Bacterial production and biophysical characterization of a hard-to-fold scFv against myeloid leukemia cell surface marker, IL-1RAP

Interleukin-1 receptor accessory protein (IL-1RAP) is one of the most promising therapeutic targets proposed for myeloid leukemia. Antibodies (Abs) specific to IL-1RAP could be valuable tools for targeted therapy of this lethal malignancy. This study is about the preparation of a difficult-to-produce single-chain variable fragment (scFv) construct against the membrane-bound isoform of human IL-1RAP using Escherichia coli (E. coli). Different approaches were examined for refolding and characterization of the scFv. Binding activities of antibody fragments were comparatively evaluated using cell-based enzyme-linked immunosorbent assay (ELISA). Homogeneity and secondary structure of selected scFv preparation were analyzed using analytical size exclusion chromatography (SEC) and circular dichroism (CD) spectroscopy, respectively. The activity of the selected preparation was evaluated after long-term storage, repeated freeze-thaw cycles, or following incubation with normal and leukemic serum. Strategies for soluble expression of the scFv failed. Even with the help of Trx, ≥ 98% of proteins were expressed as inclusion bodies (IBs). Among three different refolding methods, the highest recovery rate was obtained from the dilution method (11.2%). Trx-tag substantially enhanced the expression level (18%, considering the molecular weight (MW) differences), recovery rate (˃1.6-fold), and binding activity (˃2.6-fold increase in absorbance450nm). The produced scFv exhibited expected secondary structure as well as acceptable bio-functionality, homogeneity, and stability. We were able to produce 21 mg/L culture functional and stable anti-IL-1RAP scFv via recovering IBs by pulse dilution procedure. The produced scFv as a useful targeting agent could be used in scheming new therapeutics or diagnostics for myeloid malignancies.


Introduction
Acute myeloid leukemia (AML, the list of abbreviations provided in Table 1) is the most prevalent and fatal leukemia in adults [1].
Immunotherapeutic approaches beyond hematopoietic stem cell transplantation (HSCT), especially antibodydependent cellular cytotoxicity, (ADCC) have not been successful so far. The immunosuppressive quality of bone marrow niches in AML is the greatest impediment to developing immune-dependent therapy [2; 3]. Therefore, immune-independent approaches may be more reasonable.
Gemtuzumab ozogamicin (Mylotarg) is a good example. This drug exerts its anti-leukemic effects not through the immune effector cells but via the delivery of a potent cytotoxic molecule, calicheamicin [4]. The use of a more specific targeting agent may eventuate to better effectiveness.
Interleukin 1 receptor accessory protein (IL-1RAP), has been introduced as an excellent tumor target for targeted therapy of acute and chronic myeloid leukemia (CML). It is present or over-expressed on the surface of nearly all CML leukemic stem cells (LSCs) and most AML cells, while more than 97% of normal HSCs do not express it at all. [5][6][7]. Antibodies against IL-1RAP could be valuable tools for the targeted delivery of cytotoxic agents to myeloid leukemia cells [5].
A single-chain variable fragment (scFv) is a ~ 30 kDa artificial biomolecule comprised of only variable regions of heavy (VH domain) and light (VL domain) chains of a monoclonal antibody (mAb) [8]. These two antigen-binding domains are connected through a flexible linker of 15-25 residues, typically (Gly4Ser) n [8].
ScFvs have prominent advantages over the whole mAbs, including higher tissue/tumor penetration, quicker blood clearance, and lower immunogenicity [9]. Due to the small size and lack of glycosylation, scFvs could be produced in prokaryotic systems that have fast growth rate and inexpensive transformation and cultivation procedures. E. coli, especially, has a high popularity thanks to its well-known genetics and, therefore, availability of various engineered strains and vectors, each designed for a particular purpose [10].
Correct folding of scFvs in bacterial cytosol is usually challenging due to their mammalian origin and non-natural entity. More notably, the proper formation of two structurally important intra-domain disulfide bridges in scFvs needs a proficient folding system under oxidizing environment [11]. So, these biomolecules mostly become misfolded and form dense aggregates (inclusion bodies, IBs) after overexpression in the cytoplasm [12].
Here, we focused on the refolding of an E. coli codonoptimized scFv construct specific to the membrane-bound isoform of IL-1RAP. Our previous efforts to the soluble expression of full-length scFv from this construct repeatedly failed. Trying different strategies, including periplasmic expression (with or without sucrose supplementation), co-expression of chaperons (GroEL, GroES, and Tig), fusion with two solubility partners (Trx and eukaryotic initiation factor 2, IF-2 in Origami B context) and use of SHuffle strains, all were associated with no or very low yield.
The success in in vitro refolding mainly depends on the refolding method employed [13] and the types and balance of components in refolding buffer [12]. In the present study, three different refolding methods were comparatively investigated and then, detailed analyses of the selected scFv preparation were performed to evaluate its folding accuracy, homogeneity, binding activity, and stability.
Normal serum was voluntarily obtained from one of the authors and leukemic serum from a 59 years old male patient with relapsed AML and high leukemic burden before treatment initiation.

