Improving purity and apparent thermal stability of single-chain Fv antibodies by heat treatment in a soluble E. coli cytoplasmic expression system

Abstract

Single-chain variable fragment (scFv) antibodies outperform conventional IgG antibodies due to their lower molecular weight, reduced immunogenicity, rapid clearance and efficient bacterial expression. However, it remains difficult to achieve correct folding of scFv containing two intramolecular disulfide bonds in the reducing Escherichia coli cytoplasm. We previously achieved correct folding of a trastuzumab-derived scFv by co-expressing the disulfide isomerase DsbC and the yeast sulfhydryl oxidase Erv1p in the cytoplasm of the E. coli SHuffle T7 strain. Here, we applied this system to produce the immune checkpoint inhibitor atezolizumab-derived scFv (denoted as atz–scFv). The atz–scFv was purified through a two-step chromatography process; however, insufficient disulfide bond formation led to contamination by proteins lacking a disulfide bonds. To address this, we developed and optimized a heat-treatment method in which atz–scFv was incubated at elevated temperatures to selectively precipitate protein species lacking a disulfide bond. The resulting atz–scFv exhibited increased apparent thermal stability and retained antigen-binding affinity compared to the untreated sample. This heat-treatment method offers a cost-effective alternative to traditional purification, eliminating the need for affinity chromatography while enhancing scFv stability and purity. The present results broaden scFv purification methods.

Graphical Abstract

Graphical Abstract.

Abbreviations

Amp

ampicillin

atz

atezolizumab

CBB

Coomassie brilliant blue

DsbC

disulfide bond isomerase C

DSF

differential scanning fluorimetry

E. coli

Escherichia coli

EDTA

ethylenediaminetetraacetic acid

Erv1p

essential for respiration and vegetative growth 1 protein

Fv

fragment variable

HEPES

2-[(4-(2-hHydroxyethyl)-1-piperazinyl)] methanesulfonic acid

HER2

human epidermal growth factor receptor 2

IgG

immunoglobulin G

IPTG

isopropyl β-D-1-thiogalactopyranoside

Kan

kanamycin

Ni-NTA

nickel nitrilotriacetic acid

OD

optical density

PAGE

polyacrylamide gel electrophoresis

PD-1

programmed cell death protein 1

PD-L1

programmed death ligand-1

PEG

polyethylene glycol

RU

resonance unit; scFv, single-chain variable fragment

SDS

sodium dodecyl sulfate

SEC

size-exclusion chromatography

SOC

super optimal broth with catabolite repression

SPR

surface plasmon resonance

Str

streptomycin

Tras

trastuzumab

Tm

transition midpoint temperature

Tris

tris-(hydroxymethyl)-aminomethane

VH

variable region of immunoglobulin heavy chain

VL

variable region of immunoglobulin light chain

Single-chain variable fragment (scFv) antibodies are engineered antibody fragments composed of the variable regions of the light chain (VL) and heavy chain (VH), which together form the antigen recognition sites of an antibody and are linked by a flexible peptide linker (1–3). ScFv antibodies retain their antigen-binding activity while possessing a lower molecular weight (27,000), approximately one-fifth that of IgG antibodies (150,000). This reduced molecular weight enables the low-cost production of scFvs using Escherichia coli (4–6). However, producing scFvs in an active, correctly folded form in the E. coli cytoplasm remains challenging because scFvs contain two intramolecular disulfide bonds (7,  8).

The two disulfide bonds within the scFv molecule are essential for maintaining structural integrity and functional stability. Incomplete formation of these bonds can negatively affect thermal stability and antigen-binding activity (9–11). Meanwhile, during the production of full-length monoclonal antibodies, molecules with unpaired cysteines may arise. Their presence poses risks of reducing thermal stability and biological function, and often leading to aggregation with diminished antigen binding. Therefore, complete disulfide pairing is necessary (12–15). Typically, scFvs are expressed as inclusion bodies in E. coli BL21(DE3) strains and are solubilized with high concentrations of denaturants, such as guanidine hydrochloride. The scFv is then subjected to dialysis to gradually reduce the denaturant concentration, allowing for the formation of correct disulfide bonds and proper structure. However, this process requires large amounts of expensive denaturants, leading to high production costs and extended production times (approximately one week) (6,  16–19). Using engineered E. coli strains, such as Origami B (DE3) and SHuffle T7, which have mutations in the trxB and gor genes, creates a relatively milder reducing environment in the cytoplasm that favors disulfide bond formation, significantly enhancing soluble scFv expression in E. coli (20,  21). Additionally, co-expression of molecular chaperones such as GroEL/GroES, DnaK/DnaJ/GrpE and Trigger Factor has been shown to assist in the proper folding of scFvs (22). Combining these strategies, namely the use of engineered strains and the co-expression of chaperone proteins, can significantly increase the yield of soluble scFv in E. coli. Another approach to correct folding of scFvs is to introduce redox pathways involving DsbC and Erv1p. In our previous reports, transforming the SHuffle T7 strain with a pET–21d vector containing the DsbC and Erv1p genes led to their overexpression in the cytoplasm, resulting in successful soluble expression of various scFvs, including those derived from trastuzumab (23,  24). This method reduced the preparation time for scFvs and eliminated the need for expensive reagents compared to the inclusion body preparation method.

