Expression and purification of a native Thy1-single-chain variable fragment for use in molecular imaging

Molecular imaging using singlechain variable fragments (scFv) of antibodies targeting cancer specific antigens have been considered a non-immunogenic approach for early diagnosis in the clinic. Usually, production of proteins is performed within Escherichia coli. Recombinant proteins are either expressed in E. coli cytoplasm as insoluble inclusion bodies, that often need cumbersome denaturation and refolding processes, or secreted toward the periplasm as soluble proteins that highly reduce the overall yield. However, production of active scFvs in their native form, without any heterologous fusion, is required for clinical applications. In this study, we expressed an anti-thymocyte differentiation antigen-scFv (Thy1-scFv) as a fusion protein with a N-terminal sequence including 3 × hexa-histidines, as purification tags, together with a Trx-tag and a S-tag for enhanced-solubility. Our strategy allowed to recover ~ 35% of Thy1-scFv in the soluble cytoplasmic fraction. An enterokinase cleavage site in between Thy1-scFv and the upstream tags was used to regenerate the protein with 97.7 ± 2.3% purity without any tags. Thy1-scFv showed functionality towards its target on flow cytometry assays. Finally, in vivo molecular imaging using Thy1-scFv conjugated to an ultrasound contrast agent (MBThy1-scFv) demonstrated signal enhancement on a transgenic pancreatic ductal adenocarcinoma (PDAC) mouse model (3.1 ± 1.2 a.u.) compared to non-targeted control (0.4 ± 0.4 a.u.) suggesting potential for PDAC early diagnosis. Overall, our strategy facilitates the expression and purification of Thy1-scFv while introducing its ability for diagnostic molecular imaging of pancreatic cancer. The presented methodology could be expanded to other important eukaryotic proteins for various applications, including but not limited to molecular imaging.


Results
Engineering, expression and purification of Thy1-scFv. Construction of vectors expressing Thy1-scFv gene for efficient soluble protein production. We constructed three different expression vectors in the popular pET32b vector backbone with Trx-and S-tags and additional tandem His-Tags, giving: pET32b-1XHis-scFv, pET32b-3XHis-scFv, and pET32b-5XHis-scFv (Fig. 3a). Sequence confirmed, clones were transformed into T7 SHuffle E. coli cells to investigate the effect of each variant on the expression and purification of Thy1-scFv. In all constructs, the tagged-Thy1-scFv has a theoretical molecular weight around 50 kDa.  Figure S1) since higher temperatures usually result in a rapid decrease in protein yield due to degradation and misfolding 25 . As expected, at 37 °C, the expression rate of Thy1-scFv was very low (Supplementary Figure S2), in contrast, we observed an increase in our target protein yield when cultured at 30 °C probably because such lower temperature reduces protein degradation, improves folding efficiency, and thus reduces IB formation. Moreover, endogenous proteases have a higher turnover rate when E. coli is grown at 37 °C, thus leading to an enhanced-proteolysis of Thy1-scFv into a Thy1-scFv fragment (around 25 kDa). After purification by affinity chromatography, we detected Thy1-scFv in the elution fractions at an apparent molecular weight consistent with its theoretical mass in all three constructs. Importantly, culture expressing pET32b-3XHis-scFv with an induction temperature of 30 °C showed higher protein level compared to the other constructs (Fig. 3b, pET32b-3XHis-scFv, lanes 6 to 8), and Thy1-scFv was isolated with 47.7 ± 11.5% purity. This is in part due to higher binding of Ni-agarose with the 3X-His-tag used for purification. Apart from the full-length Thy1-scFv, a fragment of around 25 kDa was the major co-purified protein. We suspected it to come from residual endogenous proteolytic activity on Thy1-scFv on a locally weaker structure probably in the multi-histidine tag region. After transfer onto nitrocellulose membranes, an anti-hexa-histidine tag antibody was used to confirm the presence of Thy1-scFv. Analysis of both soluble and insoluble fractions indicates that ~ 35% of the protein can be recovered in the soluble fraction (Supplementary Figure S3). The tagged-scFv from pET32b-1XHis-scFv bound the column with weak affinity and therefore