxicity without the need to be internalized, HALT-1 could be advantages compared to the toxins commonly used in the construction of recombinant immunotoxins.

Recombinant α-CD64 scFv
Humanised anti-CD64-scFv sequence was obtained from Genbank with the accession number AY585869. To synthesize α-CD64-scFv, its sequence was optimized using IDT DNA's codon optimization tools for optimum expression in E. coli. The gene was synthesized by IDT DNA (USA) and then placed in pIDT-sMART vector. In order to express the recombinant scFv, the gene was digested with NheI and NdeI and subcloned into pET28a. Finally the cloned plasmid was transformed into BL21(DE3) E. coli cells.

Recombinant HALT-1
Our group had previously constructed recombinant HALT-1 in pET28a and successfully expressed it in BL21(DE3) E. coli cells and purified it with the Ni 2+ affinity chromatography [45].

Construction of recombinant immunotoxins
Fusion of HALT-1 to α-CD64-scFv as well as introducing the glycine-serine peptide linker (SGGGGSGGGG SGGGGS) were performed by overlap extension polymerase chain reaction (OE-PCR) and could result in two different orientations, HALT-1 at N-terminus followed by α-CD64-scFv or vice versa ( Fig.2a and b). Basic steps to achieve the different oriented fusion proteins are the same. Firstly, extension PCR was performed on HALT-1 and α-CD64-scFv using Pfu DNA polymerase (Nex-Bio, Malaysia) and two sets of primers for each orientation were prepared to introduce N-terminal His-tag, restriction enzyme sites and linker sequence (Supplementary Table 1). In brief, PCR started at 95°C for 5 min, one cycle of 95°C for 30 s, 50°C for 30 s, 72°C for 1 min and 30 s; and repeated for 35 cycles at 95°C for 30 s, 58°C for 30 s and 72°C for 1 min and 30 s; and finally ended with 72°C for 5 min. After extension PCR was performed, PCR products were fused in assembly PCR. Assembly PCR started at 95°C for 5 min repeated for 15 cycles at 95°C for 1 min, 60°C for 45 s and 72°C for 1 min and 30 s; and finally ended with 72°C for 5 min. After completion of assembly PCR, respective forward and reverse primers were added to the reaction tube and amplification PCR was performed immediately. Briefly, amplification PCR started at 95°C for 5 min repeated for 35 cycles at 95°C for 1 min, 58°C for 30 s and 72°C for 2 min and 30 s; and finally ended with 72°C for 5 min. PCR product was run in 1% agarose gel and desired bands were isolated, followed by purification with Wizard SV gel and PCR clean up system (Promega, USA). After the digestion of NcoI and XhoI, it was cloned into pET22b expression vector and transformed into BL21(DE3) E. coli cells. Plasmid DNA was purified using DNA-spin plasmid purification kit (iNtRON, Korea) and subjected to DNA sequencing for confirmation. In this study, we collectively called both α-CD64-scFv-HALT-1 (or scFv-HALT-1) and HALT-1-α-CD64-scFv (or HALT-1-scFv) as the recombinant immunotoxins.  Table 2). Dialysis was performed with gentle stirring on ice for 3 h each buffer change until 1 M urea buffer which was done overnight followed by three changes of 1X PBS containing 30% glycerol for 3 h each. Without the addition of glycerol, 100% of the recombinant immunotoxins aggregated in the final desalting (data not shown).

ELISA binding assays
To show α-CD64-scFv is specific for recombinant human CD64 (Sino Biological, USA), their interaction was demonstrated by the enzyme-linked immunosorbent assay (ELISA). In a 96-well microplate, wells were coated with 10 μg/mL of α-CD64-scFv overnight at 4°C. The unoccupied protein-binding site was then blocked by 3% (w/v) BSA/PBS and further incubation at room temperature for 2 h. After rinsing the wells, various concentrations of recombinant CD64 (0, 0.5, 1.0, 5.0 and 10 ng/mL) were added to the wells and incubated at room temperature for 2 h. To visualize the α-CD64-scFv and CD64 interaction, rabbit anti-CD64 polyclonal antibody (Invitrogen, USA) was added to each well, and followed by anti-rabbit HRP complex (1: 5000) and TMB substrate complex (Thermo-Fisher, USA). The signal produced was read at 370 nm by multi-mode microplate reader (TECAN, Switzerland). Four negative controls with each lacking a specific component of ELISA, either α-CD64-scFv, CD64 protein, rabbit α-CD64 polyclonal antibody or α-rabbit HRP conjugate, were prepared for each 96-well microtiter plate.
ELISA was also performed to determine binding of recombinant immunotoxins to CD64. The basic procedure was carried out as described above. Recombinant human CD64 (0.5 μg/mL) was first coated to a 96-well microtiter plate. To allow the binding of recombinant immunotoxin to CD64, the recombinant immunotoxin was added to the wells at different concentrations (0, 1, 2, 4, 8 and 10 μg/mL). By washing the wells in between incubations, rabbit anti-HALT-1 primary antibody (2.3 μg/ mL), goat anti-rabbit IgG-AP (12.5 ng/mL) and 4-MUP (4-methylumbelliferyl phosphate) (Sigma-Aldrich, USA) were added to the wells in a precise sequence of steps. The fluorescence absorbance/exciting readings at 355/ 460 nm were measured by a multi-mode microplate reader (BMG Labtech, Germany) at 5 min time-point for 30 min. The same experiment was performed without CD64 to ensure that any fluorescence detection is due to the specific binding between CD64 and recombinant immunotoxin. Three negative controls (1) CD64 coated only (2) without primary antibody and (3) without secondary antibody were also included in each set of tests.

Statistical data analysis
ELISA of HALT-1-scFv was conducted in biological and technical triplicates for each CD64-coated and CD64uncoated wells. The statistical analysis was performed by using R version 3.6.0 [47] (https://www.R-project.org/). The consistency of biological triplicate experiments was examined by using One-Way Repeated Measures ANOVA. In order to distinguish the specificity of HALT-1-scFv to CD64-coated wells from uncoated wells, one tail paired T test was conducted. For all the test conducted, p ≤ 0.05 was set to define significance of differences.

Polarization of M1-like macrophages
Cytotoxicity assay of the recombinant immunotoxins was performed on CD64 + THP-1 (ATCC TIB-202) human monocyte cell line. Cells were routinely cultured in 20 mL of Roswell Park Memorial Institute (RPMI 1640) culture medium (Nacalai tesque, Japan) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin/ streptomycin, 10 mM hepes, 1 mM pyruvate, and 50 pM 2-mercaptoethanol; and grown at 37°C with 5% CO 2 . For cytotoxicity assay, cells were seeded at 1 × 10 4 cells/ well in 96-well microtiter plate with the addition of 200 nM phorbol 12-myristate 13-acetate (PMA) and incubated for 72 h to activate the monocytes to macrophages. Activated cells were then polarized to M1-like macrophages by changing the medium to fresh RPMI medium containing 20 ng/mL of IFN-γ and 10 pg/mL of LPS, followed by incubation for another 48 h. Expression of CD64 in M1-like polarized macrophages were validated by PCR. CD64 specific primers [48] were used to determine the expression of CD64 and GAPDH specific primers [49] were used for amplifying the positive control.