In this study, we prepared and characterized the scFv derived from the anti–PD-L1 immune checkpoint antibody atezolizumab (atz–scFv) using the aforementioned E. coli co-expression system. Analysis of the produced samples revealed that a fraction of atz–scFv molecules still contained incompletely formed disulfide bonds. To remove these molecules, we developed a method that exploits the lower thermal stability of incompletely bonded molecules compared to fully bonded ones. By heating, incubating and centrifuging to separate the precipitate, we successfully removed the molecules with incompletely formed disulfide bonds. After initially treating the atz–scFv at 70°C for 10 min, the transition midpoint temperature (Tm) was clearly higher than that of the untreated sample, indicating improved thermal stability. Size-exclusion chromatography (SEC) confirmed that the samples after heat treatment maintained their molecular size and did not contain aggregates. While this method effectively removes molecules with incompletely formed disulfide bonds through heat treatment, it also leads to additional protein loss.

Our results demonstrate that heat treatment of atz–scFv produced by a soluble E. coli expression system at optimal temperatures and durations can effectively remove incompletely formed molecules while enhancing both purity and thermal stability. This method preserves the antigen affinity of scFvs, providing valuable insights for their production and purification.

Materials and Methods

Construction of expression plasmid

The DNA sequences encoding the atz–scFv (Fig. S1A) and the trastuzumab scFv (tras–scFv; Fig. S1B) were cloned into the pET-28b(+) expression vector (Novagen). A hexahistidine tag was added to the C-terminal end of each scFv to enable efficient affinity purification through nickel-chelating chromatography. The constructs were verified by sequencing to ensure accurate insertion of the scFvs into the vectors. The final vectors were named pET-28b(+)-atz–scFv and pET-28b(+)-tras–scFv (Fig. S2B).

Expression and purification of scFvs

The atz–scFv and tras–scFv were expressed using the genetically engineered E. coli strain SHuffle T7, developed by New England Biolabs with deletions in the thioredoxin reductase and glutathione reductase genes to weaken the reducing conditions of the cytoplasm. This strain was further modified though transformation with the pET–21d–Erv1p–DsbC vector, allowing for the cytoplasmic expression of both DsbC and Erv1p proteins under the control of the T7lac promoter. This E. coli strain was named SHuffle T7 (pET–21d–Erv1p–DsbC). The pET-28b(+)-atz–scFv plasmid was transformed into SHuffle T7 (pET–21d–Erv1p–DsbC). Colonies were selected from plates and inoculated into 30 ml of 2 × YT medium supplemented with 20 μg/ml streptomycin, 100 μg/ml ampicillin and 50 μg/ml kanamycin (all from Nacalai Tesque). Cells were harvested and resuspended in 500 ml of 2 × YT medium supplemented with the same antibiotics at an initial OD_{600 nm} of 0.1. The culture was then grown at 180 rpm until the OD_{600 nm} value reached 1.2–1.5. To induce protein expression, isopropyl β-d-1-thiogalactopyranoside (IPTG) (Biosynth) was added to a final concentration of 1 mM. Subsequently, the culture was maintained at 15°C and agitated at 110 rpm for 48 h. After induction, cells were harvested by centrifugation at 6000 rpm at 4°C. The cells were then resuspended in a buffer containing 50 mM Tris–HCl (pH 8.0), 150 mM NaCl and 0.2% Triton X-100 and lysed by sonication on ice. The lysate was centrifuged (12,000 rpm, 4°C, 40 min) to separate the supernatant from the cell debris. The supernatant was applied to Ni-NTA resin (Fujifilm Wako Chemicals). Non-specifically bound proteins were removed by washing the column with 10 column volumes of wash buffer containing 50 mM Tris–HCl (pH 8.0), 150 mM NaCl, 20 mM imidazole and 2% Triton X-100. Bound proteins were subsequently eluted with 3 column volumes of elution buffer containing 50 mM Tris–HCl (pH 8.0), 150 mM NaCl and 250 mM imidazole, passed through a 0.2-μm membrane filter, and further purified using SEC at 25°C with a HiLoad 16/60 Superdex 75 prep grade column (Cytiva).