was co-purified with many contaminants. The use of pET32b-5XHis-scFv generated a longer and flexible tag sequence, more prone to influence protein binding sites and improper column binding, as suggested by the affinity chromatography elution profile. Based on these results, the construct pET32b-3XHis-scFv was chosen for further experiments presented in this study. Elution fractions 6-8 from this construct were pooled and used for further enrichment. Elimination of imidazole salts resulted in a more stable protein. The sample was concentrated (Fig. 4, undigested T 0 ) yielding to 0.37 ± 0.15 mg of tagged-Thy1-scFv per liter of bacterial culture (Table 1), difficult to reach using other methods of purification.  Figure S4a). However, an incubation time > 4 h or the use of ≥ 8 U of EK during a time > 1 h triggered non-specific cleavage of Thy1-scFv (Supplementary Figure S4b) and was also not cost effective. No spontaneous hydrolysis was detected after 24 h of incubation.      (Fig. 5a) and mass spectrometry (Supplementary Figure S5). Further analysis also highlighted a preponderant monomer fraction with a small fraction of dimer (Fig. 5b). The monomer/dimer relative intensity ratio was 10:2 by SDS-PAGE and 10:3 by MALDI analyses. Contrary to the first round of purification (Fig. 6a) where the recombinant protein was recovered in the latest elution fractions, the native protein was isolated in the flow-through fraction (Fig. 6b). Recombinant EK, uncleaved recombinant Thy1-scFv, and fragments containing histidine tags were retained on the column.   In vivo US contrast agents functionalized with Thy1-scFv enhance PDAC molecular imaging in transgenic animal model. To test the functionality of Thy1-scFv in vivo, we used US contrast agents (MBs) covalently bound with Thy1-scFv on their surface using NHS-chemistry, thus giving MB Thy1-scFv . Control MBs, without ligand, were reported as MB non-targeted . Prior to imaging, both MB types were tested for size and concentration changes that may occur due to Thy1-scFv's influence on steric changes or octafluoropropane (C 3 F 8 ) gas dissipation. Incorporation of Thy1-scFv as part of the MB composition, i.e., MB Thy1-scFv , did not significantly affect the MB mean diameter or size distribution compared to MB non-targeted (mean diameter = 1.2 ± 1.4 μm and 1.1 ± 0.8 μm, respectively) nor the concentration (1.10 9 particles/mL and 1.2.10 9 particles/mL, respectively) in agreement with MBs used in the clinic (e.g., Definity (Lantheus Medical Imaging), Sonovue (Bracco Diagnostics)) (Supplementary Figure S6). Tumors were located on B-mode imaging and MBs were injected intravenously starting with MB non-targeted and then MB Thy1-scFv in this specific order (Fig. 8a). On dTE images, mice injected with MB Thy1-scFv showed tumors with significantly increased Thy1 molecular imaging signal (3.1 ± 1.2 a.u.) compared to MB non-targeted (0.4 ± 0.4 a.u.), with a quantitative outcome of ~ 7.8-fold (p < 0.03) (Fig. 8b,c) compared to control. Conversely, imaging of healthy pancreas did not produce any significant differences in imaging signal between the two MB constructs (MB Thy1-scFv = 0.1 ± 0.1 a.u. and MB non-targeted = 0.07 ± 0.06 a.u) and were significantly lower than with MB Thy1-scFv in PDAC model (39-fold, p < 0.02). Histological analysis of H&E-stained tissues confirmed presence of PDAC (Fig. 8d). Overall, these results demonstrate conserved Thy1-targeting property of Thy1-scFv in vivo and sketches the applicability for PDAC USMI.

Discussion
Molecular imaging has made a considerable contribution to oncology throughout the course of early detection and prognosis, and is an integral part of clinical trials. Biomarkers can be detected using various targeted-probes based on antibodies, peptides or proteins, oligonucleotides, or small molecules conjugated to imaging agents for suitable imaging modalities. Specifically, recombinant protein expression has become an established technique for production of cancer specific antigen-binding ligands in bacterial systems. However, conventional methods can be cumbersome toward meeting criteria for clinical applications. The purpose of this study was to engineer a production model for recombinant protein in their native form through the example of Thy1-scFv, while introducing its potential for early diagnosis of pancreatic cancer.