Sequence extraction, design, gene synthesis, and cloning procedures
The E. coli codon-optimized sequence of the anti-IL-1RAP scFv (fused to fusion format of streptavidin (SA-X), Gen-Bank accession number: OM642337) in pET28a ( Fig.  S1a) was already present in our lab. The sequence of the scFv had been formerly extracted from a patent (PCT/ US2013/077323 [14]: SEQ ID NO 13 for VH and SEQ ID NO 14 for VL) (Fig. S1b). The scFv was constructed as VH-(Gly4Ser) 5 -VL-Gly4-6xHis (Fig. S1c) by removal of downstream sequences and substitution with a Gly4 coding oligo (Fig. S1d).

Secondary structure prediction
We utilized two web servers, 2Struc [15] and PSIPRED [16], for secondary structure predictions. The scFv constructs were modeled using I-TASSER [17]. Model 1 from each protein was employed for analysis by 2Struc. For PSIPRED, the amino acid (AA) sequences were submitted.

Exploration of aggregation hotspots
Potential aggregation regions were investigated in silico by five different web servers (inspired from [18]). For AGGR-ESCAN [19], Tango [20], Waltz [21], and FoldAmyloid [22], the AA sequences of the scFv-6xHis in FASTA format were inputted. For the AGGRESCAN3D (A3D) predictor [23], I-TASSER generated 3D model of the molecule was uploaded as an input. All predictions were accomplished with the default input parameters settings of tools.

Preliminary expression of scFv
The scFv-6xHis construct was transformed to BL21 (DE3), and a single transformant was cultured overnight (ON) in a 5 ml Luria-Bertani (LB) medium at 37 ˚C. The ON culture was refreshed and induced at Abs 600nm of ~ 0.5 (after obtaining before induction (BI) sample) with 0.5 mM IPTG and continued for 4 h at 37 ˚C, 150 rpm.

Analysis of expression samples
For analysis of expression in total cell fractions (TCFs), cellular pellets of before and after induction (AI) samples were resuspended in dH 2 O. To study the soluble and insoluble fractions, the pellets of un-induced and induced (4 h, ON) cultures from three different temperature conditions were resuspended in phosphate-buffered saline (PBS) and lysed with sonication. After centrifugation (⁓11,000 g for 10 min), supernatants were collected as soluble fractions. The sediments were washed with PBS, dissolved in 0.5 ml of 8 M urea, and centrifuged. The supernatants of the second step were collected as insoluble fractions. All samples were mixed with an appropriate volume of Laemmli buffer (with 2% 2-mercaptoethanol (2ME)) and boiled for ⁓10 min before sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Refolding and purification of the scFv
We compared three different methods for refolding scFv from IBs produced in BL21 (DE3). In all, 2.5 mg denatured IB (denatured in solubilization buffer ( Table 2)) with ≥ 75% purity was introduced, and all the procedures were done in the same buffer condition (50 mM Tris-HCl, 150 mM NaCl, pH 8). Denatured protein samples in all practices were prepared from the stock solution (10 mg/ml protein in solubilization buffer plus 10 mM 2ME). We used 2 mM cysteine and 0.4 mM cystine in all methods but at different urea concentration points. After refolding and before purification, aggregates were removed using centrifugation (⁓10,000 g for 30 min). The refolded protein samples were buffer exchanged to PBS and concentrated using a 10 kDa cut-off Amicon™ Ultra-0.5 Centrifugal Filter (Merck). The final concentrations were determined using a UV spectrophotometer (Abs 280nm at the related extinction coefficient (Table S2)). The recovery rates were calculated using the formula [1].