Detection of free thiols using PEG–maleimide modification

To detect incompletely formed disulfide bonds in atz–scFv and tras–scFv, PEG–maleimide (polyethylene glycol maleimide) modification was employed. Purified atz–scFv and tras–scFv samples were adjusted to a concentration of 12 μM and mixed at a 1:2 ratio with a denaturing buffer containing 12 M urea, 50 mM HEPES (pH 7.4), 150 mM NaCl and 5 mM EDTA to expose disulfide bonds buried in the molecular core. Subsequently, PEG–maleimide (NOF Corp.; molecular weight: 2000) was added at a molar ratio of 10:1 relative to atz–scFv and tras–scFv samples, and the reaction mixture was incubated on ice for 1 h. The samples were analyzed using 15% non-reducing SDS-PAGE.

To independently confirm PEGylation, gels after SDS-PAGE were subjected to BaCl₂–I₂ staining (25). After electrophoresis, gels were rinsed with Milli-Q water and incubated for 5 min in 5% BaCl₂ prepared from barium chloride dihydrate (FUJIFILM Wako Pure Chemical Corporation), followed by washing with Milli-Q water three times for 5 min each. To visualize PEG moieties, the gels were incubated for 10 min in an iodine/iodide solution prepared from 50 mM I₂ (FUJIFILM Wako Pure Chemical Corporation), 240 mM KI (Nacalai Tesque) and 2.5 mM HCl (Nacalai Tesque). After extensive washing with Milli-Q water, images of the iodine-stained gels were acquired. The same gels were subsequently stained with Coomassie brilliant blue (CBB) for 30 min, destained and imaged to visualize total protein bands.

Heat treatment of scFvs

atz–scFv and tras–scFv samples were aliquoted into multiple PCR tubes, each containing 200 μl and subjected to independent heat treatments at different temperatures using a thermal cycler (Biometra TOne Analytik Jena). Separate samples were heated at 60, 65 and 70°C for 10 min, respectively, whereas tras–scFv samples were heated at 50, 55 and 60°C for 10 min followed by immediate cooling to 4°C. After heat treatment, each set of samples was centrifuged at 12,000 rpm and 4°C for 20 min to recover the supernatant, with the protein precipitate discarded. The supernatants containing the soluble protein were then concentrated and adjusted to 12 μM. The concentrated protein was reacted with PEG–maleimide as described in the previous section. The presence of atz–scFv and tras–scFv with incompletely formed disulfide bonds in the recovered supernatant was analyzed using 15% non-reducing SDS-PAGE.

Differential scanning fluorimetry measurements

Differential scanning fluorimetry (DSF) was employed to evaluate the thermal stability of atz–scFv and tras–scFv before and after heat treatment. All measurements were obtained at a protein concentration of 3 μM in 50 mM HEPES buffer (pH 7.4), 150 mM NaCl and 5 mM EDTA, with SYPRO-Orange dye (Thermo Fisher Scientific) added at a dilution of 1:1000. Samples were loaded into a 96-well plate, and thermal denaturation was monitored using a real-time PCR machine (BIO RAD). The temperature was increased from 25 to 95°C in increments of 0.2°C. The Tm value was determined by analyzing the inflection point of the fluorescence curve.

Size-exclusion chromatography analysis of scFvs

SEC was utilized to verify the molecular size and the presence of aggregates in atz–scFv and tras–scFv before and after heat treatment. The supernatant of heat-treated samples was centrifuged and then adjusted to the same concentration as that of the samples before heat treatment. Each sample was loaded onto a HiLoad 10/300 Superdex 75 Increase column. Chromatography was performed at room temperature using a buffer composed of 50 mM HEPES (pH 7.4), 150 mM NaCl and 5 mM EDTA.

Surface plasmon resonance measurements

Surface plasmon resonance (SPR) analyses were performed to evaluate the antigen-binding affinity of atz–scFv and tras–scFv before and after heat treatment. All SPR measurements were conducted at 25°C using a Biacore T200 system (Cytiva). An anti-human Fc antibody included in the human antibody capture kit (Cytiva) was immobilized on a CM5 sensor chip (Cytiva) via the amine coupling method using a running buffer of HBS-EP (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.005% Tween 20) (Nacalai Tesque). The sensor chip was then immobilized with human PD-L1 (hPD-L1, ACROBiosystems) containing a human Fc-tag, and different concentrations of atz–scFv were injected over the immobilized hPD-L1 at a flow rate of 30 μl/min.