In vitro refolding technology of IBs has become prevalent to recover insoluble eukaryotic proteins expressed in E. coli 10,26 . IBs are usually denatured with high concentration of chaotropes such as urea or guanidine hydrochloride 27 . The major drawbacks of this method are its complex and expensive operational process, which further needs optimization at multiple steps. In addition, use of high concentrations of denaturing agents results in complete denaturation of the secondary structure favoring re-aggregation during successive process 26 . Thus, recovery of soluble and active protein can be greatly reduced. In this study, we proposed an alternative to such extreme procedures and presented a multi-step process for enhanced-soluble protein production in SHuffle T7 E. coli cells cytoplasm using gene fusion technology. We used the well-known Trx-tag, improving disulfide bond formation, combined with S-tag. S-tag, commonly used as affinity tag, was here employed to enhance Thy1-scFv solubility thanks to its abundance in charged and polar residues 28 . Both Trx-tag and S-tag are small and do not interfere with the proper folding or function of a fused target protein. Although essential for the expression and purification of recombinant proteins, tag sequences have the potential to interfere with the structure and the function of their fusion partner. In addition, many tags have interacting partners in mammalian systems which www.nature.com/scientificreports/ can interfere with the biological applicability of recombinant proteins while generating a strong immune reaction. Therefore, tag removal should be considered, especially if the target protein is intended for pharmaceutical or therapeutic clinical applications, for crystallization, and for structural determination studies 15 . Chemical cleavage methods are usually inexpensive 29 but most systems rely on endopeptidases to separate the fusion partner from the protein of interest 30 . Serine proteases such as the activated blood coagulation factor X (Factor Xa), EK, and thrombin, have been used, as well as viral proteases such as tobacco etch virus (TEV) protease and rhinovirus 3C protease. Viral proteases have a more stringent sequence specificity due to their much slower turnover rates (catalytic rate constant (kcat)) [31][32][33][34][35] , however EK has no amino acid specificity requirement on the P' part of the scissile bond (DDDDK↓) (only proline and tryptophan should be avoided on P1' which corresponds to an alanine residue in our study). Consequently, when an affinity tag is joined to the N terminus of the protein of interest, EK is able to regenerate a native N terminus. Moreover, we used a recombinant EK presenting the same affinity tag attached to the protein of interest, i.e., hexa-histidine tag. This allowed to apply the digestion products on the same affinity chromatography for separation. Undigested fusion protein substrate, tagged-protease, cleaved tag, and any endogenous proteins that bound to the affinity resin will be separate from the untagged protein of interest in the unbound effluent. After removal of the fusion partner tag, native Thy1-scFv was recovered (0.22 mg ± 0.11) and used for further in vitro and in vivo experiments. We proved its binding functionality in vitro using Thy1expressing cells. It can be noted that our cell binding assay demonstrated a high shift in fluorescence between our scFv and the corresponding commercial full-length antibody. Knowing that Thy1-Ab-APC (Thermo Fisher, CA) was purchased as such, while Thy1-scFv was conjugated to APC following our own protocol where we used biotin-streptavidin chemistry, the number of dyes per molecules could be different depending on the conjugation www.nature.com/scientificreports/ chemistry used by the manufacturer. Moreover, full-length antibodies (MW ~ 150 kDa) will likely have more reactive amino acids available than smaller scFvs. For in vivo imaging, we used Thy1-scFv after conjugating to microbubbles by NHS-chemistry. We successfully imaged PDAC neovasculature in transgenic animals. It can be mentioned that a flexible cysteine-tag introduced on the C-terminal end of the scFv could give the possibility for multiple conjugation chemistry while providing site-specific labeling of molecules bearing maleimide functional groups for various applications. Based on our results, we anticipate that the vector design and basic strategy presented in this study should be applicable to many proteins of biological interest which are currently difficult to purify. Overall, it is generally assumed that for E. coli, a 1-L fermentation will generate ~ 150 mg of total cellular proteins. Assuming an average yield between 0.5 and 5% of total proteins, 0.75-7.5 mg of recombinant protein is available in the cells. Based on our results, the protein of interests can be recovered between 30 and 50% in the soluble fraction, hence, our strategy could allow convenient small-scale production of biologically important recombinant proteins to initiate most studies (0.1-3 mg). Scale of production could be expanded by establishing a large-scale fermentation system to produce higher amount of proteins.
Given the low median survival rate (5-year survival rate < 9%) and the low percentage of PDAC patients qualifying for tumor resection (10-20%), the need for early screening methods is globally recognized 36 . Efforts have been made to develop molecular imaging probes capable of detecting early stage PDAC 37 . Here, we produced Thy1-scFv conjugated-US contrast agent, MB Thy1-scFv , and illustrated its potential for non-invasively enhancing USMI contrast between PDAC and normal pancreatic tissues in mice consistent with related findings 24,38 . Our probe, able to detect small foci in the pancreas (> 2 mm), constitutes a promising translatable USMI agent. Naturally, the reproducibility and the sensitivity for PDAC molecular imaging will have to be further analyzed. With the success of the first and, to date, only targeted-MB in clinical trials for various cancers, BR55 (kinase insert domain receptor-targeted peptide), a rapid expansion of targeted US contrast agents is expected. Our promising pre-clinical results with MB Thy1-scFv could provide opportunities for improved PDAC prognosis, and the presented targeted US contrast agent strategy, a variable format for other biomarker targeting.

