Production and Purication of A Cell-Penetrating Peptide-Erythropoietin Fusion Protein and Optimization of an in Vitro Blood-Brain Barrier Model for Central Nervous System Drug Delivery

Drug delivery into the central nervous system (CNS) is a brilliant research eld, and the development of protein production and purication procedures for novel therapeutic proteins is crucial. Erythropoietin (EPO) is a glycoprotein with tremendous neuroprotective potential, but its bulky size prevents easy penetration across the blood-brain barrier (BBB). EPO-HBHAc is a promising cell-penetrating peptide modied protein for CNS diseases, necessitating an appropriate in vitro BBB model for further evaluation. The plasmid of EPO-HBHAc was constructed by DNA recombinant technology, and the Chinese Hamster Ovary (CHO-K1) cell expression system was selected to generate target proteins. His-tag and size exclusion purication were used to purify the target protein from the cell-conditioned medium; target proteins were further evaluated by western blotting and Coomassie blue staining. Moreover, the endothelial cells (bEnd.3) and astrocytes (CTX TNA2) were used to generate the in vitro BBB model, and transepithelial electrical resistance (TEER) and paracellular diffusion were measured to evaluate barrier integrity. The EPO-HBHAc plasmid was successfully constructed, and a stable cell line expressing EPO-HBHAc was generated. A higher protein expression level was observed in serum-containing medium than in serum-free medium. His-tag purication is not sucient to remove impurities from target proteins, and thus size exclusion purication was performed to increase the purity of the protein of interest. In contrast, a higher TEER value and lower paracellular diffusion were observed in the co-culture model than in the mono-culture model. Furthermore, the higher TEER value was observed in inserts with a larger growth area (4.67 cm 2 ) than in those with a smaller area (0.33 cm 2 ). In conclusion, we demonstrated that some critical points might impact protein production and the in vitro BBB model construction in this study. our research will in eld of CNS drug


Introduction
Neurological disorders are reportedly leading causes of disability-adjusted life-years (DALYs), with the prevalence of these diseases steeply increasing owing to an aging population [1]. Globally, neurological disorders have revealed a signi cant impact on both the economy and society. Thus, the development of therapeutics for central nervous system (CNS) disorders is urgent. Macromolecular drugs such as proteins and peptides demonstrate higher e cacy when compared with small molecular drugs, as well as present good e ciency and low toxicity owing to their target speci city [2]. Furthermore, owing to advancements in biotechnology, large-scale production of biologics is now increasingly cost-effective.
Additionally, according to a report by the U.S. Food and Drug Administration (FDA), the number of approved biologics has been growing in recent years. In 2018, 17 biologics were approved, representing a record that exceeds the number of approved biologics in previous years. Moreover, these biologics account for more than 25% of all approved drugs in the last ve years [3]. Hence, the growth potential of the therapeutic protein market is tremendous.
Erythropoietin (EPO) is a glycoprotein that has been used for the treatment of different types of anemia for decades. Apart from the erythropoiesis activity of EPO, several studies have reported the neuroprotective functions of this protein, eliciting protective effects in neurodegenerative and ischemic brain diseases [4][5][6][7]. Owing to its bulky size, EPO does not easily penetrate the blood-brain barrier (BBB), and an extremely high EPO dose is needed to attain a therapeutic concentration in the brain tissue to demonstrate a neuroprotective effect. In our previous publication, we reported that the BBB permeability of EPO was enhanced using heparin-binding hemagglutinin adhesion c (HBHAc), a cell-penetration peptide (CPP), modi cation [8].
Chinese hamster ovary (CHO-K1) cells are the primary expression system for therapeutic proteins [9] and were selected to produce EPO-HBHAc. To obtain high-quality proteins, the choice of the culture medium is a critical point in protein production. Although serum-free medium is widely used for FDA-approved products [10], the impact of serum-starvation on cell viability and protein expression remains unpredictable and depends on the cell type [11]. Moreover, the optimized puri cation process is an essential step. For puri cation of this novel CPP-modi ed protein, poly-histidine tags (His-tags) were used, coupled to a Nickle-NTA matrix. The protein can be eluted under non-denaturing conditions by adding an adequate concentration of imidazole to the elution buffer. The selected His-tag location is based on the protein of interest and is related to the post-translational modi cation of proteins, including removal of the N-terminal signal peptide [12]. Moreover, a speci c protease site added between the Histag motif and protein of interest allows the tag removal from the protein.
Conversely, the construction of the in vitro BBB model is important to develop an appropriate CNS drug delivery system. Currently, brain endothelial cells are cultured alone, co-cultured with astrocytes, or triple co-cultured with astrocytes and pericytes [13]. The advantages of transwell systems include the relative ease of setup and control. Transepithelial electrical resistance (TEER) and permeability measurements of these models provide reliable quantitative evaluations of barrier integrity [14].
The development of novel therapeutic proteins, including EPO-HBHAc, remains a complex task, and the production process requires step by step evaluation. Thus, in the present study, we demonstrated the procedure of EPO-HBHAc construction and generated an optimized in vitro BBB model.