Refolding by stepwise dialysis
In refolding by dialysis, 0.25 ml of stock solution (2.5 mg protein) was added to the 24.75 ml solubilization buffer. The resulting solution (25 ml of 100 µg/ml) was transferred into the dialysis bag with a 12 kDa cut-off. The bag was immersed in the same buffer (but with 6 M urea). The urea concentration decreased step by step (4 M, 2 M, 1 M, and 0.5 M). The refolding buffers from 2 M onward contained 1 mM 2ME, 2 mM L-cysteine, and 0.4 mM L-cystine. The buffer was then exchanged to refolding buffer B, and dialysis continued ON. The protein solution was brought out of the bag after final dialysis in binding/wash buffer (Table S1) and then applied on Ni-NTA resin (QIAGEN, Germany) pre-equilibrated with binding/wash buffer. After complete washing with the same buffer, the refolded proteins were eluted from the resin using the elution buffer (Table S1).

WB on expressed scFv
Samples were subjected to SDS-PAGE using 12% polyacrylamide gel. After electrophoresis, protein bands were transferred to the nitrocellulose membrane by the semi-dry method. The membrane was blocked (3% w/v skim milk in PBS) ON at 4 ˚C and after 4 times washing the membrane was incubated with horseradish peroxidase (HRP)conjugated anti-His-tag Ab (Sigma-Aldrich, USA). The membrane was then washed five times, and His-tagged scFv bands were developed by 3, 3'-diaminobenzidine (DAB) substrate solution.

WB on lysates of leukemic cells
K-562 and KG-1 cells were harvested (400 g, 5 min) and lysed by resuspension in minimum volumes of lysis buffer (Table S1). Total protein contents were estimated using Bradford assay, and equal protein quantities (⁓ 40 µg) from each sample were loaded on 8% polyacrylamide gel and subjected to SDS-PAGE. The protein bands were transferred to the nitrocellulose membrane. The membrane was blocked, washed, and then incubated with the anti-IL-RAP scFv solution (⁓5 µg scFv per 1ml PBS). The membrane was then washed and incubated with HRP-conjugated anti-His-tag Ab, and the experiment was continued as mentioned above.

Isolation of IBs for refolding
A single colony of transformant was cultured and induced with 0.2 mM IPTG at 37 ˚C, ON. After harvesting the culture (⁓4000 g, 15 min) the cellular pellet was resuspended in 2 ml lysis/wash buffer (Table S1) plus ⁓5 mg lysozyme and incubated for about 1 h, at RT with occasional shaking. The suspension was volumized to 30 ml with the same buffer and lysed using two rounds of sonication (10 s pulses/20 s rest intervals for 20 times at 40% amplitude). IBs were separated by centrifugation, resuspended in the buffer mentioned above, and, then sonicated. Washed IBs were isolated using centrifugation and then washed with PBS to remove residual ethylenediaminetetraacetic acid (EDTA) and detergent. High-quality IBs were stored at -20 C until solubilization.

Size exclusion chromatography
In order to analyze the homogeneity of dilution-refolded scFv preparation, size exclusion chromatography (SEC) was carried out according to the following method. 100 µL of purified Ab (0.37 µg/µl) was injected into a TSK-gel G2000 SWXL column (7.8 × 300 mm, Tosoh Bioscience, Tokyo, Japan) connected to a Shimadzu HPLC system (Kyoto, Japan). The sample was eluted with an isocratic mobile phase (Table S1) with a 0.5 ml/min flow rate (25 ºC).