Materials and methods
Ethical approval. The Administrative Panel on Laboratory Animal Care of Stanford University approved all procedures using laboratory animals used in this study, and all experiments were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals (APLAC-33828). This study was carried out following the ARRIVE guidelines.
Expression vector design. The expression vectors pET32b-1XHis-scFv, pET32b-3XHis-scFv, and pET32b-5XHis-scFv were constructed for Thy1-scFv expression. A ligation substrate featuring Thy1-scFv protein was amplified by PCR using a forward primer with NcoI restriction enzyme site and a reverse primer with XhoI restriction enzyme site. The amplified fragment was digested with NcoI and XhoI and ligated into pET-32b(+) prokaryotic expression vector digested with respective restriction enzymes to construct pET32b-1XThy1-scFv with a single inherent hexa-histidine-tag located between the Trx-and S-tags. To introduce more hexa-histidine-tags to construct pET32b-3XHis-scFv and pET32b-5XHis-scFv vectors, we inserted annealed forward and reverse primers coding for 2 and 4 additional hexa-histidine-tags (i.e., 3XHis and 5XHis total) with BglII restriction enzyme site on both the sides as overhangs with 5′-phosphate group. After ligation into pET32b-1XThy1-scFv vector previously digested with BglII restriction enzyme and dephosphorylated using Calf intestine alkaline phosphatase, we generated two additional vectors with 3X and 5X hexa-histidine-tags. All three plasmids contain: a T7 promotor; two fusion partners Trx-and S-tags for enhancing protein folding and solubility; 1, 3 or 5 hexa-histidine tag(s) for purification by immobilized metal affinity chromatography (IMAC); a DDDDK sequence on the N terminus of Thy1-scFv for tag removal using EK cleavage; and the Thy1-scFv gene. Each histidine tag was separated from each other by a few amino acid residues to increase flexible folding. The sequence confirmed, vectors were transformed into SHuffle T7 E. coli cells (New England Biolabs, Ipswich, MA) for recombinant protein expression. Oligonucleotides and recombinant protein sequences used in this study for constructing the vectors are listed in Supporting Information (Supplementary Table S1).
Thy1-scFv expression. Bacterial transformation was performed for each expression vector as follows: 50 μL of SHuffle T7 E. coli competent cells were transformed with 1 μg of expression plasmid using standard heatshock procedure; 300 μL SOC growth media was then added in each vial for cell recovery. Cells were grown at 30 °C for 1 h in a shaking incubator (100 r.p.m.), and plated on Lysogeny Broth (LB)-agar medium containing ampicillin (50 μg/mL). After overnight growth, one fresh-picked colony was inoculated in 2 mL LB-ampicillin medium (50 μg/mL) and grown overnight at 30 °C (250 r.p.m.). Bacteria were transferred into 1 L of LB-ampicillin medium and further cultured until the OD 600nm reaches 0.4. The culture was induced for protein expression by the addition of isopropyl-β-d-thiogalactoside (IPTG, 1 mM) after diluting with the addition of one-fourth volume of pre-warmed LB-ampicillin medium. Induction was allowed for 4 h at 30 °C (250 r.p.m.). The pellet was then harvested via centrifugation (5000g, 10 min, 4 °C) and stored at − 80 °C.

Fusion protein purification (IMAC1).