Cell cultures
CHO-K1 cells were maintained in Ham's F12 medium, and bEnd.3 cells (mouse brain endothelial cells) and CTX TNA2 cells (rat astrocytes) were maintained in DMEM (Dulbecco's Modi ed Eagle's medium). All media were supplemented with 10% fetal bovine serum (FBS), 50 U/mL penicillin, and 50 µg/mL streptomycin. Cells were cultured at 37 °C and in 5% CO 2 , and the subculturing procedure was performed according to American Type Culture Collection (ATCC).

Plasmid construction and stable clone selection of EPO-HBHAc
The process of plasmid construction was described in our previous study [8]. The precise plasmid encoding EPO or EPO-HBHAc was transfected into CHO-K1 cells using Lipofectamine 3000 reagent. After overnight incubation, the transfected cells were divided between 10-cm dishes and incubated for an additional 24 h. G418 was used for stable clone cell line selection, and cells with the neomycin resistance gene survived and formed colonies. Then, cells were cultured with a medium containing 1 mg/mL G418, and the medium was replaced every 2 days for 2 to 4 weeks. Until cell colony formation, the colonies were transferred into a 6-well plate (one colony/well). On reaching approximately 80% con uency, the cells were expanded. The target proteins expressed in both the cell culture supernatant and cell lysate were analyzed by western blotting with an anti-EPO antibody.

Optimization of protein expression and puri cation
To evaluate the protein expression pro le, CHO-K1 cells with EPO or EPO-HBHAc gene were seeded in a 6well plate at a density of 5 × 10 5 cells/well and incubated overnight. Next, the cells were incubated in Opti-MEM I reduced serum medium, and a medium, with 2% FBS or without FBS, was used to evaluate the effect of FBS on target protein expression. To determine the protein expression level at different incubation time intervals, the cell culture supernatant was collected after 2, 3, and 4 days incubation. Protein expression levels in different culture intervals were evaluated by western blotting using the EPO antibody.
For the large-scale production of fusion proteins, CHO-K1 cells were divided into 15-cm dishes, with the cell culture supernatant collected every 4 days. The aggregates and cell pellets were removed using a 0.22 µm lter cup and concentrated using a Tangential Flow Filtration (TFF) system (Millipore, USA). The concentrated supernatant was mixed with HisPur Ni-NTA Resin, and then the mixture was transferred to a column. The column packed with the protein-resin mixture was equilibrated with an equilibration buffer (20 mM sodium phosphate, 300 mM NaCl, 10 mM imidazole, pH 7.4). To determine the appropriate concentration of imidazole in the wash and elution steps, the column was washed with a gradient concentration of imidazole from 25 to 100 mM, and the collected fractions were analyzed by western blotting. Finally, the target proteins were puri ed under the optimized Ni-NTA puri cation process. During the elution step, fractions were collected in 1 mL/tube and analyzed by western blotting. To increase the purity of target proteins after Ni-NTA puri cation, the fractions containing target proteins were collected and further puri ed using the size exclusion column. The puri ed recombinant fusion proteins were con rmed by Coomassie blue staining and western blotting.