To evaluate the antigen-binding affinity of tras–scFv before and after heat treatment, SPR analyses were performed at 25°C using a Biacore T200 system (Cytiva) with HBS-EP as the running buffer. Recombinant GST-fused HER2 ectodomain (GST-HER2) was prepared as described previously (23). An anti-GST antibody included in the GST-capture kit (Cytiva) was immobilized on a CM5 sensor chip (Cytiva) via the amine coupling method. GST-HER2 was then captured on the active flow cell, while GST protein was captured on the reference flow cell. Tras–scFv was injected over the immobilized GST-HER2 at a flow rate of 50 μl/min.

Results

Preparation of atz–scFv proteins

The VL domain of atezolizumab was followed by a 19-residue flexible glycine-serine peptide linker and the VH domain to construct the atz–scFv. Additionally, a hexahistidine tag was attached at the C-terminus for purification using Ni–NTA affinity chromatography (Fig. S2A). The atz–scFv was expressed in the SHuffle T7 (pET–21d–Erv1p–DsbC) strain of E. coli. This strain enabled soluble expression of atz–scFv, followed by initial purification using Ni–NTA affinity chromatography and subsequent purification through SEC (Fig. 1A). Analysis of the atz–scFv using both reducing and non-reducing SDS-PAGE revealed a single band, indicating sufficient purity for physicochemical characterization (Fig. 1B).

Fig. 1.

Purification of atz–scFv protein. (A) Elution profiles of atz–scFv on size-exclusion chromatography. (B) SDS-PAGE analysis of atz–scFv purity under non-reducing (left) and reducing (right) conditions

PEG–maleimide modification reveals unbound disulfide bonds in atz–scFv molecules

PEG–maleimide is a PEG derivative functionalized with maleimide groups that selectively react with free thiol groups. This characteristic makes PEG–maleimide particularly useful for bioconjugation and PEGylation processes, where it can form stable thioether bonds with cysteine residues on proteins or other thiol-containing molecules (26,  27). To evaluate the formation of disulfide bonds within atz–scFv, we based our approach on the principle of molecular weight alteration through PEG–maleimide modification of free cysteine residues. The free cysteine residues in scFv molecules that had not formed disulfide bonds reacted with PEG–maleimide, resulting in an increase in molecular weight. This change was detectable as band shifts on the SDS-PAGE gel. In this study, the SDS-PAGE results showed distinct band shifts, indicating that PEG–maleimide was attached to free cysteine residues in the atz–scFv molecules (Fig. 2). There are four cysteine residues in atz–scFv, all of which would form disulfide bonds within the VH and VL domains. However, it can be assumed that atz–scFv produced by the SHuffle T7 E. coli expression system included a fraction of molecules with incompletely formed disulfide bonds, which could not be removed by Ni–NTA or SEC.

Fig. 2.

SDS-PAGE analyses of atz–scFv before and after PEG–maleimide modification., visualized by BaCl₂-I₂ and CBB staining. PEG-modified atz–scFv species were selectively visualized by BaCl₂-I₂ staining (left). After washing and subsequent CBB staining (right), atz–scFv before (lane 1) and after (lane 2) PEG–maleimide modification are shown.

To further confirm that the observed band shifts were indeed derived from PEG modification, we additionally performed BaCl₂–I₂ staining, which visualizes PEG moieties through the formation of an iodine–PEG complex in the presence of Ba²⁺. After SDS-PAGE, the gel was first subjected to iodine/iodide staining and imaged, then extensively washed with water to remove the iodine signal, followed by CBB staining to visualize total protein bands. Notably, the positions and migration patterns of the iodine-positive bands were consistent with those of the corresponding CBB-stained bands, demonstrating that the shifted bands represent PEG–maleimide–modified atz–scFv species (Fig. 2).

Heat treatment to remove atz–scFv with incompletely formed disulfide bonds

Disulfide bonds are responsible for the thermal stability of antibodies. Therefore, scFv antibodies with incomplete disulfide bond formation are thermally less stable than those with fully formed disulfide bonds. Based on this difference in thermal stability, we removed atz–scFv variants with incompletely formed disulfide bonds by heating atz–scFv samples to various temperatures and maintaining them for a period to precipitate proteins with poor thermal stability. The precipitate was removed by centrifugation, and thermostable atz–scFv was recovered from the supernatant. The efficacy of heat treatment at different temperatures for the removal of atz–scFv with incompletely formed disulfide bonds was assessed using SDS-PAGE analysis.