Cell pellets were resuspended in 20 mL of ice-cold lysis buffer (3 mM monosodium phosphate, 50 mM disodium phosphate, 500 mM NaCl, 5% glycerol (v/v), 5 mM CHAPS, and 20 mM imidazole containing protease inhibitors (Thermo Scientific, Rockford, IL)) and lysed by sonication (60% amplitude, 5 s on/off, 10 cycles, Branson SLPe). The soluble and insoluble fractions were then separated by centrifugation (12,000g, 10 min, 4 °C). Insoluble fractions containing cell debris and possible IBs were washed with 8 mL of the same lysis buffer. The soluble fractions were applied to a 1 mL FF His-trap column (GE www.nature.com/scientificreports/ Healthcare Biosciences, PA) in an AKTA FPLC system (GE Healthcare Biosciences) equilibrated with PBS buffer containing 20 mM imidazole to reduce non-specific binding. The tagged-Thy1-scFv was purified using a linear gradient of imidazole (from 20 to 200 mM) in PBS buffer at a flow rate of 1 mL/min. Concentration of proteins were measured by UV spectrometry in each fraction. Purity of the fractions was analyzed on 4-12% gradient SDS-PAGE followed by staining with Coomassie Blue (SimplyBlue SafeStain, Carlsbad, CA) for visualization using a BioRad Gel-Doc system. Fractions containing the protein of interest were pooled and the best expression vector was utilized for further experiments. Western blot analysis was performed using anti-His tag antibody. A 4-12% gradient SDS-PAGE of the insoluble and soluble fractions was electroblotted onto a 0.2 µm pore size nitrocellulose membrane (Bio Rad, Hercules, CA), blocked in PBS-T (PBS with 0.05% Tween 20) with 5% milk powder for overnight at 4 °C and then treated with anti-His tag antibody (BioLegend, San Diego, CA). After washing, the membrane was incubated with HRP-conjugated anti-mouse IgG antibody. Signals were visualized by the addition of enhanced-chemiluminescence (ECL) substrate and imaging using IVIS in vivo imaging system (Perkin Elmer, Santa Clara, CA).
Fusion protein cleavage by enterokinase. The  In vivo US molecular imaging of pancreas. In vivo US imaging of vascular Thy1 expression in transgenic PDAC mice and C57BL/6 mice with normal pancreas was performed using two MB constructs (MB Thy1-scFv and MB non-targeted ) by following the protocol reported previously 24 . In brief, a total of 10 8 MB Thy1-scFv or MB non-targeted (100 µL) was utilized for intravenous bolus injection via tail vein. All in vivo imaging studies were performed in contrast mode using a dedicated small animal high resolution US imaging system (Vevo 2100, FUJIFILM Visu-alSonics, Inc., Toronto, ON, Canada) with a linear transducer (MS250, VisualSonics) placed over the abdomen www.nature.com/scientificreports/ of mice, guided by B-mode imaging to detect the target tissue of interest. Contrast mode images were acquired at 18 MHz, and all imaging parameters (focal length, 10 mm; transmit power, 4%; mechanical index, 0.2; dynamic range, 40 dB) were kept constant during all imaging sessions. A total time of 5 min was allowed for MBs to attach their target before binding quantification. To differentiate the acoustic signal owing to MBs attachment to Thy1 and the signal from freely circulating MBs, the previously described destruction-replenishment technique was employed 22 . The protocol consisted of 3 steps: (1) 200 frames of images capturing blood-vessel bound and unbound MBs within the ROI, (2) a high pressure destructive pulse (1-s continuous high-power destructive pulse of 3.7 MPa, transmit power, 100%; mechanical index, 0.63) to destroy all bound and unbound MBs, and (3) an additional set of 200 frames to measure the signal magnitude from the unbound MBs flowing into the ROI immediately after the destructive pulse. The difference in US imaging signal pre-and post-destruction corresponds to the Thy1 attached contrast agents, MB Thy1-scFv or MB non-targeted . A waiting interval of 20 min was maintained between each MB injection to allow for complete clearance before subsequent imaging. Any remaining attached MBs were destroyed by applying a high-power destruction pulse (see above for acoustic parameters).
Ultrasound molecular imaging data analysis. The molecular imaging signals were quantified post image acquisition with correction for breathing motion artifacts using Vevo 2100 integrated analysis software (VevoCQ; VisualSonics). Data analysis was accomplished by manually drawing ROIs around PDAC tissues, adjacent non-PDAC tissues, as well as in the normal pancreas of control littermates. The magnitude of imaging signal from attached MBs was assessed by subtracting the average imaging signals pre-and post-destruction and expressed as the differential targeted enhancement (dTE) in arbitrary units (a.u.).
Ex vivo analysis of pancreas tissues. PDAC mice were euthanized in accordance with animal care guidelines. The pancreas was excised and fixed in 4% paraformaldehyde (Santa Cruz Biotechnology Inc., CA) at 4 °C for 24 h. Tissues were cryosectioned and stained with hematoxylin eosin before analysis using a Nanozoomer (Hamamatsu, Japan).