Analysis of secondary structure by CD
The secondary structures of three Ab preparations were figured out by far-UV CD spectroscopy (Jasco J-810 Spectropolarimeter, Japan). The ellipticity (θ) of samples was measured at 190 to 240 nm wavelength. The samples' buffer was extensively exchanged to dH 2 O using a 10 kDa cut-off Amicon™ Ultra-0.5 Centrifugal Filter prior to the measurements. The analyses were performed at RT by injection of ⁓300 µl of each sample into a 1 mm quartz cuvette.

Statistical analysis
The statistical analysis of data was accomplished using GraphPad Prism version 8.0 for Windows (GraphPad Software, La Jolla, CA, USA). The significance level between the groups was analyzed using one-way ANOVA.

Cytoplasmic expression of scFv in BL21 (DE3)
A preliminary expression experiment with the scFv-6xHis construct showed an apparent 28 kDa band absent in the "before induction" sample ( Fig. 1a). We repeated the experiment for a 200 ml LB culture (0.2 mM IPTG, 37 ˚C, 150 rpm, ON). The cells were harvested and lysed, and the clarified lysate was subjected to Ni-NTA purification to capture any soluble scFv. However, no protein was detected in the elution sample.
Next, we tested the experiment in different conditions, including soluble favorable conditions, and analyzed the induced samples by SDS-PAGE. No expression was observed in soluble fractions (Fig. 1b), and the protein was expressed entirely as IB in all conditions (Fig. 1c).

Recovery of soluble scFv from IBs
Failures in the soluble expression of intact scFv in reasonable purity or quantity (data not presented) led us to subclone

On-column refolding and purification of solubilized IBs
For on-column refolding, 0.25 ml of stock solution (2.5 mg protein) was added to a 9.75 ml solubilization buffer. The denatured protein solution (10 ml of 250 µg/ml) was loaded on the Ni-NTA column pre-equilibrated with the same buffer. The column was thoroughly washed with solubilization buffer (plus 1 mM 2ME), and then a gradient of refolding buffer B ( Table 2) was applied from 0 to 100%. Inspired by a previous report [24] since the 3 M urea point, the refolding buffer B contained 2 mM L-cysteine and 0.4 mM L-cystine. The column was then thoroughly washed with refolding buffer B, and the refolded proteins were eluted from the resin as mentioned above.

Refolding using pulse dilution method
For the dilution method, 0.25 ml of stock solution (2.5 mg protein) was added to the 0.75 ml solubilization buffer. In the first step, the resulting protein solution (1 ml of 2.5 mg/ ml) was diluted drop-by-drop into 24 ml of cold refolding buffer A ( Table 2) without stirring. The final protein concentration, if no aggregation occurred, would be 100 µg/ ml. After ˃24 h incubation at 4 ˚C, the protein solution was dialyzed (at 4 ˚C, ON) in refolding buffer B ( Table 2) without shuffling pairs and then dialyzed in binding/wash buffer (Table S1) before loading on Ni-NTA resin. The purification process was continued as mentioned above.