The samples were subjected to heat treatment at 60, 65 and 70°C for 10 min, followed by centrifugation to remove precipitated proteins. The supernatant was then treated with PEG–maleimide to modify free cysteine residues, causing a shift in molecular weight. SDS-PAGE analysis showed that only the samples heat-treated at 70°C did not exhibit this band shift, suggesting that at this temperature, atz–scFv with incompletely formed disulfide bonds was effectively precipitated and removed (Fig. 3A). In contrast, samples treated at 60 and 65°C showed band shifts, indicating the presence of free cysteine residues that had reacted with PEG–maleimide. This result highlights the importance of the 70°C treatment in ensuring the purity of atz–scFv by removing unstable molecules that have not formed proper disulfide bonds. The thermal stability of the samples after heat treatment was analyzed by DSF (Fig. 3B). The samples treated at 70°C exhibited a Tm of 74.5°C, with no variation among replicates; this value was approximately 2.1°C higher than that of the untreated samples (72.4°C) (Table I). Compared to the samples treated at 60 and 65°C, the Tm value of the 70°C treated samples was also approximately 1.0°C higher (Table I). SEC analysis of the atz–scFv before and after heat treatment at 70°C showed no peaks at the exclusion limit of 8 ml, indicating the absence of aggregates in the supernatant after heat treatment (Fig. 3C). Both before and after heat treatment, peaks corresponding to scFv dimers were observed around 12 ml (Fig. 3C). Moreover, peaks indicative of scFv monomers were found at approximately 14 ml for both treated and untreated samples, confirming that the molecular size of atz–scFv remains unchanged after heat treatment at 70°C (Fig. 3C).

Fig. 3.

Removal of atz–scFv molecules with incomplete disulfide bond formation using heat treatment method. (A) SDS-PAGE analysis of PEG–maleimide modification to detect incompletely formed disulfide bonds before and after heat treatment. Lane 1, before heat treatment and without modification; Lane 2, before heat treatment and with modification; Lane 3, 60°C heat treatment followed by modification; Lane 4, 65°C heat treatment followed by modification; Lane 5, 70°C heat treatment followed by modification. (B) DSF analysis of atz–scFv thermal stability before and after heat treatment at various temperatures. The solid line represents the thermal profile of atz–scFv before heat treatment; the rounded dashed line represents the profile after treatment at 60°C; the square-dashed line shows the profile after treatment at 65°C; the dotted line represents the profile after treatment at 70°C. (C) Elution profiles of atz–scFv before and after 70°C heat treatment on size-exclusion chromatography. The solid line represents atz–scFv before heat treatment, and the dashed line shows atz–scFv after 70°C heat treatment.

Table I.

Tm values of atz–scFv before and after heat treatment at different temperatures

	Before heat treatment	After 60°C heat treatment	After 65°C heat treatment	After 70°C heat treatment
Tm (°C)	72.4 ± 0.3	73.6 ± 0.2	73.8 ± 0.4	74.5 ± 0.1

Data are expressed as mean ± SD of three independent experiments.

Optimization of heat treatment conditions for enhanced recovery rate of atz–scFv

The concentration of atz–scFv in the supernatant after heat treatment at 70°C for 10 min was substantially reduced compared to that at 60 or 65°C for 10 min, indicating significant protein loss (Fig. 4). Therefore, the temperatures and durations of heat treatment were optimized to increase the recovery rate of the atz–scFv antibody within the range of 65 and 70°C. We examined incubation temperatures at 67, 69 and 70°C with durations of 5 min and 10 min. Protein concentrations in the supernatant, as well as the presence of atz–scFv with incompletely formed disulfide bonds in the recovered supernatant, were examined using SDS-PAGE and the PEG–maleimide reaction experiment. No band shifts were observed across the temperature range from 67 to 70°C under different time conditions in the PEG–maleimide reaction experiments (Fig. 5A). Analysis of protein concentrations in the supernatant after heat treatment under various conditions revealed that longer durations at the same temperature resulted in lower protein concentrations (Fig. 5B and C). At 67°C, the concentration was lower after 10 min than after 5 min. Similarly, at 69 and 70°C, the protein concentration decreased with an increase in treatment duration from 5 to 10 min. Higher temperatures at the same duration also led to lower protein concentrations. At a treatment duration of 10 min, the concentration at 70°C was lower than at 69°C, and that at 69°C was lower than at 67°C. These observations suggest that both higher temperatures and longer durations lead to increased protein loss during heat treatment.

Fig. 4.

Analyses of atz–scFv concentration in supernatant fraction. Concentrations of atz–scFv in supernatant fraction before and after 10 min of heat treatment at 60, 65 and 70°C are shown, respectively.