In vitro BBB model construct
For the brain endothelial cell mono-culture BBB model, bEnd.3 cells were seeded on the upper surface of the insert (cell growth area: 4.67 cm 2 ) and cultured for several days. In contrast, for the brain endothelial cell and astrocyte co-culture BBB model, CTX TNA2 cells were seeded on the lower surface of the insert and allowed to adhere overnight. The insert was ipped back and CTX TNA2 cells were cultured for an additional two days. Then, bEnd.3 cells were seeded on the upper surface of the insert and co-cultured with CTX TNA2 cells. The TEER value of each well was measured daily using the Millicell ERS Voltohmmeter (Millipore, MA, USA). Furthermore, to evaluate paracellular diffusion for both the monoculture and co-culture BBB models, FITC-4K-dextran and FITC-70K-dextran were used as tracers. FITC-4K-dextran or FITC-70K-dextran was added to the BBB models when the TEER value reached the plateau, and the basolateral medium was collected 3, 6, 12, and 24 h after dosing. The uorescence intensity in the collected medium was detected using a uorescence microplate reader. To compare the effect of the growth area of the insert on the BBB model, bEnd.3 and CTX TNA2 cells were additionally cocultured on the insert with smaller growth areas (0.33 cm 2 ), using the previously described cell culture process. Fluorescence-labeled-70K-dextran was added to the model as the TEER value reached the plateau, and the uorescence intensity in the basolateral chamber was detected 2 h after dosing.

Statistics
All values are presented as mean ± SD. A signi cant difference was evaluated by ANOVA, followed by the Bonferroni modi ed t-test. A p-value of < 0.05 was considered statistically signi cant.

Stable clone selection
After con rming the correctness of constructed plasmids, the plasmids containing the gene of EPO and EPO-HBHAc were transfected into CHO-K1 cells, respectively. The colonies were formed and selected from the cells transfected with speci c plasmids after G418 selection, and the target protein expression in cell culture supernatant and cell lysate were detected by western blotting. Proteins with the precise molecular weight were observed in the cell culture supernatant of both the EPO (Fig. 2 Land 2) and EPO-HBHAc groups ( Fig. 2 Lane 4). Low molecular products were detected in the cell lysate of EPO (Fig. 2 Land 1) and EPO-HBHAc groups (Fig. 2 Lane 3). This result indicated that the plasmids with the target gene were successfully transfected into CHO-K1 cells, and the target proteins were mainly released into the cell culture medium.