Cell-based ELISA
All solutions were prepared with Hank's balanced salt solution (HBSS) (Table S1), and the plate was kept cold throughout the process. At first, the wells were pre-coated with high molecular weight (HMW) polyethyleneimine (50-100 kDa PEI, MP Biomedicals (FKN ICN BIOMED) the USA, CAS number: 9002-98-6) 100 µl of PEI solution (0.1 mg/ml in dH 2 O) was added to the wells. The plate was incubated at 37 ˚C to be completely dried. After pre-coating, 100 µl of washed cells were added to the wells (≥ 5 × 10 4 /well) and allowed (4 h) to be settled. The cells were then fixed by slow addition of 0.5% v/v glutaraldehyde solution (100 µl/ well). After 1 h, the wells were washed and blocked with 1% w/v bovine serum albumin (BSA) (200 µl/well) for 4 h. After washing, 50 µl of the produced scFvs (50 µg/ml) were added to the wells. The plate was incubated at 4 ˚C, ON, and after 4 times washing, 100 µl of mouse HRP-conjugated anti-His-tag Ab (1:1000) was added to the wells. Ab solution was removed and wells thoroughly washed before adding 3,3',5,5'-tetramethylbenzidine (TMB) substrate solution. The reactions were stopped with sulfuric acid, and the plate was read at 450 nm wavelength. protein was confirmed by WB (Fig. S3b). We compared three different methods to recover soluble Abs from IBs: dialysis, dilution, and on-column refolding.
The dilution method showed the highest recovery rate (Fig. 2a). The recovery rates were calculated after concentration and buffer exchange of elution samples and the actual values might be higher than those reported because of probable protein loss during Amicon filtration.
For comparison, we repeated the dilution method on the insoluble fraction of both scFv-6xHis and Trx-fusion Ab. The results showed that the recovery rate was remarkably reproducible (11.2% in both experiments, Fig. 2a and b). The recovery value obtained from Trx-6xHis-scFv fusion was considerably higher than scFv-6xHis ( Fig. 2b and c), perhaps due to the in vitro solubility enhancement effect of the Trx-tag. the scFv coding sequence downstream of the Trx tag in pET32a. The Trx-6xHis-scFv construct was transformed to Origami strain, the optimal host for soluble expression using Trx tag. However, soluble expression (˃5 mg/L) was only observed in truncated form. Although there was no stop codon or frame-shift mutation in the sequencing result of the Trx-6xHis-scFv construct (Fig. S2), the eluted protein had an apparent MW of ⁓30 kDa (Fig. S3a), 15 kDa lighter than the expected MW (Table S2). So we decided to analyze the insoluble fraction of induced culture. The SDS-PAGE showed an intense band of ⁓45 kDa (Fig. S3a) with a positive reaction in WB (Fig. S3b).
After solubilization of isolated IBs (insoluble fractions of induced cultures of both scFv-6xHis(pET28a)/BL21DE3 and Trx-6xHis-scFv(pET32a)/Origami B transformants), the supernatant of denatured protein solutions had ≥ 75% purity on SDS-PAGE (Fig. S4). The presence of His-tagged was in agreement with the SDS-PAGE data. Based on the chromatogram analysis, the main peak (monomeric form) encompassed ⁓60% of the protein content (Fig. 3c).

Physical characterizations of produced scFv
We used SDS-PAGE and SEC-HPLC to analyze our Abs' purity, aggregation, and monomer/multimer status. All preparations (except for on-column refolded preparation) were subjected to SDS-PAGE in two different conditions: reduced/boiled and non-reduced/unboiled. The expected band size for Trx-free and Trx-fused preparations in both conditions is ⁓28 kDa and ⁓43.6 kDa respectively (Table S2). The results showed that the dialysis-refolded scFv-6xHis and the dilution-refolded Trx-6xHis-scFv preparations had the highest and the lowest homogeneity, respectively. There seemed to be some multimeric species in dilution preparations ( Fig. 3a and b).

Discussion
Designing a new delivery system for myeloid leukemia targeting IL-1RAP as an LSCs-specific marker is a promising therapeutic approach. Due to its small size, we chose the scFv format of an anti-IL-1RAP mAb and utilized E. coli as a suitable host for the inexpensive production of our scFv.
We first focused on different soluble expression strategies, which all were disappointing. Almost all the scFv molecules formed IB when expressed cytoplasmically. Even with the help of Trx, more than 98% of recombinant protein was expressed as IB, emphasizing our scFv molecule's aggregation tendency.
The high aggregation propensity of the scFv made us explore the aggregation hotspots of the construct in silico. We found several areas inside and outside of hydrophobic complementary determining regions (CDRs) that may be responsible for the low solubility of our hard-to-fold scFv (Fig. 3e). Accordingly, it may be possible to get higher levels of soluble scFv through soluble expression strategies by rational mutagenesis of aggregation-prone regions in this construct.
For IB denaturation, high concentrations of chaotropic agents are usually used. A reducing agent is also needed in the denaturation step of proteins with disulfide bonds to reduce all S-S bridges that may have been formed in vivo or during cell lysis [13]. Unfortunately, high concentrations of these agents can deteriorate the Ni-NTA resin by reducing Ni 2+ ions and leaving brownish precipitates in the column. To prepare the denatured protein solution, we first dissolved IBs in a solubilization buffer containing 10 mM 2ME and then diluted this stock solution in a 2ME-free solubilization buffer just before the refolding. We also performed the purification step after the refolding. Through these strategies, we get the advantage of reducing power of 2ME while keeping away from the deteriorating effect on nickel resin. We also used 1 mM 2ME in refolding buffers in dialysis and dilution methods. 2ME in such a low concentration does not affect the nickel resin and has been reported to improve the refolding yield significantly, perhaps through its anti-aggregation activity [12].
We used cysteine in excess and employed a slightly alkaline pH for all refolding procedures, a well-known condition for fast disulfide exchange reactions [27].
The structural analysis of the dilution-refolded scFv-6xHis was accomplished using the CD. We also analyzed the dilution-refolded Trx-6xHis-scFv preparation for comparison. The results ( Fig. 3d; Table 3) showed that the protein was mainly composed of β-sheet structures (43%) while in the Trx-6xHis-scFv molecule, coils, and turns were the dominant structures. This observation was in accordance with previous reports [8; 25] and in silico predictions (Table 3).
We also further extend the structural characterization of scFv-6xHis construct via in silico analysis of potential aggregation hotspots using different webservers (Fig. 3e).