Fig. 5.

Optimization of heat treatment conditions to improve the recovery rate of atz–scFv. (A) SDS-PAGE analysis of atz–scFv modified with PEG–maleimide to detect incompletely formed disulfide bonds before and after various heat treatments. All lanes except Lane 1 were modified with PEG–maleimide. Lane 1, before heat treatment and without modification; Lane 2, before heat treatment; Lane 3, 67°C for 5 min; Lane 4, 67°C for 10 min; Lane 5, 69°C for 5 min; Lane 6, 69°C for 10 min; Lane 7, 70°C for 5 min; Lane 8, 70°C for 10 min. (B) Concentrations of atz–scFv in the supernatant fraction before and after heat treatment at 67, 69 and 70°C for 5 min. (C) Concentrations of atz–scFv in the supernatant fraction before and after heat treatment at 67, 69 and 70°C for 10 min. (D) DSF analysis of atz–scFv before and after heat treatment at various temperatures and durations. Black line, before heat treatment; dotted line, after 67°C for 5 min; square-cornered dash-dot line, after 67°C for 10 min; square-cornered dashed line, after 69°C for 5 min; rounded-cornered dashed line, after 69°C for 10 min; rounded dash-dot line, after 70°C for 5 min; rounded dash-dot line with two dots between dashes, after 70°C for 10 min.

The thermal stability of atz–scFv was examined by DSF (Fig. 5D), revealing that samples treated at 69°C for 10 min and at 70°C for 10 min exhibited similar Tm values (Table III). Although other conditions likewise showed no detectable migration shifts in the PEG–maleimide reaction experiments, lower Tm values were observed (Tables II and III). This suggests the presence of atz–scFv with incompletely formed disulfide bonds that were undetectable by the PEG–maleimide assay. Atz–scFv samples heat-treated at 69 and 70°C for 10 min exhibited almost identical Tm values, differing only by approximately 0.2°C (Table III). However, protein recovery at 69°C was approximately 10% higher compared to that at 70°C. Moreover, SPR analyses revealed no significant difference in antigen-binding affinity to hPD-L1 between samples before and after heat treatment at both 69 and 70°C for 10 min (Fig. S4, Table IV). Therefore, considering the comparable thermal stability and antigen-binding activity, along with improved protein recovery, the heat treatment providing the best balance between thermal stability and protein yield was determined to be 69°C for 10 min.

Table III.

Tm values of atz–scFv after 10 min heat treatment at various temperatures

	Before heat treatment	After 67°C 10 min	After 69°C 10 min	After 70°C 10 min
Tm (°C)	72.3 ± 0.1	73.4 ± 0.3	74.7 ± 0.3	74.9 ± 0.4

Data are expressed as mean ± SD of three independent experiments.

Table II.

Tm values of atz–scFv after 5 min heat treatment at various temperatures

	Before heat treatment	After 67°C 5 min	After 69°C 5 min	After 70°C 5 min
Tm (°C)	72.3 ± 0.1	73.2 ± 0.2	73.6 ± 0.1	73.2 ± 0.3

Data are expressed as mean ± SD of three independent experiments.

Table IV.

SPR analyses of atz–scFv before and after heat treatment at 69 and 70°C for 10 min

	k ass × 106 (1/Ms)	k diss × 10−4 (1/s)	K D × 10−11 (M)
Before heat treatment	3.5 ± 0.2	2.3 ± 0.0	6.5 ± 0.4
After 69°C 10 min heat treatment	4.1 ± 0.1	2.3 ± 0.1	5.6 ± 0.1
After 70°C 10 min heat treatment	3.4 ± 0.1	2.3 ± 0.2	6.8 ± 0.6

Data are expressed as mean ± SD of three independent experiments.

We quantified the bands in Fig. 5A for lane 1 (sample before heat treatment without PEG) and lane 2 (sample before heat treatment with PEG) using ImageJ. The analysis indicated that approximately 89.6% of the initial sample consisted of species with complete disulfide bonds. Under the optimized condition (69°C for 10 min), the overall recovery yield was approximately 33.1% (Fig. 5C). Species with incomplete disulfide bonds were almost completely removed after heat treatment, while approximately 36.9% of the correctly folded species were retained.