Large-scale protein expression and puri cation
As shown in Fig. 3A, the intensity of the target protein increased with increasing incubation intervals in both 2% FBS and FBS-free groups. Furthermore, the highest EPO expression level was observed in cells treated with serum-free medium over a 4-day incubation (Fig. 3A Lane 7). However, in the FBS-free cultured group, the protein expression level decreased after the replacement of fresh medium for an additional 4 days of incubation ( Fig. 3A Lane 8), demonstrating weak cell adherence. Consistent protein expression was observed in the 2% FBS cultured group after replacement with fresh medium for an additional 4 days incubation (Fig. 3A Lane 4). Conversely, the highest level of EPO-HBHAc fusion protein expression was detected in the cell group incubated in medium supplied with 2% FBS for 4 days (Fig. 3B  Lane 4), and the proteins were consistently expressed even after replacement with fresh medium for an additional 4 days of incubation ( Fig. 3B Lane 8).
For large-scale of production of the target protein, the transfected CHO-K1 cells were divided between 15cm cell culture dishes and incubated with the medium containing 2% FBS. According to the evaluated protein expression pro le, the cell culture supernatant containing target proteins was collected every 4 days. The collected supernatant was ltered and the volume of ltered supernatant was reduced using the TFF system before puri cation using the Ni-NTA column. To determine the suitable concentration of imidazole for the wash process, gradient concentrations from 25 to 100 mM imidazole were used. The chromatogram is shown in Fig. 4A and selected fractions were further evaluated by western blotting. Only small amounts of protein were detected in the equilibration step and at the beginning of wash step ( imidazole. Therefore, the buffer containing 30 mM of imidazole was used for the wash step during puri cation. The Ni-NTA chromatogram is shown in Fig. 5A and collected fractions were evaluated by Coomassie blue staining. Although several impurities could be removed by washing with 30 mM of imidazole, the purity of fusion proteins remained low after Ni-NTA puri cation (Fig. 5B). EPO fusion proteins were further puri ed using the size exclusion column, and major impurities were separated from our target protein (Fig. 6). The puri cation process for EPO-HBHAc fusion proteins followed the same procedure of EPO fusion proteins, and the chromatography of Ni-NTA and size exclusion puri cation are shown in Fig. 7A and 7B, respectively. Unfortunately, the major impurities could not be precisely separated from EPO-HBHAc owing to the small molecular weight difference between them (Fig. 7C Lane 4 to Lane 6). However, the purity of EPO-HBHAc improved after puri cation by size exclusion column. The purity of the puri ed proteins was analyzed by Coomassie blue staining. EPO and EPO-HBHAc demonstrated a purity of > 80% and > 60%, respectively.

In vitro BBB model construct
To establish a suitable in vitro BBB model, the endothelial cell mono-culture and endothelial cell and astrocyte co-culture models were evaluated. The TEER values 7 days after bEnd.3 cells were seeded are shown in Fig. 8A. The co-culture model demonstrated higher TEER values than the mono-culture model (208 ± 4.6 Ω*cm 2 vs. 150 ± 9.1 Ω*cm 2 ). Additionally, the paracellular diffusion assay was performed on day 7 after bEnd.3 cells were seeded, and FITC-4K-dextran and FITC-70K-dextran were used as tracers. As shown in Fig. 8B, 51.1 ± 0.5% and 37.5 ± 1.7% of FITC-4K-dextran were transported from the upper chamber to the basolateral chamber in the mono-culture model and co-culture model after 24 h of incubation, respectively. In contrast, only 7.3 ± 1.1% and 2.8 ± 0.1% of FITC-70K-dextran was transported from the upper chamber to the basolateral chamber in the mono-culture model and co-culture model after 24 h of incubation, respectively. In the co-culture model, lower paracellular diffusion was observed for both 4K and 70K dextran, with higher TEER values. These results indicated that the CTX TNA2 and bEnd.3 co-culture model would be a superior BBB model to evaluate EPO-HBHAc transcytosis in vitro. Furthermore, bEnd.3 cells and CTX TNA2 cells were co-cultured on the smaller inserts (growth area: 0.33 cm 2 ), and the TEER value is shown in Fig. 9. The TEER value reached the plateau 4 days after bEnd.3 cells were seeded. The uorescence intensity of 70K-dextran in the basolateral chamber was detected 2 h after dosing, and 14.9 ± 2.2% of 70K-dextran was transported from the upper chamber to the basolateral chamber in the co-culture model (0.33 cm 2 ).