Binding assessments of produced ab fragments
We next, compared the binding activity of four Ab preparations to leukemic cells (K-562 and KG-1) using cell-based ELISA. ScFv from on-column refolding was not analyzed due to insufficient quantity. The highest activity belonged to dilution-refolded preparations (Abs 450nm up to 2.5, Fig. 4a). It seemed that the binding activities of refolded Abs were proportional to their recovery rates.
The binding activity of scFv to IL-1RAP was further investigated by WB analysis on leukemic cell lysates (K-562 lysate as test and KG-1 lysate as control). The WB data (Fig. 4b) showed that the produced scFv can successfully recognize the membrane-bound isoform of IL-1RAP (MW: ⁓66 kDa). As expected the band intensity in the lane pertained to KG1 lysate was weaker.

Stability analysis
The decrease in binding activity of dilution-refolded scFv-6xHis was investigated following three freeze-thaw cycles or after one to ten days of incubation at 37 ˚C. The freeze-thaw process decreased the binding activity by 37% (Fig. 4c). Likewise, incubation at 37 ˚C decreased the binding activity until day 4 (by ⁓50% compared to control). Interestingly, the activity seemed to recover since day 4, reaching 76% of its original activity on day 10. Overall, the dilution-refolded scFv exhibited good stability against long-term incubation in body temperature.

Investigation of binding activity of dilutionrefolded scFv-6xHis in serum environment
To evaluate the binding ability in the serum environment, ⁓80 µl of dilution-refolded scFv-6xHis (0.260 mg/ml) was added to 320 µl of PBS, normal or leukemic serum. After mixing, the samples were incubated at 37 ˚C for ⁓3 h and then ON at RT. The samples were then used for ELISA on K-562 cells seeded on 96 well plate (⁓67,000 cells/well).
Pulse dilution refolding was found to be the most effective refolding method for our Ab. The main benefit of this method is its cost-effectiveness because it does not need massive amounts of salt and urea (in contrast to the dialysis Numerous reports have successfully employed the oncolumn method with relatively high refolding yields (for example, [25]). However, except for rare reports [25; 28] in the case of scFvs, the recoveries have generally not been satisfactory [24; 29; 30]. The recovery rate from the on-column method in our experience was similarly disappointing Binding activity of the scFv (25 µg/ml concentration or 1.25 µg/well) to K-562 cells after repeated freeze-thaw or long-term incubation at 37 ˚C. Data analyses were performed, and the graph was plotted using GraphPad Prism version 8.0. ns: not significant, *P ˂ 0.05, **P ˂ 0.01, ***P ˂ 0.001 and ****P ˂ 0.0001 unknown structurally similar soluble factor(s)) is increased in this malignancy. The exact identity of the unknown factor(s) can be elucidated by two-dimensional western blot (2D WB) analysis of AML serum with the produced scFv and sequencing of the positive dot(s) using mass spectrometry. Anyway, this is beyond the scope of this study.
From the clinical perspective, the results of the serum interference experiment demonstrated that this antibody is not a suitable anti-leukemic agent for use in high leukemic burden. On the other hand, the proliferation inhibition assay showed that the described scFv cannot prevent leukemia progression through blocking of IL-1 signaling (data not shown). Instead, its conjugated or fusion formats with cytotoxic molecules may offer therapeutic potential in eradicating minimum residual disease (MRD) during complete remission. We're now designing a new gene delivery system using this targeting agent in our lab.
It is worth noting that alternative approaches for soluble expression may also be helpful, such as the utilization of other chaperons (e.g., RNA-based chaperons) or other solubility tags (e.g., MBP or SUMO) and the use of different prokaryotic (e.g., Bacillus strains) or eukaryotic (mammalian or yeast) hosts.