Application of heat treatment for preparation of other scFv proteins

To verify the versatility of the heat treatment for removing scFv species with incompletely formed disulfide bonds, we investigated its application to tras–scFv. To prepare samples containing incompletely disulfide-bonded scFv species, the culture temperature was lowered to 15°C, significantly below the 37°C used in our previous study (23). Tras–scFv was purified by Ni–NTA affinity chromatography followed by SEC, and the purity of the resulting preparation was confirmed by SDS-PAGE (Fig. S3A and B). The presence of scFv species with incompletely formed disulfide bonds in the 15°C culture was confirmed by PEG–maleimide assay and BaCl₂–I₂ staining (Fig. 6A). Optimization of the heat treatment parameters revealed that incubation at 54°C for 10 min were the optimal conditions for the selective removal of scFv species with incompletely formed disulfide bonds (Fig. 6B). Analysis of protein concentrations in the supernatant showed that, consistent with the results for atz–scFv, the recovery rate depended on both temperature and duration, with higher temperatures and longer heating times leading to lower recovery (Fig. 6C and D). The thermal stability of tras–scFv was examined by DSF. Although the Tm values did not differ markedly among the tested conditions, the low-temperature heterogeneity-associated shoulder progressively decreased as the treatment temperature increased, indicating improved sample homogeneity. The melting curves of the samples treated at 54°C for 10 min and at 55°C for 10 min were nearly superimposable (Fig. 6E) (Tables V and VI). Together with the absence of band shifts in the PEG–maleimide assay, these results suggest that both conditions effectively removed incompletely disulfide-bonded species. Notably, the recovery rate at 54°C for 10 min was approximately 8% higher than that at 55°C for 10 min, indicating a more favorable balance between sample quality and protein yield (Fig. 6C and D). Moreover, SPR analyses revealed no significant difference in antigen-binding affinity to HER2 between tras–scFv samples before and after heat treatment at both 54 and 55°C for 10 min (Fig. S5) (Table VII).

Fig. 6.

Application and optimization of heat treatment for purification of tras–scFv by removal of incompletely disulfide-bonded species. (A) PEG-modified tras–scFv species were selectively visualized by BaCl₂-I₂ staining (left). After washing and subsequent CBB staining (right), tras–scFv before (lane 1) and after (lane 2) PEG–maleimide modification are shown. (B) SDS-PAGE analysis of tras–scFv modified with PEG–maleimide to detect incompletely formed disulfide bonds before and after various heat treatments. All lanes except Lane 1 were modified with PEG–maleimide. Lane 1, before heat treatment and without modification. Lane 2, before heat treatment. Lane 3, 52°C for 5 min. Lane 4, 52°C for 10 min. Lane 5, 54°C for 5 min. Lane 6, 54°C for 10 min. Lane 7, 55°C for 5 min. Lane 8, 55°C for 10 min. (C) Concentrations of tras–scFv in the supernatant fraction before and after heat treatment at 52, 54 and 55°C for 5 min. (D) Concentrations of tras–scFv in the supernatant fraction before and after heat treatment at 52, 54 and 55°C for 10 min. (E) DSF analysis of tras–scFv before and after heat treatment at various temperatures and durations. Black line, before heat treatment. Dotted line, after 52°C for 5 min. Square-cornered dash-dot line, after 52°C for 10 min. Square-cornered dashed line, after 54°C for 5 min. Rounded-cornered dashed line, after 54°C for 10 min. Rounded dash-dot line, after 55°C for 5 min. Rounded dash-dot line with two dots between dashes, after 55°C for 10 min.

Table V.

Tm values of tras–scFv after 5 min heat treatment at various temperatures

	Before heat treatment	After 52°C 5 min	After 54°C 5 min	After 55°C 5 min
Tm (°C)	63.3 ± 0.1	63.6 ± 0.4	63.8 ± 0.3	63.4 ± 0.0

Data are expressed as mean ± SD of three independent experiments.

Table VI.

Tm values of tras–scFv after 10 min heat treatment at various temperatures

	Before heat treatment	After 52°C 10 min	After 54°C 10 min	After 55°C 10 min
Tm (°C)	63.3 ± 0.1	63.6 ± 0.2	63.5 ± 0.1	63.9 ± 0.3

Data are expressed as mean ± SD of three independent experiments.

Table VII.

SPR analyses of tras–scFv before and after heat treatment at 54 and 55°C for 10 min

	k ass × 105 (1/Ms)	k diss × 10−4 (1/s)	K D × 10−10 (M)
Before heat treatment	4.9 ± 0.0	4.0 ± 0.0	8.3 ± 0.2
After 54°C 10 min heat treatment	4.4 ± 0.1	4.0 ± 0.1	9.0 ± 0.1
After 55°C 10 min heat treatment	9.2 ± 0.1	3.9 ± 0.1	4.3 ± 0.1

Data are expressed as mean ± SD of three independent experiments.