Discussion
CHO-K1 cells were chosen to generate EPO and EPO-HBHAc fusion proteins, and stable clones expressing EPO and EPO-HBHAc were selected and cryopreserved. Furthermore, the target proteins were puri ed under a two-step puri cation process. Chemical conjugation is another approach to generate EPO-HBHAc; however, it is challenging to control the speci c site of conjugation between EPO and HBHAc. A heterogeneous mixture would be generated, and EPO bioactivity could be affected if critical sites on EPO were occupied by HBHAc. The production of proteins in mammalian cells is an important tool not only for basic research but also in the biotech industry. Mammalian proteins, including tissue plasminogen activator and EPO, require a mammalian cell production system for better bioactivity, and this could be attributed to the ability of mammalian cells to generate proteins with appropriate molecular structures and biochemical properties [15]. Furthermore, compared with transient transfection, stable expression of the transgene is generally more desirable for large-scale production, demonstrating higher protein quality and homogeneity [16]. The adherent expression system is a traditional method for protein expression and was selected to generate EPO-HBHAc on a bench-scale in this study. However, the surface area of the ask might limit the yield of target proteins [17]. This is a proof-of-concept study, evaluating the protein production and puri cation process for a novel CPP-modi ed protein, and bench-scale production was su cient for this purpose. However, a scale-up protein expression system for target proteins is necessary for further in vivo evaluations, with the suspension-adapted cell culture system providing an alternative to increasing the yield of this novel CPP-modi ed protein. In previous study investigating EPO production, the conditioned medium was collected every 2 days [18]; however, in the present study, the highest protein expression was observed after 4 days of incubation. Protein expression levels can vary among the different groups, and thus, we suggest that evaluating the protein expression pro le is crucial when setting up a new protein production platform.
In our in vitro co-culture BBB model, the TEER value was signi cantly higher than that observed in the mono-culture model, and the penetration of 70K-dextran was signi cantly blocked from the upper chamber to the basolateral chamber. It remains a challenge to design an optimized in vitro experimental model to mimic the physiological and functional characteristics of the BBB. High junctional tightness measured as TEER is an important feature for an appropriate model mimicking the BBB in vitro. However, the optimal value of TEER for experiments could vary when obtained from different studies. This could be attributed to differences in measuring equipment and temperature, as well as the handling of the cells [14]. Hence, the tightness of in vitro BBB models was validated using permeability studies with hydrophilic tracers [19].
Additionally, our mono-culture model showed lower TEER values and higher paracellular diffusion when compared with the co-culture model, which is consistent with previous studies. Owing to inadequate tight junctions in endothelial cell mono-culture models, several research groups have attempted to reduce paracellular diffusion in these mono-culture models. One approach is to co-culture the endothelial cells with astrocytes [20,21]. Previous studies have reported that cultured astrocytes implanted into areas with normally leaky vessels were able to induce the tightening of the endothelium, demonstrating that astrocytes play a major role in inducing barrier properties [22]. To better mimic the physiological structure of the BBB, endothelial cells are co-cultured with astrocytes, and the interaction between endothelial cells and astrocytes increases the expression of transporters, as well as that of tight junctions in endothelial cells. Furthermore, this interaction induces the formation of cell polarity in endothelial cells. Collectively, these advantages reveal that the endothelium-astrocyte co-culture model is more representative of the BBB [23][24][25][26][27].
A large amount of 4K-dextran was transported from the upper chamber to the basolateral chamber in both the monoculture and co-culture BBB models. This demonstrates that our models failed to afford su cient barrier properties to small peptides/proteins with a molecular weight of approximately 4,000 Dalton. In contrast, only a limited amount of 70K-dextran was transported from the upper chamber to the basolateral chamber in both the monoculture and co-culture BBB models. Based on our results, the extents of EPO and EPO-HBHAc transportation differed in the co-culture model, suggesting that our co-culture model could be applied in BBB penetrating investigations for substances with a molecular weight larger than that of EPO.

Conclusion
The optimized protein production and puri cation for EPO-HBHAc were successfully established, and the protein expression level was affected by the composition of the culture medium and incubation period. Conversely, higher barrier integrity was observed in the co-culture BBB model than in the mono-culture model, and the integrity was affected by the growth area of the insert. This comprehensive study revealed some critical factors for CPP-modi ed protein generation and BBB model construction, providing valuable information in the eld of CNS drug delivery.