Conclusion
Briefly, in this study, we produced and characterized an scFv against human IL-1RAP. A comparison of various approaches showed that recovering IBs through dilution refolding is the most fruitful method for bacterial production of the molecule. The productivity value reported here ( 21 mg/L culture in LB medium) is sufficient for preclinical studies and can be improved by further optimizations.
Funding This study was extracted from the Ph.D. thesis of the first author and financially supported by the Pasteur Institute of Iran (grant number: BD-9365). This work was also partly supported by Iran National Science Foundation (INSF) (grants number 96011491).

Declarations
Ethical statement The authors are responsible for the correctness of the statements provided in the manuscript. All procedures performed method). The higher recovery is likely due to the wellknown anti-aggregation agent, L-arginine (L-Arg). Despite the previous reports based on direct loading of L-Arg-containing protein solution on Ni-NTA column or washing by L-Arg containing washing buffer [13], we realized that the presence of L-Arg strongly interfered with protein capturing by nickel resin. Our observation is in accordance with the 2009 report [31]. This made us buffer exchange the refolded scFv solution in two steps: (1) dialysis in 50 mM Tris-HCL, 150 mM NaCl, pH 8, for removing all refolding additives, especially L-Arg, and (2) dialysis in 50 mM NaH 2 PO 4 , 300 mM NaCl, 25 mM Imidazole, pH 8, for equilibration in manufacturer-recommended binding buffer. The workflow diagram of bacterial production of the selected scFv preparation has been depicted in Fig. S5.
The scFv recovered by dilution refolding was mostly composed of β-sheets as previously reported for other scFvs [32; 33]. The high stability of this preparation may be related to its correct 2D and 3D structure.
There is some evidence that the use of a longer linker (≥ 25 AA) favors monomer formation [34] and the solubility of scFv molecules [35]. Correspondingly, we incorporated a 25 residue flexible linker between the VH and VL domain (S1C, the underlined sequence) to avoid the formation of higher-order species. As showed by SEC analysis, our selected scFv preparation seems to be empty of soluble aggregates and HMW entities.
Fusing with Trx significantly improved final yield (63.5 vs. 21 mg/L, i.e., 1.88-fold increase considering the MW difference) (Fig. 2c). The presence of N-terminal Trx did not interfere with binding to surface antigen and seemingly improved it. Moreover, the presence of Trx resulted in lowered aggregation and enhanced recovery rate (similar to results obtained from MPB [11; 26]). Although the Trxtagged scFv could be potentially employed in diagnostic and prognostic schemes, we focused on the dilution-refolded scFv-6xHis preparation due to its higher homogeneity and no need for expensive and yield-reductive tag removal step that is essential for the molecule to be used in therapeutic applications.
The parent antibody had been originally developed against a 12-residues peptide (VPAPRYTVELAC) [14] located in proximity to the transmembrane domain and is exclusively present in the membrane-bound isoform of IL-1RAP. It has been shown by developers that the parent whole antibody could not bind to the secreted isoform of IL-1RAP [14]. Likewise, we demonstrated that the soluble isoforms present in high concentrations in normal human serum (≥ 300 ng/ml [36]) do not affect the binding of the antibody. However, the produced scFv exhibited a feeble binding activity in AML serum. The possible explanation for our observation is that the serum level of membrane isoform (or an