Discussion

In the present study, we successfully enhanced the purity and apparent thermal stability of scFv derived from the anti-PD-L1 immune checkpoint atz–scFv by employing a heat treatment method. We utilized the engineered SHuffle T7 strain of E. coli, which co-expresses the molecular chaperones DsbC and Erv1p, to achieve soluble expression of atz–scFv, although some atz–scFv molecules still contained incompletely formed disulfide bonds. Tras–scFv species with incompletely formed disulfide bonds—generated by reducing the cultivation temperature to 15°C from the 37°C utilized in our previous work (23)—were removable by heat treatment, consistent with our observations for atz–scFv. These findings demonstrate that even introducing redox systems into an E. coli strain with a milder cytoplasmic reducing environment was insufficient for the correct production of certain scFvs. Although our work centres on E. coli-expressed scFv, similar unpaired-cysteine molecules have been reported in CHO-derived IgG, indicating that disulfide control is a general consideration in antibody development (28–30).

Our results indicated that heat treatment of the produced atz–scFv samples was an effective method for removing molecules with incompletely formed disulfide bonds. The DSF results showed a steeper denaturation curve of the heat-treated atz–scFv, indicating increased homogeneity of the atz–scFv samples. Additionally, the Tm value increased by approximately 2.4°C, demonstrating that the heat treatment effectively removed thermally unstable or misfolded protein molecules. Therefore, while heat treatment allowed for the removal of improperly folded proteins, it also had the drawback of protein loss. Thus, the heat treatment temperature and time need to be optimized precisely. For atz–scFv, heating at 69°C for 10 min maximized the recovery rate while effectively removing molecules with incompletely formed disulfide bonds. Under conditions of 67°C for 5 or 10 min, 69°C for 5 min and 70°C for 5 min, no band shift was observed on SDS-PAGE, while the Tm values did not reach the same level as those observed with 69°C for 10 min or 70°C for 10 min, suggesting the presence of molecules with incompletely formed disulfide bonds that were not fully removed. On the other hand, the Tm values for 69°C for 10 min and 70°C for 10 min were of similar magnitude, indicating that both conditions effectively removed the incompletely bonded molecules. Under the 5-min heat treatment condition, the recovery rates at 69 and 70°C were almost identical. However, when the treatment time was extended to 10 min, the recovery rate at 70°C dropped by approximately 10% compared to that at 69°C. These results suggest that even a 1°C difference can exceed a critical threshold at which structurally distinct proteins begin to denature. Thus, the optimal balance between recovery rate and thermal stability is often found near the denaturation threshold. The samples before and after heat treatment exhibited nearly identical antigen-binding affinity, indicating that even with incompletely formed disulfide bonds, the atz–scFv molecules retained the necessary structure for antigen binding.

In our previous work (23), we successfully expressed a trastuzumab-derived scFv (denoted as tras–scFv) using the SHuffle T7 (pET–21d–Erv1p–DsbC) system at a cultivation temperature of 37°C; in contrast, the present study used a cultivation temperature of 15°C, resulting in contamination with scFv species containing incompletely formed disulfide bonds (Fig. S6A). In the case of atz–scFv, increasing the cultivation temperature from 15 to 37°C did not result in the disappearance of atz–scFv species with incompletely formed disulfide bonds (Fig. S6B). The two scFv proteins exhibited different intrinsic thermal stabilities: the Tm of tras–scFv before heat treatment was approximately 62°C, whereas that of atz–scFv was approximately 72°C. The higher stability of atz–scFv suppressed the partitioning of scFv species with incompletely formed disulfide bonds into inclusion bodies during cultivation at 37°C. It could be assumed that increasing the cultivation temperature would fully sequester molecular species with incompletely formed disulfide bonds into inclusion bodies, rendering subsequent heat treatment unnecessary. However, for highly thermostable scFvs, increasing the temperature may not fully sequester these incomplete species into inclusion bodies, necessitating heat treatment to remove them.

In conclusion, this study demonstrated that heat treatment provides an efficient method for improving the purity and thermal stability of scFv antibodies. The strategy offers a practical alternative to conventional purification techniques and supports broader applications of scFvs in therapeutic and diagnostic contexts.

Supplementary Data

Supplementary Data are available at JB Online.

Funding

This study was partly supported by JSPS KAKENHI grant number 24 K02165 to H.M. and AMED under grant numbers 23ak0101193 and 24ak0101193 to H.M., Y.K. and T.S.

Conflict of Interest

None declared.

Author Contributions

Y.K. and H.M. designed the experiments. S.D., K.O., S.K. and T.S. performed the experiments and analyzed the data. S.D., Y.K. and H.M. wrote the paper. All authors discussed the results.